wine/wine-staging
0
0
Fork 0
mirror of https://gitlab.winehq.org/wine/wine-staging.git synced 2024-11-21 16:46:54 -08:00
Code Issues Packages Projects Releases Wiki Activity
2a2c8b5228
wine-staging/patches/vkd3d-latest/0002-Updated-vkd3d-to-2a3413e0f01524f2068bce12100906eb220.patch
Alistair Leslie-Hughes dab36ebe1e Updated vkd3d-latest patchset
2023-08-03 12:06:16 +10:00

9441 lines
357 KiB
Diff
Raw Blame History

From ef302da9a91d5d7d4456506303ce3f3964f9081f Mon Sep 17 00:00:00 2001
From: Alistair Leslie-Hughes <leslie_alistair@hotmail.com>
Date: Wed, 28 Jun 2023 16:27:03 +1000
Subject: [PATCH 2/3] Updated vkd3d to
2a3413e0f01524f2068bce12100906eb2200c965.
---
include/d3d12.idl | 4 +-
libs/vkd3d/Makefile.in | 1 +
libs/vkd3d/include/private/vkd3d_common.h | 7 +
.../include/private/vkd3d_shader_utils.h | 63 +
libs/vkd3d/include/vkd3d.h | 35 +
libs/vkd3d/include/vkd3d_shader.h | 116 +-
libs/vkd3d/libs/vkd3d-common/debug.c | 17 +-
libs/vkd3d/libs/vkd3d-shader/d3d_asm.c | 48 +-
libs/vkd3d/libs/vkd3d-shader/d3dbc.c | 311 ++-
libs/vkd3d/libs/vkd3d-shader/dxbc.c | 16 +-
libs/vkd3d/libs/vkd3d-shader/dxil.c | 2313 +++++++++++++++++
libs/vkd3d/libs/vkd3d-shader/hlsl.c | 161 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.h | 32 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.y | 1323 ++++++----
libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c | 384 ++-
.../libs/vkd3d-shader/hlsl_constant_ops.c | 363 ++-
libs/vkd3d/libs/vkd3d-shader/spirv.c | 26 +-
libs/vkd3d/libs/vkd3d-shader/tpf.c | 372 ++-
.../libs/vkd3d-shader/vkd3d_shader_main.c | 164 +-
.../libs/vkd3d-shader/vkd3d_shader_private.h | 29 +
libs/vkd3d/libs/vkd3d/command.c | 222 +-
libs/vkd3d/libs/vkd3d/device.c | 2 +
libs/vkd3d/libs/vkd3d/resource.c | 51 +-
libs/vkd3d/libs/vkd3d/state.c | 10 +-
libs/vkd3d/libs/vkd3d/vkd3d_private.h | 47 +-
25 files changed, 5088 insertions(+), 1029 deletions(-)
create mode 100644 libs/vkd3d/include/private/vkd3d_shader_utils.h
create mode 100644 libs/vkd3d/libs/vkd3d-shader/dxil.c
diff --git a/include/d3d12.idl b/include/d3d12.idl
index 4fec32d2656..c6064939e1f 100644
--- a/include/d3d12.idl
+++ b/include/d3d12.idl
@@ -2243,8 +2243,8 @@ interface ID3D12CommandQueue : ID3D12Pageable
ID3D12Heap *heap,
UINT range_count,
const D3D12_TILE_RANGE_FLAGS *range_flags,
- UINT *heap_range_offsets,
- UINT *range_tile_counts,
+ const UINT *heap_range_offsets,
+ const UINT *range_tile_counts,
D3D12_TILE_MAPPING_FLAGS flags);
void CopyTileMappings(ID3D12Resource *dst_resource,
diff --git a/libs/vkd3d/Makefile.in b/libs/vkd3d/Makefile.in
index 1ba0e9f71e1..f647af11d07 100644
--- a/libs/vkd3d/Makefile.in
+++ b/libs/vkd3d/Makefile.in
@@ -17,6 +17,7 @@ SOURCES = \
libs/vkd3d-shader/d3d_asm.c \
libs/vkd3d-shader/d3dbc.c \
libs/vkd3d-shader/dxbc.c \
+ libs/vkd3d-shader/dxil.c \
libs/vkd3d-shader/glsl.c \
libs/vkd3d-shader/hlsl.c \
libs/vkd3d-shader/hlsl.l \
diff --git a/libs/vkd3d/include/private/vkd3d_common.h b/libs/vkd3d/include/private/vkd3d_common.h
index 1ac23b4a085..0263fc47297 100644
--- a/libs/vkd3d/include/private/vkd3d_common.h
+++ b/libs/vkd3d/include/private/vkd3d_common.h
@@ -20,6 +20,7 @@
#define __VKD3D_COMMON_H
#include "config.h"
+#define WIN32_LEAN_AND_MEAN
#include "windows.h"
#include "vkd3d_types.h"
@@ -28,6 +29,7 @@
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
+#include <stdlib.h>
#ifdef _MSC_VER
#include <intrin.h>
@@ -171,6 +173,11 @@ static inline bool vkd3d_bound_range(size_t start, size_t count, size_t limit)
#endif
}
+static inline bool vkd3d_object_range_overflow(size_t start, size_t count, size_t size)
+{
+ return (~(size_t)0 - start) / size < count;
+}
+
static inline uint16_t vkd3d_make_u16(uint8_t low, uint8_t high)
{
return low | ((uint16_t)high << 8);
diff --git a/libs/vkd3d/include/private/vkd3d_shader_utils.h b/libs/vkd3d/include/private/vkd3d_shader_utils.h
new file mode 100644
index 00000000000..00052a89988
--- /dev/null
+++ b/libs/vkd3d/include/private/vkd3d_shader_utils.h
@@ -0,0 +1,63 @@
+/*
+ * Copyright 2023 Conor McCarthy for CodeWeavers
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
+ */
+
+#ifndef __VKD3D_SHADER_UTILS_H
+#define __VKD3D_SHADER_UTILS_H
+
+#include "vkd3d_shader.h"
+
+#define TAG_DXIL VKD3D_MAKE_TAG('D', 'X', 'I', 'L')
+#define TAG_SHDR VKD3D_MAKE_TAG('S', 'H', 'D', 'R')
+#define TAG_SHEX VKD3D_MAKE_TAG('S', 'H', 'E', 'X')
+
+static inline enum vkd3d_result vkd3d_shader_parse_dxbc_source_type(const struct vkd3d_shader_code *dxbc,
+ enum vkd3d_shader_source_type *type, char **messages)
+{
+ struct vkd3d_shader_dxbc_desc desc;
+ enum vkd3d_result ret;
+ unsigned int i;
+
+ *type = VKD3D_SHADER_SOURCE_NONE;
+
+ if ((ret = vkd3d_shader_parse_dxbc(dxbc, 0, &desc, messages)) < 0)
+ return ret;
+
+ for (i = 0; i < desc.section_count; ++i)
+ {
+ uint32_t tag = desc.sections[i].tag;
+ if (tag == TAG_SHDR || tag == TAG_SHEX)
+ {
+ *type = VKD3D_SHADER_SOURCE_DXBC_TPF;
+ }
+ else if (tag == TAG_DXIL)
+ {
+ *type = VKD3D_SHADER_SOURCE_DXBC_DXIL;
+ /* Default to DXIL if both are present. */
+ break;
+ }
+ }
+
+ vkd3d_shader_free_dxbc(&desc);
+
+ if (*type == VKD3D_SHADER_SOURCE_NONE)
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ return VKD3D_OK;
+}
+
+#endif /* __VKD3D_SHADER_UTILS_H */
diff --git a/libs/vkd3d/include/vkd3d.h b/libs/vkd3d/include/vkd3d.h
index 72ed3ced671..2ccda47248a 100644
--- a/libs/vkd3d/include/vkd3d.h
+++ b/libs/vkd3d/include/vkd3d.h
@@ -207,7 +207,42 @@ VKD3D_API VkPhysicalDevice vkd3d_get_vk_physical_device(ID3D12Device *device);
VKD3D_API struct vkd3d_instance *vkd3d_instance_from_device(ID3D12Device *device);
VKD3D_API uint32_t vkd3d_get_vk_queue_family_index(ID3D12CommandQueue *queue);
+
+/**
+ * Acquire the Vulkan queue backing a command queue.
+ *
+ * While a queue is acquired by the client, it is locked so that
+ * neither the vkd3d library nor other threads can submit work to
+ * it. For that reason it should be released as soon as possible with
+ * vkd3d_release_vk_queue(). The lock is not reentrant, so the same
+ * queue must not be acquired more than once by the same thread.
+ *
+ * Work submitted through the Direct3D 12 API exposed by vkd3d is not
+ * always immediately submitted to the Vulkan queue; sometimes it is
+ * kept in another internal queue, which might not necessarily be
+ * empty at the time vkd3d_acquire_vk_queue() is called. For this
+ * reason, work submitted directly to the Vulkan queue might appear to
+ * the Vulkan driver as being submitted before other work submitted
+ * though the Direct3D 12 API. If this is not desired, it is
+ * recommended to synchronize work submission using an ID3D12Fence
+ * object, by submitting to the queue a signal operation after all the
+ * Direct3D 12 work is submitted and waiting for it before calling
+ * vkd3d_acquire_vk_queue().
+ *
+ * \since 1.0
+ */
VKD3D_API VkQueue vkd3d_acquire_vk_queue(ID3D12CommandQueue *queue);
+
+/**
+ * Release the Vulkan queue backing a command queue.
+ *
+ * This must be paired to an earlier corresponding
+ * vkd3d_acquire_vk_queue(). After this function is called, the Vulkan
+ * queue returned by vkd3d_acquire_vk_queue() must not be used any
+ * more.
+ *
+ * \since 1.0
+ */
VKD3D_API void vkd3d_release_vk_queue(ID3D12CommandQueue *queue);
VKD3D_API HRESULT vkd3d_create_image_resource(ID3D12Device *device,
diff --git a/libs/vkd3d/include/vkd3d_shader.h b/libs/vkd3d/include/vkd3d_shader.h
index 274241546ea..6c17a07b9d2 100644
--- a/libs/vkd3d/include/vkd3d_shader.h
+++ b/libs/vkd3d/include/vkd3d_shader.h
@@ -85,6 +85,11 @@ enum vkd3d_shader_structure_type
* \since 1.3
*/
VKD3D_SHADER_STRUCTURE_TYPE_DESCRIPTOR_OFFSET_INFO,
+ /**
+ * The structure is a vkd3d_shader_scan_signature_info structure.
+ * \since 1.9
+ */
+ VKD3D_SHADER_STRUCTURE_TYPE_SCAN_SIGNATURE_INFO,
VKD3D_FORCE_32_BIT_ENUM(VKD3D_SHADER_STRUCTURE_TYPE),
};
@@ -611,6 +616,11 @@ enum vkd3d_shader_source_type
* model 1, 2, and 3 shaders. \since 1.3
*/
VKD3D_SHADER_SOURCE_D3D_BYTECODE,
+ /**
+ * A 'DirectX Intermediate Language' shader embedded in a DXBC container. This is
+ * the format used for Direct3D shader model 6 shaders. \since 1.9
+ */
+ VKD3D_SHADER_SOURCE_DXBC_DXIL,
VKD3D_FORCE_32_BIT_ENUM(VKD3D_SHADER_SOURCE_TYPE),
};
@@ -620,7 +630,7 @@ enum vkd3d_shader_target_type
{
/**
* The shader has no type or is to be ignored. This is not a valid value
- * for vkd3d_shader_compile() or vkd3d_shader_scan().
+ * for vkd3d_shader_compile().
*/
VKD3D_SHADER_TARGET_NONE,
/**
@@ -1551,6 +1561,64 @@ static inline uint32_t vkd3d_shader_create_swizzle(enum vkd3d_shader_swizzle_com
| ((w & VKD3D_SHADER_SWIZZLE_MASK) << VKD3D_SHADER_SWIZZLE_SHIFT(3));
}
+/**
+ * A chained structure containing descriptions of shader inputs and outputs.
+ *
+ * This structure is currently implemented only for DXBC and legacy D3D bytecode
+ * source types.
+ * For DXBC shaders, the returned information is parsed directly from the
+ * signatures embedded in the DXBC shader.
+ * For legacy D3D shaders, the returned information is synthesized based on
+ * registers declared or used by shader instructions.
+ * For all other shader types, the structure is zeroed.
+ *
+ * All members (except for \ref type and \ref next) are output-only.
+ *
+ * This structure is passed to vkd3d_shader_scan() and extends
+ * vkd3d_shader_compile_info.
+ *
+ * Members of this structure are allocated by vkd3d-shader and should be freed
+ * with vkd3d_shader_free_scan_signature_info() when no longer needed.
+ *
+ * All signatures may contain pointers into the input shader, and should only
+ * be accessed while the input shader remains valid.
+ *
+ * Signature elements are synthesized from legacy Direct3D bytecode as follows:
+ * - The \ref vkd3d_shader_signature_element.semantic_name field is set to an
+ * uppercase string corresponding to the HLSL name for the usage, e.g.
+ * "POSITION", "BLENDWEIGHT", "COLOR", "PSIZE", etc.
+ * - The \ref vkd3d_shader_signature_element.semantic_index field is set to the
+ * usage index.
+ * - The \ref vkd3d_shader_signature_element.stream_index is always 0.
+ *
+ * Signature elements are synthesized for any input or output register declared
+ * or used in a legacy Direct3D bytecode shader, including the following:
+ * - Shader model 1 and 2 colour and texture coordinate registers.
+ * - The shader model 1 pixel shader output register.
+ * - Shader model 1 and 2 vertex shader output registers (position, fog, and
+ * point size).
+ * - Shader model 3 pixel shader system value input registers (pixel position
+ * and face).
+ *
+ * \since 1.9
+ */
+struct vkd3d_shader_scan_signature_info
+{
+ /** Must be set to VKD3D_SHADER_STRUCTURE_TYPE_SCAN_SIGNATURE_INFO. */
+ enum vkd3d_shader_structure_type type;
+ /** Optional pointer to a structure containing further parameters. */
+ const void *next;
+
+ /** The shader input varyings. */
+ struct vkd3d_shader_signature input;
+
+ /** The shader output varyings. */
+ struct vkd3d_shader_signature output;
+
+ /** The shader patch constant varyings. */
+ struct vkd3d_shader_signature patch_constant;
+};
+
#ifdef LIBVKD3D_SHADER_SOURCE
# define VKD3D_SHADER_API VKD3D_EXPORT
#else
@@ -1625,6 +1693,7 @@ VKD3D_SHADER_API const enum vkd3d_shader_target_type *vkd3d_shader_get_supported
* following chained structures:
* - vkd3d_shader_interface_info
* - vkd3d_shader_scan_descriptor_info
+ * - vkd3d_shader_scan_signature_info
* - vkd3d_shader_spirv_domain_shader_target_info
* - vkd3d_shader_spirv_target_info
* - vkd3d_shader_transform_feedback_info
@@ -1784,6 +1853,26 @@ VKD3D_SHADER_API int vkd3d_shader_convert_root_signature(struct vkd3d_shader_ver
* Parse shader source code or byte code, returning various types of requested
* information.
*
+ * The \a source_type member of \a compile_info must be set to the type of the
+ * shader.
+ *
+ * The \a target_type member may be set to VKD3D_SHADER_TARGET_NONE, in which
+ * case vkd3d_shader_scan() will return information about the shader in
+ * isolation. Alternatively, it may be set to a valid compilation target for the
+ * shader, in which case vkd3d_shader_scan() will return information that
+ * reflects the interface for a shader as it will be compiled to that target.
+ * In this case other chained structures may be appended to \a compile_info as
+ * they would be passed to vkd3d_shader_compile(), and interpreted accordingly,
+ * such as vkd3d_shader_spirv_target_info.
+ *
+ * (For a hypothetical example, suppose the source shader distinguishes float
+ * and integer texture data, but the target environment does not support integer
+ * textures. In this case vkd3d_shader_compile() might translate integer
+ * operations to float. Accordingly using VKD3D_SHADER_TARGET_NONE would
+ * accurately report whether the texture expects integer or float data, but
+ * using the relevant specific target type would report
+ * VKD3D_SHADER_RESOURCE_DATA_FLOAT.)
+ *
* Currently this function supports the following code types:
* - VKD3D_SHADER_SOURCE_DXBC_TPF
*
@@ -1791,6 +1880,7 @@ VKD3D_SHADER_API int vkd3d_shader_convert_root_signature(struct vkd3d_shader_ver
* \n
* The DXBC_TPF scanner supports the following chained structures:
* - vkd3d_shader_scan_descriptor_info
+ * - vkd3d_shader_scan_signature_info
* \n
* Although the \a compile_info parameter is read-only, chained structures
* passed to this function need not be, and may serve as output parameters,
@@ -1827,12 +1917,18 @@ VKD3D_SHADER_API void vkd3d_shader_free_scan_descriptor_info(
struct vkd3d_shader_scan_descriptor_info *scan_descriptor_info);
/**
- * Read the input signature of a compiled shader, returning a structural
+ * Read the input signature of a compiled DXBC shader, returning a structural
* description which can be easily parsed by C code.
*
* This function parses a compiled shader. To parse a standalone root signature,
* use vkd3d_shader_parse_root_signature().
*
+ * This function only parses DXBC shaders, and only retrieves the input
+ * signature. To retrieve signatures from other shader types, or other signature
+ * types, use vkd3d_shader_scan() and struct vkd3d_shader_scan_signature_info.
+ * This function returns the same input signature that is returned in
+ * struct vkd3d_shader_scan_signature_info.
+ *
* \param dxbc Compiled byte code, in DXBC format.
*
* \param signature Output location in which the parsed root signature will be
@@ -2022,6 +2118,19 @@ VKD3D_SHADER_API int vkd3d_shader_parse_dxbc(const struct vkd3d_shader_code *dxb
VKD3D_SHADER_API int vkd3d_shader_serialize_dxbc(size_t section_count,
const struct vkd3d_shader_dxbc_section_desc *sections, struct vkd3d_shader_code *dxbc, char **messages);
+/**
+ * Free members of struct vkd3d_shader_scan_signature_info allocated by
+ * vkd3d_shader_scan().
+ *
+ * This function may free members of vkd3d_shader_scan_signature_info, but
+ * does not free the structure itself.
+ *
+ * \param info Scan information to free.
+ *
+ * \since 1.9
+ */
+VKD3D_SHADER_API void vkd3d_shader_free_scan_signature_info(struct vkd3d_shader_scan_signature_info *info);
+
#endif /* VKD3D_SHADER_NO_PROTOTYPES */
/** Type of vkd3d_shader_get_version(). */
@@ -2087,6 +2196,9 @@ typedef int (*PFN_vkd3d_shader_parse_dxbc)(const struct vkd3d_shader_code *dxbc,
typedef int (*PFN_vkd3d_shader_serialize_dxbc)(size_t section_count,
const struct vkd3d_shader_dxbc_section_desc *sections, struct vkd3d_shader_code *dxbc, char **messages);
+/** Type of vkd3d_shader_free_scan_signature_info(). \since 1.9 */
+typedef void (*PFN_vkd3d_shader_free_scan_signature_info)(struct vkd3d_shader_scan_signature_info *info);
+
#ifdef __cplusplus
}
#endif /* __cplusplus */
diff --git a/libs/vkd3d/libs/vkd3d-common/debug.c b/libs/vkd3d/libs/vkd3d-common/debug.c
index b363efbd360..aa7df5bd764 100644
--- a/libs/vkd3d/libs/vkd3d-common/debug.c
+++ b/libs/vkd3d/libs/vkd3d-common/debug.c
@@ -31,6 +31,7 @@
#include <stdlib.h>
#include <stdbool.h>
#include <string.h>
+#include <unistd.h>
#ifdef HAVE_PTHREAD_H
#include <pthread.h>
#endif
@@ -44,11 +45,11 @@ extern const char *const vkd3d_dbg_env_name;
static const char *const debug_level_names[] =
{
- /* VKD3D_DBG_LEVEL_NONE */ "none",
- /* VKD3D_DBG_LEVEL_ERR */ "err",
- /* VKD3D_DBG_LEVEL_FIXME */ "fixme",
- /* VKD3D_DBG_LEVEL_WARN */ "warn",
- /* VKD3D_DBG_LEVEL_TRACE */ "trace",
+ [VKD3D_DBG_LEVEL_NONE ] = "none",
+ [VKD3D_DBG_LEVEL_ERR ] = "err",
+ [VKD3D_DBG_LEVEL_FIXME] = "fixme",
+ [VKD3D_DBG_LEVEL_WARN ] = "warn",
+ [VKD3D_DBG_LEVEL_TRACE] = "trace",
};
enum vkd3d_dbg_level vkd3d_dbg_get_level(void)
@@ -105,7 +106,13 @@ void vkd3d_dbg_printf(enum vkd3d_dbg_level level, const char *function, const ch
assert(level < ARRAY_SIZE(debug_level_names));
+#ifdef _WIN32
+ vkd3d_dbg_output("vkd3d:%04lx:%s:%s ", GetCurrentThreadId(), debug_level_names[level], function);
+#elif HAVE_GETTID
+ vkd3d_dbg_output("vkd3d:%u:%s:%s ", gettid(), debug_level_names[level], function);
+#else
vkd3d_dbg_output("vkd3d:%s:%s ", debug_level_names[level], function);
+#endif
va_start(args, fmt);
vkd3d_dbg_voutput(fmt, args);
va_end(args);
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
index 0a821b5c878..d72402eb250 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
@@ -578,17 +578,17 @@ static void shader_dump_resource_type(struct vkd3d_d3d_asm_compiler *compiler, e
{
static const char *const resource_type_names[] =
{
- /* VKD3D_SHADER_RESOURCE_NONE */ "none",
- /* VKD3D_SHADER_RESOURCE_BUFFER */ "buffer",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_1D */ "texture1d",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_2D */ "texture2d",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_2DMS */ "texture2dms",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_3D */ "texture3d",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_CUBE */ "texturecube",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_1DARRAY */ "texture1darray",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_2DARRAY */ "texture2darray",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY */ "texture2dmsarray",
- /* VKD3D_SHADER_RESOURCE_TEXTURE_CUBEARRAY */ "texturecubearray",
+ [VKD3D_SHADER_RESOURCE_NONE ] = "none",
+ [VKD3D_SHADER_RESOURCE_BUFFER ] = "buffer",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_1D ] = "texture1d",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_2D ] = "texture2d",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_2DMS ] = "texture2dms",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_3D ] = "texture3d",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_CUBE ] = "texturecube",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_1DARRAY ] = "texture1darray",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_2DARRAY ] = "texture2darray",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_2DMSARRAY] = "texture2dmsarray",
+ [VKD3D_SHADER_RESOURCE_TEXTURE_CUBEARRAY] = "texturecubearray",
};
if (type < ARRAY_SIZE(resource_type_names))
@@ -601,19 +601,19 @@ static void shader_dump_data_type(struct vkd3d_d3d_asm_compiler *compiler, const
{
static const char *const data_type_names[] =
{
- /* VKD3D_DATA_FLOAT */ "float",
- /* VKD3D_DATA_INT */ "int",
- /* VKD3D_DATA_RESOURCE */ "resource",
- /* VKD3D_DATA_SAMPLER */ "sampler",
- /* VKD3D_DATA_UAV */ "uav",
- /* VKD3D_DATA_UINT */ "uint",
- /* VKD3D_DATA_UNORM */ "unorm",
- /* VKD3D_DATA_SNORM */ "snorm",
- /* VKD3D_DATA_OPAQUE */ "opaque",
- /* VKD3D_DATA_MIXED */ "mixed",
- /* VKD3D_DATA_DOUBLE */ "double",
- /* VKD3D_DATA_CONTINUED */ "<continued>",
- /* VKD3D_DATA_UNUSED */ "<unused>",
+ [VKD3D_DATA_FLOAT ] = "float",
+ [VKD3D_DATA_INT ] = "int",
+ [VKD3D_DATA_RESOURCE ] = "resource",
+ [VKD3D_DATA_SAMPLER ] = "sampler",
+ [VKD3D_DATA_UAV ] = "uav",
+ [VKD3D_DATA_UINT ] = "uint",
+ [VKD3D_DATA_UNORM ] = "unorm",
+ [VKD3D_DATA_SNORM ] = "snorm",
+ [VKD3D_DATA_OPAQUE ] = "opaque",
+ [VKD3D_DATA_MIXED ] = "mixed",
+ [VKD3D_DATA_DOUBLE ] = "double",
+ [VKD3D_DATA_CONTINUED] = "<continued>",
+ [VKD3D_DATA_UNUSED ] = "<unused>",
};
const char *name;
int i;
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
index 712613ac13b..369112ce18d 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
@@ -260,9 +260,9 @@ static const struct vkd3d_sm1_opcode_info vs_opcode_table[] =
/* Declarations */
{VKD3D_SM1_OP_DCL, 0, 2, VKD3DSIH_DCL},
/* Constant definitions */
- {VKD3D_SM1_OP_DEF, 1, 4, VKD3DSIH_DEF},
+ {VKD3D_SM1_OP_DEF, 1, 1, VKD3DSIH_DEF},
{VKD3D_SM1_OP_DEFB, 1, 1, VKD3DSIH_DEFB},
- {VKD3D_SM1_OP_DEFI, 1, 4, VKD3DSIH_DEFI},
+ {VKD3D_SM1_OP_DEFI, 1, 1, VKD3DSIH_DEFI},
/* Control flow */
{VKD3D_SM1_OP_REP, 0, 1, VKD3DSIH_REP, {2, 0}, {~0u, ~0u}},
{VKD3D_SM1_OP_ENDREP, 0, 0, VKD3DSIH_ENDREP, {2, 0}, {~0u, ~0u}},
@@ -327,9 +327,9 @@ static const struct vkd3d_sm1_opcode_info ps_opcode_table[] =
/* Declarations */
{VKD3D_SM1_OP_DCL, 0, 2, VKD3DSIH_DCL},
/* Constant definitions */
- {VKD3D_SM1_OP_DEF, 1, 4, VKD3DSIH_DEF},
+ {VKD3D_SM1_OP_DEF, 1, 1, VKD3DSIH_DEF},
{VKD3D_SM1_OP_DEFB, 1, 1, VKD3DSIH_DEFB},
- {VKD3D_SM1_OP_DEFI, 1, 4, VKD3DSIH_DEFI},
+ {VKD3D_SM1_OP_DEFI, 1, 1, VKD3DSIH_DEFI},
/* Control flow */
{VKD3D_SM1_OP_REP, 0, 1, VKD3DSIH_REP, {2, 1}, {~0u, ~0u}},
{VKD3D_SM1_OP_ENDREP, 0, 0, VKD3DSIH_ENDREP, {2, 1}, {~0u, ~0u}},
@@ -490,6 +490,255 @@ static void shader_sm1_parse_dst_param(uint32_t param, const struct vkd3d_shader
dst->shift = (param & VKD3D_SM1_DSTSHIFT_MASK) >> VKD3D_SM1_DSTSHIFT_SHIFT;
}
+static struct signature_element *find_signature_element(const struct shader_signature *signature,
+ const char *semantic_name, unsigned int semantic_index)
+{
+ struct signature_element *e = signature->elements;
+ unsigned int i;
+
+ for (i = 0; i < signature->element_count; ++i)
+ {
+ if (!ascii_strcasecmp(e[i].semantic_name, semantic_name)
+ && e[i].semantic_index == semantic_index)
+ return &e[i];
+ }
+
+ return NULL;
+}
+
+static struct signature_element *find_signature_element_by_register_index(
+ const struct shader_signature *signature, unsigned int register_index)
+{
+ struct signature_element *e = signature->elements;
+ unsigned int i;
+
+ for (i = 0; i < signature->element_count; ++i)
+ {
+ if (e[i].register_index == register_index)
+ return &e[i];
+ }
+
+ return NULL;
+}
+
+static bool add_signature_element(struct vkd3d_shader_sm1_parser *sm1, bool output,
+ const char *name, unsigned int index, enum vkd3d_shader_sysval_semantic sysval,
+ unsigned int register_index, bool is_dcl, unsigned int mask)
+{
+ struct shader_signature *signature;
+ struct signature_element *element;
+
+ if (output)
+ signature = &sm1->p.shader_desc.output_signature;
+ else
+ signature = &sm1->p.shader_desc.input_signature;
+
+ if ((element = find_signature_element(signature, name, index)))
+ {
+ element->mask |= mask;
+ if (!is_dcl)
+ element->used_mask |= mask;
+ return true;
+ }
+
+ if (!vkd3d_array_reserve((void **)&signature->elements, &signature->elements_capacity,
+ signature->element_count + 1, sizeof(*signature->elements)))
+ return false;
+ element = &signature->elements[signature->element_count++];
+
+ element->semantic_name = name;
+ element->semantic_index = index;
+ element->stream_index = 0;
+ element->sysval_semantic = sysval;
+ element->component_type = VKD3D_SHADER_COMPONENT_FLOAT;
+ element->register_index = register_index;
+ element->register_count = 1;
+ element->mask = mask;
+ element->used_mask = is_dcl ? 0 : mask;
+ element->min_precision = VKD3D_SHADER_MINIMUM_PRECISION_NONE;
+
+ return true;
+}
+
+static void add_signature_mask(struct vkd3d_shader_sm1_parser *sm1, bool output,
+ unsigned int register_index, unsigned int mask)
+{
+ struct shader_signature *signature;
+ struct signature_element *element;
+
+ if (output)
+ signature = &sm1->p.shader_desc.output_signature;
+ else
+ signature = &sm1->p.shader_desc.input_signature;
+
+ if (!(element = find_signature_element_by_register_index(signature, register_index)))
+ {
+ vkd3d_shader_parser_error(&sm1->p, VKD3D_SHADER_ERROR_D3DBC_UNDECLARED_SEMANTIC,
+ "%s register %u was used without being declared.", output ? "Output" : "Input", register_index);
+ return;
+ }
+
+ element->used_mask |= mask;
+}
+
+static bool add_signature_element_from_register(struct vkd3d_shader_sm1_parser *sm1,
+ const struct vkd3d_shader_register *reg, bool is_dcl, unsigned int mask)
+{
+ unsigned int register_index = reg->idx[0].offset;
+
+ switch (reg->type)
+ {
+ case VKD3DSPR_TEMP:
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_PIXEL
+ && sm1->p.shader_version.major == 1 && !register_index)
+ return add_signature_element(sm1, true, "COLOR", 0, VKD3D_SHADER_SV_NONE, 0, is_dcl, mask);
+ return true;
+
+ case VKD3DSPR_INPUT:
+ /* For vertex shaders or sm3 pixel shaders, we should have already
+ * had a DCL instruction. Otherwise, this is a colour input. */
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_VERTEX || sm1->p.shader_version.major == 3)
+ {
+ add_signature_mask(sm1, false, register_index, mask);
+ return true;
+ }
+ return add_signature_element(sm1, false, "COLOR", register_index,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, mask);
+
+ case VKD3DSPR_TEXTURE:
+ /* For vertex shaders, this is ADDR. */
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_VERTEX)
+ return true;
+ return add_signature_element(sm1, false, "TEXCOORD", register_index,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, mask);
+
+ case VKD3DSPR_OUTPUT:
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_VERTEX)
+ {
+ /* For sm < 2 vertex shaders, this is TEXCRDOUT.
+ *
+ * For sm3 vertex shaders, this is OUTPUT, but we already
+ * should have had a DCL instruction. */
+ if (sm1->p.shader_version.major == 3)
+ {
+ add_signature_mask(sm1, true, register_index, mask);
+ return true;
+ }
+ return add_signature_element(sm1, true, "TEXCOORD", register_index,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, mask);
+ }
+ /* fall through */
+
+ case VKD3DSPR_ATTROUT:
+ case VKD3DSPR_COLOROUT:
+ return add_signature_element(sm1, true, "COLOR", register_index,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, mask);
+
+ case VKD3DSPR_DEPTHOUT:
+ return add_signature_element(sm1, true, "DEPTH", 0,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, 0x1);
+
+ case VKD3DSPR_RASTOUT:
+ switch (register_index)
+ {
+ case 0:
+ return add_signature_element(sm1, true, "POSITION", 0,
+ VKD3D_SHADER_SV_POSITION, register_index, is_dcl, mask);
+
+ case 1:
+ return add_signature_element(sm1, true, "FOG", 0,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, 0x1);
+
+ case 2:
+ return add_signature_element(sm1, true, "PSIZE", 0,
+ VKD3D_SHADER_SV_NONE, register_index, is_dcl, 0x1);
+
+ default:
+ vkd3d_shader_parser_error(&sm1->p, VKD3D_SHADER_ERROR_D3DBC_INVALID_REGISTER_INDEX,
+ "Invalid rasterizer output index %u.", register_index);
+ return true;
+ }
+
+ case VKD3DSPR_MISCTYPE:
+ switch (register_index)
+ {
+ case 0:
+ return add_signature_element(sm1, false, "VPOS", 0,
+ VKD3D_SHADER_SV_POSITION, register_index, is_dcl, mask);
+
+ case 1:
+ return add_signature_element(sm1, false, "VFACE", 0,
+ VKD3D_SHADER_SV_IS_FRONT_FACE, register_index, is_dcl, 0x1);
+
+ default:
+ vkd3d_shader_parser_error(&sm1->p, VKD3D_SHADER_ERROR_D3DBC_INVALID_REGISTER_INDEX,
+ "Invalid miscellaneous fragment input index %u.", register_index);
+ return true;
+ }
+
+ default:
+ return true;
+ }
+}
+
+static bool add_signature_element_from_semantic(struct vkd3d_shader_sm1_parser *sm1,
+ const struct vkd3d_shader_semantic *semantic)
+{
+ const struct vkd3d_shader_register *reg = &semantic->resource.reg.reg;
+ enum vkd3d_shader_sysval_semantic sysval = VKD3D_SHADER_SV_NONE;
+ unsigned int mask = semantic->resource.reg.write_mask;
+ bool output;
+
+ static const char sm1_semantic_names[][13] =
+ {
+ [VKD3D_DECL_USAGE_POSITION ] = "POSITION",
+ [VKD3D_DECL_USAGE_BLEND_WEIGHT ] = "BLENDWEIGHT",
+ [VKD3D_DECL_USAGE_BLEND_INDICES] = "BLENDINDICES",
+ [VKD3D_DECL_USAGE_NORMAL ] = "NORMAL",
+ [VKD3D_DECL_USAGE_PSIZE ] = "PSIZE",
+ [VKD3D_DECL_USAGE_TEXCOORD ] = "TEXCOORD",
+ [VKD3D_DECL_USAGE_TANGENT ] = "TANGENT",
+ [VKD3D_DECL_USAGE_BINORMAL ] = "BINORMAL",
+ [VKD3D_DECL_USAGE_TESS_FACTOR ] = "TESSFACTOR",
+ [VKD3D_DECL_USAGE_POSITIONT ] = "POSITIONT",
+ [VKD3D_DECL_USAGE_COLOR ] = "COLOR",
+ [VKD3D_DECL_USAGE_FOG ] = "FOG",
+ [VKD3D_DECL_USAGE_DEPTH ] = "DEPTH",
+ [VKD3D_DECL_USAGE_SAMPLE ] = "SAMPLE",
+ };
+
+ if (reg->type == VKD3DSPR_OUTPUT)
+ output = true;
+ else if (reg->type == VKD3DSPR_INPUT || reg->type == VKD3DSPR_TEXTURE)
+ output = false;
+ else /* vpos and vface don't have a semantic. */
+ return add_signature_element_from_register(sm1, reg, true, mask);
+
+ /* sm2 pixel shaders use DCL but don't provide a semantic. */
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_PIXEL && sm1->p.shader_version.major == 2)
+ return add_signature_element_from_register(sm1, reg, true, mask);
+
+ /* With the exception of vertex POSITION output, none of these are system
+ * values. Pixel POSITION input is not equivalent to SV_Position; the closer
+ * equivalent is VPOS, which is not declared as a semantic. */
+ if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_VERTEX
+ && output && semantic->usage == VKD3D_DECL_USAGE_POSITION)
+ sysval = VKD3D_SHADER_SV_POSITION;
+
+ return add_signature_element(sm1, output, sm1_semantic_names[semantic->usage],
+ semantic->usage_idx, sysval, reg->idx[0].offset, true, mask);
+}
+
+static void shader_sm1_scan_register(struct vkd3d_shader_sm1_parser *sm1, const struct vkd3d_shader_register *reg, unsigned int mask)
+{
+ uint32_t register_index = reg->idx[0].offset;
+
+ if (reg->type == VKD3DSPR_TEMP)
+ sm1->p.shader_desc.temp_count = max(sm1->p.shader_desc.temp_count, register_index + 1);
+
+ add_signature_element_from_register(sm1, reg, false, mask);
+}
+
/* Read a parameter token from the input stream, and possibly a relative
* addressing token. */
static void shader_sm1_read_param(struct vkd3d_shader_sm1_parser *sm1,
@@ -640,6 +889,8 @@ static void shader_sm1_read_semantic(struct vkd3d_shader_sm1_parser *sm1,
range = &semantic->resource.range;
range->space = 0;
range->first = range->last = semantic->resource.reg.reg.idx[0].offset;
+
+ add_signature_element_from_semantic(sm1, semantic);
}
static void shader_sm1_read_immconst(struct vkd3d_shader_sm1_parser *sm1, const uint32_t **ptr,
@@ -744,6 +995,14 @@ static void shader_sm1_validate_instruction(struct vkd3d_shader_sm1_parser *sm1,
}
}
+static unsigned int mask_from_swizzle(unsigned int swizzle)
+{
+ return (1u << vkd3d_swizzle_get_component(swizzle, 0))
+ | (1u << vkd3d_swizzle_get_component(swizzle, 1))
+ | (1u << vkd3d_swizzle_get_component(swizzle, 2))
+ | (1u << vkd3d_swizzle_get_component(swizzle, 3));
+}
+
static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, struct vkd3d_shader_instruction *ins)
{
struct vkd3d_shader_src_param *src_params, *predicate;
@@ -832,7 +1091,10 @@ static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, str
{
/* Destination token */
if (ins->dst_count)
+ {
shader_sm1_read_dst_param(sm1, &p, dst_param);
+ shader_sm1_scan_register(sm1, &dst_param->reg, dst_param->write_mask);
+ }
/* Predication token */
if (ins->predicate)
@@ -840,7 +1102,10 @@ static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, str
/* Other source tokens */
for (i = 0; i < ins->src_count; ++i)
+ {
shader_sm1_read_src_param(sm1, &p, &src_params[i]);
+ shader_sm1_scan_register(sm1, &src_params[i].reg, mask_from_swizzle(src_params[i].swizzle));
+ }
}
if (sm1->abort)
@@ -1553,12 +1818,13 @@ static void write_sm1_constant_defs(struct hlsl_ctx *ctx, struct vkd3d_bytecode_
for (i = 0; i < ctx->constant_defs.count; ++i)
{
+ const struct hlsl_constant_register *constant_reg = &ctx->constant_defs.regs[i];
uint32_t token = D3DSIO_DEF;
const struct sm1_dst_register reg =
{
.type = D3DSPR_CONST,
.writemask = VKD3DSP_WRITEMASK_ALL,
- .reg = i,
+ .reg = constant_reg->index,
};
if (ctx->profile->major_version > 1)
@@ -1567,7 +1833,7 @@ static void write_sm1_constant_defs(struct hlsl_ctx *ctx, struct vkd3d_bytecode_
write_sm1_dst_register(buffer, &reg);
for (x = 0; x < 4; ++x)
- put_f32(buffer, ctx->constant_defs.values[i].f[x]);
+ put_f32(buffer, constant_reg->value.f[x]);
}
}
@@ -1844,6 +2110,35 @@ static void write_sm1_expr(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b
}
}
+static void write_sm1_jump(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr)
+{
+ const struct hlsl_ir_jump *jump = hlsl_ir_jump(instr);
+
+ switch (jump->type)
+ {
+ case HLSL_IR_JUMP_DISCARD_NEG:
+ {
+ struct hlsl_reg *reg = &jump->condition.node->reg;
+
+ struct sm1_instruction instr =
+ {
+ .opcode = VKD3D_SM1_OP_TEXKILL,
+
+ .dst.type = D3DSPR_TEMP,
+ .dst.reg = reg->id,
+ .dst.writemask = reg->writemask,
+ .has_dst = 1,
+ };
+
+ write_sm1_instruction(ctx, buffer, &instr);
+ break;
+ }
+
+ default:
+ hlsl_fixme(ctx, &jump->node.loc, "Jump type %s.\n", hlsl_jump_type_to_string(jump->type));
+ }
+}
+
static void write_sm1_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr)
{
const struct hlsl_ir_load *load = hlsl_ir_load(instr);
@@ -2038,6 +2333,10 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
write_sm1_expr(ctx, buffer, instr);
break;
+ case HLSL_IR_JUMP:
+ write_sm1_jump(ctx, buffer, instr);
+ break;
+
case HLSL_IR_LOAD:
write_sm1_load(ctx, buffer, instr);
break;
diff --git a/libs/vkd3d/libs/vkd3d-shader/dxbc.c b/libs/vkd3d/libs/vkd3d-shader/dxbc.c
index 3e3f06faeb5..716b7bdb721 100644
--- a/libs/vkd3d/libs/vkd3d-shader/dxbc.c
+++ b/libs/vkd3d/libs/vkd3d-shader/dxbc.c
@@ -493,8 +493,14 @@ static int shdr_handler(const struct vkd3d_shader_dxbc_section_desc *section,
return ret;
break;
+ case TAG_DXIL:
case TAG_SHDR:
case TAG_SHEX:
+ if ((section->tag == TAG_DXIL) != desc->is_dxil)
+ {
+ TRACE("Skipping chunk %#x.\n", section->tag);
+ break;
+ }
if (desc->byte_code)
FIXME("Multiple shader code chunks.\n");
desc->byte_code = section->data.code;
@@ -505,10 +511,6 @@ static int shdr_handler(const struct vkd3d_shader_dxbc_section_desc *section,
TRACE("Skipping AON9 shader code chunk.\n");
break;
- case TAG_DXIL:
- FIXME("Skipping DXIL shader model 6+ code chunk.\n");
- break;
-
default:
TRACE("Skipping chunk %#x.\n", section->tag);
break;
@@ -529,12 +531,6 @@ int shader_extract_from_dxbc(const struct vkd3d_shader_code *dxbc,
{
int ret;
- desc->byte_code = NULL;
- desc->byte_code_size = 0;
- memset(&desc->input_signature, 0, sizeof(desc->input_signature));
- memset(&desc->output_signature, 0, sizeof(desc->output_signature));
- memset(&desc->patch_constant_signature, 0, sizeof(desc->patch_constant_signature));
-
ret = for_each_dxbc_section(dxbc, message_context, source_name, shdr_handler, desc);
if (!desc->byte_code)
ret = VKD3D_ERROR_INVALID_ARGUMENT;
diff --git a/libs/vkd3d/libs/vkd3d-shader/dxil.c b/libs/vkd3d/libs/vkd3d-shader/dxil.c
new file mode 100644
index 00000000000..53a4c2da4ba
--- /dev/null
+++ b/libs/vkd3d/libs/vkd3d-shader/dxil.c
@@ -0,0 +1,2313 @@
+/*
+ * Copyright 2023 Conor McCarthy for CodeWeavers
+ *
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * This library is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA
+ */
+
+#include "vkd3d_shader_private.h"
+
+#define VKD3D_SM6_VERSION_MAJOR(version) (((version) >> 4) & 0xf)
+#define VKD3D_SM6_VERSION_MINOR(version) (((version) >> 0) & 0xf)
+
+#define BITCODE_MAGIC VKD3D_MAKE_TAG('B', 'C', 0xc0, 0xde)
+
+enum bitcode_block_id
+{
+ BLOCKINFO_BLOCK = 0,
+ MODULE_BLOCK = 8,
+ PARAMATTR_BLOCK = 9,
+ PARAMATTR_GROUP_BLOCK = 10,
+ CONSTANTS_BLOCK = 11,
+ FUNCTION_BLOCK = 12,
+ VALUE_SYMTAB_BLOCK = 14,
+ METADATA_BLOCK = 15,
+ METADATA_ATTACHMENT_BLOCK = 16,
+ TYPE_BLOCK = 17,
+ USELIST_BLOCK = 18,
+};
+
+enum bitcode_blockinfo_code
+{
+ SETBID = 1,
+ BLOCKNAME = 2,
+ SETRECORDNAME = 3,
+};
+
+enum bitcode_block_abbreviation
+{
+ END_BLOCK = 0,
+ ENTER_SUBBLOCK = 1,
+ DEFINE_ABBREV = 2,
+ UNABBREV_RECORD = 3,
+};
+
+enum bitcode_abbrev_type
+{
+ ABBREV_FIXED = 1,
+ ABBREV_VBR = 2,
+ ABBREV_ARRAY = 3,
+ ABBREV_CHAR = 4,
+ ABBREV_BLOB = 5,
+};
+
+enum bitcode_address_space
+{
+ ADDRESS_SPACE_DEFAULT,
+ ADDRESS_SPACE_DEVICEMEM,
+ ADDRESS_SPACE_CBUFFER,
+ ADDRESS_SPACE_GROUPSHARED,
+};
+
+enum bitcode_module_code
+{
+ MODULE_CODE_VERSION = 1,
+ MODULE_CODE_GLOBALVAR = 7,
+ MODULE_CODE_FUNCTION = 8,
+};
+
+enum bitcode_constant_code
+{
+ CST_CODE_SETTYPE = 1,
+ CST_CODE_NULL = 2,
+ CST_CODE_UNDEF = 3,
+ CST_CODE_INTEGER = 4,
+ CST_CODE_FLOAT = 6,
+ CST_CODE_STRING = 8,
+ CST_CODE_CE_GEP = 12,
+ CST_CODE_CE_INBOUNDS_GEP = 20,
+ CST_CODE_DATA = 22,
+};
+
+enum bitcode_function_code
+{
+ FUNC_CODE_DECLAREBLOCKS = 1,
+ FUNC_CODE_INST_BINOP = 2,
+ FUNC_CODE_INST_CAST = 3,
+ FUNC_CODE_INST_RET = 10,
+ FUNC_CODE_INST_BR = 11,
+ FUNC_CODE_INST_SWITCH = 12,
+ FUNC_CODE_INST_PHI = 16,
+ FUNC_CODE_INST_ALLOCA = 19,
+ FUNC_CODE_INST_LOAD = 20,
+ FUNC_CODE_INST_EXTRACTVAL = 26,
+ FUNC_CODE_INST_CMP2 = 28,
+ FUNC_CODE_INST_VSELECT = 29,
+ FUNC_CODE_INST_CALL = 34,
+ FUNC_CODE_INST_ATOMICRMW = 38,
+ FUNC_CODE_INST_LOADATOMIC = 41,
+ FUNC_CODE_INST_GEP = 43,
+ FUNC_CODE_INST_STORE = 44,
+ FUNC_CODE_INST_STOREATOMIC = 45,
+ FUNC_CODE_INST_CMPXCHG = 46,
+};
+
+enum bitcode_type_code
+{
+ TYPE_CODE_NUMENTRY = 1,
+ TYPE_CODE_VOID = 2,
+ TYPE_CODE_FLOAT = 3,
+ TYPE_CODE_DOUBLE = 4,
+ TYPE_CODE_LABEL = 5,
+ TYPE_CODE_INTEGER = 7,
+ TYPE_CODE_POINTER = 8,
+ TYPE_CODE_HALF = 10,
+ TYPE_CODE_ARRAY = 11,
+ TYPE_CODE_VECTOR = 12,
+ TYPE_CODE_METADATA = 16,
+ TYPE_CODE_STRUCT_ANON = 18,
+ TYPE_CODE_STRUCT_NAME = 19,
+ TYPE_CODE_STRUCT_NAMED = 20,
+ TYPE_CODE_FUNCTION = 21,
+};
+
+enum bitcode_value_symtab_code
+{
+ VST_CODE_ENTRY = 1,
+ VST_CODE_BBENTRY = 2,
+};
+
+struct sm6_pointer_info
+{
+ const struct sm6_type *type;
+ enum bitcode_address_space addr_space;
+};
+
+struct sm6_struct_info
+{
+ const char *name;
+ unsigned int elem_count;
+ const struct sm6_type *elem_types[];
+};
+
+struct sm6_function_info
+{
+ const struct sm6_type *ret_type;
+ unsigned int param_count;
+ const struct sm6_type *param_types[];
+};
+
+struct sm6_array_info
+{
+ unsigned int count;
+ const struct sm6_type *elem_type;
+};
+
+enum sm6_type_class
+{
+ TYPE_CLASS_VOID,
+ TYPE_CLASS_INTEGER,
+ TYPE_CLASS_FLOAT,
+ TYPE_CLASS_POINTER,
+ TYPE_CLASS_STRUCT,
+ TYPE_CLASS_FUNCTION,
+ TYPE_CLASS_VECTOR,
+ TYPE_CLASS_ARRAY,
+ TYPE_CLASS_LABEL,
+ TYPE_CLASS_METADATA,
+};
+
+struct sm6_type
+{
+ enum sm6_type_class class;
+ union
+ {
+ unsigned int width;
+ struct sm6_pointer_info pointer;
+ struct sm6_struct_info *struc;
+ struct sm6_function_info *function;
+ struct sm6_array_info array;
+ } u;
+};
+
+enum sm6_value_type
+{
+ VALUE_TYPE_FUNCTION,
+ VALUE_TYPE_REG,
+};
+
+struct sm6_function_data
+{
+ const char *name;
+ bool is_prototype;
+ unsigned int attribs_id;
+};
+
+struct sm6_value
+{
+ const struct sm6_type *type;
+ enum sm6_value_type value_type;
+ union
+ {
+ struct sm6_function_data function;
+ struct vkd3d_shader_register reg;
+ } u;
+};
+
+struct dxil_record
+{
+ unsigned int code;
+ unsigned int operand_count;
+ uint64_t operands[];
+};
+
+struct sm6_symbol
+{
+ unsigned int id;
+ const char *name;
+};
+
+struct sm6_block
+{
+ struct vkd3d_shader_instruction *instructions;
+ size_t instruction_capacity;
+ size_t instruction_count;
+};
+
+struct sm6_function
+{
+ const struct sm6_value *declaration;
+
+ struct sm6_block *blocks[1];
+ size_t block_count;
+};
+
+struct dxil_block
+{
+ const struct dxil_block *parent;
+ enum bitcode_block_id id;
+ unsigned int abbrev_len;
+ unsigned int start;
+ unsigned int length;
+ unsigned int level;
+
+ /* The abbrev, block and record structs are not relocatable. */
+ struct dxil_abbrev **abbrevs;
+ size_t abbrev_capacity;
+ size_t abbrev_count;
+ unsigned int blockinfo_bid;
+ bool has_bid;
+
+ struct dxil_block **child_blocks;
+ size_t child_block_capacity;
+ size_t child_block_count;
+
+ struct dxil_record **records;
+ size_t record_capacity;
+ size_t record_count;
+};
+
+struct sm6_parser
+{
+ const uint32_t *ptr, *start, *end;
+ unsigned int bitpos;
+
+ struct dxil_block root_block;
+ struct dxil_block *current_block;
+
+ struct dxil_global_abbrev **abbrevs;
+ size_t abbrev_capacity;
+ size_t abbrev_count;
+
+ struct sm6_type *types;
+ size_t type_count;
+
+ struct sm6_symbol *global_symbols;
+ size_t global_symbol_count;
+
+ struct sm6_function *functions;
+ size_t function_count;
+
+ struct sm6_value *values;
+ size_t value_count;
+ size_t value_capacity;
+
+ struct vkd3d_shader_parser p;
+};
+
+struct dxil_abbrev_operand
+{
+ uint64_t context;
+ bool (*read_operand)(struct sm6_parser *sm6, uint64_t context, uint64_t *operand);
+};
+
+struct dxil_abbrev
+{
+ unsigned int count;
+ bool is_array;
+ struct dxil_abbrev_operand operands[];
+};
+
+struct dxil_global_abbrev
+{
+ unsigned int block_id;
+ struct dxil_abbrev abbrev;
+};
+
+static size_t size_add_with_overflow_check(size_t a, size_t b)
+{
+ size_t i = a + b;
+ return (i < a) ? SIZE_MAX : i;
+}
+
+static struct sm6_parser *sm6_parser(struct vkd3d_shader_parser *parser)
+{
+ return CONTAINING_RECORD(parser, struct sm6_parser, p);
+}
+
+static bool sm6_parser_is_end(struct sm6_parser *sm6)
+{
+ return sm6->ptr == sm6->end;
+}
+
+static uint32_t sm6_parser_read_uint32(struct sm6_parser *sm6)
+{
+ if (sm6_parser_is_end(sm6))
+ {
+ sm6->p.failed = true;
+ return 0;
+ }
+ return *sm6->ptr++;
+}
+
+static uint32_t sm6_parser_read_bits(struct sm6_parser *sm6, unsigned int length)
+{
+ unsigned int l, prev_len = 0;
+ uint32_t bits;
+
+ if (!length)
+ return 0;
+
+ assert(length < 32);
+
+ if (sm6_parser_is_end(sm6))
+ {
+ sm6->p.failed = true;
+ return 0;
+ }
+
+ assert(sm6->bitpos < 32);
+ bits = *sm6->ptr >> sm6->bitpos;
+ l = 32 - sm6->bitpos;
+ if (l <= length)
+ {
+ ++sm6->ptr;
+ if (sm6_parser_is_end(sm6) && l < length)
+ {
+ sm6->p.failed = true;
+ return bits;
+ }
+ sm6->bitpos = 0;
+ bits |= *sm6->ptr << l;
+ prev_len = l;
+ }
+ sm6->bitpos += length - prev_len;
+
+ return bits & ((1 << length) - 1);
+}
+
+static uint64_t sm6_parser_read_vbr(struct sm6_parser *sm6, unsigned int length)
+{
+ unsigned int bits, flag, mask, shift = 0;
+ uint64_t result = 0;
+
+ if (!length)
+ return 0;
+
+ if (sm6_parser_is_end(sm6))
+ {
+ sm6->p.failed = true;
+ return 0;
+ }
+
+ flag = 1 << (length - 1);
+ mask = flag - 1;
+ do
+ {
+ bits = sm6_parser_read_bits(sm6, length);
+ result |= (uint64_t)(bits & mask) << shift;
+ shift += length - 1;
+ } while ((bits & flag) && !sm6->p.failed && shift < 64);
+
+ sm6->p.failed |= !!(bits & flag);
+
+ return result;
+}
+
+static void sm6_parser_align_32(struct sm6_parser *sm6)
+{
+ if (!sm6->bitpos)
+ return;
+
+ if (sm6_parser_is_end(sm6))
+ {
+ sm6->p.failed = true;
+ return;
+ }
+
+ ++sm6->ptr;
+ sm6->bitpos = 0;
+}
+
+static bool dxil_block_handle_blockinfo_record(struct dxil_block *block, struct dxil_record *record)
+{
+ /* BLOCKINFO blocks must only occur immediately below the module root block. */
+ if (block->level > 1)
+ {
+ WARN("Invalid blockinfo block level %u.\n", block->level);
+ return false;
+ }
+
+ switch (record->code)
+ {
+ case SETBID:
+ if (!record->operand_count)
+ {
+ WARN("Missing id operand.\n");
+ return false;
+ }
+ if (record->operands[0] > UINT_MAX)
+ WARN("Truncating block id %"PRIu64".\n", record->operands[0]);
+ block->blockinfo_bid = record->operands[0];
+ block->has_bid = true;
+ break;
+ case BLOCKNAME:
+ case SETRECORDNAME:
+ break;
+ default:
+ FIXME("Unhandled BLOCKINFO record type %u.\n", record->code);
+ break;
+ }
+
+ return true;
+}
+
+static enum vkd3d_result dxil_block_add_record(struct dxil_block *block, struct dxil_record *record)
+{
+ unsigned int reserve;
+
+ switch (block->id)
+ {
+ /* Rough initial reserve sizes for small shaders. */
+ case CONSTANTS_BLOCK: reserve = 32; break;
+ case FUNCTION_BLOCK: reserve = 128; break;
+ case METADATA_BLOCK: reserve = 32; break;
+ case TYPE_BLOCK: reserve = 32; break;
+ default: reserve = 8; break;
+ }
+ reserve = max(reserve, block->record_count + 1);
+ if (!vkd3d_array_reserve((void **)&block->records, &block->record_capacity, reserve, sizeof(*block->records)))
+ {
+ ERR("Failed to allocate %u records.\n", reserve);
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if (block->id == BLOCKINFO_BLOCK && !dxil_block_handle_blockinfo_record(block, record))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ block->records[block->record_count++] = record;
+
+ return VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_read_unabbrev_record(struct sm6_parser *sm6)
+{
+ struct dxil_block *block = sm6->current_block;
+ enum vkd3d_result ret = VKD3D_OK;
+ unsigned int code, count, i;
+ struct dxil_record *record;
+
+ code = sm6_parser_read_vbr(sm6, 6);
+
+ count = sm6_parser_read_vbr(sm6, 6);
+ if (!(record = vkd3d_malloc(sizeof(*record) + count * sizeof(record->operands[0]))))
+ {
+ ERR("Failed to allocate record with %u operands.\n", count);
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ record->code = code;
+ record->operand_count = count;
+
+ for (i = 0; i < count; ++i)
+ record->operands[i] = sm6_parser_read_vbr(sm6, 6);
+ if (sm6->p.failed)
+ ret = VKD3D_ERROR_INVALID_SHADER;
+
+ if (ret < 0 || (ret = dxil_block_add_record(block, record)) < 0)
+ vkd3d_free(record);
+
+ return ret;
+}
+
+static bool sm6_parser_read_literal_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op)
+{
+ *op = context;
+ return !sm6->p.failed;
+}
+
+static bool sm6_parser_read_fixed_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op)
+{
+ *op = sm6_parser_read_bits(sm6, context);
+ return !sm6->p.failed;
+}
+
+static bool sm6_parser_read_vbr_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op)
+{
+ *op = sm6_parser_read_vbr(sm6, context);
+ return !sm6->p.failed;
+}
+
+static bool sm6_parser_read_char6_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op)
+{
+ *op = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789._"[sm6_parser_read_bits(sm6, 6)];
+ return !sm6->p.failed;
+}
+
+static bool sm6_parser_read_blob_operand(struct sm6_parser *sm6, uint64_t context, uint64_t *op)
+{
+ int count = sm6_parser_read_vbr(sm6, 6);
+ sm6_parser_align_32(sm6);
+ for (; count > 0; count -= 4)
+ sm6_parser_read_uint32(sm6);
+ FIXME("Unhandled blob operand.\n");
+ return false;
+}
+
+static enum vkd3d_result dxil_abbrev_init(struct dxil_abbrev *abbrev, unsigned int count, struct sm6_parser *sm6)
+{
+ enum bitcode_abbrev_type prev_type, type;
+ unsigned int i;
+
+ abbrev->is_array = false;
+
+ for (i = 0, prev_type = 0; i < count && !sm6->p.failed; ++i)
+ {
+ if (sm6_parser_read_bits(sm6, 1))
+ {
+ if (prev_type == ABBREV_ARRAY)
+ {
+ WARN("Unexpected literal abbreviation after array.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ abbrev->operands[i].context = sm6_parser_read_vbr(sm6, 8);
+ abbrev->operands[i].read_operand = sm6_parser_read_literal_operand;
+ continue;
+ }
+
+ switch (type = sm6_parser_read_bits(sm6, 3))
+ {
+ case ABBREV_FIXED:
+ case ABBREV_VBR:
+ abbrev->operands[i].context = sm6_parser_read_vbr(sm6, 5);
+ abbrev->operands[i].read_operand = (type == ABBREV_FIXED) ? sm6_parser_read_fixed_operand
+ : sm6_parser_read_vbr_operand;
+ break;
+
+ case ABBREV_ARRAY:
+ if (prev_type == ABBREV_ARRAY || i != count - 2)
+ {
+ WARN("Unexpected array abbreviation.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ abbrev->is_array = true;
+ --i;
+ --count;
+ break;
+
+ case ABBREV_CHAR:
+ abbrev->operands[i].read_operand = sm6_parser_read_char6_operand;
+ break;
+
+ case ABBREV_BLOB:
+ if (prev_type == ABBREV_ARRAY || i != count - 1)
+ {
+ WARN("Unexpected blob abbreviation.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ abbrev->operands[i].read_operand = sm6_parser_read_blob_operand;
+ break;
+ }
+
+ prev_type = type;
+ }
+
+ abbrev->count = count;
+
+ return sm6->p.failed ? VKD3D_ERROR_INVALID_SHADER : VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_add_global_abbrev(struct sm6_parser *sm6)
+{
+ struct dxil_block *block = sm6->current_block;
+ unsigned int count = sm6_parser_read_vbr(sm6, 5);
+ struct dxil_global_abbrev *global_abbrev;
+ enum vkd3d_result ret;
+
+ assert(block->id == BLOCKINFO_BLOCK);
+
+ if (!vkd3d_array_reserve((void **)&sm6->abbrevs, &sm6->abbrev_capacity, sm6->abbrev_count + 1, sizeof(*sm6->abbrevs))
+ || !(global_abbrev = vkd3d_malloc(sizeof(*global_abbrev) + count * sizeof(global_abbrev->abbrev.operands[0]))))
+ {
+ ERR("Failed to allocate global abbreviation.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if ((ret = dxil_abbrev_init(&global_abbrev->abbrev, count, sm6)) < 0)
+ {
+ vkd3d_free(global_abbrev);
+ return ret;
+ }
+
+ if (!block->has_bid)
+ {
+ WARN("Missing blockinfo block id.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (block->blockinfo_bid == MODULE_BLOCK)
+ {
+ FIXME("Unhandled global abbreviation for module block.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ global_abbrev->block_id = block->blockinfo_bid;
+
+ sm6->abbrevs[sm6->abbrev_count++] = global_abbrev;
+
+ return VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_add_block_abbrev(struct sm6_parser *sm6)
+{
+ struct dxil_block *block = sm6->current_block;
+ struct dxil_abbrev *abbrev;
+ enum vkd3d_result ret;
+ unsigned int count;
+
+ if (block->id == BLOCKINFO_BLOCK)
+ return sm6_parser_add_global_abbrev(sm6);
+
+ count = sm6_parser_read_vbr(sm6, 5);
+ if (!vkd3d_array_reserve((void **)&block->abbrevs, &block->abbrev_capacity, block->abbrev_count + 1, sizeof(*block->abbrevs))
+ || !(abbrev = vkd3d_malloc(sizeof(*abbrev) + count * sizeof(abbrev->operands[0]))))
+ {
+ ERR("Failed to allocate block abbreviation.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if ((ret = dxil_abbrev_init(abbrev, count, sm6)) < 0)
+ {
+ vkd3d_free(abbrev);
+ return ret;
+ }
+
+ block->abbrevs[block->abbrev_count++] = abbrev;
+
+ return VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_read_abbrev_record(struct sm6_parser *sm6, unsigned int abbrev_id)
+{
+ enum vkd3d_result ret = VKD3D_ERROR_INVALID_SHADER;
+ struct dxil_block *block = sm6->current_block;
+ struct dxil_record *temp, *record;
+ unsigned int i, count, array_len;
+ struct dxil_abbrev *abbrev;
+ uint64_t code;
+
+ if (abbrev_id >= block->abbrev_count)
+ {
+ WARN("Invalid abbreviation id %u.\n", abbrev_id);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ abbrev = block->abbrevs[abbrev_id];
+ if (!(count = abbrev->count))
+ return VKD3D_OK;
+ if (count == 1 && abbrev->is_array)
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ /* First operand is the record code. The array is included in the count, but will be done separately. */
+ count -= abbrev->is_array + 1;
+ if (!(record = vkd3d_malloc(sizeof(*record) + count * sizeof(record->operands[0]))))
+ {
+ ERR("Failed to allocate record with %u operands.\n", count);
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if (!abbrev->operands[0].read_operand(sm6, abbrev->operands[0].context, &code))
+ goto fail;
+ if (code > UINT_MAX)
+ FIXME("Truncating 64-bit record code %#"PRIx64".\n", code);
+ record->code = code;
+
+ for (i = 0; i < count; ++i)
+ if (!abbrev->operands[i + 1].read_operand(sm6, abbrev->operands[i + 1].context, &record->operands[i]))
+ goto fail;
+ record->operand_count = count;
+
+ /* An array can occur only as the last operand. */
+ if (abbrev->is_array)
+ {
+ array_len = sm6_parser_read_vbr(sm6, 6);
+ if (!(temp = vkd3d_realloc(record, sizeof(*record) + (count + array_len) * sizeof(record->operands[0]))))
+ {
+ ERR("Failed to allocate record with %u operands.\n", count + array_len);
+ ret = VKD3D_ERROR_OUT_OF_MEMORY;
+ goto fail;
+ }
+ record = temp;
+
+ for (i = 0; i < array_len; ++i)
+ {
+ if (!abbrev->operands[count + 1].read_operand(sm6, abbrev->operands[count + 1].context,
+ &record->operands[count + i]))
+ {
+ goto fail;
+ }
+ }
+ record->operand_count += array_len;
+ }
+
+ if ((ret = dxil_block_add_record(block, record)) < 0)
+ goto fail;
+
+ return VKD3D_OK;
+
+fail:
+ vkd3d_free(record);
+ return ret;
+}
+
+static enum vkd3d_result dxil_block_init(struct dxil_block *block, const struct dxil_block *parent,
+ struct sm6_parser *sm6);
+
+static enum vkd3d_result dxil_block_read(struct dxil_block *parent, struct sm6_parser *sm6)
+{
+ unsigned int reserve = (parent->id == MODULE_BLOCK) ? 12 : 2;
+ struct dxil_block *block;
+ enum vkd3d_result ret;
+
+ sm6->current_block = parent;
+
+ do
+ {
+ unsigned int abbrev_id = sm6_parser_read_bits(sm6, parent->abbrev_len);
+
+ switch (abbrev_id)
+ {
+ case END_BLOCK:
+ sm6_parser_align_32(sm6);
+ return VKD3D_OK;
+
+ case ENTER_SUBBLOCK:
+ if (parent->id != MODULE_BLOCK && parent->id != FUNCTION_BLOCK)
+ {
+ WARN("Invalid subblock parent id %u.\n", parent->id);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (!vkd3d_array_reserve((void **)&parent->child_blocks, &parent->child_block_capacity,
+ max(reserve, parent->child_block_count + 1), sizeof(*parent->child_blocks))
+ || !(block = vkd3d_calloc(1, sizeof(*block))))
+ {
+ ERR("Failed to allocate block.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if ((ret = dxil_block_init(block, parent, sm6)) < 0)
+ {
+ vkd3d_free(block);
+ return ret;
+ }
+
+ parent->child_blocks[parent->child_block_count++] = block;
+ sm6->current_block = parent;
+ break;
+
+ case DEFINE_ABBREV:
+ if ((ret = sm6_parser_add_block_abbrev(sm6)) < 0)
+ return ret;
+ break;
+
+ case UNABBREV_RECORD:
+ if ((ret = sm6_parser_read_unabbrev_record(sm6)) < 0)
+ {
+ WARN("Failed to read unabbreviated record.\n");
+ return ret;
+ }
+ break;
+
+ default:
+ if ((ret = sm6_parser_read_abbrev_record(sm6, abbrev_id - 4)) < 0)
+ {
+ WARN("Failed to read abbreviated record.\n");
+ return ret;
+ }
+ break;
+ }
+ } while (!sm6->p.failed);
+
+ return VKD3D_ERROR_INVALID_SHADER;
+}
+
+static size_t sm6_parser_compute_global_abbrev_count_for_block_id(struct sm6_parser *sm6,
+ unsigned int block_id)
+{
+ size_t i, count;
+
+ for (i = 0, count = 0; i < sm6->abbrev_count; ++i)
+ count += sm6->abbrevs[i]->block_id == block_id;
+
+ return count;
+}
+
+static void dxil_block_destroy(struct dxil_block *block)
+{
+ size_t i;
+
+ for (i = 0; i < block->record_count; ++i)
+ vkd3d_free(block->records[i]);
+ vkd3d_free(block->records);
+
+ for (i = 0; i < block->child_block_count; ++i)
+ {
+ dxil_block_destroy(block->child_blocks[i]);
+ vkd3d_free(block->child_blocks[i]);
+ }
+ vkd3d_free(block->child_blocks);
+
+ block->records = NULL;
+ block->record_count = 0;
+ block->child_blocks = NULL;
+ block->child_block_count = 0;
+}
+
+static enum vkd3d_result dxil_block_init(struct dxil_block *block, const struct dxil_block *parent,
+ struct sm6_parser *sm6)
+{
+ size_t i, abbrev_count = 0;
+ enum vkd3d_result ret;
+
+ block->parent = parent;
+ block->level = parent ? parent->level + 1 : 0;
+ block->id = sm6_parser_read_vbr(sm6, 8);
+ block->abbrev_len = sm6_parser_read_vbr(sm6, 4);
+ sm6_parser_align_32(sm6);
+ block->length = sm6_parser_read_uint32(sm6);
+ block->start = sm6->ptr - sm6->start;
+
+ if (sm6->p.failed)
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ if ((block->abbrev_count = sm6_parser_compute_global_abbrev_count_for_block_id(sm6, block->id)))
+ {
+ if (!vkd3d_array_reserve((void **)&block->abbrevs, &block->abbrev_capacity,
+ block->abbrev_count, sizeof(*block->abbrevs)))
+ {
+ ERR("Failed to allocate block abbreviations.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ for (i = 0; i < sm6->abbrev_count; ++i)
+ if (sm6->abbrevs[i]->block_id == block->id)
+ block->abbrevs[abbrev_count++] = &sm6->abbrevs[i]->abbrev;
+
+ assert(abbrev_count == block->abbrev_count);
+ }
+
+ if ((ret = dxil_block_read(block, sm6)) < 0)
+ dxil_block_destroy(block);
+
+ for (i = abbrev_count; i < block->abbrev_count; ++i)
+ vkd3d_free(block->abbrevs[i]);
+ vkd3d_free(block->abbrevs);
+ block->abbrevs = NULL;
+ block->abbrev_count = 0;
+
+ return ret;
+}
+
+static size_t dxil_block_compute_function_count(const struct dxil_block *root)
+{
+ size_t i, count;
+
+ for (i = 0, count = 0; i < root->child_block_count; ++i)
+ count += root->child_blocks[i]->id == FUNCTION_BLOCK;
+
+ return count;
+}
+
+static size_t dxil_block_compute_module_decl_count(const struct dxil_block *block)
+{
+ size_t i, count;
+
+ for (i = 0, count = 0; i < block->record_count; ++i)
+ count += block->records[i]->code == MODULE_CODE_FUNCTION;
+ return count;
+}
+
+static size_t dxil_block_compute_constants_count(const struct dxil_block *block)
+{
+ size_t i, count;
+
+ for (i = 0, count = 0; i < block->record_count; ++i)
+ count += block->records[i]->code != CST_CODE_SETTYPE;
+ return count;
+}
+
+static void dxil_global_abbrevs_cleanup(struct dxil_global_abbrev **abbrevs, size_t count)
+{
+ size_t i;
+
+ for (i = 0; i < count; ++i)
+ vkd3d_free(abbrevs[i]);
+ vkd3d_free(abbrevs);
+}
+
+static const struct dxil_block *sm6_parser_get_level_one_block(const struct sm6_parser *sm6,
+ enum bitcode_block_id id, bool *is_unique)
+{
+ const struct dxil_block *block, *found = NULL;
+ size_t i;
+
+ for (i = 0, *is_unique = true; i < sm6->root_block.child_block_count; ++i)
+ {
+ block = sm6->root_block.child_blocks[i];
+ if (block->id != id)
+ continue;
+
+ if (!found)
+ found = block;
+ else
+ *is_unique = false;
+ }
+
+ return found;
+}
+
+static char *dxil_record_to_string(const struct dxil_record *record, unsigned int offset)
+{
+ unsigned int i;
+ char *str;
+
+ assert(offset <= record->operand_count);
+ if (!(str = vkd3d_calloc(record->operand_count - offset + 1, 1)))
+ return NULL;
+
+ for (i = offset; i < record->operand_count; ++i)
+ str[i - offset] = record->operands[i];
+
+ return str;
+}
+
+static bool dxil_record_validate_operand_min_count(const struct dxil_record *record, unsigned int min_count,
+ struct sm6_parser *sm6)
+{
+ if (record->operand_count >= min_count)
+ return true;
+
+ WARN("Invalid operand count %u for code %u.\n", record->operand_count, record->code);
+ vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND_COUNT,
+ "Invalid operand count %u for record code %u.", record->operand_count, record->code);
+ return false;
+}
+
+static void dxil_record_validate_operand_max_count(const struct dxil_record *record, unsigned int max_count,
+ struct sm6_parser *sm6)
+{
+ if (record->operand_count <= max_count)
+ return;
+
+ WARN("Ignoring %u extra operands for code %u.\n", record->operand_count - max_count, record->code);
+ vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_IGNORING_OPERANDS,
+ "Ignoring %u extra operands for record code %u.", record->operand_count - max_count, record->code);
+}
+
+static bool dxil_record_validate_operand_count(const struct dxil_record *record, unsigned int min_count,
+ unsigned int max_count, struct sm6_parser *sm6)
+{
+ dxil_record_validate_operand_max_count(record, max_count, sm6);
+ return dxil_record_validate_operand_min_count(record, min_count, sm6);
+}
+
+static enum vkd3d_result sm6_parser_type_table_init(struct sm6_parser *sm6)
+{
+ const struct dxil_record *record;
+ size_t i, type_count, type_index;
+ const struct dxil_block *block;
+ char *struct_name = NULL;
+ unsigned int j, count;
+ struct sm6_type *type;
+ uint64_t type_id;
+ bool is_unique;
+
+ sm6->p.location.line = 0;
+ sm6->p.location.column = 0;
+
+ if (!(block = sm6_parser_get_level_one_block(sm6, TYPE_BLOCK, &is_unique)))
+ {
+ WARN("No type definitions found.\n");
+ return VKD3D_OK;
+ }
+ if (!is_unique)
+ WARN("Ignoring invalid extra type table(s).\n");
+
+ sm6->p.location.line = block->id;
+
+ type_count = 0;
+ for (i = 0; i < block->record_count; ++i)
+ type_count += block->records[i]->code != TYPE_CODE_NUMENTRY && block->records[i]->code != TYPE_CODE_STRUCT_NAME;
+
+ /* The type array must not be relocated. */
+ if (!(sm6->types = vkd3d_calloc(type_count, sizeof(*sm6->types))))
+ {
+ ERR("Failed to allocate type array.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ for (i = 0; i < block->record_count; ++i)
+ {
+ sm6->p.location.column = i;
+ record = block->records[i];
+
+ type = &sm6->types[sm6->type_count];
+ type_index = sm6->type_count;
+
+ switch (record->code)
+ {
+ case TYPE_CODE_ARRAY:
+ case TYPE_CODE_VECTOR:
+ if (!dxil_record_validate_operand_count(record, 2, 2, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ type->class = record->code == TYPE_CODE_ARRAY ? TYPE_CLASS_ARRAY : TYPE_CLASS_VECTOR;
+
+ if (!(type->u.array.count = record->operands[0]))
+ {
+ TRACE("Setting unbounded for type %zu.\n", type_index);
+ type->u.array.count = UINT_MAX;
+ }
+
+ if ((type_id = record->operands[1]) >= type_count)
+ {
+ WARN("Invalid contained type id %"PRIu64" for type %zu.\n", type_id, type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ type->u.array.elem_type = &sm6->types[type_id];
+ break;
+
+ case TYPE_CODE_DOUBLE:
+ dxil_record_validate_operand_max_count(record, 0, sm6);
+ type->class = TYPE_CLASS_FLOAT;
+ type->u.width = 64;
+ break;
+
+ case TYPE_CODE_FLOAT:
+ dxil_record_validate_operand_max_count(record, 0, sm6);
+ type->class = TYPE_CLASS_FLOAT;
+ type->u.width = 32;
+ break;
+
+ case TYPE_CODE_FUNCTION:
+ if (!dxil_record_validate_operand_min_count(record, 2, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+ if (record->operands[0])
+ FIXME("Unhandled vararg function type %zu.\n", type_index);
+
+ type->class = TYPE_CLASS_FUNCTION;
+
+ if ((type_id = record->operands[1]) >= type_count)
+ {
+ WARN("Invalid return type id %"PRIu64" for type %zu.\n", type_id, type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ count = record->operand_count - 2;
+ if (vkd3d_object_range_overflow(sizeof(type->u.function), count, sizeof(type->u.function->param_types[0]))
+ || !(type->u.function = vkd3d_malloc(offsetof(struct sm6_function_info, param_types[count]))))
+ {
+ ERR("Failed to allocate function parameter types.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ type->u.function->ret_type = &sm6->types[type_id];
+ type->u.function->param_count = count;
+ for (j = 0; j < count; ++j)
+ {
+ if ((type_id = record->operands[j + 2]) >= type_count)
+ {
+ WARN("Invalid parameter type id %"PRIu64" for type %zu.\n", type_id, type_index);
+ vkd3d_free(type->u.function);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ type->u.function->param_types[j] = &sm6->types[type_id];
+ }
+ break;
+
+ case TYPE_CODE_HALF:
+ dxil_record_validate_operand_max_count(record, 0, sm6);
+ type->class = TYPE_CLASS_FLOAT;
+ type->u.width = 16;
+ break;
+
+ case TYPE_CODE_INTEGER:
+ {
+ uint64_t width;
+
+ if (!dxil_record_validate_operand_count(record, 1, 1, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ type->class = TYPE_CLASS_INTEGER;
+
+ switch ((width = record->operands[0]))
+ {
+ case 1:
+ case 8:
+ case 16:
+ case 32:
+ case 64:
+ break;
+ default:
+ WARN("Invalid integer width %"PRIu64" for type %zu.\n", width, type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ type->u.width = width;
+ break;
+ }
+
+ case TYPE_CODE_LABEL:
+ type->class = TYPE_CLASS_LABEL;
+ break;
+
+ case TYPE_CODE_METADATA:
+ type->class = TYPE_CLASS_METADATA;
+ break;
+
+ case TYPE_CODE_NUMENTRY:
+ continue;
+
+ case TYPE_CODE_POINTER:
+ if (!dxil_record_validate_operand_count(record, 1, 2, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ type->class = TYPE_CLASS_POINTER;
+
+ if ((type_id = record->operands[0]) >= type_count)
+ {
+ WARN("Invalid pointee type id %"PRIu64" for type %zu.\n", type_id, type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ type->u.pointer.type = &sm6->types[type_id];
+ type->u.pointer.addr_space = (record->operand_count > 1) ? record->operands[1] : ADDRESS_SPACE_DEFAULT;
+ break;
+
+ case TYPE_CODE_STRUCT_ANON:
+ case TYPE_CODE_STRUCT_NAMED:
+ if (!dxil_record_validate_operand_min_count(record, 2, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+ if (record->code == TYPE_CODE_STRUCT_NAMED && !struct_name)
+ {
+ WARN("Missing struct name before struct type %zu.\n", type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ type->class = TYPE_CLASS_STRUCT;
+
+ count = record->operand_count - 1;
+ if (vkd3d_object_range_overflow(sizeof(type->u.struc), count, sizeof(type->u.struc->elem_types[0]))
+ || !(type->u.struc = vkd3d_malloc(offsetof(struct sm6_struct_info, elem_types[count]))))
+ {
+ ERR("Failed to allocate struct element types.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if (record->operands[0])
+ FIXME("Ignoring struct packed attribute.\n");
+
+ type->u.struc->elem_count = count;
+ for (j = 0; j < count; ++j)
+ {
+ if ((type_id = record->operands[j + 1]) >= type_count)
+ {
+ WARN("Invalid contained type id %"PRIu64" for type %zu.\n", type_id, type_index);
+ vkd3d_free(type->u.struc);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ type->u.struc->elem_types[j] = &sm6->types[type_id];
+ }
+
+ if (record->code == TYPE_CODE_STRUCT_ANON)
+ {
+ type->u.struc->name = NULL;
+ break;
+ }
+
+ type->u.struc->name = struct_name;
+ struct_name = NULL;
+ break;
+
+ case TYPE_CODE_STRUCT_NAME:
+ if (!(struct_name = dxil_record_to_string(record, 0)))
+ {
+ ERR("Failed to allocate struct name.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+ if (!struct_name[0])
+ WARN("Struct name is empty for type %zu.\n", type_index);
+ continue;
+
+ case TYPE_CODE_VOID:
+ dxil_record_validate_operand_max_count(record, 0, sm6);
+ type->class = TYPE_CLASS_VOID;
+ break;
+
+ default:
+ FIXME("Unhandled type %u at index %zu.\n", record->code, type_index);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ ++sm6->type_count;
+ }
+
+ assert(sm6->type_count == type_count);
+
+ if (struct_name)
+ {
+ WARN("Unused struct name %s.\n", struct_name);
+ vkd3d_free(struct_name);
+ }
+
+ return VKD3D_OK;
+}
+
+static inline bool sm6_type_is_void(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_VOID;
+}
+
+static inline bool sm6_type_is_integer(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_INTEGER;
+}
+
+static inline bool sm6_type_is_floating_point(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_FLOAT;
+}
+
+static inline bool sm6_type_is_numeric(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_INTEGER || type->class == TYPE_CLASS_FLOAT;
+}
+
+static inline bool sm6_type_is_pointer(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_POINTER;
+}
+
+static bool sm6_type_is_numeric_aggregate(const struct sm6_type *type)
+{
+ unsigned int i;
+
+ switch (type->class)
+ {
+ case TYPE_CLASS_ARRAY:
+ case TYPE_CLASS_VECTOR:
+ return sm6_type_is_numeric(type->u.array.elem_type);
+
+ case TYPE_CLASS_STRUCT:
+ /* Do not handle nested structs. Support can be added if they show up. */
+ for (i = 0; i < type->u.struc->elem_count; ++i)
+ if (!sm6_type_is_numeric(type->u.struc->elem_types[i]))
+ return false;
+ return true;
+
+ default:
+ return false;
+ }
+}
+
+static inline bool sm6_type_is_struct(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_STRUCT;
+}
+
+static inline bool sm6_type_is_function(const struct sm6_type *type)
+{
+ return type->class == TYPE_CLASS_FUNCTION;
+}
+
+static inline bool sm6_type_is_function_pointer(const struct sm6_type *type)
+{
+ return sm6_type_is_pointer(type) && sm6_type_is_function(type->u.pointer.type);
+}
+
+static inline bool sm6_type_is_handle(const struct sm6_type *type)
+{
+ return sm6_type_is_struct(type) && !strcmp(type->u.struc->name, "dx.types.Handle");
+}
+
+static inline const struct sm6_type *sm6_type_get_element_type(const struct sm6_type *type)
+{
+ return (type->class == TYPE_CLASS_ARRAY || type->class == TYPE_CLASS_VECTOR) ? type->u.array.elem_type : type;
+}
+
+static const struct sm6_type *sm6_type_get_pointer_to_type(const struct sm6_type *type,
+ enum bitcode_address_space addr_space, struct sm6_parser *sm6)
+{
+ size_t i, start = type - sm6->types;
+ const struct sm6_type *pointer_type;
+
+ /* DXC seems usually to place the pointer type immediately after its pointee. */
+ for (i = (start + 1) % sm6->type_count; i != start; i = (i + 1) % sm6->type_count)
+ {
+ pointer_type = &sm6->types[i];
+ if (sm6_type_is_pointer(pointer_type) && pointer_type->u.pointer.type == type
+ && pointer_type->u.pointer.addr_space == addr_space)
+ return pointer_type;
+ }
+
+ return NULL;
+}
+
+static const struct sm6_type *sm6_parser_get_type(struct sm6_parser *sm6, uint64_t type_id)
+{
+ if (type_id >= sm6->type_count)
+ {
+ WARN("Invalid type index %"PRIu64" at %zu.\n", type_id, sm6->value_count);
+ vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_ID,
+ "DXIL type id %"PRIu64" is invalid.", type_id);
+ return NULL;
+ }
+ return &sm6->types[type_id];
+}
+
+static int global_symbol_compare(const void *a, const void *b)
+{
+ return vkd3d_u32_compare(((const struct sm6_symbol *)a)->id, ((const struct sm6_symbol *)b)->id);
+}
+
+static enum vkd3d_result sm6_parser_symtab_init(struct sm6_parser *sm6)
+{
+ const struct dxil_record *record;
+ const struct dxil_block *block;
+ struct sm6_symbol *symbol;
+ size_t i, count;
+ bool is_unique;
+
+ sm6->p.location.line = 0;
+ sm6->p.location.column = 0;
+
+ if (!(block = sm6_parser_get_level_one_block(sm6, VALUE_SYMTAB_BLOCK, &is_unique)))
+ {
+ /* There should always be at least one symbol: the name of the entry point function. */
+ WARN("No value symtab block found.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (!is_unique)
+ FIXME("Ignoring extra value symtab block(s).\n");
+
+ sm6->p.location.line = block->id;
+
+ for (i = 0, count = 0; i < block->record_count; ++i)
+ count += block->records[i]->code == VST_CODE_ENTRY;
+
+ if (!(sm6->global_symbols = vkd3d_calloc(count, sizeof(*sm6->global_symbols))))
+ {
+ ERR("Failed to allocate global symbols.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ for (i = 0; i < block->record_count; ++i)
+ {
+ sm6->p.location.column = i;
+ record = block->records[i];
+
+ if (record->code != VST_CODE_ENTRY)
+ {
+ FIXME("Unhandled symtab code %u.\n", record->code);
+ continue;
+ }
+ if (!dxil_record_validate_operand_min_count(record, 1, sm6))
+ continue;
+
+ symbol = &sm6->global_symbols[sm6->global_symbol_count];
+ symbol->id = record->operands[0];
+ if (!(symbol->name = dxil_record_to_string(record, 1)))
+ {
+ ERR("Failed to allocate symbol name.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+ ++sm6->global_symbol_count;
+ }
+
+ sm6->p.location.column = block->record_count;
+
+ qsort(sm6->global_symbols, sm6->global_symbol_count, sizeof(*sm6->global_symbols), global_symbol_compare);
+ for (i = 1; i < sm6->global_symbol_count; ++i)
+ {
+ if (sm6->global_symbols[i].id == sm6->global_symbols[i - 1].id)
+ {
+ WARN("Invalid duplicate symbol id %u.\n", sm6->global_symbols[i].id);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ }
+
+ return VKD3D_OK;
+}
+
+static const char *sm6_parser_get_global_symbol_name(const struct sm6_parser *sm6, size_t id)
+{
+ size_t i, start;
+
+ /* id == array index is normally true */
+ i = start = id % sm6->global_symbol_count;
+ do
+ {
+ if (sm6->global_symbols[i].id == id)
+ return sm6->global_symbols[i].name;
+ i = (i + 1) % sm6->global_symbol_count;
+ } while (i != start);
+
+ return NULL;
+}
+
+static inline bool sm6_value_is_dx_intrinsic_dcl(const struct sm6_value *fn)
+{
+ assert(fn->value_type == VALUE_TYPE_FUNCTION);
+ return fn->u.function.is_prototype && !strncmp(fn->u.function.name, "dx.op.", 6);
+}
+
+static inline struct sm6_value *sm6_parser_get_current_value(const struct sm6_parser *sm6)
+{
+ assert(sm6->value_count < sm6->value_capacity);
+ return &sm6->values[sm6->value_count];
+}
+
+static enum vkd3d_data_type vkd3d_data_type_from_sm6_type(const struct sm6_type *type)
+{
+ if (type->class == TYPE_CLASS_INTEGER)
+ {
+ switch (type->u.width)
+ {
+ case 8:
+ return VKD3D_DATA_UINT8;
+ case 32:
+ return VKD3D_DATA_UINT;
+ default:
+ FIXME("Unhandled width %u.\n", type->u.width);
+ return VKD3D_DATA_UINT;
+ }
+ }
+ else if (type->class == TYPE_CLASS_FLOAT)
+ {
+ switch (type->u.width)
+ {
+ case 32:
+ return VKD3D_DATA_FLOAT;
+ case 64:
+ return VKD3D_DATA_DOUBLE;
+ default:
+ FIXME("Unhandled width %u.\n", type->u.width);
+ return VKD3D_DATA_FLOAT;
+ }
+ }
+
+ FIXME("Unhandled type %u.\n", type->class);
+ return VKD3D_DATA_UINT;
+}
+
+/* Recurse through the block tree while maintaining a current value count. The current
+ * count is the sum of the global count plus all declarations within the current function.
+ * Store into value_capacity the highest count seen. */
+static size_t sm6_parser_compute_max_value_count(struct sm6_parser *sm6,
+ const struct dxil_block *block, size_t value_count)
+{
+ size_t i, old_value_count = value_count;
+
+ if (block->id == MODULE_BLOCK)
+ value_count = size_add_with_overflow_check(value_count, dxil_block_compute_module_decl_count(block));
+
+ for (i = 0; i < block->child_block_count; ++i)
+ value_count = sm6_parser_compute_max_value_count(sm6, block->child_blocks[i], value_count);
+
+ switch (block->id)
+ {
+ case CONSTANTS_BLOCK:
+ /* Function local constants are contained in a child block of the function block. */
+ value_count = size_add_with_overflow_check(value_count, dxil_block_compute_constants_count(block));
+ break;
+ case FUNCTION_BLOCK:
+ /* A function must start with a block count, which emits no value. This formula is likely to
+ * overestimate the value count somewhat, but this should be no problem. */
+ value_count = size_add_with_overflow_check(value_count, max(block->record_count, 1u) - 1);
+ sm6->value_capacity = max(sm6->value_capacity, value_count);
+ /* The value count returns to its previous value after handling a function. */
+ if (value_count < SIZE_MAX)
+ value_count = old_value_count;
+ break;
+ default:
+ break;
+ }
+
+ return value_count;
+}
+
+static bool sm6_parser_declare_function(struct sm6_parser *sm6, const struct dxil_record *record)
+{
+ const unsigned int max_count = 15;
+ const struct sm6_type *ret_type;
+ struct sm6_value *fn;
+ unsigned int i, j;
+
+ if (!dxil_record_validate_operand_count(record, 8, max_count, sm6))
+ return false;
+
+ fn = sm6_parser_get_current_value(sm6);
+ fn->value_type = VALUE_TYPE_FUNCTION;
+ if (!(fn->u.function.name = sm6_parser_get_global_symbol_name(sm6, sm6->value_count)))
+ {
+ WARN("Missing symbol name for function %zu.\n", sm6->value_count);
+ fn->u.function.name = "";
+ }
+
+ if (!(fn->type = sm6_parser_get_type(sm6, record->operands[0])))
+ return false;
+ if (!sm6_type_is_function(fn->type))
+ {
+ WARN("Type is not a function.\n");
+ return false;
+ }
+ ret_type = fn->type->u.function->ret_type;
+
+ if (!(fn->type = sm6_type_get_pointer_to_type(fn->type, ADDRESS_SPACE_DEFAULT, sm6)))
+ {
+ WARN("Failed to get pointer type for type %u.\n", fn->type->class);
+ return false;
+ }
+
+ if (record->operands[1])
+ WARN("Ignoring calling convention %#"PRIx64".\n", record->operands[1]);
+
+ fn->u.function.is_prototype = !!record->operands[2];
+
+ if (record->operands[3])
+ WARN("Ignoring linkage %#"PRIx64".\n", record->operands[3]);
+
+ if (record->operands[4] > UINT_MAX)
+ WARN("Invalid attributes id %#"PRIx64".\n", record->operands[4]);
+ /* 1-based index. */
+ if ((fn->u.function.attribs_id = record->operands[4]))
+ TRACE("Ignoring function attributes.\n");
+
+ /* These always seem to be zero. */
+ for (i = 5, j = 0; i < min(record->operand_count, max_count); ++i)
+ j += !!record->operands[i];
+ if (j)
+ WARN("Ignoring %u operands.\n", j);
+
+ if (sm6_value_is_dx_intrinsic_dcl(fn) && !sm6_type_is_void(ret_type) && !sm6_type_is_numeric(ret_type)
+ && !sm6_type_is_numeric_aggregate(ret_type) && !sm6_type_is_handle(ret_type))
+ {
+ WARN("Unexpected return type for dx intrinsic function '%s'.\n", fn->u.function.name);
+ }
+
+ ++sm6->value_count;
+
+ return true;
+}
+
+static inline uint64_t decode_rotated_signed_value(uint64_t value)
+{
+ if (value != 1)
+ {
+ bool neg = value & 1;
+ value >>= 1;
+ return neg ? -value : value;
+ }
+ return value << 63;
+}
+
+static inline float bitcast_uint64_to_float(uint64_t value)
+{
+ union
+ {
+ uint32_t uint32_value;
+ float float_value;
+ } u;
+
+ u.uint32_value = value;
+ return u.float_value;
+}
+
+static inline double bitcast_uint64_to_double(uint64_t value)
+{
+ union
+ {
+ uint64_t uint64_value;
+ double double_value;
+ } u;
+
+ u.uint64_value = value;
+ return u.double_value;
+}
+
+static enum vkd3d_result sm6_parser_constants_init(struct sm6_parser *sm6, const struct dxil_block *block)
+{
+ enum vkd3d_shader_register_type reg_type = VKD3DSPR_INVALID;
+ const struct sm6_type *type, *elem_type;
+ enum vkd3d_data_type reg_data_type;
+ const struct dxil_record *record;
+ struct sm6_value *dst;
+ size_t i, value_idx;
+ uint64_t value;
+
+ for (i = 0, type = NULL; i < block->record_count; ++i)
+ {
+ sm6->p.location.column = i;
+ record = block->records[i];
+ value_idx = sm6->value_count;
+
+ if (record->code == CST_CODE_SETTYPE)
+ {
+ if (!dxil_record_validate_operand_count(record, 1, 1, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ if (!(type = sm6_parser_get_type(sm6, record->operands[0])))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ elem_type = sm6_type_get_element_type(type);
+ if (sm6_type_is_numeric(elem_type))
+ {
+ reg_data_type = vkd3d_data_type_from_sm6_type(elem_type);
+ reg_type = elem_type->u.width > 32 ? VKD3DSPR_IMMCONST64 : VKD3DSPR_IMMCONST;
+ }
+ else
+ {
+ reg_data_type = VKD3D_DATA_UNUSED;
+ reg_type = VKD3DSPR_INVALID;
+ }
+
+ if (i == block->record_count - 1)
+ WARN("Unused SETTYPE record.\n");
+
+ continue;
+ }
+
+ if (!type)
+ {
+ WARN("Constant record %zu has no type.\n", value_idx);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ dst = sm6_parser_get_current_value(sm6);
+ dst->type = type;
+ dst->value_type = VALUE_TYPE_REG;
+ dst->u.reg.type = reg_type;
+ dst->u.reg.immconst_type = VKD3D_IMMCONST_SCALAR;
+ dst->u.reg.data_type = reg_data_type;
+
+ switch (record->code)
+ {
+ case CST_CODE_NULL:
+ /* Register constant data is already zero-filled. */
+ break;
+
+ case CST_CODE_INTEGER:
+ if (!dxil_record_validate_operand_count(record, 1, 1, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ if (!sm6_type_is_integer(type))
+ {
+ WARN("Invalid integer of non-integer type %u at constant idx %zu.\n", type->class, value_idx);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ value = decode_rotated_signed_value(record->operands[0]);
+ if (type->u.width <= 32)
+ dst->u.reg.u.immconst_uint[0] = value & ((1ull << type->u.width) - 1);
+ else
+ dst->u.reg.u.immconst_uint64[0] = value;
+
+ break;
+
+ case CST_CODE_FLOAT:
+ if (!dxil_record_validate_operand_count(record, 1, 1, sm6))
+ return VKD3D_ERROR_INVALID_SHADER;
+
+ if (!sm6_type_is_floating_point(type))
+ {
+ WARN("Invalid float of non-fp type %u at constant idx %zu.\n", type->class, value_idx);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (type->u.width == 16)
+ FIXME("Half float type is not supported yet.\n");
+ else if (type->u.width == 32)
+ dst->u.reg.u.immconst_float[0] = bitcast_uint64_to_float(record->operands[0]);
+ else if (type->u.width == 64)
+ dst->u.reg.u.immconst_double[0] = bitcast_uint64_to_double(record->operands[0]);
+ else
+ vkd3d_unreachable();
+
+ break;
+
+ case CST_CODE_DATA:
+ WARN("Unhandled constant array.\n");
+ break;
+
+ default:
+ FIXME("Unhandled constant code %u.\n", record->code);
+ break;
+ }
+
+ ++sm6->value_count;
+ }
+
+ return VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6)
+{
+ const struct dxil_block *block = &sm6->root_block;
+ const struct dxil_record *record;
+ uint64_t version;
+ size_t i;
+
+ sm6->p.location.line = block->id;
+ sm6->p.location.column = 0;
+
+ for (i = 0; i < block->record_count; ++i)
+ {
+ sm6->p.location.column = i;
+ record = block->records[i];
+ switch (record->code)
+ {
+ case MODULE_CODE_FUNCTION:
+ if (!sm6_parser_declare_function(sm6, record))
+ {
+ vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_INVALID_FUNCTION_DCL,
+ "A DXIL function declaration is invalid.");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ break;
+
+ case MODULE_CODE_GLOBALVAR:
+ FIXME("Global variables are not implemented yet.\n");
+ break;
+
+ case MODULE_CODE_VERSION:
+ dxil_record_validate_operand_count(record, 1, 1, sm6);
+ if ((version = record->operands[0]) != 1)
+ {
+ FIXME("Unsupported format version %#"PRIx64".\n", version);
+ vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED_BITCODE_FORMAT,
+ "Bitcode format version %#"PRIx64" is unsupported.", version);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ break;
+
+ default:
+ break;
+ }
+ }
+
+ return VKD3D_OK;
+}
+
+static const struct sm6_value *sm6_parser_next_function_definition(struct sm6_parser *sm6)
+{
+ size_t i, count = sm6->function_count;
+
+ for (i = 0; i < sm6->value_count; ++i)
+ {
+ if (sm6_type_is_function_pointer(sm6->values[i].type) && !sm6->values[i].u.function.is_prototype && !count--)
+ break;
+ }
+ if (i == sm6->value_count)
+ return NULL;
+
+ ++sm6->function_count;
+ return &sm6->values[i];
+}
+
+static struct sm6_block *sm6_block_create()
+{
+ struct sm6_block *block = vkd3d_calloc(1, sizeof(*block));
+ return block;
+}
+
+static void sm6_parser_emit_ret(struct sm6_parser *sm6, const struct dxil_record *record,
+ struct sm6_block *code_block, struct vkd3d_shader_instruction *ins)
+{
+ if (!dxil_record_validate_operand_count(record, 0, 1, sm6))
+ return;
+
+ if (record->operand_count)
+ FIXME("Non-void return is not implemented.\n");
+
+ ins->handler_idx = VKD3DSIH_NOP;
+}
+
+static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const struct dxil_block *block,
+ struct sm6_function *function)
+{
+ struct vkd3d_shader_instruction *ins;
+ const struct dxil_record *record;
+ struct sm6_block *code_block;
+ struct sm6_value *dst;
+ size_t i, block_idx;
+ bool ret_found;
+ enum
+ {
+ RESULT_VALUE,
+ RESULT_TERMINATE,
+ } result_type;
+
+ if (sm6->function_count)
+ {
+ FIXME("Multiple functions are not supported yet.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (!(function->declaration = sm6_parser_next_function_definition(sm6)))
+ {
+ WARN("Failed to find definition to match function body.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (block->record_count < 2)
+ {
+ /* It should contain at least a block count and a RET instruction. */
+ WARN("Invalid function block record count %zu.\n", block->record_count);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (block->records[0]->code != FUNC_CODE_DECLAREBLOCKS || !block->records[0]->operand_count
+ || block->records[0]->operands[0] > UINT_MAX)
+ {
+ WARN("Block count declaration not found or invalid.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (!(function->block_count = block->records[0]->operands[0]))
+ {
+ WARN("Function contains no blocks.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (function->block_count > 1)
+ {
+ FIXME("Branched shaders are not supported yet.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (!(function->blocks[0] = sm6_block_create()))
+ {
+ ERR("Failed to allocate code block.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+ code_block = function->blocks[0];
+
+ for (i = 1, block_idx = 0, ret_found = false; i < block->record_count; ++i)
+ {
+ sm6->p.location.column = i;
+
+ /* block->record_count - 1 is the instruction count, but some instructions
+ * can emit >1 IR instruction, so extra may be used. */
+ if (!vkd3d_array_reserve((void **)&code_block->instructions, &code_block->instruction_capacity,
+ max(code_block->instruction_count + 1, block->record_count), sizeof(*code_block->instructions)))
+ {
+ ERR("Failed to allocate instructions.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ ins = &code_block->instructions[code_block->instruction_count];
+ ins->handler_idx = VKD3DSIH_INVALID;
+
+ dst = sm6_parser_get_current_value(sm6);
+ dst->type = NULL;
+ dst->value_type = VALUE_TYPE_REG;
+ result_type = RESULT_VALUE;
+
+ record = block->records[i];
+ switch (record->code)
+ {
+ case FUNC_CODE_INST_RET:
+ sm6_parser_emit_ret(sm6, record, code_block, ins);
+ result_type = RESULT_TERMINATE;
+ ret_found = true;
+ break;
+ default:
+ FIXME("Unhandled dxil instruction %u.\n", record->code);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (result_type == RESULT_TERMINATE)
+ {
+ ++block_idx;
+ code_block = (block_idx < function->block_count) ? function->blocks[block_idx] : NULL;
+ }
+ if (code_block)
+ code_block->instruction_count += ins->handler_idx != VKD3DSIH_NOP;
+ else
+ assert(ins->handler_idx == VKD3DSIH_NOP);
+ sm6->value_count += !!dst->type;
+ }
+
+ if (!ret_found)
+ {
+ WARN("Function contains no RET instruction.\n");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ return VKD3D_OK;
+}
+
+static enum vkd3d_result sm6_parser_module_init(struct sm6_parser *sm6, const struct dxil_block *block,
+ unsigned int level)
+{
+ size_t i, old_value_count = sm6->value_count;
+ struct sm6_function *function;
+ enum vkd3d_result ret;
+
+ for (i = 0; i < block->child_block_count; ++i)
+ {
+ if ((ret = sm6_parser_module_init(sm6, block->child_blocks[i], level + 1)) < 0)
+ return ret;
+ }
+
+ sm6->p.location.line = block->id;
+ sm6->p.location.column = 0;
+
+ switch (block->id)
+ {
+ case CONSTANTS_BLOCK:
+ return sm6_parser_constants_init(sm6, block);
+
+ case FUNCTION_BLOCK:
+ function = &sm6->functions[sm6->function_count];
+ if ((ret = sm6_parser_function_init(sm6, block, function)) < 0)
+ return ret;
+ /* The value index returns to its previous value after handling a function. It's usually nonzero
+ * at the start because of global constants/variables/function declarations. Function constants
+ * occur in a child block, so value_count is already saved before they are emitted. */
+ memset(&sm6->values[old_value_count], 0, (sm6->value_count - old_value_count) * sizeof(*sm6->values));
+ sm6->value_count = old_value_count;
+ break;
+
+ case BLOCKINFO_BLOCK:
+ case MODULE_BLOCK:
+ case PARAMATTR_BLOCK:
+ case PARAMATTR_GROUP_BLOCK:
+ case VALUE_SYMTAB_BLOCK:
+ case METADATA_BLOCK:
+ case METADATA_ATTACHMENT_BLOCK:
+ case TYPE_BLOCK:
+ break;
+
+ default:
+ FIXME("Unhandled block id %u.\n", block->id);
+ break;
+ }
+
+ return VKD3D_OK;
+}
+
+static void sm6_type_table_cleanup(struct sm6_type *types, size_t count)
+{
+ size_t i;
+
+ if (!types)
+ return;
+
+ for (i = 0; i < count; ++i)
+ {
+ switch (types[i].class)
+ {
+ case TYPE_CLASS_STRUCT:
+ vkd3d_free((void *)types[i].u.struc->name);
+ vkd3d_free(types[i].u.struc);
+ break;
+ case TYPE_CLASS_FUNCTION:
+ vkd3d_free(types[i].u.function);
+ break;
+ default:
+ break;
+ }
+ }
+
+ vkd3d_free(types);
+}
+
+static void sm6_symtab_cleanup(struct sm6_symbol *symbols, size_t count)
+{
+ size_t i;
+
+ for (i = 0; i < count; ++i)
+ vkd3d_free((void *)symbols[i].name);
+ vkd3d_free(symbols);
+}
+
+static void sm6_block_destroy(struct sm6_block *block)
+{
+ vkd3d_free(block->instructions);
+ vkd3d_free(block);
+}
+
+static void sm6_functions_cleanup(struct sm6_function *functions, size_t count)
+{
+ size_t i, j;
+
+ for (i = 0; i < count; ++i)
+ {
+ for (j = 0; j < functions[i].block_count; ++j)
+ sm6_block_destroy(functions[i].blocks[j]);
+ }
+ vkd3d_free(functions);
+}
+
+static void sm6_parser_destroy(struct vkd3d_shader_parser *parser)
+{
+ struct sm6_parser *sm6 = sm6_parser(parser);
+
+ dxil_block_destroy(&sm6->root_block);
+ dxil_global_abbrevs_cleanup(sm6->abbrevs, sm6->abbrev_count);
+ shader_instruction_array_destroy(&parser->instructions);
+ sm6_type_table_cleanup(sm6->types, sm6->type_count);
+ sm6_symtab_cleanup(sm6->global_symbols, sm6->global_symbol_count);
+ sm6_functions_cleanup(sm6->functions, sm6->function_count);
+ vkd3d_free(sm6->values);
+ free_shader_desc(&parser->shader_desc);
+ vkd3d_free(sm6);
+}
+
+static const struct vkd3d_shader_parser_ops sm6_parser_ops =
+{
+ .parser_destroy = sm6_parser_destroy,
+};
+
+static enum vkd3d_result sm6_parser_init(struct sm6_parser *sm6, const uint32_t *byte_code, size_t byte_code_size,
+ const char *source_name, struct vkd3d_shader_message_context *message_context)
+{
+ const struct vkd3d_shader_location location = {.source_name = source_name};
+ uint32_t version_token, dxil_version, token_count, magic;
+ unsigned int chunk_offset, chunk_size;
+ size_t count, length, function_count;
+ enum bitcode_block_abbreviation abbr;
+ struct vkd3d_shader_version version;
+ struct dxil_block *block;
+ enum vkd3d_result ret;
+
+ count = byte_code_size / sizeof(*byte_code);
+ if (count < 6)
+ {
+ WARN("Invalid data size %zu.\n", byte_code_size);
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_SIZE,
+ "DXIL chunk size %zu is smaller than the DXIL header size.", byte_code_size);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ version_token = byte_code[0];
+ TRACE("Compiler version: 0x%08x.\n", version_token);
+ token_count = byte_code[1];
+ TRACE("Token count: %u.\n", token_count);
+
+ if (token_count < 6 || count < token_count)
+ {
+ WARN("Invalid token count %u (word count %zu).\n", token_count, count);
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_CHUNK_SIZE,
+ "DXIL chunk token count %#x is invalid (word count %zu).", token_count, count);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ if (byte_code[2] != TAG_DXIL)
+ WARN("Unknown magic number 0x%08x.\n", byte_code[2]);
+
+ dxil_version = byte_code[3];
+ if (dxil_version > 0x102)
+ WARN("Unknown DXIL version: 0x%08x.\n", dxil_version);
+ else
+ TRACE("DXIL version: 0x%08x.\n", dxil_version);
+
+ chunk_offset = byte_code[4];
+ if (chunk_offset < 16 || chunk_offset >= byte_code_size)
+ {
+ WARN("Invalid bitcode chunk offset %#x (data size %zu).\n", chunk_offset, byte_code_size);
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_CHUNK_OFFSET,
+ "DXIL bitcode chunk has invalid offset %#x (data size %#zx).", chunk_offset, byte_code_size);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ chunk_size = byte_code[5];
+ if (chunk_size > byte_code_size - chunk_offset)
+ {
+ WARN("Invalid bitcode chunk size %#x (data size %zu, chunk offset %#x).\n",
+ chunk_size, byte_code_size, chunk_offset);
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_CHUNK_SIZE,
+ "DXIL bitcode chunk has invalid size %#x (data size %#zx, chunk offset %#x).",
+ chunk_size, byte_code_size, chunk_offset);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ sm6->start = (const uint32_t *)((const char*)&byte_code[2] + chunk_offset);
+ if ((magic = sm6->start[0]) != BITCODE_MAGIC)
+ {
+ WARN("Unknown magic number 0x%08x.\n", magic);
+ vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_UNKNOWN_MAGIC_NUMBER,
+ "DXIL bitcode chunk magic number 0x%08x is not the expected 0x%08x.", magic, BITCODE_MAGIC);
+ }
+
+ sm6->end = &sm6->start[(chunk_size + sizeof(*sm6->start) - 1) / sizeof(*sm6->start)];
+
+ if ((version.type = version_token >> 16) >= VKD3D_SHADER_TYPE_COUNT)
+ {
+ FIXME("Unknown shader type %#x.\n", version.type);
+ vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_UNKNOWN_SHADER_TYPE,
+ "Unknown shader type %#x.", version.type);
+ }
+
+ version.major = VKD3D_SM6_VERSION_MAJOR(version_token);
+ version.minor = VKD3D_SM6_VERSION_MINOR(version_token);
+
+ if ((abbr = sm6->start[1] & 3) != ENTER_SUBBLOCK)
+ {
+ WARN("Initial block abbreviation %u is not ENTER_SUBBLOCK.\n", abbr);
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_BITCODE,
+ "DXIL bitcode chunk has invalid initial block abbreviation %u.", abbr);
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+
+ /* Estimate instruction count to avoid reallocation in most shaders. */
+ count = max(token_count, 400) - 400;
+ vkd3d_shader_parser_init(&sm6->p, message_context, source_name, &version, &sm6_parser_ops,
+ (count + (count >> 2)) / 2u + 10);
+ sm6->ptr = &sm6->start[1];
+ sm6->bitpos = 2;
+
+ block = &sm6->root_block;
+ if ((ret = dxil_block_init(block, NULL, sm6)) < 0)
+ {
+ if (ret == VKD3D_ERROR_OUT_OF_MEMORY)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory parsing DXIL bitcode chunk.");
+ else if (ret == VKD3D_ERROR_INVALID_SHADER)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_BITCODE,
+ "DXIL bitcode chunk has invalid bitcode.");
+ else
+ vkd3d_unreachable();
+ return ret;
+ }
+
+ dxil_global_abbrevs_cleanup(sm6->abbrevs, sm6->abbrev_count);
+ sm6->abbrevs = NULL;
+ sm6->abbrev_count = 0;
+
+ length = sm6->ptr - sm6->start - block->start;
+ if (length != block->length)
+ {
+ WARN("Invalid block length %zu; expected %u.\n", length, block->length);
+ vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_INVALID_BLOCK_LENGTH,
+ "Root block ends with length %zu but indicated length is %u.", length, block->length);
+ }
+ if (sm6->ptr != sm6->end)
+ {
+ size_t expected_length = sm6->end - sm6->start;
+ length = sm6->ptr - sm6->start;
+ WARN("Invalid module length %zu; expected %zu.\n", length, expected_length);
+ vkd3d_shader_parser_warning(&sm6->p, VKD3D_SHADER_WARNING_DXIL_INVALID_MODULE_LENGTH,
+ "Module ends with length %zu but indicated length is %zu.", length, expected_length);
+ }
+
+ if ((ret = sm6_parser_type_table_init(sm6)) < 0)
+ {
+ if (ret == VKD3D_ERROR_OUT_OF_MEMORY)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory parsing DXIL type table.");
+ else if (ret == VKD3D_ERROR_INVALID_SHADER)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_TABLE,
+ "DXIL type table is invalid.");
+ else
+ vkd3d_unreachable();
+ return ret;
+ }
+
+ if ((ret = sm6_parser_symtab_init(sm6)) < 0)
+ {
+ if (ret == VKD3D_ERROR_OUT_OF_MEMORY)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory parsing DXIL value symbol table.");
+ else if (ret == VKD3D_ERROR_INVALID_SHADER)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_VALUE_SYMTAB,
+ "DXIL value symbol table is invalid.");
+ else
+ vkd3d_unreachable();
+ return ret;
+ }
+
+ function_count = dxil_block_compute_function_count(&sm6->root_block);
+ if (!(sm6->functions = vkd3d_calloc(function_count, sizeof(*sm6->functions))))
+ {
+ ERR("Failed to allocate function array.\n");
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory allocating DXIL function array.");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if (sm6_parser_compute_max_value_count(sm6, &sm6->root_block, 0) == SIZE_MAX)
+ {
+ WARN("Value array count overflowed.\n");
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_MODULE,
+ "Overflow occurred in the DXIL module value count.");
+ return VKD3D_ERROR_INVALID_SHADER;
+ }
+ if (!(sm6->values = vkd3d_calloc(sm6->value_capacity, sizeof(*sm6->values))))
+ {
+ ERR("Failed to allocate value array.\n");
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory allocating DXIL value array.");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ if ((ret = sm6_parser_globals_init(sm6)) < 0)
+ {
+ WARN("Failed to load global declarations.\n");
+ return ret;
+ }
+
+ if ((ret = sm6_parser_module_init(sm6, &sm6->root_block, 0)) < 0)
+ {
+ if (ret == VKD3D_ERROR_OUT_OF_MEMORY)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY,
+ "Out of memory parsing DXIL module.");
+ else if (ret == VKD3D_ERROR_INVALID_SHADER)
+ vkd3d_shader_error(message_context, &location, VKD3D_SHADER_ERROR_DXIL_INVALID_MODULE,
+ "DXIL module is invalid.");
+ else
+ vkd3d_unreachable();
+ return ret;
+ }
+
+ dxil_block_destroy(&sm6->root_block);
+
+ return VKD3D_OK;
+}
+
+int vkd3d_shader_sm6_parser_create(const struct vkd3d_shader_compile_info *compile_info,
+ struct vkd3d_shader_message_context *message_context, struct vkd3d_shader_parser **parser)
+{
+ struct vkd3d_shader_desc *shader_desc;
+ uint32_t *byte_code = NULL;
+ struct sm6_parser *sm6;
+ int ret;
+
+ if (!(sm6 = vkd3d_calloc(1, sizeof(*sm6))))
+ {
+ ERR("Failed to allocate parser.\n");
+ return VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+
+ shader_desc = &sm6->p.shader_desc;
+ shader_desc->is_dxil = true;
+ if ((ret = shader_extract_from_dxbc(&compile_info->source, message_context, compile_info->source_name,
+ shader_desc)) < 0)
+ {
+ WARN("Failed to extract shader, vkd3d result %d.\n", ret);
+ vkd3d_free(sm6);
+ return ret;
+ }
+
+ sm6->p.shader_desc = *shader_desc;
+ shader_desc = &sm6->p.shader_desc;
+
+ if (((uintptr_t)shader_desc->byte_code & (VKD3D_DXBC_CHUNK_ALIGNMENT - 1)))
+ {
+ /* LLVM bitcode should be 32-bit aligned, but before dxc v1.7.2207 this was not always the case in the DXBC
+ * container due to missing padding after signature names. Get an aligned copy to prevent unaligned access. */
+ if (!(byte_code = vkd3d_malloc(align(shader_desc->byte_code_size, VKD3D_DXBC_CHUNK_ALIGNMENT))))
+ ERR("Failed to allocate aligned chunk. Unaligned access will occur.\n");
+ else
+ memcpy(byte_code, shader_desc->byte_code, shader_desc->byte_code_size);
+ }
+
+ ret = sm6_parser_init(sm6, byte_code ? byte_code : shader_desc->byte_code, shader_desc->byte_code_size,
+ compile_info->source_name, message_context);
+ vkd3d_free(byte_code);
+
+ if (ret < 0)
+ {
+ WARN("Failed to initialise shader parser.\n");
+ sm6_parser_destroy(&sm6->p);
+ return ret;
+ }
+
+ *parser = &sm6->p;
+
+ return ret;
+}
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.c b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
index ba5bcfbfaf0..ab508502623 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
@@ -430,6 +430,51 @@ struct hlsl_type *hlsl_type_get_component_type(struct hlsl_ctx *ctx, struct hlsl
return type;
}
+unsigned int hlsl_type_get_component_offset(struct hlsl_ctx *ctx, struct hlsl_type *type,
+ enum hlsl_regset regset, unsigned int index)
+{
+ struct hlsl_type *next_type;
+ unsigned int offset = 0;
+ unsigned int idx;
+
+ while (!type_is_single_component(type))
+ {
+ next_type = type;
+ idx = traverse_path_from_component_index(ctx, &next_type, &index);
+
+ switch (type->class)
+ {
+ case HLSL_CLASS_SCALAR:
+ case HLSL_CLASS_VECTOR:
+ case HLSL_CLASS_MATRIX:
+ if (regset == HLSL_REGSET_NUMERIC)
+ offset += idx;
+ break;
+
+ case HLSL_CLASS_STRUCT:
+ offset += type->e.record.fields[idx].reg_offset[regset];
+ break;
+
+ case HLSL_CLASS_ARRAY:
+ if (regset == HLSL_REGSET_NUMERIC)
+ offset += idx * align(type->e.array.type->reg_size[regset], 4);
+ else
+ offset += idx * type->e.array.type->reg_size[regset];
+ break;
+
+ case HLSL_CLASS_OBJECT:
+ assert(idx == 0);
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+ type = next_type;
+ }
+
+ return offset;
+}
+
static bool init_deref(struct hlsl_ctx *ctx, struct hlsl_deref *deref, struct hlsl_ir_var *var,
unsigned int path_len)
{
@@ -524,7 +569,9 @@ struct hlsl_type *hlsl_deref_get_type(struct hlsl_ctx *ctx, const struct hlsl_de
unsigned int i;
assert(deref);
- assert(!deref->offset.node);
+
+ if (deref->offset.node)
+ return deref->data_type;
type = deref->var->data_type;
for (i = 0; i < deref->path_len; ++i)
@@ -626,6 +673,7 @@ struct hlsl_type *hlsl_new_array_type(struct hlsl_ctx *ctx, struct hlsl_type *ba
type->e.array.type = basic_type;
type->dimx = basic_type->dimx;
type->dimy = basic_type->dimy;
+ type->sampler_dim = basic_type->sampler_dim;
hlsl_type_calculate_reg_size(ctx, type);
list_add_tail(&ctx->types, &type->entry);
@@ -992,20 +1040,31 @@ struct hlsl_ir_var *hlsl_new_synthetic_var(struct hlsl_ctx *ctx, const char *tem
struct vkd3d_string_buffer *string;
struct hlsl_ir_var *var;
static LONG counter;
- const char *name;
if (!(string = hlsl_get_string_buffer(ctx)))
return NULL;
vkd3d_string_buffer_printf(string, "<%s-%u>", template, InterlockedIncrement(&counter));
- if (!(name = hlsl_strdup(ctx, string->buffer)))
- {
- hlsl_release_string_buffer(ctx, string);
- return NULL;
- }
- var = hlsl_new_var(ctx, name, type, loc, NULL, 0, NULL);
+ var = hlsl_new_synthetic_var_named(ctx, string->buffer, type, loc, true);
hlsl_release_string_buffer(ctx, string);
+ return var;
+}
+
+struct hlsl_ir_var *hlsl_new_synthetic_var_named(struct hlsl_ctx *ctx, const char *name,
+ struct hlsl_type *type, const struct vkd3d_shader_location *loc, bool dummy_scope)
+{
+ struct hlsl_ir_var *var;
+ const char *name_copy;
+
+ if (!(name_copy = hlsl_strdup(ctx, name)))
+ return NULL;
+ var = hlsl_new_var(ctx, name_copy, type, loc, NULL, 0, NULL);
if (var)
- list_add_tail(&ctx->dummy_scope->vars, &var->scope_entry);
+ {
+ if (dummy_scope)
+ list_add_tail(&ctx->dummy_scope->vars, &var->scope_entry);
+ else
+ list_add_tail(&ctx->globals->vars, &var->scope_entry);
+ }
return var;
}
@@ -1432,7 +1491,7 @@ struct hlsl_ir_node *hlsl_new_index(struct hlsl_ctx *ctx, struct hlsl_ir_node *v
}
struct hlsl_ir_node *hlsl_new_jump(struct hlsl_ctx *ctx, enum hlsl_ir_jump_type type,
- const struct vkd3d_shader_location *loc)
+ struct hlsl_ir_node *condition, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_jump *jump;
@@ -1440,6 +1499,7 @@ struct hlsl_ir_node *hlsl_new_jump(struct hlsl_ctx *ctx, enum hlsl_ir_jump_type
return NULL;
init_node(&jump->node, HLSL_IR_JUMP, NULL, loc);
jump->type = type;
+ hlsl_src_from_node(&jump->condition, condition);
return &jump->node;
}
@@ -1585,9 +1645,9 @@ static struct hlsl_ir_node *clone_if(struct hlsl_ctx *ctx, struct clone_instr_ma
return dst;
}
-static struct hlsl_ir_node *clone_jump(struct hlsl_ctx *ctx, struct hlsl_ir_jump *src)
+static struct hlsl_ir_node *clone_jump(struct hlsl_ctx *ctx, struct clone_instr_map *map, struct hlsl_ir_jump *src)
{
- return hlsl_new_jump(ctx, src->type, &src->node.loc);
+ return hlsl_new_jump(ctx, src->type, map_instr(map, src->condition.node), &src->node.loc);
}
static struct hlsl_ir_node *clone_load(struct hlsl_ctx *ctx, struct clone_instr_map *map, struct hlsl_ir_load *src)
@@ -1728,7 +1788,7 @@ static struct hlsl_ir_node *clone_instr(struct hlsl_ctx *ctx,
return clone_index(ctx, map, hlsl_ir_index(instr));
case HLSL_IR_JUMP:
- return clone_jump(ctx, hlsl_ir_jump(instr));
+ return clone_jump(ctx, map, hlsl_ir_jump(instr));
case HLSL_IR_LOAD:
return clone_load(ctx, map, hlsl_ir_load(instr));
@@ -2065,6 +2125,31 @@ struct vkd3d_string_buffer *hlsl_type_to_string(struct hlsl_ctx *ctx, const stru
}
}
+struct vkd3d_string_buffer *hlsl_component_to_string(struct hlsl_ctx *ctx, const struct hlsl_ir_var *var,
+ unsigned int index)
+{
+ struct hlsl_type *type = var->data_type, *current_type;
+ struct vkd3d_string_buffer *buffer;
+ unsigned int element_index;
+
+ if (!(buffer = hlsl_get_string_buffer(ctx)))
+ return NULL;
+
+ vkd3d_string_buffer_printf(buffer, "%s", var->name);
+
+ while (!type_is_single_component(type))
+ {
+ current_type = type;
+ element_index = traverse_path_from_component_index(ctx, &type, &index);
+ if (current_type->class == HLSL_CLASS_STRUCT)
+ vkd3d_string_buffer_printf(buffer, ".%s", current_type->e.record.fields[element_index].name);
+ else
+ vkd3d_string_buffer_printf(buffer, "[%u]", element_index);
+ }
+
+ return buffer;
+}
+
const char *debug_hlsl_type(struct hlsl_ctx *ctx, const struct hlsl_type *type)
{
struct vkd3d_string_buffer *string;
@@ -2123,18 +2208,18 @@ const char *hlsl_node_type_to_string(enum hlsl_ir_node_type type)
{
static const char * const names[] =
{
- "HLSL_IR_CALL",
- "HLSL_IR_CONSTANT",
- "HLSL_IR_EXPR",
- "HLSL_IR_IF",
- "HLSL_IR_INDEX",
- "HLSL_IR_LOAD",
- "HLSL_IR_LOOP",
- "HLSL_IR_JUMP",
- "HLSL_IR_RESOURCE_LOAD",
- "HLSL_IR_RESOURCE_STORE",
- "HLSL_IR_STORE",
- "HLSL_IR_SWIZZLE",
+ [HLSL_IR_CALL ] = "HLSL_IR_CALL",
+ [HLSL_IR_CONSTANT ] = "HLSL_IR_CONSTANT",
+ [HLSL_IR_EXPR ] = "HLSL_IR_EXPR",
+ [HLSL_IR_IF ] = "HLSL_IR_IF",
+ [HLSL_IR_INDEX ] = "HLSL_IR_INDEX",
+ [HLSL_IR_LOAD ] = "HLSL_IR_LOAD",
+ [HLSL_IR_LOOP ] = "HLSL_IR_LOOP",
+ [HLSL_IR_JUMP ] = "HLSL_IR_JUMP",
+ [HLSL_IR_RESOURCE_LOAD ] = "HLSL_IR_RESOURCE_LOAD",
+ [HLSL_IR_RESOURCE_STORE] = "HLSL_IR_RESOURCE_STORE",
+ [HLSL_IR_STORE ] = "HLSL_IR_STORE",
+ [HLSL_IR_SWIZZLE ] = "HLSL_IR_SWIZZLE",
};
if (type >= ARRAY_SIZE(names))
@@ -2146,10 +2231,11 @@ const char *hlsl_jump_type_to_string(enum hlsl_ir_jump_type type)
{
static const char * const names[] =
{
- "HLSL_IR_JUMP_BREAK",
- "HLSL_IR_JUMP_CONTINUE",
- "HLSL_IR_JUMP_DISCARD",
- "HLSL_IR_JUMP_RETURN",
+ [HLSL_IR_JUMP_BREAK] = "HLSL_IR_JUMP_BREAK",
+ [HLSL_IR_JUMP_CONTINUE] = "HLSL_IR_JUMP_CONTINUE",
+ [HLSL_IR_JUMP_DISCARD_NEG] = "HLSL_IR_JUMP_DISCARD_NEG",
+ [HLSL_IR_JUMP_DISCARD_NZ] = "HLSL_IR_JUMP_DISCARD_NZ",
+ [HLSL_IR_JUMP_RETURN] = "HLSL_IR_JUMP_RETURN",
};
assert(type < ARRAY_SIZE(names));
@@ -2337,7 +2423,11 @@ const char *debug_hlsl_expr_op(enum hlsl_ir_expr_op op)
[HLSL_OP1_COS] = "cos",
[HLSL_OP1_COS_REDUCED] = "cos_reduced",
[HLSL_OP1_DSX] = "dsx",
+ [HLSL_OP1_DSX_COARSE] = "dsx_coarse",
+ [HLSL_OP1_DSX_FINE] = "dsx_fine",
[HLSL_OP1_DSY] = "dsy",
+ [HLSL_OP1_DSY_COARSE] = "dsy_coarse",
+ [HLSL_OP1_DSY_FINE] = "dsy_fine",
[HLSL_OP1_EXP2] = "exp2",
[HLSL_OP1_FRACT] = "fract",
[HLSL_OP1_LOG2] = "log2",
@@ -2418,8 +2508,12 @@ static void dump_ir_jump(struct vkd3d_string_buffer *buffer, const struct hlsl_i
vkd3d_string_buffer_printf(buffer, "continue");
break;
- case HLSL_IR_JUMP_DISCARD:
- vkd3d_string_buffer_printf(buffer, "discard");
+ case HLSL_IR_JUMP_DISCARD_NEG:
+ vkd3d_string_buffer_printf(buffer, "discard_neg");
+ break;
+
+ case HLSL_IR_JUMP_DISCARD_NZ:
+ vkd3d_string_buffer_printf(buffer, "discard_nz");
break;
case HLSL_IR_JUMP_RETURN:
@@ -2703,6 +2797,7 @@ static void free_ir_if(struct hlsl_ir_if *if_node)
static void free_ir_jump(struct hlsl_ir_jump *jump)
{
+ hlsl_src_remove(&jump->condition);
vkd3d_free(jump);
}
@@ -3127,8 +3222,8 @@ static void declare_predefined_types(struct hlsl_ctx *ctx)
for (bt = 0; bt <= HLSL_TYPE_LAST_SCALAR; ++bt)
{
- unsigned int n_variants = 0;
const char *const *variants;
+ unsigned int n_variants;
switch (bt)
{
@@ -3148,6 +3243,8 @@ static void declare_predefined_types(struct hlsl_ctx *ctx)
break;
default:
+ n_variants = 0;
+ variants = NULL;
break;
}
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.h b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
index bce48e94b24..1a4b995abbf 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.h
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
@@ -422,6 +422,7 @@ struct hlsl_ir_var
uint32_t is_output_semantic : 1;
uint32_t is_uniform : 1;
uint32_t is_param : 1;
+ uint32_t is_separated_resource : 1;
};
/* Sized array of variables representing a function's parameters. */
@@ -502,7 +503,11 @@ enum hlsl_ir_expr_op
HLSL_OP1_COS,
HLSL_OP1_COS_REDUCED, /* Reduced range [-pi, pi] */
HLSL_OP1_DSX,
+ HLSL_OP1_DSX_COARSE,
+ HLSL_OP1_DSX_FINE,
HLSL_OP1_DSY,
+ HLSL_OP1_DSY_COARSE,
+ HLSL_OP1_DSY_FINE,
HLSL_OP1_EXP2,
HLSL_OP1_FLOOR,
HLSL_OP1_FRACT,
@@ -558,7 +563,8 @@ enum hlsl_ir_jump_type
{
HLSL_IR_JUMP_BREAK,
HLSL_IR_JUMP_CONTINUE,
- HLSL_IR_JUMP_DISCARD,
+ HLSL_IR_JUMP_DISCARD_NEG,
+ HLSL_IR_JUMP_DISCARD_NZ,
HLSL_IR_JUMP_RETURN,
};
@@ -566,6 +572,8 @@ struct hlsl_ir_jump
{
struct hlsl_ir_node node;
enum hlsl_ir_jump_type type;
+ /* Argument used for HLSL_IR_JUMP_DISCARD_NZ and HLSL_IR_JUMP_DISCARD_NEG. */
+ struct hlsl_src condition;
};
struct hlsl_ir_swizzle
@@ -600,9 +608,11 @@ struct hlsl_deref
* components, within the pertaining regset), from the start of the variable, of the part
* referenced.
* The path is lowered to this single offset -- whose value may vary between SM1 and SM4 --
- * before writing the bytecode. */
+ * before writing the bytecode.
+ * Since the type information cannot longer be retrieved from the offset alone, the type is
+ * stored in the data_type field. */
struct hlsl_src offset;
- enum hlsl_regset offset_regset;
+ struct hlsl_type *data_type;
};
struct hlsl_ir_load
@@ -803,7 +813,11 @@ struct hlsl_ctx
* Only used for SM1 profiles. */
struct hlsl_constant_defs
{
- struct hlsl_vec4 *values;
+ struct hlsl_constant_register
+ {
+ uint32_t index;
+ struct hlsl_vec4 value;
+ } *regs;
size_t count, size;
} constant_defs;
/* Number of temp. registers required for the shader to run, i.e. the largest temp register
@@ -1055,6 +1069,8 @@ const char *debug_hlsl_writemask(unsigned int writemask);
const char *debug_hlsl_swizzle(unsigned int swizzle, unsigned int count);
struct vkd3d_string_buffer *hlsl_type_to_string(struct hlsl_ctx *ctx, const struct hlsl_type *type);
+struct vkd3d_string_buffer *hlsl_component_to_string(struct hlsl_ctx *ctx, const struct hlsl_ir_var *var,
+ unsigned int index);
struct vkd3d_string_buffer *hlsl_modifiers_to_string(struct hlsl_ctx *ctx, unsigned int modifiers);
const char *hlsl_node_type_to_string(enum hlsl_ir_node_type type);
@@ -1120,7 +1136,7 @@ struct hlsl_ir_node *hlsl_new_if(struct hlsl_ctx *ctx, struct hlsl_ir_node *cond
struct hlsl_block *then_block, struct hlsl_block *else_block, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_int_constant(struct hlsl_ctx *ctx, int32_t n, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_jump(struct hlsl_ctx *ctx,
- enum hlsl_ir_jump_type type, const struct vkd3d_shader_location *loc);
+ enum hlsl_ir_jump_type type, struct hlsl_ir_node *condition, const struct vkd3d_shader_location *loc);
void hlsl_init_simple_deref_from_var(struct hlsl_deref *deref, struct hlsl_ir_var *var);
@@ -1132,6 +1148,8 @@ struct hlsl_ir_load *hlsl_new_load_parent(struct hlsl_ctx *ctx, const struct hls
const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_load_component(struct hlsl_ctx *ctx, struct hlsl_block *block,
const struct hlsl_deref *deref, unsigned int comp, const struct vkd3d_shader_location *loc);
+struct hlsl_ir_node *hlsl_add_load_component(struct hlsl_ctx *ctx, struct list *instrs,
+ struct hlsl_ir_node *var_instr, unsigned int comp, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_simple_store(struct hlsl_ctx *ctx, struct hlsl_ir_var *lhs, struct hlsl_ir_node *rhs);
struct hlsl_ir_node *hlsl_new_store_index(struct hlsl_ctx *ctx, const struct hlsl_deref *lhs,
@@ -1156,6 +1174,8 @@ struct hlsl_ir_node *hlsl_new_swizzle(struct hlsl_ctx *ctx, DWORD s, unsigned in
struct hlsl_ir_node *val, const struct vkd3d_shader_location *loc);
struct hlsl_ir_var *hlsl_new_synthetic_var(struct hlsl_ctx *ctx, const char *template,
struct hlsl_type *type, const struct vkd3d_shader_location *loc);
+struct hlsl_ir_var *hlsl_new_synthetic_var_named(struct hlsl_ctx *ctx, const char *name,
+ struct hlsl_type *type, const struct vkd3d_shader_location *loc, bool dummy_scope);
struct hlsl_type *hlsl_new_texture_type(struct hlsl_ctx *ctx, enum hlsl_sampler_dim dim, struct hlsl_type *format,
unsigned int sample_count);
struct hlsl_type *hlsl_new_uav_type(struct hlsl_ctx *ctx, enum hlsl_sampler_dim dim, struct hlsl_type *format);
@@ -1187,6 +1207,8 @@ unsigned int hlsl_type_component_count(const struct hlsl_type *type);
unsigned int hlsl_type_get_array_element_reg_size(const struct hlsl_type *type, enum hlsl_regset regset);
struct hlsl_type *hlsl_type_get_component_type(struct hlsl_ctx *ctx, struct hlsl_type *type,
unsigned int index);
+unsigned int hlsl_type_get_component_offset(struct hlsl_ctx *ctx, struct hlsl_type *type,
+ enum hlsl_regset regset, unsigned int index);
bool hlsl_type_is_row_major(const struct hlsl_type *type);
unsigned int hlsl_type_minor_size(const struct hlsl_type *type);
unsigned int hlsl_type_major_size(const struct hlsl_type *type);
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.y b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
index 0e07fe578e1..6bf87f8f916 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
@@ -53,7 +53,7 @@ struct parse_initializer
{
struct hlsl_ir_node **args;
unsigned int args_count;
- struct list *instrs;
+ struct hlsl_block *instrs;
bool braces;
};
@@ -73,6 +73,10 @@ struct parse_variable_def
struct hlsl_semantic semantic;
struct hlsl_reg_reservation reg_reservation;
struct parse_initializer initializer;
+
+ struct hlsl_type *basic_type;
+ unsigned int modifiers;
+ struct vkd3d_shader_location modifiers_loc;
};
struct parse_function
@@ -85,8 +89,8 @@ struct parse_function
struct parse_if_body
{
- struct list *then_block;
- struct list *else_block;
+ struct hlsl_block *then_block;
+ struct hlsl_block *else_block;
};
enum parse_assign_op
@@ -129,9 +133,34 @@ static void yyerror(YYLTYPE *loc, void *scanner, struct hlsl_ctx *ctx, const cha
hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, "%s", s);
}
-static struct hlsl_ir_node *node_from_list(struct list *list)
+static struct hlsl_ir_node *node_from_block(struct hlsl_block *block)
{
- return LIST_ENTRY(list_tail(list), struct hlsl_ir_node, entry);
+ return LIST_ENTRY(list_tail(&block->instrs), struct hlsl_ir_node, entry);
+}
+
+static struct list *block_to_list(struct hlsl_block *block)
+{
+ /* This is a temporary hack to ease the transition from lists to blocks.
+ * It takes advantage of the fact that an allocated hlsl_block pointer is
+ * byte-compatible with an allocated list pointer. */
+ return &block->instrs;
+}
+
+static struct hlsl_block *list_to_block(struct list *list)
+{
+ /* This is a temporary hack to ease the transition from lists to blocks.
+ * It takes advantage of the fact that an allocated hlsl_block pointer is
+ * byte-compatible with an allocated list pointer. */
+ return CONTAINING_RECORD(list, struct hlsl_block, instrs);
+}
+
+static struct hlsl_block *make_empty_block(struct hlsl_ctx *ctx)
+{
+ struct hlsl_block *block;
+
+ if ((block = hlsl_alloc(ctx, sizeof(*block))))
+ hlsl_block_init(block);
+ return block;
}
static struct list *make_empty_list(struct hlsl_ctx *ctx)
@@ -149,6 +178,12 @@ static void destroy_instr_list(struct list *list)
vkd3d_free(list);
}
+static void destroy_block(struct hlsl_block *block)
+{
+ hlsl_block_cleanup(block);
+ vkd3d_free(block);
+}
+
static bool hlsl_types_are_componentwise_compatible(struct hlsl_ctx *ctx, struct hlsl_type *src,
struct hlsl_type *dst)
{
@@ -273,9 +308,6 @@ static bool implicit_compatible_data_types(struct hlsl_ctx *ctx, struct hlsl_typ
return hlsl_types_are_componentwise_equal(ctx, src, dst);
}
-static struct hlsl_ir_node *add_load_component(struct hlsl_ctx *ctx, struct list *instrs, struct hlsl_ir_node *var_instr,
- unsigned int comp, const struct vkd3d_shader_location *loc);
-
static struct hlsl_ir_node *add_cast(struct hlsl_ctx *ctx, struct list *instrs,
struct hlsl_ir_node *node, struct hlsl_type *dst_type, const struct vkd3d_shader_location *loc)
{
@@ -333,7 +365,7 @@ static struct hlsl_ir_node *add_cast(struct hlsl_ctx *ctx, struct list *instrs,
dst_comp_type = hlsl_type_get_component_type(ctx, dst_type, dst_idx);
- if (!(component_load = add_load_component(ctx, instrs, node, src_idx, loc)))
+ if (!(component_load = hlsl_add_load_component(ctx, instrs, node, src_idx, loc)))
return NULL;
if (!(cast = hlsl_new_cast(ctx, component_load, dst_comp_type, loc)))
@@ -405,29 +437,29 @@ static DWORD add_modifiers(struct hlsl_ctx *ctx, DWORD modifiers, DWORD mod,
return modifiers | mod;
}
-static bool append_conditional_break(struct hlsl_ctx *ctx, struct list *cond_list)
+static bool append_conditional_break(struct hlsl_ctx *ctx, struct hlsl_block *cond_block)
{
struct hlsl_ir_node *condition, *not, *iff, *jump;
struct hlsl_block then_block;
/* E.g. "for (i = 0; ; ++i)". */
- if (list_empty(cond_list))
+ if (list_empty(&cond_block->instrs))
return true;
- condition = node_from_list(cond_list);
+ condition = node_from_block(cond_block);
if (!(not = hlsl_new_unary_expr(ctx, HLSL_OP1_LOGIC_NOT, condition, &condition->loc)))
return false;
- list_add_tail(cond_list, &not->entry);
+ hlsl_block_add_instr(cond_block, not);
hlsl_block_init(&then_block);
- if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_BREAK, &condition->loc)))
+ if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_BREAK, NULL, &condition->loc)))
return false;
hlsl_block_add_instr(&then_block, jump);
if (!(iff = hlsl_new_if(ctx, not, &then_block, NULL, &condition->loc)))
return false;
- list_add_tail(cond_list, &iff->entry);
+ hlsl_block_add_instr(cond_block, iff);
return true;
}
@@ -454,10 +486,10 @@ static bool attribute_list_has_duplicates(const struct parse_attribute_list *att
return false;
}
-static struct list *create_loop(struct hlsl_ctx *ctx, enum loop_type type, const struct parse_attribute_list *attributes, struct list *init, struct list *cond,
- struct list *iter, struct list *body, const struct vkd3d_shader_location *loc)
+static struct hlsl_block *create_loop(struct hlsl_ctx *ctx, enum loop_type type,
+ const struct parse_attribute_list *attributes, struct hlsl_block *init, struct hlsl_block *cond,
+ struct hlsl_block *iter, struct hlsl_block *body, const struct vkd3d_shader_location *loc)
{
- struct hlsl_block body_block;
struct hlsl_ir_node *loop;
unsigned int i;
@@ -476,7 +508,7 @@ static struct list *create_loop(struct hlsl_ctx *ctx, enum loop_type type, const
}
else
{
- hlsl_warning(ctx, loc, VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED, "Loop unrolling is not implemented.\n");
+ hlsl_warning(ctx, loc, VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED, "Loop unrolling is not implemented.");
}
}
else if (!strcmp(attr->name, "loop")
@@ -491,38 +523,34 @@ static struct list *create_loop(struct hlsl_ctx *ctx, enum loop_type type, const
}
}
- if (!init && !(init = make_empty_list(ctx)))
+ if (!init && !(init = make_empty_block(ctx)))
goto oom;
if (!append_conditional_break(ctx, cond))
goto oom;
- hlsl_block_init(&body_block);
-
- if (type != LOOP_DO_WHILE)
- list_move_tail(&body_block.instrs, cond);
-
- list_move_tail(&body_block.instrs, body);
-
if (iter)
- list_move_tail(&body_block.instrs, iter);
+ hlsl_block_add_block(body, iter);
if (type == LOOP_DO_WHILE)
- list_move_tail(&body_block.instrs, cond);
+ list_move_tail(&body->instrs, &cond->instrs);
+ else
+ list_move_head(&body->instrs, &cond->instrs);
- if (!(loop = hlsl_new_loop(ctx, &body_block, loc)))
+ if (!(loop = hlsl_new_loop(ctx, body, loc)))
goto oom;
- list_add_tail(init, &loop->entry);
+ hlsl_block_add_instr(init, loop);
- vkd3d_free(cond);
- vkd3d_free(body);
+ destroy_block(cond);
+ destroy_block(body);
+ destroy_block(iter);
return init;
oom:
- destroy_instr_list(init);
- destroy_instr_list(cond);
- destroy_instr_list(iter);
- destroy_instr_list(body);
+ destroy_block(init);
+ destroy_block(cond);
+ destroy_block(iter);
+ destroy_block(body);
return NULL;
}
@@ -539,7 +567,7 @@ static unsigned int initializer_size(const struct parse_initializer *initializer
static void free_parse_initializer(struct parse_initializer *initializer)
{
- destroy_instr_list(initializer->instrs);
+ destroy_block(initializer->instrs);
vkd3d_free(initializer->args);
}
@@ -625,7 +653,7 @@ static struct hlsl_ir_node *get_swizzle(struct hlsl_ctx *ctx, struct hlsl_ir_nod
return NULL;
}
-static bool add_return(struct hlsl_ctx *ctx, struct list *instrs,
+static bool add_return(struct hlsl_ctx *ctx, struct hlsl_block *block,
struct hlsl_ir_node *return_value, const struct vkd3d_shader_location *loc)
{
struct hlsl_type *return_type = ctx->cur_function->return_type;
@@ -637,7 +665,7 @@ static bool add_return(struct hlsl_ctx *ctx, struct list *instrs,
{
struct hlsl_ir_node *store;
- if (!(return_value = add_implicit_conversion(ctx, instrs, return_value, return_type, loc)))
+ if (!(return_value = add_implicit_conversion(ctx, block_to_list(block), return_value, return_type, loc)))
return false;
if (!(store = hlsl_new_simple_store(ctx, ctx->cur_function->return_var, return_value)))
@@ -656,14 +684,14 @@ static bool add_return(struct hlsl_ctx *ctx, struct list *instrs,
hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RETURN, "Void functions cannot return a value.");
}
- if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_RETURN, loc)))
+ if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_RETURN, NULL, loc)))
return false;
- list_add_tail(instrs, &jump->entry);
+ hlsl_block_add_instr(block, jump);
return true;
}
-static struct hlsl_ir_node *add_load_component(struct hlsl_ctx *ctx, struct list *instrs, struct hlsl_ir_node *var_instr,
+struct hlsl_ir_node *hlsl_add_load_component(struct hlsl_ctx *ctx, struct list *instrs, struct hlsl_ir_node *var_instr,
unsigned int comp, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_node *load, *store;
@@ -686,7 +714,7 @@ static struct hlsl_ir_node *add_load_component(struct hlsl_ctx *ctx, struct list
return load;
}
-static bool add_record_access(struct hlsl_ctx *ctx, struct list *instrs, struct hlsl_ir_node *record,
+static bool add_record_access(struct hlsl_ctx *ctx, struct hlsl_block *block, struct hlsl_ir_node *record,
unsigned int idx, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_node *index, *c;
@@ -695,16 +723,16 @@ static bool add_record_access(struct hlsl_ctx *ctx, struct list *instrs, struct
if (!(c = hlsl_new_uint_constant(ctx, idx, loc)))
return false;
- list_add_tail(instrs, &c->entry);
+ hlsl_block_add_instr(block, c);
if (!(index = hlsl_new_index(ctx, record, c, loc)))
return false;
- list_add_tail(instrs, &index->entry);
+ hlsl_block_add_instr(block, index);
return true;
}
-static struct hlsl_ir_node *add_binary_arithmetic_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_arithmetic_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc);
@@ -830,6 +858,16 @@ static bool shader_is_sm_5_1(const struct hlsl_ctx *ctx)
return ctx->profile->major_version == 5 && ctx->profile->minor_version >= 1;
}
+static bool shader_profile_version_ge(const struct hlsl_ctx *ctx, unsigned int major, unsigned int minor)
+{
+ return ctx->profile->major_version > major || (ctx->profile->major_version == major && ctx->profile->minor_version >= minor);
+}
+
+static bool shader_profile_version_lt(const struct hlsl_ctx *ctx, unsigned int major, unsigned int minor)
+{
+ return !shader_profile_version_ge(ctx, major, minor);
+}
+
static bool gen_struct_fields(struct hlsl_ctx *ctx, struct parse_fields *fields,
struct hlsl_type *type, unsigned int modifiers, struct list *defs)
{
@@ -1020,7 +1058,7 @@ static struct hlsl_reg_reservation parse_packoffset(struct hlsl_ctx *ctx, const
struct hlsl_reg_reservation reservation = {0};
char *endptr;
- if (ctx->profile->major_version < 4)
+ if (shader_profile_version_lt(ctx, 4, 0))
return reservation;
reservation.offset_index = strtoul(reg_string + 1, &endptr, 10);
@@ -1097,20 +1135,50 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str
{
struct hlsl_ir_constant *constant;
struct hlsl_ir_node *node;
+ struct hlsl_block expr;
unsigned int ret = 0;
bool progress;
- if (!add_implicit_conversion(ctx, &block->instrs, node_from_list(&block->instrs),
+ LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry)
+ {
+ switch (node->type)
+ {
+ case HLSL_IR_CONSTANT:
+ case HLSL_IR_EXPR:
+ case HLSL_IR_SWIZZLE:
+ case HLSL_IR_LOAD:
+ case HLSL_IR_INDEX:
+ continue;
+ case HLSL_IR_CALL:
+ case HLSL_IR_IF:
+ case HLSL_IR_LOOP:
+ case HLSL_IR_JUMP:
+ case HLSL_IR_RESOURCE_LOAD:
+ case HLSL_IR_RESOURCE_STORE:
+ case HLSL_IR_STORE:
+ hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
+ "Expected literal expression.");
+ }
+ }
+
+ if (!hlsl_clone_block(ctx, &expr, &ctx->static_initializers))
+ return 0;
+ hlsl_block_add_block(&expr, block);
+
+ if (!add_implicit_conversion(ctx, &expr.instrs, node_from_block(&expr),
hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), loc))
+ {
+ hlsl_block_cleanup(&expr);
return 0;
+ }
do
{
- progress = hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, block, NULL);
- progress |= hlsl_copy_propagation_execute(ctx, block);
+ progress = hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, &expr, NULL);
+ progress |= hlsl_copy_propagation_execute(ctx, &expr);
} while (progress);
- node = node_from_list(&block->instrs);
+ node = node_from_block(&expr);
if (node->type == HLSL_IR_CONSTANT)
{
constant = hlsl_ir_constant(node);
@@ -1119,9 +1187,11 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str
else
{
hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
- "Failed to evaluate constant expression %d.", node->type);
+ "Failed to evaluate constant expression.");
}
+ hlsl_block_cleanup(&expr);
+
return ret;
}
@@ -1284,7 +1354,7 @@ static struct hlsl_ir_node *add_expr(struct hlsl_ctx *ctx, struct list *instrs,
{
if (operands[j])
{
- if (!(load = add_load_component(ctx, instrs, operands[j], i, loc)))
+ if (!(load = hlsl_add_load_component(ctx, instrs, operands[j], i, loc)))
return NULL;
cell_operands[j] = load;
@@ -1334,23 +1404,23 @@ static void check_integer_type(struct hlsl_ctx *ctx, const struct hlsl_ir_node *
}
}
-static struct hlsl_ir_node *add_unary_arithmetic_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_unary_arithmetic_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_node *args[HLSL_MAX_OPERANDS] = {arg};
- return add_expr(ctx, instrs, op, args, arg->data_type, loc);
+ return add_expr(ctx, block_to_list(block), op, args, arg->data_type, loc);
}
-static struct hlsl_ir_node *add_unary_bitwise_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_unary_bitwise_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg, const struct vkd3d_shader_location *loc)
{
check_integer_type(ctx, arg);
- return add_unary_arithmetic_expr(ctx, instrs, op, arg, loc);
+ return add_unary_arithmetic_expr(ctx, block, op, arg, loc);
}
-static struct hlsl_ir_node *add_unary_logical_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_unary_logical_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_node *args[HLSL_MAX_OPERANDS] = {0};
@@ -1359,10 +1429,10 @@ static struct hlsl_ir_node *add_unary_logical_expr(struct hlsl_ctx *ctx, struct
bool_type = hlsl_get_numeric_type(ctx, arg->data_type->class, HLSL_TYPE_BOOL,
arg->data_type->dimx, arg->data_type->dimy);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg, bool_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(block), arg, bool_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, bool_type, loc);
+ return add_expr(ctx, block_to_list(block), op, args, bool_type, loc);
}
static struct hlsl_type *get_common_numeric_type(struct hlsl_ctx *ctx, const struct hlsl_ir_node *arg1,
@@ -1378,7 +1448,7 @@ static struct hlsl_type *get_common_numeric_type(struct hlsl_ctx *ctx, const str
return hlsl_get_numeric_type(ctx, type, base, dimx, dimy);
}
-static struct hlsl_ir_node *add_binary_arithmetic_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_arithmetic_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
@@ -1387,49 +1457,26 @@ static struct hlsl_ir_node *add_binary_arithmetic_expr(struct hlsl_ctx *ctx, str
common_type = get_common_numeric_type(ctx, arg1, arg2, loc);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg1, common_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(block), arg1, common_type, loc)))
return NULL;
- if (!(args[1] = add_implicit_conversion(ctx, instrs, arg2, common_type, loc)))
+ if (!(args[1] = add_implicit_conversion(ctx, block_to_list(block), arg2, common_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, common_type, loc);
+ return add_expr(ctx, block_to_list(block), op, args, common_type, loc);
}
-static struct list *add_binary_arithmetic_expr_merge(struct hlsl_ctx *ctx, struct list *list1, struct list *list2,
- enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
-{
- struct hlsl_ir_node *arg1 = node_from_list(list1), *arg2 = node_from_list(list2);
-
- list_move_tail(list1, list2);
- vkd3d_free(list2);
- add_binary_arithmetic_expr(ctx, list1, op, arg1, arg2, loc);
- return list1;
-}
-
-static struct hlsl_ir_node *add_binary_bitwise_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_bitwise_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
check_integer_type(ctx, arg1);
check_integer_type(ctx, arg2);
- return add_binary_arithmetic_expr(ctx, instrs, op, arg1, arg2, loc);
-}
-
-static struct list *add_binary_bitwise_expr_merge(struct hlsl_ctx *ctx, struct list *list1, struct list *list2,
- enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
-{
- struct hlsl_ir_node *arg1 = node_from_list(list1), *arg2 = node_from_list(list2);
-
- list_move_tail(list1, list2);
- vkd3d_free(list2);
- add_binary_bitwise_expr(ctx, list1, op, arg1, arg2, loc);
-
- return list1;
+ return add_binary_arithmetic_expr(ctx, block, op, arg1, arg2, loc);
}
-static struct hlsl_ir_node *add_binary_comparison_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_comparison_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
@@ -1445,27 +1492,16 @@ static struct hlsl_ir_node *add_binary_comparison_expr(struct hlsl_ctx *ctx, str
common_type = hlsl_get_numeric_type(ctx, type, base, dimx, dimy);
return_type = hlsl_get_numeric_type(ctx, type, HLSL_TYPE_BOOL, dimx, dimy);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg1, common_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(block), arg1, common_type, loc)))
return NULL;
- if (!(args[1] = add_implicit_conversion(ctx, instrs, arg2, common_type, loc)))
+ if (!(args[1] = add_implicit_conversion(ctx, block_to_list(block), arg2, common_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, return_type, loc);
-}
-
-static struct list *add_binary_comparison_expr_merge(struct hlsl_ctx *ctx, struct list *list1, struct list *list2,
- enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
-{
- struct hlsl_ir_node *arg1 = node_from_list(list1), *arg2 = node_from_list(list2);
-
- list_move_tail(list1, list2);
- vkd3d_free(list2);
- add_binary_comparison_expr(ctx, list1, op, arg1, arg2, loc);
- return list1;
+ return add_expr(ctx, block_to_list(block), op, args, return_type, loc);
}
-static struct hlsl_ir_node *add_binary_logical_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_logical_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
@@ -1479,28 +1515,16 @@ static struct hlsl_ir_node *add_binary_logical_expr(struct hlsl_ctx *ctx, struct
common_type = hlsl_get_numeric_type(ctx, type, HLSL_TYPE_BOOL, dimx, dimy);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg1, common_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(block), arg1, common_type, loc)))
return NULL;
- if (!(args[1] = add_implicit_conversion(ctx, instrs, arg2, common_type, loc)))
+ if (!(args[1] = add_implicit_conversion(ctx, block_to_list(block), arg2, common_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, common_type, loc);
+ return add_expr(ctx, block_to_list(block), op, args, common_type, loc);
}
-static struct list *add_binary_logical_expr_merge(struct hlsl_ctx *ctx, struct list *list1, struct list *list2,
- enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
-{
- struct hlsl_ir_node *arg1 = node_from_list(list1), *arg2 = node_from_list(list2);
-
- list_move_tail(list1, list2);
- vkd3d_free(list2);
- add_binary_logical_expr(ctx, list1, op, arg1, arg2, loc);
-
- return list1;
-}
-
-static struct hlsl_ir_node *add_binary_shift_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_shift_expr(struct hlsl_ctx *ctx, struct hlsl_block *block,
enum hlsl_ir_expr_op op, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
@@ -1522,28 +1546,16 @@ static struct hlsl_ir_node *add_binary_shift_expr(struct hlsl_ctx *ctx, struct l
return_type = hlsl_get_numeric_type(ctx, type, base, dimx, dimy);
integer_type = hlsl_get_numeric_type(ctx, type, HLSL_TYPE_INT, dimx, dimy);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg1, return_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(block), arg1, return_type, loc)))
return NULL;
- if (!(args[1] = add_implicit_conversion(ctx, instrs, arg2, integer_type, loc)))
+ if (!(args[1] = add_implicit_conversion(ctx, block_to_list(block), arg2, integer_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, return_type, loc);
+ return add_expr(ctx, block_to_list(block), op, args, return_type, loc);
}
-static struct list *add_binary_shift_expr_merge(struct hlsl_ctx *ctx, struct list *list1, struct list *list2,
- enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
-{
- struct hlsl_ir_node *arg1 = node_from_list(list1), *arg2 = node_from_list(list2);
-
- list_move_tail(list1, list2);
- vkd3d_free(list2);
- add_binary_shift_expr(ctx, list1, op, arg1, arg2, loc);
-
- return list1;
-}
-
-static struct hlsl_ir_node *add_binary_dot_expr(struct hlsl_ctx *ctx, struct list *instrs,
+static struct hlsl_ir_node *add_binary_dot_expr(struct hlsl_ctx *ctx, struct hlsl_block *instrs,
struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2, const struct vkd3d_shader_location *loc)
{
enum hlsl_base_type base = expr_common_base_type(arg1->data_type->base_type, arg2->data_type->base_type);
@@ -1557,8 +1569,7 @@ static struct hlsl_ir_node *add_binary_dot_expr(struct hlsl_ctx *ctx, struct lis
struct vkd3d_string_buffer *string;
if ((string = hlsl_type_to_string(ctx, arg1->data_type)))
- hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Invalid type %s.\n", string->buffer);
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE, "Invalid type %s.", string->buffer);
hlsl_release_string_buffer(ctx, string);
return NULL;
}
@@ -1568,8 +1579,7 @@ static struct hlsl_ir_node *add_binary_dot_expr(struct hlsl_ctx *ctx, struct lis
struct vkd3d_string_buffer *string;
if ((string = hlsl_type_to_string(ctx, arg2->data_type)))
- hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Invalid type %s.\n", string->buffer);
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE, "Invalid type %s.", string->buffer);
hlsl_release_string_buffer(ctx, string);
return NULL;
}
@@ -1589,13 +1599,60 @@ static struct hlsl_ir_node *add_binary_dot_expr(struct hlsl_ctx *ctx, struct lis
common_type = hlsl_get_vector_type(ctx, base, dim);
ret_type = hlsl_get_scalar_type(ctx, base);
- if (!(args[0] = add_implicit_conversion(ctx, instrs, arg1, common_type, loc)))
+ if (!(args[0] = add_implicit_conversion(ctx, block_to_list(instrs), arg1, common_type, loc)))
return NULL;
- if (!(args[1] = add_implicit_conversion(ctx, instrs, arg2, common_type, loc)))
+ if (!(args[1] = add_implicit_conversion(ctx, block_to_list(instrs), arg2, common_type, loc)))
return NULL;
- return add_expr(ctx, instrs, op, args, ret_type, loc);
+ return add_expr(ctx, block_to_list(instrs), op, args, ret_type, loc);
+}
+
+static struct hlsl_block *add_binary_expr_merge(struct hlsl_ctx *ctx, struct hlsl_block *block1,
+ struct hlsl_block *block2, enum hlsl_ir_expr_op op, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *arg1 = node_from_block(block1), *arg2 = node_from_block(block2);
+
+ hlsl_block_add_block(block1, block2);
+ destroy_block(block2);
+
+ switch (op)
+ {
+ case HLSL_OP2_ADD:
+ case HLSL_OP2_DIV:
+ case HLSL_OP2_MOD:
+ case HLSL_OP2_MUL:
+ add_binary_arithmetic_expr(ctx, block1, op, arg1, arg2, loc);
+ break;
+
+ case HLSL_OP2_BIT_AND:
+ case HLSL_OP2_BIT_OR:
+ case HLSL_OP2_BIT_XOR:
+ add_binary_bitwise_expr(ctx, block1, op, arg1, arg2, loc);
+ break;
+
+ case HLSL_OP2_LESS:
+ case HLSL_OP2_GEQUAL:
+ case HLSL_OP2_EQUAL:
+ case HLSL_OP2_NEQUAL:
+ add_binary_comparison_expr(ctx, block1, op, arg1, arg2, loc);
+ break;
+
+ case HLSL_OP2_LOGIC_AND:
+ case HLSL_OP2_LOGIC_OR:
+ add_binary_logical_expr(ctx, block1, op, arg1, arg2, loc);
+ break;
+
+ case HLSL_OP2_LSHIFT:
+ case HLSL_OP2_RSHIFT:
+ add_binary_shift_expr(ctx, block1, op, arg1, arg2, loc);
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+
+ return block1;
}
static enum hlsl_ir_expr_op op_from_assignment(enum parse_assign_op op)
@@ -1663,7 +1720,7 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct list *in
if (assign_op == ASSIGN_OP_SUB)
{
- if (!(rhs = add_unary_arithmetic_expr(ctx, instrs, HLSL_OP1_NEG, rhs, &rhs->loc)))
+ if (!(rhs = add_unary_arithmetic_expr(ctx, list_to_block(instrs), HLSL_OP1_NEG, rhs, &rhs->loc)))
return NULL;
assign_op = ASSIGN_OP_ADD;
}
@@ -1672,7 +1729,7 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct list *in
enum hlsl_ir_expr_op op = op_from_assignment(assign_op);
assert(op);
- if (!(rhs = add_binary_arithmetic_expr(ctx, instrs, op, lhs, rhs, &rhs->loc)))
+ if (!(rhs = add_binary_arithmetic_expr(ctx, list_to_block(instrs), op, lhs, rhs, &rhs->loc)))
return NULL;
}
@@ -1779,7 +1836,7 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct list *in
return NULL;
list_add_tail(instrs, &cell->entry);
- if (!(load = add_load_component(ctx, instrs, rhs, k++, &rhs->loc)))
+ if (!(load = hlsl_add_load_component(ctx, instrs, rhs, k++, &rhs->loc)))
return NULL;
if (!hlsl_init_deref_from_index_chain(ctx, &deref, cell))
@@ -1820,10 +1877,10 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct list *in
return copy;
}
-static bool add_increment(struct hlsl_ctx *ctx, struct list *instrs, bool decrement, bool post,
+static bool add_increment(struct hlsl_ctx *ctx, struct hlsl_block *block, bool decrement, bool post,
const struct vkd3d_shader_location *loc)
{
- struct hlsl_ir_node *lhs = node_from_list(instrs);
+ struct hlsl_ir_node *lhs = node_from_block(block);
struct hlsl_ir_node *one;
if (lhs->data_type->modifiers & HLSL_MODIFIER_CONST)
@@ -1832,9 +1889,9 @@ static bool add_increment(struct hlsl_ctx *ctx, struct list *instrs, bool decrem
if (!(one = hlsl_new_int_constant(ctx, 1, loc)))
return false;
- list_add_tail(instrs, &one->entry);
+ hlsl_block_add_instr(block, one);
- if (!add_assignment(ctx, instrs, lhs, decrement ? ASSIGN_OP_SUB : ASSIGN_OP_ADD, one))
+ if (!add_assignment(ctx, block_to_list(block), lhs, decrement ? ASSIGN_OP_SUB : ASSIGN_OP_ADD, one))
return false;
if (post)
@@ -1843,7 +1900,7 @@ static bool add_increment(struct hlsl_ctx *ctx, struct list *instrs, bool decrem
if (!(copy = hlsl_new_copy(ctx, lhs)))
return false;
- list_add_tail(instrs, &copy->entry);
+ hlsl_block_add_instr(block, copy);
/* Post increment/decrement expressions are considered const. */
if (!(copy->data_type = hlsl_type_clone(ctx, copy->data_type, 0, HLSL_MODIFIER_CONST)))
@@ -1853,7 +1910,7 @@ static bool add_increment(struct hlsl_ctx *ctx, struct list *instrs, bool decrem
return true;
}
-static void initialize_var_components(struct hlsl_ctx *ctx, struct list *instrs,
+static void initialize_var_components(struct hlsl_ctx *ctx, struct hlsl_block *instrs,
struct hlsl_ir_var *dst, unsigned int *store_index, struct hlsl_ir_node *src)
{
unsigned int src_comp_count = hlsl_type_component_count(src->data_type);
@@ -1868,17 +1925,17 @@ static void initialize_var_components(struct hlsl_ctx *ctx, struct list *instrs,
struct hlsl_type *dst_comp_type;
struct hlsl_block block;
- if (!(load = add_load_component(ctx, instrs, src, k, &src->loc)))
+ if (!(load = hlsl_add_load_component(ctx, block_to_list(instrs), src, k, &src->loc)))
return;
dst_comp_type = hlsl_type_get_component_type(ctx, dst->data_type, *store_index);
- if (!(conv = add_implicit_conversion(ctx, instrs, load, dst_comp_type, &src->loc)))
+ if (!(conv = add_implicit_conversion(ctx, block_to_list(instrs), load, dst_comp_type, &src->loc)))
return;
if (!hlsl_new_store_component(ctx, &block, &dst_deref, *store_index, conv))
return;
- list_move_tail(instrs, &block.instrs);
+ hlsl_block_add_block(instrs, &block);
++*store_index;
}
@@ -1924,211 +1981,234 @@ static bool type_has_numeric_components(struct hlsl_type *type)
return false;
}
-static struct list *declare_vars(struct hlsl_ctx *ctx, struct hlsl_type *basic_type,
- unsigned int modifiers, const struct vkd3d_shader_location *modifiers_loc, struct list *var_list)
+static void check_invalid_in_out_modifiers(struct hlsl_ctx *ctx, unsigned int modifiers,
+ const struct vkd3d_shader_location *loc)
{
- struct parse_variable_def *v, *v_next;
+ modifiers &= (HLSL_STORAGE_IN | HLSL_STORAGE_OUT);
+ if (modifiers)
+ {
+ struct vkd3d_string_buffer *string;
+
+ if ((string = hlsl_modifiers_to_string(ctx, modifiers)))
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
+ "Modifiers '%s' are not allowed on non-parameter variables.", string->buffer);
+ hlsl_release_string_buffer(ctx, string);
+ }
+}
+
+static void declare_var(struct hlsl_ctx *ctx, struct parse_variable_def *v)
+{
+ struct hlsl_type *basic_type = v->basic_type;
struct hlsl_ir_function_decl *func;
- unsigned int invalid_modifiers;
- struct list *statements_list;
+ struct hlsl_semantic new_semantic;
+ uint32_t modifiers = v->modifiers;
+ bool unbounded_res_array = false;
struct hlsl_ir_var *var;
struct hlsl_type *type;
bool local = true;
+ char *var_name;
+ unsigned int i;
+
+ assert(basic_type);
if (basic_type->class == HLSL_CLASS_MATRIX)
assert(basic_type->modifiers & HLSL_MODIFIERS_MAJORITY_MASK);
- if (!(statements_list = make_empty_list(ctx)))
- {
- LIST_FOR_EACH_ENTRY_SAFE(v, v_next, var_list, struct parse_variable_def, entry)
- free_parse_variable_def(v);
- vkd3d_free(var_list);
- return NULL;
- }
-
- if (!var_list)
- return statements_list;
+ type = basic_type;
- invalid_modifiers = modifiers & (HLSL_STORAGE_IN | HLSL_STORAGE_OUT);
- if (invalid_modifiers)
+ if (shader_is_sm_5_1(ctx) && type->class == HLSL_CLASS_OBJECT)
{
- struct vkd3d_string_buffer *string;
-
- if ((string = hlsl_modifiers_to_string(ctx, invalid_modifiers)))
- hlsl_error(ctx, modifiers_loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
- "Modifiers '%s' are not allowed on non-parameter variables.", string->buffer);
- hlsl_release_string_buffer(ctx, string);
+ for (i = 0; i < v->arrays.count; ++i)
+ unbounded_res_array |= (v->arrays.sizes[i] == HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT);
}
- LIST_FOR_EACH_ENTRY_SAFE(v, v_next, var_list, struct parse_variable_def, entry)
+ if (unbounded_res_array)
{
- bool unbounded_res_array = false;
- unsigned int i;
-
- type = basic_type;
-
- if (shader_is_sm_5_1(ctx) && type->class == HLSL_CLASS_OBJECT)
+ if (v->arrays.count == 1)
{
- for (i = 0; i < v->arrays.count; ++i)
- unbounded_res_array |= (v->arrays.sizes[i] == HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT);
+ hlsl_fixme(ctx, &v->loc, "Unbounded resource arrays.");
+ return;
}
-
- if (unbounded_res_array)
+ else
{
- if (v->arrays.count == 1)
- {
- hlsl_fixme(ctx, &v->loc, "Unbounded resource arrays.");
- free_parse_variable_def(v);
- continue;
- }
- else
- {
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Unbounded resource arrays cannot be multi-dimensional.");
- }
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Unbounded resource arrays cannot be multi-dimensional.");
}
- else
+ }
+ else
+ {
+ for (i = 0; i < v->arrays.count; ++i)
{
- for (i = 0; i < v->arrays.count; ++i)
+ if (v->arrays.sizes[i] == HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT)
{
- if (v->arrays.sizes[i] == HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT)
- {
- unsigned int size = initializer_size(&v->initializer);
- unsigned int elem_components = hlsl_type_component_count(type);
-
- if (i < v->arrays.count - 1)
- {
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Only innermost array size can be implicit.");
- free_parse_initializer(&v->initializer);
- v->initializer.args_count = 0;
- }
- else if (elem_components == 0)
- {
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Cannot declare an implicit size array of a size 0 type.");
- free_parse_initializer(&v->initializer);
- v->initializer.args_count = 0;
- }
- else if (size == 0)
- {
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Implicit size arrays need to be initialized.");
- free_parse_initializer(&v->initializer);
- v->initializer.args_count = 0;
+ unsigned int size = initializer_size(&v->initializer);
+ unsigned int elem_components = hlsl_type_component_count(type);
- }
- else if (size % elem_components != 0)
- {
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_WRONG_PARAMETER_COUNT,
- "Cannot initialize implicit size array with %u components, expected a multiple of %u.",
- size, elem_components);
- free_parse_initializer(&v->initializer);
- v->initializer.args_count = 0;
- }
- else
- {
- v->arrays.sizes[i] = size / elem_components;
- }
+ if (i < v->arrays.count - 1)
+ {
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Only innermost array size can be implicit.");
+ v->initializer.args_count = 0;
+ }
+ else if (elem_components == 0)
+ {
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Cannot declare an implicit size array of a size 0 type.");
+ v->initializer.args_count = 0;
+ }
+ else if (size == 0)
+ {
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Implicit size arrays need to be initialized.");
+ v->initializer.args_count = 0;
+ }
+ else if (size % elem_components != 0)
+ {
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_WRONG_PARAMETER_COUNT,
+ "Cannot initialize implicit size array with %u components, expected a multiple of %u.",
+ size, elem_components);
+ v->initializer.args_count = 0;
+ }
+ else
+ {
+ v->arrays.sizes[i] = size / elem_components;
}
- type = hlsl_new_array_type(ctx, type, v->arrays.sizes[i]);
}
+ type = hlsl_new_array_type(ctx, type, v->arrays.sizes[i]);
}
- vkd3d_free(v->arrays.sizes);
-
- if (!(var = hlsl_new_var(ctx, v->name, type, &v->loc, &v->semantic, modifiers, &v->reg_reservation)))
- {
- free_parse_variable_def(v);
- continue;
- }
+ }
- var->buffer = ctx->cur_buffer;
+ if (!(var_name = vkd3d_strdup(v->name)))
+ return;
- if (var->buffer == ctx->globals_buffer)
+ new_semantic = v->semantic;
+ if (v->semantic.name)
+ {
+ if (!(new_semantic.name = vkd3d_strdup(v->semantic.name)))
{
- if (var->reg_reservation.offset_type)
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION,
- "packoffset() is only allowed inside constant buffer declarations.");
+ vkd3d_free(var_name);
+ return;
}
+ }
- if (ctx->cur_scope == ctx->globals)
- {
- local = false;
+ if (!(var = hlsl_new_var(ctx, var_name, type, &v->loc, &new_semantic, modifiers, &v->reg_reservation)))
+ {
+ hlsl_cleanup_semantic(&new_semantic);
+ vkd3d_free(var_name);
+ return;
+ }
- if ((modifiers & HLSL_STORAGE_UNIFORM) && (modifiers & HLSL_STORAGE_STATIC))
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
- "Variable '%s' is declared as both \"uniform\" and \"static\".", var->name);
+ var->buffer = ctx->cur_buffer;
- /* Mark it as uniform. We need to do this here since synthetic
- * variables also get put in the global scope, but shouldn't be
- * considered uniforms, and we have no way of telling otherwise. */
- if (!(modifiers & HLSL_STORAGE_STATIC))
- var->storage_modifiers |= HLSL_STORAGE_UNIFORM;
+ if (var->buffer == ctx->globals_buffer)
+ {
+ if (var->reg_reservation.offset_type)
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION,
+ "packoffset() is only allowed inside constant buffer declarations.");
+ }
- if (ctx->profile->major_version < 5 && (var->storage_modifiers & HLSL_STORAGE_UNIFORM) &&
- type_has_object_components(var->data_type, true))
- {
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Target profile doesn't support objects as struct members in uniform variables.\n");
- }
+ if (ctx->cur_scope == ctx->globals)
+ {
+ local = false;
- if ((func = hlsl_get_func_decl(ctx, var->name)))
- {
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_REDEFINED,
- "'%s' is already defined as a function.", var->name);
- hlsl_note(ctx, &func->loc, VKD3D_SHADER_LOG_ERROR,
- "'%s' was previously defined here.", var->name);
- }
- }
- else
- {
- static const unsigned int invalid = HLSL_STORAGE_EXTERN | HLSL_STORAGE_SHARED
- | HLSL_STORAGE_GROUPSHARED | HLSL_STORAGE_UNIFORM;
+ if ((modifiers & HLSL_STORAGE_UNIFORM) && (modifiers & HLSL_STORAGE_STATIC))
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
+ "Variable '%s' is declared as both \"uniform\" and \"static\".", var->name);
- if (modifiers & invalid)
- {
- struct vkd3d_string_buffer *string;
+ /* Mark it as uniform. We need to do this here since synthetic
+ * variables also get put in the global scope, but shouldn't be
+ * considered uniforms, and we have no way of telling otherwise. */
+ if (!(modifiers & HLSL_STORAGE_STATIC))
+ var->storage_modifiers |= HLSL_STORAGE_UNIFORM;
- if ((string = hlsl_modifiers_to_string(ctx, modifiers & invalid)))
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
- "Modifiers '%s' are not allowed on local variables.", string->buffer);
- hlsl_release_string_buffer(ctx, string);
- }
- if (var->semantic.name)
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SEMANTIC,
- "Semantics are not allowed on local variables.");
+ if (ctx->profile->major_version < 5 && (var->storage_modifiers & HLSL_STORAGE_UNIFORM) &&
+ type_has_object_components(var->data_type, true))
+ {
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Target profile doesn't support objects as struct members in uniform variables.");
}
- if ((var->storage_modifiers & HLSL_STORAGE_STATIC) && type_has_numeric_components(var->data_type)
- && type_has_object_components(var->data_type, false))
+ if ((func = hlsl_get_func_decl(ctx, var->name)))
{
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Static variables cannot have both numeric and resource components.");
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_REDEFINED,
+ "'%s' is already defined as a function.", var->name);
+ hlsl_note(ctx, &func->loc, VKD3D_SHADER_LOG_ERROR,
+ "'%s' was previously defined here.", var->name);
}
+ }
+ else
+ {
+ static const unsigned int invalid = HLSL_STORAGE_EXTERN | HLSL_STORAGE_SHARED
+ | HLSL_STORAGE_GROUPSHARED | HLSL_STORAGE_UNIFORM;
- if ((type->modifiers & HLSL_MODIFIER_CONST) && !v->initializer.args_count
- && !(modifiers & (HLSL_STORAGE_STATIC | HLSL_STORAGE_UNIFORM)))
+ if (modifiers & invalid)
{
- hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_MISSING_INITIALIZER,
- "Const variable \"%s\" is missing an initializer.", var->name);
- hlsl_free_var(var);
- free_parse_initializer(&v->initializer);
- vkd3d_free(v);
- continue;
+ struct vkd3d_string_buffer *string;
+
+ if ((string = hlsl_modifiers_to_string(ctx, modifiers & invalid)))
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
+ "Modifiers '%s' are not allowed on local variables.", string->buffer);
+ hlsl_release_string_buffer(ctx, string);
}
+ if (var->semantic.name)
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SEMANTIC,
+ "Semantics are not allowed on local variables.");
+ }
- if (!hlsl_add_var(ctx, var, local))
+ if ((var->storage_modifiers & HLSL_STORAGE_STATIC) && type_has_numeric_components(var->data_type)
+ && type_has_object_components(var->data_type, false))
+ {
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Static variables cannot have both numeric and resource components.");
+ }
+
+ if ((type->modifiers & HLSL_MODIFIER_CONST) && !v->initializer.args_count
+ && !(modifiers & (HLSL_STORAGE_STATIC | HLSL_STORAGE_UNIFORM)))
+ {
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_MISSING_INITIALIZER,
+ "Const variable \"%s\" is missing an initializer.", var->name);
+ hlsl_free_var(var);
+ return;
+ }
+
+ if (!hlsl_add_var(ctx, var, local))
+ {
+ struct hlsl_ir_var *old = hlsl_get_var(ctx->cur_scope, var->name);
+
+ hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_REDEFINED,
+ "Variable \"%s\" was already declared in this scope.", var->name);
+ hlsl_note(ctx, &old->loc, VKD3D_SHADER_LOG_ERROR, "\"%s\" was previously declared here.", old->name);
+ hlsl_free_var(var);
+ return;
+ }
+}
+
+static struct list *initialize_vars(struct hlsl_ctx *ctx, struct list *var_list)
+{
+ struct parse_variable_def *v, *v_next;
+ struct list *statements_list;
+ struct hlsl_ir_var *var;
+ struct hlsl_type *type;
+
+ if (!(statements_list = make_empty_list(ctx)))
+ {
+ LIST_FOR_EACH_ENTRY_SAFE(v, v_next, var_list, struct parse_variable_def, entry)
{
- struct hlsl_ir_var *old = hlsl_get_var(ctx->cur_scope, var->name);
+ free_parse_variable_def(v);
+ }
+ vkd3d_free(var_list);
+ return NULL;
+ }
- hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_REDEFINED,
- "Variable \"%s\" was already declared in this scope.", var->name);
- hlsl_note(ctx, &old->loc, VKD3D_SHADER_LOG_ERROR, "\"%s\" was previously declared here.", old->name);
- hlsl_free_var(var);
- free_parse_initializer(&v->initializer);
- vkd3d_free(v);
+ LIST_FOR_EACH_ENTRY_SAFE(v, v_next, var_list, struct parse_variable_def, entry)
+ {
+ /* If this fails, the variable failed to be declared. */
+ if (!(var = hlsl_get_var(ctx->cur_scope, v->name)))
+ {
+ free_parse_variable_def(v);
continue;
}
+ type = var->data_type;
if (v->initializer.args_count)
{
@@ -2143,8 +2223,7 @@ static struct list *declare_vars(struct hlsl_ctx *ctx, struct hlsl_type *basic_t
hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_WRONG_PARAMETER_COUNT,
"Expected %u components in initializer, but got %u.",
hlsl_type_component_count(type), size);
- free_parse_initializer(&v->initializer);
- vkd3d_free(v);
+ free_parse_variable_def(v);
continue;
}
@@ -2159,16 +2238,14 @@ static struct list *declare_vars(struct hlsl_ctx *ctx, struct hlsl_type *basic_t
struct hlsl_ir_load *load = hlsl_new_var_load(ctx, var, &var->loc);
assert(v->initializer.args_count == 1);
- list_add_tail(v->initializer.instrs, &load->node.entry);
- add_assignment(ctx, v->initializer.instrs, &load->node, ASSIGN_OP_ASSIGN, v->initializer.args[0]);
+ hlsl_block_add_instr(v->initializer.instrs, &load->node);
+ add_assignment(ctx, block_to_list(v->initializer.instrs), &load->node, ASSIGN_OP_ASSIGN, v->initializer.args[0]);
}
- if (modifiers & HLSL_STORAGE_STATIC)
- list_move_tail(&ctx->static_initializers.instrs, v->initializer.instrs);
+ if (var->storage_modifiers & HLSL_STORAGE_STATIC)
+ hlsl_block_add_block(&ctx->static_initializers, v->initializer.instrs);
else
- list_move_tail(statements_list, v->initializer.instrs);
- vkd3d_free(v->initializer.args);
- vkd3d_free(v->initializer.instrs);
+ list_move_tail(statements_list, &v->initializer.instrs->instrs);
}
else if (var->storage_modifiers & HLSL_STORAGE_STATIC)
{
@@ -2178,32 +2255,33 @@ static struct list *declare_vars(struct hlsl_ctx *ctx, struct hlsl_type *basic_t
if (type_has_object_components(var->data_type, false))
{
- vkd3d_free(v);
+ free_parse_variable_def(v);
continue;
}
if (!(zero = hlsl_new_uint_constant(ctx, 0, &var->loc)))
{
- vkd3d_free(v);
+ free_parse_variable_def(v);
continue;
}
hlsl_block_add_instr(&ctx->static_initializers, zero);
if (!(cast = add_cast(ctx, &ctx->static_initializers.instrs, zero, var->data_type, &var->loc)))
{
- vkd3d_free(v);
+ free_parse_variable_def(v);
continue;
}
if (!(store = hlsl_new_simple_store(ctx, var, cast)))
{
- vkd3d_free(v);
+ free_parse_variable_def(v);
continue;
}
hlsl_block_add_instr(&ctx->static_initializers, store);
}
- vkd3d_free(v);
+ free_parse_variable_def(v);
}
+
vkd3d_free(var_list);
return statements_list;
}
@@ -2286,7 +2364,7 @@ static struct hlsl_ir_node *intrinsic_float_convert_arg(struct hlsl_ctx *ctx,
return arg;
type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_FLOAT, type->dimx, type->dimy);
- return add_implicit_conversion(ctx, params->instrs, arg, type, loc);
+ return add_implicit_conversion(ctx, block_to_list(params->instrs), arg, type, loc);
}
static bool convert_args(struct hlsl_ctx *ctx, const struct parse_initializer *params,
@@ -2298,7 +2376,7 @@ static bool convert_args(struct hlsl_ctx *ctx, const struct parse_initializer *p
{
struct hlsl_ir_node *new_arg;
- if (!(new_arg = add_implicit_conversion(ctx, params->instrs, params->args[i], type, loc)))
+ if (!(new_arg = add_implicit_conversion(ctx, block_to_list(params->instrs), params->args[i], type, loc)))
return false;
params->args[i] = new_arg;
}
@@ -2394,18 +2472,18 @@ static bool intrinsic_all(struct hlsl_ctx *ctx,
if (!(one = hlsl_new_float_constant(ctx, 1.0f, loc)))
return false;
- list_add_tail(params->instrs, &one->entry);
+ hlsl_block_add_instr(params->instrs, one);
if (!(zero = hlsl_new_float_constant(ctx, 0.0f, loc)))
return false;
- list_add_tail(params->instrs, &zero->entry);
+ hlsl_block_add_instr(params->instrs, zero);
mul = one;
count = hlsl_type_component_count(arg->data_type);
for (i = 0; i < count; ++i)
{
- if (!(load = add_load_component(ctx, params->instrs, arg, i, loc)))
+ if (!(load = hlsl_add_load_component(ctx, block_to_list(params->instrs), arg, i, loc)))
return false;
if (!(mul = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, load, mul, loc)))
@@ -2431,7 +2509,7 @@ static bool intrinsic_any(struct hlsl_ctx *ctx,
{
if (!(zero = hlsl_new_float_constant(ctx, 0.0f, loc)))
return false;
- list_add_tail(params->instrs, &zero->entry);
+ hlsl_block_add_instr(params->instrs, zero);
if (!(dot = add_binary_dot_expr(ctx, params->instrs, arg, arg, loc)))
return false;
@@ -2442,14 +2520,14 @@ static bool intrinsic_any(struct hlsl_ctx *ctx,
{
if (!(bfalse = hlsl_new_bool_constant(ctx, false, loc)))
return false;
- list_add_tail(params->instrs, &bfalse->entry);
+ hlsl_block_add_instr(params->instrs, bfalse);
or = bfalse;
count = hlsl_type_component_count(arg->data_type);
for (i = 0; i < count; ++i)
{
- if (!(load = add_load_component(ctx, params->instrs, arg, i, loc)))
+ if (!(load = hlsl_add_load_component(ctx, block_to_list(params->instrs), arg, i, loc)))
return false;
if (!(or = add_binary_bitwise_expr(ctx, params->instrs, HLSL_OP2_BIT_OR, or, load, loc)))
@@ -2491,7 +2569,7 @@ static bool intrinsic_asfloat(struct hlsl_ctx *ctx,
data_type = convert_numeric_type(ctx, data_type, HLSL_TYPE_FLOAT);
operands[0] = params->args[0];
- return add_expr(ctx, params->instrs, HLSL_OP1_REINTERPRET, operands, data_type, loc);
+ return add_expr(ctx, block_to_list(params->instrs), HLSL_OP1_REINTERPRET, operands, data_type, loc);
}
static bool intrinsic_asuint(struct hlsl_ctx *ctx,
@@ -2527,7 +2605,7 @@ static bool intrinsic_asuint(struct hlsl_ctx *ctx,
data_type = convert_numeric_type(ctx, data_type, HLSL_TYPE_UINT);
operands[0] = params->args[0];
- return add_expr(ctx, params->instrs, HLSL_OP1_REINTERPRET, operands, data_type, loc);
+ return add_expr(ctx, block_to_list(params->instrs), HLSL_OP1_REINTERPRET, operands, data_type, loc);
}
static bool intrinsic_clamp(struct hlsl_ctx *ctx,
@@ -2544,6 +2622,34 @@ static bool intrinsic_clamp(struct hlsl_ctx *ctx,
return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MIN, max, params->args[2], loc);
}
+static bool intrinsic_clip(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *condition, *jump;
+
+ if (!elementwise_intrinsic_float_convert_args(ctx, params, loc))
+ return false;
+
+ condition = params->args[0];
+
+ if (ctx->profile->major_version < 4 && hlsl_type_component_count(condition->data_type) > 4)
+ {
+ struct vkd3d_string_buffer *string;
+
+ if ((string = hlsl_type_to_string(ctx, condition->data_type)))
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Argument type cannot exceed 4 components, got type \"%s\".", string->buffer);
+ hlsl_release_string_buffer(ctx, string);
+ return false;
+ }
+
+ if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_DISCARD_NEG, condition, loc)))
+ return false;
+ hlsl_block_add_instr(params->instrs, jump);
+
+ return true;
+}
+
static bool intrinsic_cos(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
@@ -2571,34 +2677,34 @@ static bool intrinsic_cross(struct hlsl_ctx *ctx,
cast_type = hlsl_get_vector_type(ctx, base, 3);
- if (!(arg1_cast = add_implicit_conversion(ctx, params->instrs, arg1, cast_type, loc)))
+ if (!(arg1_cast = add_implicit_conversion(ctx, block_to_list(params->instrs), arg1, cast_type, loc)))
return false;
- if (!(arg2_cast = add_implicit_conversion(ctx, params->instrs, arg2, cast_type, loc)))
+ if (!(arg2_cast = add_implicit_conversion(ctx, block_to_list(params->instrs), arg2, cast_type, loc)))
return false;
if (!(arg1_swzl1 = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(Z, X, Y, Z), 3, arg1_cast, loc)))
return false;
- list_add_tail(params->instrs, &arg1_swzl1->entry);
+ hlsl_block_add_instr(params->instrs, arg1_swzl1);
if (!(arg2_swzl1 = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(Y, Z, X, Y), 3, arg2_cast, loc)))
return false;
- list_add_tail(params->instrs, &arg2_swzl1->entry);
+ hlsl_block_add_instr(params->instrs, arg2_swzl1);
if (!(mul1 = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, arg1_swzl1, arg2_swzl1, loc)))
return false;
if (!(mul1_neg = hlsl_new_unary_expr(ctx, HLSL_OP1_NEG, mul1, loc)))
return false;
- list_add_tail(params->instrs, &mul1_neg->entry);
+ hlsl_block_add_instr(params->instrs, mul1_neg);
if (!(arg1_swzl2 = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(Y, Z, X, Y), 3, arg1_cast, loc)))
return false;
- list_add_tail(params->instrs, &arg1_swzl2->entry);
+ hlsl_block_add_instr(params->instrs, arg1_swzl2);
if (!(arg2_swzl2 = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(Z, X, Y, Z), 3, arg2_cast, loc)))
return false;
- list_add_tail(params->instrs, &arg2_swzl2->entry);
+ hlsl_block_add_instr(params->instrs, arg2_swzl2);
if (!(mul2 = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, arg1_swzl2, arg2_swzl2, loc)))
return false;
@@ -2617,6 +2723,28 @@ static bool intrinsic_ddx(struct hlsl_ctx *ctx,
return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSX, arg, loc);
}
+static bool intrinsic_ddx_coarse(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *arg;
+
+ if (!(arg = intrinsic_float_convert_arg(ctx, params, params->args[0], loc)))
+ return false;
+
+ return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSX_COARSE, arg, loc);
+}
+
+static bool intrinsic_ddx_fine(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *arg;
+
+ if (!(arg = intrinsic_float_convert_arg(ctx, params, params->args[0], loc)))
+ return false;
+
+ return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSX_FINE, arg, loc);
+}
+
static bool intrinsic_ddy(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
@@ -2628,6 +2756,28 @@ static bool intrinsic_ddy(struct hlsl_ctx *ctx,
return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSY, arg, loc);
}
+static bool intrinsic_ddy_coarse(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *arg;
+
+ if (!(arg = intrinsic_float_convert_arg(ctx, params, params->args[0], loc)))
+ return false;
+
+ return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSY_COARSE, arg, loc);
+}
+
+static bool intrinsic_ddy_fine(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *arg;
+
+ if (!(arg = intrinsic_float_convert_arg(ctx, params, params->args[0], loc)))
+ return false;
+
+ return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_DSY_FINE, arg, loc);
+}
+
static bool intrinsic_distance(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
@@ -2668,7 +2818,7 @@ static bool intrinsic_exp(struct hlsl_ctx *ctx,
/* 1/ln(2) */
if (!(coeff = hlsl_new_float_constant(ctx, 1.442695f, loc)))
return false;
- list_add_tail(params->instrs, &coeff->entry);
+ hlsl_block_add_instr(params->instrs, coeff);
if (!(mul = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, coeff, params->args[0], loc)))
return false;
@@ -2715,7 +2865,7 @@ static bool intrinsic_fmod(struct hlsl_ctx *ctx, const struct parse_initializer
if (!(zero = hlsl_new_constant(ctx, div->data_type, &zero_value, loc)))
return false;
- list_add_tail(params->instrs, &zero->entry);
+ hlsl_block_add_instr(params->instrs, zero);
if (!(abs = add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_ABS, div, loc)))
return false;
@@ -2729,7 +2879,7 @@ static bool intrinsic_fmod(struct hlsl_ctx *ctx, const struct parse_initializer
if (!(ge = add_binary_comparison_expr(ctx, params->instrs, HLSL_OP2_GEQUAL, div, zero, loc)))
return false;
- if (!(select = hlsl_add_conditional(ctx, params->instrs, ge, frac, neg_frac)))
+ if (!(select = hlsl_add_conditional(ctx, block_to_list(params->instrs), ge, frac, neg_frac)))
return false;
return !!add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, select, y, loc);
@@ -2806,7 +2956,7 @@ static bool intrinsic_lerp(struct hlsl_ctx *ctx,
}
static struct hlsl_ir_node * add_pow_expr(struct hlsl_ctx *ctx,
- struct list *instrs, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
+ struct hlsl_block *instrs, struct hlsl_ir_node *arg1, struct hlsl_ir_node *arg2,
const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_node *log, *mul;
@@ -2861,15 +3011,15 @@ static bool intrinsic_lit(struct hlsl_ctx *ctx,
init_value.u[3].f = 1.0f;
if (!(init = hlsl_new_constant(ctx, ret_type, &init_value, loc)))
return false;
- list_add_tail(params->instrs, &init->entry);
+ hlsl_block_add_instr(params->instrs, init);
if (!(store = hlsl_new_simple_store(ctx, var, init)))
return false;
- list_add_tail(params->instrs, &store->entry);
+ hlsl_block_add_instr(params->instrs, store);
if (!(zero = hlsl_new_float_constant(ctx, 0.0f, loc)))
return false;
- list_add_tail(params->instrs, &zero->entry);
+ hlsl_block_add_instr(params->instrs, zero);
/* Diffuse component. */
if (!(diffuse = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MAX, n_l, zero, loc)))
@@ -2877,7 +3027,7 @@ static bool intrinsic_lit(struct hlsl_ctx *ctx,
if (!hlsl_new_store_component(ctx, &block, &var_deref, 1, diffuse))
return false;
- list_move_tail(params->instrs, &block.instrs);
+ hlsl_block_add_block(params->instrs, &block);
/* Specular component. */
if (!(n_h_neg = add_binary_comparison_expr(ctx, params->instrs, HLSL_OP2_LESS, n_h, zero, loc)))
@@ -2892,16 +3042,16 @@ static bool intrinsic_lit(struct hlsl_ctx *ctx,
if (!(specular_pow = add_pow_expr(ctx, params->instrs, n_h, m, loc)))
return false;
- if (!(load = hlsl_add_conditional(ctx, params->instrs, specular_or, zero, specular_pow)))
+ if (!(load = hlsl_add_conditional(ctx, block_to_list(params->instrs), specular_or, zero, specular_pow)))
return false;
if (!hlsl_new_store_component(ctx, &block, &var_deref, 2, load))
return false;
- list_move_tail(params->instrs, &block.instrs);
+ hlsl_block_add_block(params->instrs, &block);
if (!(var_load = hlsl_new_var_load(ctx, var, loc)))
return false;
- list_add_tail(params->instrs, &var_load->node.entry);
+ hlsl_block_add_instr(params->instrs, &var_load->node);
return true;
}
@@ -3013,10 +3163,10 @@ static bool intrinsic_mul(struct hlsl_ctx *ctx,
ret_type = hlsl_get_scalar_type(ctx, base);
}
- if (!(cast1 = add_implicit_conversion(ctx, params->instrs, arg1, cast_type1, loc)))
+ if (!(cast1 = add_implicit_conversion(ctx, block_to_list(params->instrs), arg1, cast_type1, loc)))
return false;
- if (!(cast2 = add_implicit_conversion(ctx, params->instrs, arg2, cast_type2, loc)))
+ if (!(cast2 = add_implicit_conversion(ctx, block_to_list(params->instrs), arg2, cast_type2, loc)))
return false;
if (!(var = hlsl_new_synthetic_var(ctx, "mul", matrix_type, loc)))
@@ -3034,10 +3184,12 @@ static bool intrinsic_mul(struct hlsl_ctx *ctx,
{
struct hlsl_ir_node *value1, *value2, *mul;
- if (!(value1 = add_load_component(ctx, params->instrs, cast1, j * cast1->data_type->dimx + k, loc)))
+ if (!(value1 = hlsl_add_load_component(ctx, block_to_list(params->instrs),
+ cast1, j * cast1->data_type->dimx + k, loc)))
return false;
- if (!(value2 = add_load_component(ctx, params->instrs, cast2, k * cast2->data_type->dimx + i, loc)))
+ if (!(value2 = hlsl_add_load_component(ctx, block_to_list(params->instrs),
+ cast2, k * cast2->data_type->dimx + i, loc)))
return false;
if (!(mul = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, value1, value2, loc)))
@@ -3056,15 +3208,15 @@ static bool intrinsic_mul(struct hlsl_ctx *ctx,
if (!hlsl_new_store_component(ctx, &block, &var_deref, j * matrix_type->dimx + i, instr))
return false;
- list_move_tail(params->instrs, &block.instrs);
+ hlsl_block_add_block(params->instrs, &block);
}
}
if (!(load = hlsl_new_var_load(ctx, var, loc)))
return false;
- list_add_tail(params->instrs, &load->node.entry);
+ hlsl_block_add_instr(params->instrs, &load->node);
- return !!add_implicit_conversion(ctx, params->instrs, &load->node, ret_type, loc);
+ return !!add_implicit_conversion(ctx, block_to_list(params->instrs), &load->node, ret_type, loc);
}
static bool intrinsic_normalize(struct hlsl_ctx *ctx,
@@ -3169,14 +3321,14 @@ static bool intrinsic_sign(struct hlsl_ctx *ctx,
if (!(zero = hlsl_new_constant(ctx, hlsl_get_scalar_type(ctx, arg->data_type->base_type), &zero_value, loc)))
return false;
- list_add_tail(params->instrs, &zero->entry);
+ hlsl_block_add_instr(params->instrs, zero);
/* Check if 0 < arg, cast bool to int */
if (!(lt = add_binary_comparison_expr(ctx, params->instrs, HLSL_OP2_LESS, zero, arg, loc)))
return false;
- if (!(op1 = add_implicit_conversion(ctx, params->instrs, lt, int_type, loc)))
+ if (!(op1 = add_implicit_conversion(ctx, block_to_list(params->instrs), lt, int_type, loc)))
return false;
/* Check if arg < 0, cast bool to int and invert (meaning true is -1) */
@@ -3184,7 +3336,7 @@ static bool intrinsic_sign(struct hlsl_ctx *ctx,
if (!(lt = add_binary_comparison_expr(ctx, params->instrs, HLSL_OP2_LESS, arg, zero, loc)))
return false;
- if (!(op2 = add_implicit_conversion(ctx, params->instrs, lt, int_type, loc)))
+ if (!(op2 = add_implicit_conversion(ctx, block_to_list(params->instrs), lt, int_type, loc)))
return false;
if (!(neg = add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_NEG, op2, loc)))
@@ -3229,7 +3381,7 @@ static bool intrinsic_smoothstep(struct hlsl_ctx *ctx,
if (!(one = hlsl_new_float_constant(ctx, 1.0, loc)))
return false;
- list_add_tail(params->instrs, &one->entry);
+ hlsl_block_add_instr(params->instrs, one);
if (!(p_denom = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_DIV, one, p_denom, loc)))
return false;
@@ -3242,11 +3394,11 @@ static bool intrinsic_smoothstep(struct hlsl_ctx *ctx,
if (!(minus_two = hlsl_new_float_constant(ctx, -2.0, loc)))
return false;
- list_add_tail(params->instrs, &minus_two->entry);
+ hlsl_block_add_instr(params->instrs, minus_two);
if (!(three = hlsl_new_float_constant(ctx, 3.0, loc)))
return false;
- list_add_tail(params->instrs, &three->entry);
+ hlsl_block_add_instr(params->instrs, three);
if (!(res = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, minus_two, p, loc)))
return false;
@@ -3289,7 +3441,7 @@ static bool intrinsic_step(struct hlsl_ctx *ctx,
type = ge->data_type;
type = hlsl_get_numeric_type(ctx, type->class, HLSL_TYPE_FLOAT, type->dimx, type->dimy);
- return !!add_implicit_conversion(ctx, params->instrs, ge, type, loc);
+ return !!add_implicit_conversion(ctx, block_to_list(params->instrs), ge, type, loc);
}
static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *params,
@@ -3308,7 +3460,7 @@ static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *
if (params->args_count == 4)
{
- hlsl_fixme(ctx, loc, "Samples with gradients are not implemented.\n");
+ hlsl_fixme(ctx, loc, "Samples with gradients are not implemented.");
}
sampler_type = params->args[0]->data_type;
@@ -3324,7 +3476,7 @@ static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *
hlsl_release_string_buffer(ctx, string);
}
- if (!(coords = add_implicit_conversion(ctx, params->instrs, params->args[1],
+ if (!(coords = add_implicit_conversion(ctx, block_to_list(params->instrs), params->args[1],
hlsl_get_vector_type(ctx, HLSL_TYPE_FLOAT, hlsl_sampler_dim_count(dim)), loc)))
coords = params->args[1];
@@ -3335,7 +3487,7 @@ static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *
if (!(load = hlsl_new_resource_load(ctx, &load_params, loc)))
return false;
- list_add_tail(params->instrs, &load->entry);
+ hlsl_block_add_instr(params->instrs, load);
return true;
}
@@ -3369,7 +3521,7 @@ static bool intrinsic_transpose(struct hlsl_ctx *ctx,
if ((string = hlsl_type_to_string(ctx, arg_type)))
hlsl_error(ctx, &arg->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
- "Wrong type for argument 1 of transpose(): expected a matrix or scalar type, but got '%s'.\n",
+ "Wrong type for argument 1 of transpose(): expected a matrix or scalar type, but got '%s'.",
string->buffer);
hlsl_release_string_buffer(ctx, string);
return false;
@@ -3377,7 +3529,7 @@ static bool intrinsic_transpose(struct hlsl_ctx *ctx,
if (arg_type->class == HLSL_CLASS_SCALAR)
{
- list_add_tail(params->instrs, &arg->entry);
+ hlsl_block_add_instr(params->instrs, arg);
return true;
}
@@ -3393,18 +3545,18 @@ static bool intrinsic_transpose(struct hlsl_ctx *ctx,
{
struct hlsl_block block;
- if (!(load = add_load_component(ctx, params->instrs, arg, j * arg->data_type->dimx + i, loc)))
+ if (!(load = hlsl_add_load_component(ctx, block_to_list(params->instrs), arg, j * arg->data_type->dimx + i, loc)))
return false;
if (!hlsl_new_store_component(ctx, &block, &var_deref, i * var->data_type->dimx + j, load))
return false;
- list_move_tail(params->instrs, &block.instrs);
+ hlsl_block_add_block(params->instrs, &block);
}
}
if (!(var_load = hlsl_new_var_load(ctx, var, loc)))
return false;
- list_add_tail(params->instrs, &var_load->node.entry);
+ hlsl_block_add_instr(params->instrs, &var_load->node);
return true;
}
@@ -3444,13 +3596,13 @@ static bool intrinsic_d3dcolor_to_ubyte4(struct hlsl_ctx *ctx,
if (!(c = hlsl_new_float_constant(ctx, 255.0f + (0.5f / 256.0f), loc)))
return false;
- list_add_tail(params->instrs, &c->entry);
+ hlsl_block_add_instr(params->instrs, c);
if (arg_type->class == HLSL_CLASS_VECTOR)
{
if (!(swizzle = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(Z, Y, X, W), 4, arg, loc)))
return false;
- list_add_tail(params->instrs, &swizzle->entry);
+ hlsl_block_add_instr(params->instrs, swizzle);
arg = swizzle;
}
@@ -3458,7 +3610,7 @@ static bool intrinsic_d3dcolor_to_ubyte4(struct hlsl_ctx *ctx,
if (!(ret = add_binary_arithmetic_expr(ctx, params->instrs, HLSL_OP2_MUL, arg, c, loc)))
return false;
- if (ctx->profile->major_version >= 4)
+ if (shader_profile_version_ge(ctx, 4, 0))
return !!add_unary_arithmetic_expr(ctx, params->instrs, HLSL_OP1_TRUNC, ret, loc);
return true;
@@ -3482,10 +3634,15 @@ intrinsic_functions[] =
{"asfloat", 1, true, intrinsic_asfloat},
{"asuint", -1, true, intrinsic_asuint},
{"clamp", 3, true, intrinsic_clamp},
+ {"clip", 1, true, intrinsic_clip},
{"cos", 1, true, intrinsic_cos},
{"cross", 2, true, intrinsic_cross},
{"ddx", 1, true, intrinsic_ddx},
+ {"ddx_coarse", 1, true, intrinsic_ddx_coarse},
+ {"ddx_fine", 1, true, intrinsic_ddx_fine},
{"ddy", 1, true, intrinsic_ddy},
+ {"ddy_coarse", 1, true, intrinsic_ddy_coarse},
+ {"ddy_fine", 1, true, intrinsic_ddy_fine},
{"distance", 2, true, intrinsic_distance},
{"dot", 2, true, intrinsic_dot},
{"exp", 1, true, intrinsic_exp},
@@ -3549,7 +3706,7 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
{
struct hlsl_ir_node *cast;
- if (!(cast = add_cast(ctx, args->instrs, arg, param->data_type, &arg->loc)))
+ if (!(cast = add_cast(ctx, block_to_list(args->instrs), arg, param->data_type, &arg->loc)))
goto fail;
args->args[i] = cast;
arg = cast;
@@ -3561,13 +3718,13 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
if (!(store = hlsl_new_simple_store(ctx, param, arg)))
goto fail;
- list_add_tail(args->instrs, &store->entry);
+ hlsl_block_add_instr(args->instrs, store);
}
}
if (!(call = hlsl_new_call(ctx, decl, loc)))
goto fail;
- list_add_tail(args->instrs, &call->entry);
+ hlsl_block_add_instr(args->instrs, call);
for (i = 0; i < decl->parameters.count; ++i)
{
@@ -3584,9 +3741,9 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
if (!(load = hlsl_new_var_load(ctx, param, &arg->loc)))
goto fail;
- list_add_tail(args->instrs, &load->node.entry);
+ hlsl_block_add_instr(args->instrs, &load->node);
- if (!add_assignment(ctx, args->instrs, arg, ASSIGN_OP_ASSIGN, &load->node))
+ if (!add_assignment(ctx, block_to_list(args->instrs), arg, ASSIGN_OP_ASSIGN, &load->node))
goto fail;
}
}
@@ -3597,7 +3754,7 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
if (!(load = hlsl_new_var_load(ctx, decl->return_var, loc)))
goto fail;
- list_add_tail(args->instrs, &load->node.entry);
+ hlsl_block_add_instr(args->instrs, &load->node);
}
else
{
@@ -3606,7 +3763,7 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
if (!(expr = hlsl_new_expr(ctx, HLSL_OP0_VOID, operands, ctx->builtin_types.Void, loc)))
goto fail;
- list_add_tail(args->instrs, &expr->entry);
+ hlsl_block_add_instr(args->instrs, expr);
}
}
else if ((intrinsic = bsearch(name, intrinsic_functions, ARRAY_SIZE(intrinsic_functions),
@@ -3655,14 +3812,14 @@ static struct list *add_call(struct hlsl_ctx *ctx, const char *name,
goto fail;
}
vkd3d_free(args->args);
- return args->instrs;
+ return block_to_list(args->instrs);
fail:
free_parse_initializer(args);
return NULL;
}
-static struct list *add_constructor(struct hlsl_ctx *ctx, struct hlsl_type *type,
+static struct hlsl_block *add_constructor(struct hlsl_ctx *ctx, struct hlsl_type *type,
struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_load *load;
@@ -3692,7 +3849,7 @@ static struct list *add_constructor(struct hlsl_ctx *ctx, struct hlsl_type *type
if (!(load = hlsl_new_var_load(ctx, var, loc)))
return NULL;
- list_add_tail(params->instrs, &load->node.entry);
+ hlsl_block_add_instr(params->instrs, &load->node);
vkd3d_free(params->args);
return params->instrs;
@@ -4272,6 +4429,7 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type
char *name;
DWORD modifiers;
struct hlsl_ir_node *instr;
+ struct hlsl_block *block;
struct list *list;
struct parse_fields fields;
struct parse_function function;
@@ -4399,38 +4557,13 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type
%token <intval> C_INTEGER
%token <intval> PRE_LINE
-%type <list> add_expr
-%type <list> assignment_expr
-%type <list> bitand_expr
-%type <list> bitor_expr
-%type <list> bitxor_expr
-%type <list> compound_statement
-%type <list> conditional_expr
%type <list> declaration
%type <list> declaration_statement
-%type <list> discard_statement
-%type <list> equality_expr
-%type <list> expr
-%type <list> expr_optional
-%type <list> expr_statement
-%type <list> initializer_expr
-%type <list> jump_statement
-%type <list> logicand_expr
-%type <list> logicor_expr
-%type <list> loop_statement
-%type <list> mul_expr
-%type <list> postfix_expr
%type <list> primary_expr
-%type <list> relational_expr
-%type <list> selection_statement
-%type <list> shift_expr
-%type <list> statement
-%type <list> statement_list
-%type <list> struct_declaration
+%type <list> struct_declaration_without_vars
%type <list> type_specs
-%type <list> unary_expr
%type <list> variables_def
-%type <list> variables_def_optional
+%type <list> variables_def_typed
%token <name> VAR_IDENTIFIER
%token <name> NEW_IDENTIFIER
@@ -4446,6 +4579,31 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type
%type <attr_list> attribute_list
%type <attr_list> attribute_list_optional
+%type <block> add_expr
+%type <block> assignment_expr
+%type <block> bitand_expr
+%type <block> bitor_expr
+%type <block> bitxor_expr
+%type <block> compound_statement
+%type <block> conditional_expr
+%type <block> equality_expr
+%type <block> expr
+%type <block> expr_optional
+%type <block> expr_statement
+%type <block> initializer_expr
+%type <block> jump_statement
+%type <block> logicand_expr
+%type <block> logicor_expr
+%type <block> loop_statement
+%type <block> mul_expr
+%type <block> postfix_expr
+%type <block> relational_expr
+%type <block> shift_expr
+%type <block> selection_statement
+%type <block> statement
+%type <block> statement_list
+%type <block> unary_expr
+
%type <boolval> boolean
%type <buffer_type> buffer_type
@@ -4493,6 +4651,7 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type
%type <variable_def> type_spec
%type <variable_def> variable_decl
%type <variable_def> variable_def
+%type <variable_def> variable_def_typed
%%
@@ -4561,25 +4720,19 @@ preproc_directive:
}
}
-struct_declaration:
- var_modifiers struct_spec variables_def_optional ';'
+struct_declaration_without_vars:
+ var_modifiers struct_spec ';'
{
- struct hlsl_type *type;
- unsigned int modifiers = $1;
+ if (!$2->name)
+ hlsl_error(ctx, &@2, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
+ "Anonymous struct type must declare a variable.");
- if (!$3)
- {
- if (!$2->name)
- hlsl_error(ctx, &@2, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
- "Anonymous struct type must declare a variable.");
- if (modifiers)
- hlsl_error(ctx, &@1, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
- "Modifiers are not allowed on struct type declarations.");
- }
+ if ($1)
+ hlsl_error(ctx, &@1, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER,
+ "Modifiers are not allowed on struct type declarations.");
- if (!(type = apply_type_modifiers(ctx, $2, &modifiers, true, &@1)))
+ if (!($$ = make_empty_list(ctx)))
YYABORT;
- $$ = declare_vars(ctx, type, modifiers, &@1, $3);
}
struct_spec:
@@ -4702,7 +4855,7 @@ attribute:
}
$$->name = $2;
list_init(&$$->instrs);
- list_move_tail(&$$->instrs, $4.instrs);
+ list_move_tail(&$$->instrs, &$4.instrs->instrs);
vkd3d_free($4.instrs);
$$->loc = @$;
$$->args_count = $4.args_count;
@@ -4758,15 +4911,15 @@ func_declaration:
"Function \"%s\" is already defined.", decl->func->name);
hlsl_note(ctx, &decl->loc, VKD3D_SHADER_LOG_ERROR,
"\"%s\" was previously defined here.", decl->func->name);
- hlsl_free_instr_list($2);
+ destroy_block($2);
}
else
{
size_t i;
decl->has_body = true;
- list_move_tail(&decl->body.instrs, $2);
- vkd3d_free($2);
+ hlsl_block_add_block(&decl->body, $2);
+ destroy_block($2);
/* Semantics are taken from whichever definition has a body.
* We can't just replace the hlsl_ir_var pointers, though: if
@@ -4943,7 +5096,7 @@ func_prototype:
compound_statement:
'{' '}'
{
- if (!($$ = make_empty_list(ctx)))
+ if (!($$ = make_empty_block(ctx)))
YYABORT;
}
| '{' scope_start statement_list '}'
@@ -5261,7 +5414,12 @@ type_no_void:
{
validate_texture_format_type(ctx, $3, &@3);
- /* TODO: unspecified sample count is not allowed for all targets */
+ if (shader_profile_version_lt(ctx, 4, 1))
+ {
+ hlsl_error(ctx, &@1, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "Multisampled texture object declaration needs sample count for profile %s.", ctx->profile->name);
+ }
+
$$ = hlsl_new_texture_type(ctx, $1, $3, 0);
}
| texture_ms_type '<' type ',' shift_expr '>'
@@ -5270,7 +5428,7 @@ type_no_void:
struct hlsl_block block;
hlsl_block_init(&block);
- list_move_tail(&block.instrs, $5);
+ hlsl_block_add_block(&block, $5);
sample_count = evaluate_static_expression_as_uint(ctx, &block, &@5);
@@ -5325,7 +5483,7 @@ type_no_void:
$$ = hlsl_get_type(ctx->cur_scope, $1, true, true);
if ($$->is_minimum_precision)
{
- if (ctx->profile->major_version < 4)
+ if (shader_profile_version_lt(ctx, 4, 0))
{
hlsl_error(ctx, &@1, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
"Target profile doesn't support minimum-precision types.");
@@ -5354,7 +5512,7 @@ type:
declaration_statement:
declaration
- | struct_declaration
+ | struct_declaration_without_vars
| typedef
{
if (!($$ = make_empty_list(ctx)))
@@ -5416,23 +5574,12 @@ type_spec:
}
declaration:
- var_modifiers type variables_def ';'
+ variables_def_typed ';'
{
- struct hlsl_type *type;
- unsigned int modifiers = $1;
-
- if (!(type = apply_type_modifiers(ctx, $2, &modifiers, true, &@1)))
+ if (!($$ = initialize_vars(ctx, $1)))
YYABORT;
- $$ = declare_vars(ctx, type, modifiers, &@1, $3);
}
-variables_def_optional:
- %empty
- {
- $$ = NULL;
- }
- | variables_def
-
variables_def:
variable_def
{
@@ -5446,6 +5593,33 @@ variables_def:
list_add_tail($$, &$3->entry);
}
+variables_def_typed:
+ variable_def_typed
+ {
+ if (!($$ = make_empty_list(ctx)))
+ YYABORT;
+ list_add_head($$, &$1->entry);
+
+ declare_var(ctx, $1);
+ }
+ | variables_def_typed ',' variable_def
+ {
+ struct parse_variable_def *head_def;
+
+ assert(!list_empty($1));
+ head_def = LIST_ENTRY(list_head($1), struct parse_variable_def, entry);
+
+ assert(head_def->basic_type);
+ $3->basic_type = head_def->basic_type;
+ $3->modifiers = head_def->modifiers;
+ $3->modifiers_loc = head_def->modifiers_loc;
+
+ declare_var(ctx, $3);
+
+ $$ = $1;
+ list_add_tail($$, &$3->entry);
+ }
+
variable_decl:
any_identifier arrays colon_attribute
{
@@ -5461,7 +5635,7 @@ state:
any_identifier '=' expr ';'
{
vkd3d_free($1);
- hlsl_free_instr_list($3);
+ destroy_block($3);
}
state_block_start:
@@ -5487,6 +5661,38 @@ variable_def:
ctx->in_state_block = 0;
}
+variable_def_typed:
+ var_modifiers struct_spec variable_def
+ {
+ unsigned int modifiers = $1;
+ struct hlsl_type *type;
+
+ if (!(type = apply_type_modifiers(ctx, $2, &modifiers, true, &@1)))
+ YYABORT;
+
+ check_invalid_in_out_modifiers(ctx, modifiers, &@1);
+
+ $$ = $3;
+ $$->basic_type = type;
+ $$->modifiers = modifiers;
+ $$->modifiers_loc = @1;
+ }
+ | var_modifiers type variable_def
+ {
+ unsigned int modifiers = $1;
+ struct hlsl_type *type;
+
+ if (!(type = apply_type_modifiers(ctx, $2, &modifiers, true, &@1)))
+ YYABORT;
+
+ check_invalid_in_out_modifiers(ctx, modifiers, &@1);
+
+ $$ = $3;
+ $$->basic_type = type;
+ $$->modifiers = modifiers;
+ $$->modifiers_loc = @1;
+ }
+
arrays:
%empty
{
@@ -5495,17 +5701,12 @@ arrays:
}
| '[' expr ']' arrays
{
- struct hlsl_block block;
uint32_t *new_array;
unsigned int size;
- hlsl_clone_block(ctx, &block, &ctx->static_initializers);
- list_move_tail(&block.instrs, $2);
+ size = evaluate_static_expression_as_uint(ctx, $2, &@2);
- size = evaluate_static_expression_as_uint(ctx, &block, &@2);
-
- hlsl_block_cleanup(&block);
- vkd3d_free($2);
+ destroy_block($2);
$$ = $4;
@@ -5618,10 +5819,10 @@ complex_initializer:
$$.args_count = 1;
if (!($$.args = hlsl_alloc(ctx, sizeof(*$$.args))))
{
- destroy_instr_list($1);
+ destroy_block($1);
YYABORT;
}
- $$.args[0] = node_from_list($1);
+ $$.args[0] = node_from_block($1);
$$.instrs = $1;
$$.braces = false;
}
@@ -5653,7 +5854,7 @@ complex_initializer_list:
$$.args = new_args;
for (i = 0; i < $3.args_count; ++i)
$$.args[$$.args_count++] = $3.args[i];
- list_move_tail($$.instrs, $3.instrs);
+ hlsl_block_add_block($$.instrs, $3.instrs);
free_parse_initializer(&$3);
}
@@ -5666,10 +5867,10 @@ initializer_expr_list:
$$.args_count = 1;
if (!($$.args = hlsl_alloc(ctx, sizeof(*$$.args))))
{
- destroy_instr_list($1);
+ destroy_block($1);
YYABORT;
}
- $$.args[0] = node_from_list($1);
+ $$.args[0] = node_from_block($1);
$$.instrs = $1;
$$.braces = false;
}
@@ -5681,13 +5882,13 @@ initializer_expr_list:
if (!(new_args = hlsl_realloc(ctx, $$.args, ($$.args_count + 1) * sizeof(*$$.args))))
{
free_parse_initializer(&$$);
- destroy_instr_list($3);
+ destroy_block($3);
YYABORT;
}
$$.args = new_args;
- $$.args[$$.args_count++] = node_from_list($3);
- list_move_tail($$.instrs, $3);
- vkd3d_free($3);
+ $$.args[$$.args_count++] = node_from_block($3);
+ hlsl_block_add_block($$.instrs, $3);
+ destroy_block($3);
}
boolean:
@@ -5705,15 +5906,17 @@ statement_list:
| statement_list statement
{
$$ = $1;
- list_move_tail($$, $2);
- vkd3d_free($2);
+ hlsl_block_add_block($$, $2);
+ destroy_block($2);
}
statement:
declaration_statement
+ {
+ $$ = list_to_block($1);
+ }
| expr_statement
| compound_statement
- | discard_statement
| jump_statement
| selection_statement
| loop_statement
@@ -5721,47 +5924,47 @@ statement:
jump_statement:
KW_RETURN expr ';'
{
- if (!add_return(ctx, $2, node_from_list($2), &@1))
- YYABORT;
$$ = $2;
+ if (!add_return(ctx, $$, node_from_block($$), &@1))
+ YYABORT;
}
| KW_RETURN ';'
{
- if (!($$ = make_empty_list(ctx)))
+ if (!($$ = make_empty_block(ctx)))
YYABORT;
if (!add_return(ctx, $$, NULL, &@1))
YYABORT;
}
-
-discard_statement:
- KW_DISCARD ';'
+ | KW_DISCARD ';'
{
- struct hlsl_ir_node *discard;
+ struct hlsl_ir_node *discard, *c;
- if (!($$ = make_empty_list(ctx)))
+ if (!($$ = make_empty_block(ctx)))
YYABORT;
- if (!(discard = hlsl_new_jump(ctx, HLSL_IR_JUMP_DISCARD, &@1)))
+
+ if (!(c = hlsl_new_uint_constant(ctx, ~0u, &@1)))
+ return false;
+ hlsl_block_add_instr($$, c);
+
+ if (!(discard = hlsl_new_jump(ctx, HLSL_IR_JUMP_DISCARD_NZ, c, &@1)))
return false;
- list_add_tail($$, &discard->entry);
+ hlsl_block_add_instr($$, discard);
}
selection_statement:
KW_IF '(' expr ')' if_body
{
- struct hlsl_ir_node *condition = node_from_list($3);
- struct hlsl_block then_block, else_block;
+ struct hlsl_ir_node *condition = node_from_block($3);
struct hlsl_ir_node *instr;
- hlsl_block_init(&then_block);
- list_move_tail(&then_block.instrs, $5.then_block);
- hlsl_block_init(&else_block);
- if ($5.else_block)
- list_move_tail(&else_block.instrs, $5.else_block);
- vkd3d_free($5.then_block);
- vkd3d_free($5.else_block);
-
- if (!(instr = hlsl_new_if(ctx, condition, &then_block, &else_block, &@1)))
+ if (!(instr = hlsl_new_if(ctx, condition, $5.then_block, $5.else_block, &@1)))
+ {
+ destroy_block($5.then_block);
+ destroy_block($5.else_block);
YYABORT;
+ }
+ destroy_block($5.then_block);
+ destroy_block($5.else_block);
if (condition->data_type->dimx > 1 || condition->data_type->dimy > 1)
{
struct vkd3d_string_buffer *string;
@@ -5772,7 +5975,7 @@ selection_statement:
hlsl_release_string_buffer(ctx, string);
}
$$ = $3;
- list_add_tail($$, &instr->entry);
+ hlsl_block_add_instr($$, instr);
}
if_body:
@@ -5803,14 +6006,14 @@ loop_statement:
}
| attribute_list_optional KW_FOR '(' scope_start declaration expr_statement expr_optional ')' statement
{
- $$ = create_loop(ctx, LOOP_FOR, &$1, $5, $6, $7, $9, &@2);
+ $$ = create_loop(ctx, LOOP_FOR, &$1, list_to_block($5), $6, $7, $9, &@2);
hlsl_pop_scope(ctx);
}
expr_optional:
%empty
{
- if (!($$ = make_empty_list(ctx)))
+ if (!($$ = make_empty_block(ctx)))
YYABORT;
}
| expr
@@ -5826,7 +6029,7 @@ func_arguments:
{
$$.args = NULL;
$$.args_count = 0;
- if (!($$.instrs = make_empty_list(ctx)))
+ if (!($$.instrs = make_empty_block(ctx)))
YYABORT;
$$.braces = false;
}
@@ -5880,7 +6083,7 @@ primary_expr:
}
| '(' expr ')'
{
- $$ = $2;
+ $$ = block_to_list($2);
}
| var_identifier '(' func_arguments ')'
{
@@ -5915,11 +6118,14 @@ primary_expr:
postfix_expr:
primary_expr
+ {
+ $$ = list_to_block($1);
+ }
| postfix_expr OP_INC
{
if (!add_increment(ctx, $1, false, true, &@2))
{
- destroy_instr_list($1);
+ destroy_block($1);
YYABORT;
}
$$ = $1;
@@ -5928,14 +6134,14 @@ postfix_expr:
{
if (!add_increment(ctx, $1, true, true, &@2))
{
- destroy_instr_list($1);
+ destroy_block($1);
YYABORT;
}
$$ = $1;
}
| postfix_expr '.' any_identifier
{
- struct hlsl_ir_node *node = node_from_list($1);
+ struct hlsl_ir_node *node = node_from_block($1);
if (node->data_type->class == HLSL_CLASS_STRUCT)
{
@@ -5963,7 +6169,7 @@ postfix_expr:
hlsl_error(ctx, &@3, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, "Invalid swizzle \"%s\".", $3);
YYABORT;
}
- list_add_tail($1, &swizzle->entry);
+ hlsl_block_add_instr($1, swizzle);
$$ = $1;
}
else
@@ -5974,17 +6180,17 @@ postfix_expr:
}
| postfix_expr '[' expr ']'
{
- struct hlsl_ir_node *array = node_from_list($1), *index = node_from_list($3);
+ struct hlsl_ir_node *array = node_from_block($1), *index = node_from_block($3);
- list_move_head($1, $3);
- vkd3d_free($3);
+ hlsl_block_add_block($3, $1);
+ destroy_block($1);
- if (!add_array_access(ctx, $1, array, index, &@2))
+ if (!add_array_access(ctx, block_to_list($3), array, index, &@2))
{
- destroy_instr_list($1);
+ destroy_block($3);
YYABORT;
}
- $$ = $1;
+ $$ = $3;
}
/* var_modifiers is necessary to avoid shift/reduce conflicts. */
@@ -6025,14 +6231,14 @@ postfix_expr:
}
| postfix_expr '.' any_identifier '(' func_arguments ')'
{
- struct hlsl_ir_node *object = node_from_list($1);
+ struct hlsl_ir_node *object = node_from_block($1);
- list_move_tail($1, $5.instrs);
+ hlsl_block_add_block($1, $5.instrs);
vkd3d_free($5.instrs);
- if (!add_method_call(ctx, $1, object, $3, &$5, &@3))
+ if (!add_method_call(ctx, block_to_list($1), object, $3, &$5, &@3))
{
- hlsl_free_instr_list($1);
+ destroy_block($1);
vkd3d_free($5.args);
YYABORT;
}
@@ -6046,7 +6252,7 @@ unary_expr:
{
if (!add_increment(ctx, $2, false, false, &@1))
{
- destroy_instr_list($2);
+ destroy_block($2);
YYABORT;
}
$$ = $2;
@@ -6055,7 +6261,7 @@ unary_expr:
{
if (!add_increment(ctx, $2, true, false, &@1))
{
- destroy_instr_list($2);
+ destroy_block($2);
YYABORT;
}
$$ = $2;
@@ -6066,23 +6272,23 @@ unary_expr:
}
| '-' unary_expr
{
- add_unary_arithmetic_expr(ctx, $2, HLSL_OP1_NEG, node_from_list($2), &@1);
+ add_unary_arithmetic_expr(ctx, $2, HLSL_OP1_NEG, node_from_block($2), &@1);
$$ = $2;
}
| '~' unary_expr
{
- add_unary_bitwise_expr(ctx, $2, HLSL_OP1_BIT_NOT, node_from_list($2), &@1);
+ add_unary_bitwise_expr(ctx, $2, HLSL_OP1_BIT_NOT, node_from_block($2), &@1);
$$ = $2;
}
| '!' unary_expr
{
- add_unary_logical_expr(ctx, $2, HLSL_OP1_LOGIC_NOT, node_from_list($2), &@1);
+ add_unary_logical_expr(ctx, $2, HLSL_OP1_LOGIC_NOT, node_from_block($2), &@1);
$$ = $2;
}
/* var_modifiers is necessary to avoid shift/reduce conflicts. */
| '(' var_modifiers type arrays ')' unary_expr
{
- struct hlsl_type *src_type = node_from_list($6)->data_type;
+ struct hlsl_type *src_type = node_from_block($6)->data_type;
struct hlsl_type *dst_type;
unsigned int i;
@@ -6118,9 +6324,9 @@ unary_expr:
YYABORT;
}
- if (!add_cast(ctx, $6, node_from_list($6), dst_type, &@3))
+ if (!add_cast(ctx, block_to_list($6), node_from_block($6), dst_type, &@3))
{
- hlsl_free_instr_list($6);
+ destroy_block($6);
YYABORT;
}
$$ = $6;
@@ -6130,131 +6336,132 @@ mul_expr:
unary_expr
| mul_expr '*' unary_expr
{
- $$ = add_binary_arithmetic_expr_merge(ctx, $1, $3, HLSL_OP2_MUL, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_MUL, &@2);
}
| mul_expr '/' unary_expr
{
- $$ = add_binary_arithmetic_expr_merge(ctx, $1, $3, HLSL_OP2_DIV, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_DIV, &@2);
}
| mul_expr '%' unary_expr
{
- $$ = add_binary_arithmetic_expr_merge(ctx, $1, $3, HLSL_OP2_MOD, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_MOD, &@2);
}
add_expr:
mul_expr
| add_expr '+' mul_expr
{
- $$ = add_binary_arithmetic_expr_merge(ctx, $1, $3, HLSL_OP2_ADD, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_ADD, &@2);
}
| add_expr '-' mul_expr
{
struct hlsl_ir_node *neg;
- if (!(neg = hlsl_new_unary_expr(ctx, HLSL_OP1_NEG, node_from_list($3), &@2)))
+ if (!(neg = add_unary_arithmetic_expr(ctx, $3, HLSL_OP1_NEG, node_from_block($3), &@2)))
YYABORT;
- list_add_tail($3, &neg->entry);
- $$ = add_binary_arithmetic_expr_merge(ctx, $1, $3, HLSL_OP2_ADD, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_ADD, &@2);
}
shift_expr:
add_expr
| shift_expr OP_LEFTSHIFT add_expr
{
- $$ = add_binary_shift_expr_merge(ctx, $1, $3, HLSL_OP2_LSHIFT, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_LSHIFT, &@2);
}
| shift_expr OP_RIGHTSHIFT add_expr
{
- $$ = add_binary_shift_expr_merge(ctx, $1, $3, HLSL_OP2_RSHIFT, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_RSHIFT, &@2);
}
relational_expr:
shift_expr
| relational_expr '<' shift_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $1, $3, HLSL_OP2_LESS, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_LESS, &@2);
}
| relational_expr '>' shift_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $3, $1, HLSL_OP2_LESS, &@2);
+ $$ = add_binary_expr_merge(ctx, $3, $1, HLSL_OP2_LESS, &@2);
}
| relational_expr OP_LE shift_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $3, $1, HLSL_OP2_GEQUAL, &@2);
+ $$ = add_binary_expr_merge(ctx, $3, $1, HLSL_OP2_GEQUAL, &@2);
}
| relational_expr OP_GE shift_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $1, $3, HLSL_OP2_GEQUAL, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_GEQUAL, &@2);
}
equality_expr:
relational_expr
| equality_expr OP_EQ relational_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $1, $3, HLSL_OP2_EQUAL, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_EQUAL, &@2);
}
| equality_expr OP_NE relational_expr
{
- $$ = add_binary_comparison_expr_merge(ctx, $1, $3, HLSL_OP2_NEQUAL, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_NEQUAL, &@2);
}
bitand_expr:
equality_expr
| bitand_expr '&' equality_expr
{
- $$ = add_binary_bitwise_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_AND, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_AND, &@2);
}
bitxor_expr:
bitand_expr
| bitxor_expr '^' bitand_expr
{
- $$ = add_binary_bitwise_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_XOR, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_XOR, &@2);
}
bitor_expr:
bitxor_expr
| bitor_expr '|' bitxor_expr
{
- $$ = add_binary_bitwise_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_OR, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_BIT_OR, &@2);
}
logicand_expr:
bitor_expr
| logicand_expr OP_AND bitor_expr
{
- $$ = add_binary_logical_expr_merge(ctx, $1, $3, HLSL_OP2_LOGIC_AND, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_LOGIC_AND, &@2);
}
logicor_expr:
logicand_expr
| logicor_expr OP_OR logicand_expr
{
- $$ = add_binary_logical_expr_merge(ctx, $1, $3, HLSL_OP2_LOGIC_OR, &@2);
+ $$ = add_binary_expr_merge(ctx, $1, $3, HLSL_OP2_LOGIC_OR, &@2);
}
conditional_expr:
logicor_expr
| logicor_expr '?' expr ':' assignment_expr
{
- struct hlsl_ir_node *cond = node_from_list($1), *first = node_from_list($3), *second = node_from_list($5);
+ struct hlsl_ir_node *cond = node_from_block($1);
+ struct hlsl_ir_node *first = node_from_block($3);
+ struct hlsl_ir_node *second = node_from_block($5);
struct hlsl_type *common_type;
- list_move_tail($1, $3);
- list_move_tail($1, $5);
- vkd3d_free($3);
- vkd3d_free($5);
+ hlsl_block_add_block($1, $3);
+ hlsl_block_add_block($1, $5);
+ destroy_block($3);
+ destroy_block($5);
if (!(common_type = get_common_numeric_type(ctx, first, second, &@3)))
YYABORT;
- if (!(first = add_implicit_conversion(ctx, $1, first, common_type, &@3)))
+ if (!(first = add_implicit_conversion(ctx, block_to_list($1), first, common_type, &@3)))
YYABORT;
- if (!(second = add_implicit_conversion(ctx, $1, second, common_type, &@5)))
+ if (!(second = add_implicit_conversion(ctx, block_to_list($1), second, common_type, &@5)))
YYABORT;
- if (!hlsl_add_conditional(ctx, $1, cond, first, second))
+ if (!hlsl_add_conditional(ctx, block_to_list($1), cond, first, second))
YYABORT;
$$ = $1;
}
@@ -6264,16 +6471,16 @@ assignment_expr:
conditional_expr
| unary_expr assign_op assignment_expr
{
- struct hlsl_ir_node *lhs = node_from_list($1), *rhs = node_from_list($3);
+ struct hlsl_ir_node *lhs = node_from_block($1), *rhs = node_from_block($3);
if (lhs->data_type->modifiers & HLSL_MODIFIER_CONST)
{
hlsl_error(ctx, &@2, VKD3D_SHADER_ERROR_HLSL_MODIFIES_CONST, "Statement modifies a const expression.");
YYABORT;
}
- list_move_tail($3, $1);
- vkd3d_free($1);
- if (!add_assignment(ctx, $3, lhs, $2, rhs))
+ hlsl_block_add_block($3, $1);
+ destroy_block($1);
+ if (!add_assignment(ctx, block_to_list($3), lhs, $2, rhs))
YYABORT;
$$ = $3;
}
@@ -6329,6 +6536,6 @@ expr:
| expr ',' assignment_expr
{
$$ = $1;
- list_move_tail($$, $3);
- vkd3d_free($3);
+ hlsl_block_add_block($$, $3);
+ destroy_block($3);
}
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
index 765b1907426..09a3ea4ca08 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
@@ -97,6 +97,7 @@ static struct hlsl_ir_node *new_offset_from_path_index(struct hlsl_ctx *ctx, str
static struct hlsl_ir_node *new_offset_instr_from_deref(struct hlsl_ctx *ctx, struct hlsl_block *block,
const struct hlsl_deref *deref, const struct vkd3d_shader_location *loc)
{
+ enum hlsl_regset regset = hlsl_type_get_regset(deref->data_type);
struct hlsl_ir_node *offset = NULL;
struct hlsl_type *type;
unsigned int i;
@@ -111,7 +112,7 @@ static struct hlsl_ir_node *new_offset_instr_from_deref(struct hlsl_ctx *ctx, st
struct hlsl_block idx_block;
if (!(offset = new_offset_from_path_index(ctx, &idx_block, type, offset, deref->path[i].node,
- deref->offset_regset, loc)))
+ regset, loc)))
return NULL;
hlsl_block_add_block(block, &idx_block);
@@ -126,7 +127,7 @@ static struct hlsl_ir_node *new_offset_instr_from_deref(struct hlsl_ctx *ctx, st
static bool replace_deref_path_with_offset(struct hlsl_ctx *ctx, struct hlsl_deref *deref,
struct hlsl_ir_node *instr)
{
- const struct hlsl_type *type;
+ struct hlsl_type *type;
struct hlsl_ir_node *offset;
struct hlsl_block block;
@@ -145,7 +146,7 @@ static bool replace_deref_path_with_offset(struct hlsl_ctx *ctx, struct hlsl_der
return true;
}
- deref->offset_regset = hlsl_type_get_regset(type);
+ deref->data_type = type;
if (!(offset = new_offset_instr_from_deref(ctx, &block, deref, &instr->loc)))
return false;
@@ -666,7 +667,7 @@ static void insert_early_return_break(struct hlsl_ctx *ctx,
return;
list_add_after(&cf_instr->entry, &load->node.entry);
- if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_BREAK, &cf_instr->loc)))
+ if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_BREAK, NULL, &cf_instr->loc)))
return;
hlsl_block_add_instr(&then_block, jump);
@@ -1689,7 +1690,7 @@ static bool validate_static_object_references(struct hlsl_ctx *ctx, struct hlsl_
{
struct hlsl_ir_resource_load *load = hlsl_ir_resource_load(instr);
- if (!(load->resource.var->storage_modifiers & HLSL_STORAGE_UNIFORM))
+ if (!load->resource.var->is_uniform)
{
hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
"Loaded resource must have a single uniform source.");
@@ -1704,7 +1705,7 @@ static bool validate_static_object_references(struct hlsl_ctx *ctx, struct hlsl_
if (load->sampler.var)
{
- if (!(load->sampler.var->storage_modifiers & HLSL_STORAGE_UNIFORM))
+ if (!load->sampler.var->is_uniform)
{
hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
"Resource load sampler must have a single uniform source.");
@@ -1722,7 +1723,7 @@ static bool validate_static_object_references(struct hlsl_ctx *ctx, struct hlsl_
{
struct hlsl_ir_resource_store *store = hlsl_ir_resource_store(instr);
- if (!(store->resource.var->storage_modifiers & HLSL_STORAGE_UNIFORM))
+ if (!store->resource.var->is_uniform)
{
hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
"Accessed resource must have a single uniform source.");
@@ -1889,7 +1890,7 @@ static bool split_matrix_copies(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr
if (rhs->type != HLSL_IR_LOAD)
{
- hlsl_fixme(ctx, &instr->loc, "Copying from unsupported node type.\n");
+ hlsl_fixme(ctx, &instr->loc, "Copying from unsupported node type.");
return false;
}
@@ -2066,6 +2067,97 @@ static bool lower_nonconstant_vector_derefs(struct hlsl_ctx *ctx, struct hlsl_ir
return false;
}
+/* Lower combined samples and sampler variables to synthesized separated textures and samplers.
+ * That is, translate SM1-style samples in the source to SM4-style samples in the bytecode. */
+static bool lower_combined_samples(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
+{
+ struct hlsl_ir_resource_load *load;
+ struct vkd3d_string_buffer *name;
+ struct hlsl_ir_var *var;
+ unsigned int i;
+
+ if (instr->type != HLSL_IR_RESOURCE_LOAD)
+ return false;
+ load = hlsl_ir_resource_load(instr);
+
+ switch (load->load_type)
+ {
+ case HLSL_RESOURCE_LOAD:
+ case HLSL_RESOURCE_GATHER_RED:
+ case HLSL_RESOURCE_GATHER_GREEN:
+ case HLSL_RESOURCE_GATHER_BLUE:
+ case HLSL_RESOURCE_GATHER_ALPHA:
+ case HLSL_RESOURCE_SAMPLE_CMP:
+ case HLSL_RESOURCE_SAMPLE_CMP_LZ:
+ case HLSL_RESOURCE_SAMPLE_GRAD:
+ return false;
+
+ case HLSL_RESOURCE_SAMPLE:
+ case HLSL_RESOURCE_SAMPLE_LOD:
+ case HLSL_RESOURCE_SAMPLE_LOD_BIAS:
+ break;
+ }
+ if (load->sampler.var)
+ return false;
+
+ if (!hlsl_type_is_resource(load->resource.var->data_type))
+ {
+ hlsl_fixme(ctx, &instr->loc, "Lower combined samplers within structs.");
+ return false;
+ }
+
+ assert(hlsl_type_get_regset(load->resource.var->data_type) == HLSL_REGSET_SAMPLERS);
+
+ if (!(name = hlsl_get_string_buffer(ctx)))
+ return false;
+ vkd3d_string_buffer_printf(name, "<resource>%s", load->resource.var->name);
+
+ TRACE("Lowering to separate resource %s.\n", debugstr_a(name->buffer));
+
+ if (!(var = hlsl_get_var(ctx->globals, name->buffer)))
+ {
+ struct hlsl_type *texture_array_type = hlsl_new_texture_type(ctx, load->sampling_dim,
+ hlsl_get_vector_type(ctx, HLSL_TYPE_FLOAT, 4), 0);
+
+ /* Create (possibly multi-dimensional) texture array type with the same dims as the sampler array. */
+ struct hlsl_type *arr_type = load->resource.var->data_type;
+ for (i = 0; i < load->resource.path_len; ++i)
+ {
+ assert(arr_type->class == HLSL_CLASS_ARRAY);
+ texture_array_type = hlsl_new_array_type(ctx, texture_array_type, arr_type->e.array.elements_count);
+ arr_type = arr_type->e.array.type;
+ }
+
+ if (!(var = hlsl_new_synthetic_var_named(ctx, name->buffer, texture_array_type, &instr->loc, false)))
+ {
+ hlsl_release_string_buffer(ctx, name);
+ return false;
+ }
+ var->is_uniform = 1;
+ var->is_separated_resource = true;
+
+ list_add_tail(&ctx->extern_vars, &var->extern_entry);
+ }
+ hlsl_release_string_buffer(ctx, name);
+
+ if (load->sampling_dim != var->data_type->sampler_dim)
+ {
+ hlsl_error(ctx, &load->node.loc, VKD3D_SHADER_ERROR_HLSL_INCONSISTENT_SAMPLER,
+ "Cannot split combined samplers from \"%s\" if they have different usage dimensions.",
+ load->resource.var->name);
+ hlsl_note(ctx, &var->loc, VKD3D_SHADER_LOG_ERROR, "First use as combined sampler is here.");
+ return false;
+
+ }
+
+ hlsl_copy_deref(ctx, &load->sampler, &load->resource);
+ load->resource.var = var;
+ assert(hlsl_deref_get_type(ctx, &load->resource)->base_type == HLSL_TYPE_TEXTURE);
+ assert(hlsl_deref_get_type(ctx, &load->sampler)->base_type == HLSL_TYPE_SAMPLER);
+
+ return true;
+}
+
/* Lower DIV to RCP + MUL. */
static bool lower_division(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
{
@@ -2584,6 +2676,61 @@ static bool lower_float_modulus(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr
return true;
}
+static bool lower_discard_neg(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
+{
+ struct hlsl_ir_node *zero, *bool_false, *or, *cmp, *load;
+ static const struct hlsl_constant_value zero_value;
+ struct hlsl_type *arg_type, *cmp_type;
+ struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = { 0 };
+ struct hlsl_ir_jump *jump;
+ unsigned int i, count;
+ struct list instrs;
+
+ if (instr->type != HLSL_IR_JUMP)
+ return false;
+ jump = hlsl_ir_jump(instr);
+ if (jump->type != HLSL_IR_JUMP_DISCARD_NEG)
+ return false;
+
+ list_init(&instrs);
+
+ arg_type = jump->condition.node->data_type;
+ if (!(zero = hlsl_new_constant(ctx, arg_type, &zero_value, &instr->loc)))
+ return false;
+ list_add_tail(&instrs, &zero->entry);
+
+ operands[0] = jump->condition.node;
+ operands[1] = zero;
+ cmp_type = hlsl_get_numeric_type(ctx, arg_type->class, HLSL_TYPE_BOOL, arg_type->dimx, arg_type->dimy);
+ if (!(cmp = hlsl_new_expr(ctx, HLSL_OP2_LESS, operands, cmp_type, &instr->loc)))
+ return false;
+ list_add_tail(&instrs, &cmp->entry);
+
+ if (!(bool_false = hlsl_new_constant(ctx, hlsl_get_scalar_type(ctx, HLSL_TYPE_BOOL), &zero_value, &instr->loc)))
+ return false;
+ list_add_tail(&instrs, &bool_false->entry);
+
+ or = bool_false;
+
+ count = hlsl_type_component_count(cmp_type);
+ for (i = 0; i < count; ++i)
+ {
+ if (!(load = hlsl_add_load_component(ctx, &instrs, cmp, i, &instr->loc)))
+ return false;
+
+ if (!(or = hlsl_new_binary_expr(ctx, HLSL_OP2_LOGIC_OR, or, load)))
+ return NULL;
+ list_add_tail(&instrs, &or->entry);
+ }
+
+ list_move_tail(&instr->entry, &instrs);
+ hlsl_src_remove(&jump->condition);
+ hlsl_src_from_node(&jump->condition, or);
+ jump->type = HLSL_IR_JUMP_DISCARD_NZ;
+
+ return true;
+}
+
static bool dce(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
{
switch (instr->type)
@@ -2848,8 +2995,15 @@ static void compute_liveness_recurse(struct hlsl_block *block, unsigned int loop
index->idx.node->last_read = last_read;
break;
}
- case HLSL_IR_CONSTANT:
case HLSL_IR_JUMP:
+ {
+ struct hlsl_ir_jump *jump = hlsl_ir_jump(instr);
+
+ if (jump->condition.node)
+ jump->condition.node->last_read = last_read;
+ break;
+ }
+ case HLSL_IR_CONSTANT:
break;
}
}
@@ -3034,7 +3188,7 @@ static const char *debug_register(char class, struct hlsl_reg reg, const struct
return vkd3d_dbg_sprintf("%c%u%s", class, reg.id, debug_hlsl_writemask(reg.writemask));
}
-static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
+static bool track_object_components_sampler_dim(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
{
struct hlsl_ir_resource_load *load;
struct hlsl_ir_var *var;
@@ -3046,15 +3200,16 @@ static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_n
load = hlsl_ir_resource_load(instr);
var = load->resource.var;
+
regset = hlsl_type_get_regset(hlsl_deref_get_type(ctx, &load->resource));
+ if (!hlsl_regset_index_from_deref(ctx, &load->resource, regset, &index))
+ return false;
if (regset == HLSL_REGSET_SAMPLERS)
{
enum hlsl_sampler_dim dim;
assert(!load->sampler.var);
- if (!hlsl_regset_index_from_deref(ctx, &load->resource, regset, &index))
- return false;
dim = var->objects_usage[regset][index].sampler_dim;
if (dim != load->sampling_dim)
@@ -3072,25 +3227,37 @@ static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_n
return false;
}
}
- var->objects_usage[regset][index].used = true;
- var->objects_usage[regset][index].sampler_dim = load->sampling_dim;
}
- else
- {
- if (!hlsl_regset_index_from_deref(ctx, &load->resource, regset, &index))
- return false;
+ var->objects_usage[regset][index].sampler_dim = load->sampling_dim;
- var->objects_usage[regset][index].used = true;
- var->objects_usage[regset][index].sampler_dim = load->sampling_dim;
+ return false;
+}
- if (load->sampler.var)
- {
- var = load->sampler.var;
- if (!hlsl_regset_index_from_deref(ctx, &load->sampler, HLSL_REGSET_SAMPLERS, &index))
- return false;
+static bool track_object_components_usage(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context)
+{
+ struct hlsl_ir_resource_load *load;
+ struct hlsl_ir_var *var;
+ enum hlsl_regset regset;
+ unsigned int index;
- var->objects_usage[HLSL_REGSET_SAMPLERS][index].used = true;
- }
+ if (instr->type != HLSL_IR_RESOURCE_LOAD)
+ return false;
+
+ load = hlsl_ir_resource_load(instr);
+ var = load->resource.var;
+
+ regset = hlsl_type_get_regset(hlsl_deref_get_type(ctx, &load->resource));
+ if (!hlsl_regset_index_from_deref(ctx, &load->resource, regset, &index))
+ return false;
+
+ var->objects_usage[regset][index].used = true;
+ if (load->sampler.var)
+ {
+ var = load->sampler.var;
+ if (!hlsl_regset_index_from_deref(ctx, &load->sampler, HLSL_REGSET_SAMPLERS, &index))
+ return false;
+
+ var->objects_usage[HLSL_REGSET_SAMPLERS][index].used = true;
}
return false;
@@ -3110,9 +3277,12 @@ static void calculate_resource_register_counts(struct hlsl_ctx *ctx)
{
for (i = 0; i < type->reg_size[k]; ++i)
{
- /* Samplers are only allocated until the last used one. */
+ bool is_separated = var->is_separated_resource;
+
+ /* Samplers (and textures separated from them) are only allocated until the last
+ * used one. */
if (var->objects_usage[k][i].used)
- var->regs[k].bind_count = (k == HLSL_REGSET_SAMPLERS) ? i + 1 : type->reg_size[k];
+ var->regs[k].bind_count = (k == HLSL_REGSET_SAMPLERS || is_separated) ? i + 1 : type->reg_size[k];
}
}
}
@@ -3192,10 +3362,33 @@ static void allocate_temp_registers_recurse(struct hlsl_ctx *ctx,
}
}
+static void record_constant(struct hlsl_ctx *ctx, unsigned int component_index, float f)
+{
+ struct hlsl_constant_defs *defs = &ctx->constant_defs;
+ struct hlsl_constant_register *reg;
+ size_t i;
+
+ for (i = 0; i < defs->count; ++i)
+ {
+ reg = &defs->regs[i];
+ if (reg->index == (component_index / 4))
+ {
+ reg->value.f[component_index % 4] = f;
+ return;
+ }
+ }
+
+ if (!hlsl_array_reserve(ctx, (void **)&defs->regs, &defs->size, defs->count + 1, sizeof(*defs->regs)))
+ return;
+ reg = &defs->regs[defs->count++];
+ memset(reg, 0, sizeof(*reg));
+ reg->index = component_index / 4;
+ reg->value.f[component_index % 4] = f;
+}
+
static void allocate_const_registers_recurse(struct hlsl_ctx *ctx,
struct hlsl_block *block, struct register_allocator *allocator)
{
- struct hlsl_constant_defs *defs = &ctx->constant_defs;
struct hlsl_ir_node *instr;
LIST_FOR_EACH_ENTRY(instr, &block->instrs, struct hlsl_ir_node, entry)
@@ -3206,66 +3399,52 @@ static void allocate_const_registers_recurse(struct hlsl_ctx *ctx,
{
struct hlsl_ir_constant *constant = hlsl_ir_constant(instr);
const struct hlsl_type *type = instr->data_type;
- unsigned int x, y, i, writemask, end_reg;
- unsigned int reg_size = type->reg_size[HLSL_REGSET_NUMERIC];
+ unsigned int x, i;
constant->reg = allocate_numeric_registers_for_type(ctx, allocator, 1, UINT_MAX, type);
TRACE("Allocated constant @%u to %s.\n", instr->index, debug_register('c', constant->reg, type));
- if (!hlsl_array_reserve(ctx, (void **)&defs->values, &defs->size,
- constant->reg.id + reg_size / 4, sizeof(*defs->values)))
- return;
- end_reg = constant->reg.id + reg_size / 4;
- if (end_reg > defs->count)
- {
- memset(&defs->values[defs->count], 0, sizeof(*defs->values) * (end_reg - defs->count));
- defs->count = end_reg;
- }
-
assert(type->class <= HLSL_CLASS_LAST_NUMERIC);
+ assert(type->dimy == 1);
+ assert(constant->reg.writemask);
- if (!(writemask = constant->reg.writemask))
- writemask = (1u << type->dimx) - 1;
-
- for (y = 0; y < type->dimy; ++y)
+ for (x = 0, i = 0; x < 4; ++x)
{
- for (x = 0, i = 0; x < 4; ++x)
+ const union hlsl_constant_value_component *value;
+ float f;
+
+ if (!(constant->reg.writemask & (1u << x)))
+ continue;
+ value = &constant->value.u[i++];
+
+ switch (type->base_type)
{
- const union hlsl_constant_value_component *value;
- float f;
-
- if (!(writemask & (1u << x)))
- continue;
- value = &constant->value.u[i++];
-
- switch (type->base_type)
- {
- case HLSL_TYPE_BOOL:
- f = !!value->u;
- break;
-
- case HLSL_TYPE_FLOAT:
- case HLSL_TYPE_HALF:
- f = value->f;
- break;
-
- case HLSL_TYPE_INT:
- f = value->i;
- break;
-
- case HLSL_TYPE_UINT:
- f = value->u;
- break;
-
- case HLSL_TYPE_DOUBLE:
- FIXME("Double constant.\n");
- return;
-
- default:
- vkd3d_unreachable();
- }
- defs->values[constant->reg.id + y].f[x] = f;
+ case HLSL_TYPE_BOOL:
+ f = !!value->u;
+ break;
+
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ f = value->f;
+ break;
+
+ case HLSL_TYPE_INT:
+ f = value->i;
+ break;
+
+ case HLSL_TYPE_UINT:
+ f = value->u;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ FIXME("Double constant.\n");
+ return;
+
+ default:
+ vkd3d_unreachable();
}
+
+ record_constant(ctx, constant->reg.id * 4 + x, f);
}
break;
@@ -3297,8 +3476,6 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_ir_functi
struct register_allocator allocator = {0};
struct hlsl_ir_var *var;
- allocate_const_registers_recurse(ctx, &entry_func->body, &allocator);
-
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
if (var->is_uniform && var->last_read)
@@ -3315,6 +3492,8 @@ static void allocate_const_registers(struct hlsl_ctx *ctx, struct hlsl_ir_functi
}
}
+ allocate_const_registers_recurse(ctx, &entry_func->body, &allocator);
+
vkd3d_free(allocator.allocations);
}
@@ -3497,7 +3676,7 @@ static void validate_buffer_offsets(struct hlsl_ctx *ctx)
LIST_FOR_EACH_ENTRY(var1, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
- if (!var1->is_uniform || var1->data_type->class == HLSL_CLASS_OBJECT)
+ if (!var1->is_uniform || hlsl_type_is_resource(var1->data_type))
continue;
buffer = var1->buffer;
@@ -3508,7 +3687,7 @@ static void validate_buffer_offsets(struct hlsl_ctx *ctx)
{
unsigned int var1_reg_size, var2_reg_size;
- if (!var2->is_uniform || var2->data_type->class == HLSL_CLASS_OBJECT)
+ if (!var2->is_uniform || hlsl_type_is_resource(var2->data_type))
continue;
if (var1 == var2 || var1->buffer != var2->buffer)
@@ -3558,7 +3737,7 @@ static void allocate_buffers(struct hlsl_ctx *ctx)
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
- if (var->is_uniform && var->data_type->class != HLSL_CLASS_OBJECT)
+ if (var->is_uniform && !hlsl_type_is_resource(var->data_type))
{
if (var->is_param)
var->buffer = ctx->params_buffer;
@@ -3618,7 +3797,7 @@ static void allocate_buffers(struct hlsl_ctx *ctx)
}
static const struct hlsl_ir_var *get_allocated_object(struct hlsl_ctx *ctx, enum hlsl_regset regset,
- uint32_t index)
+ uint32_t index, bool allocated_only)
{
const struct hlsl_ir_var *var;
unsigned int start, count;
@@ -3632,6 +3811,9 @@ static const struct hlsl_ir_var *get_allocated_object(struct hlsl_ctx *ctx, enum
* bound there even if the reserved vars aren't used. */
start = var->reg_reservation.reg_index;
count = var->data_type->reg_size[regset];
+
+ if (!var->regs[regset].allocated && allocated_only)
+ continue;
}
else if (var->regs[regset].allocated)
{
@@ -3672,6 +3854,7 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
if (count == 0)
continue;
+ /* The variable was already allocated if it has a reservation. */
if (var->regs[regset].allocated)
{
const struct hlsl_ir_var *reserved_object, *last_reported = NULL;
@@ -3690,7 +3873,10 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
{
index = var->regs[regset].id + i;
- reserved_object = get_allocated_object(ctx, regset, index);
+ /* get_allocated_object() may return "var" itself, but we
+ * actually want that, otherwise we'll end up reporting the
+ * same conflict between the same two variables twice. */
+ reserved_object = get_allocated_object(ctx, regset, index, true);
if (reserved_object && reserved_object != var && reserved_object != last_reported)
{
hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS,
@@ -3709,7 +3895,7 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
while (available < count)
{
- if (get_allocated_object(ctx, regset, index))
+ if (get_allocated_object(ctx, regset, index, false))
available = 0;
else
++available;
@@ -3853,6 +4039,7 @@ bool hlsl_regset_index_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref
bool hlsl_offset_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref, unsigned int *offset)
{
struct hlsl_ir_node *offset_node = deref->offset.node;
+ enum hlsl_regset regset;
unsigned int size;
if (!offset_node)
@@ -3869,8 +4056,9 @@ bool hlsl_offset_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref
return false;
*offset = hlsl_ir_constant(offset_node)->value.u[0].u;
+ regset = hlsl_type_get_regset(deref->data_type);
- size = deref->var->data_type->reg_size[deref->offset_regset];
+ size = deref->var->data_type->reg_size[regset];
if (*offset >= size)
{
hlsl_error(ctx, &deref->offset.node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
@@ -3900,7 +4088,8 @@ struct hlsl_reg hlsl_reg_from_deref(struct hlsl_ctx *ctx, const struct hlsl_dere
struct hlsl_reg ret = var->regs[HLSL_REGSET_NUMERIC];
unsigned int offset = hlsl_offset_from_deref_safe(ctx, deref);
- assert(deref->offset_regset == HLSL_REGSET_NUMERIC);
+ assert(deref->data_type);
+ assert(deref->data_type->class <= HLSL_CLASS_LAST_NUMERIC);
ret.id += offset / 4;
@@ -4062,6 +4251,10 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
hlsl_error(ctx, &entry_func->loc, VKD3D_SHADER_ERROR_HLSL_MISSING_ATTRIBUTE,
"Entry point \"%s\" is missing a [numthreads] attribute.", entry_func->func->name);
+ if (profile->major_version >= 4)
+ {
+ hlsl_transform_ir(ctx, lower_discard_neg, body, NULL);
+ }
hlsl_transform_ir(ctx, lower_broadcasts, body, NULL);
while (hlsl_transform_ir(ctx, fold_redundant_casts, body, NULL));
do
@@ -4094,6 +4287,12 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
hlsl_transform_ir(ctx, lower_casts_to_bool, body, NULL);
hlsl_transform_ir(ctx, lower_int_dot, body, NULL);
+ hlsl_transform_ir(ctx, validate_static_object_references, body, NULL);
+ hlsl_transform_ir(ctx, track_object_components_sampler_dim, body, NULL);
+ if (profile->major_version >= 4)
+ hlsl_transform_ir(ctx, lower_combined_samples, body, NULL);
+ hlsl_transform_ir(ctx, track_object_components_usage, body, NULL);
+
if (profile->major_version < 4)
{
hlsl_transform_ir(ctx, lower_division, body, NULL);
@@ -4107,9 +4306,6 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
hlsl_transform_ir(ctx, lower_abs, body, NULL);
}
- hlsl_transform_ir(ctx, validate_static_object_references, body, NULL);
- hlsl_transform_ir(ctx, track_object_components_usage, body, NULL);
-
/* TODO: move forward, remove when no longer needed */
transform_derefs(ctx, replace_deref_path_with_offset, body);
while (hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, body, NULL));
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c
index 301113c8477..01c438ae212 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_constant_ops.c
@@ -80,7 +80,7 @@ static bool fold_cast(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst,
return false;
}
- for (k = 0; k < 4; ++k)
+ for (k = 0; k < dst_type->dimx; ++k)
{
switch (src->node.data_type->base_type)
{
@@ -160,7 +160,7 @@ static bool fold_neg(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst,
assert(type == src->node.data_type->base_type);
- for (k = 0; k < 4; ++k)
+ for (k = 0; k < dst_type->dimx; ++k)
{
switch (type)
{
@@ -186,6 +186,51 @@ static bool fold_neg(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst,
return true;
}
+static bool fold_rcp(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src, const struct vkd3d_shader_location *loc)
+{
+ enum hlsl_base_type type = dst_type->base_type;
+ unsigned int k;
+
+ assert(type == src->node.data_type->base_type);
+
+ for (k = 0; k < 4; ++k)
+ {
+ switch (type)
+ {
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ if (ctx->profile->major_version >= 4 && src->value.u[k].f == 0.0f)
+ {
+ hlsl_warning(ctx, loc, VKD3D_SHADER_WARNING_HLSL_DIVISION_BY_ZERO,
+ "Floating point division by zero.");
+ }
+ dst->u[k].f = 1.0f / src->value.u[k].f;
+ if (ctx->profile->major_version < 4 && !isfinite(dst->u[k].f))
+ {
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_DIVISION_BY_ZERO,
+ "Infinities and NaNs are not allowed by the shader model.");
+ }
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ if (src->value.u[k].d == 0.0)
+ {
+ hlsl_warning(ctx, loc, VKD3D_SHADER_WARNING_HLSL_DIVISION_BY_ZERO,
+ "Floating point division by zero.");
+ }
+ dst->u[k].d = 1.0 / src->value.u[k].d;
+ break;
+
+ default:
+ FIXME("Fold 'rcp' for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ return false;
+ }
+ }
+
+ return true;
+}
+
static bool fold_add(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
@@ -195,7 +240,7 @@ static bool fold_add(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
assert(type == src1->node.data_type->base_type);
assert(type == src2->node.data_type->base_type);
- for (k = 0; k < 4; ++k)
+ for (k = 0; k < dst_type->dimx; ++k)
{
switch (type)
{
@@ -223,7 +268,7 @@ static bool fold_add(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
return true;
}
-static bool fold_mul(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+static bool fold_and(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
enum hlsl_base_type type = dst_type->base_type;
@@ -232,64 +277,73 @@ static bool fold_mul(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
assert(type == src1->node.data_type->base_type);
assert(type == src2->node.data_type->base_type);
- for (k = 0; k < 4; ++k)
+ for (k = 0; k < dst_type->dimx; ++k)
{
switch (type)
{
- case HLSL_TYPE_FLOAT:
- case HLSL_TYPE_HALF:
- dst->u[k].f = src1->value.u[k].f * src2->value.u[k].f;
- break;
-
- case HLSL_TYPE_DOUBLE:
- dst->u[k].d = src1->value.u[k].d * src2->value.u[k].d;
- break;
-
case HLSL_TYPE_INT:
case HLSL_TYPE_UINT:
- dst->u[k].u = src1->value.u[k].u * src2->value.u[k].u;
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u & src2->value.u[k].u;
break;
default:
- FIXME("Fold multiplication for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ FIXME("Fold bit/logic and for type %s.\n", debug_hlsl_type(ctx, dst_type));
return false;
}
}
return true;
}
-static bool fold_nequal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+static bool fold_or(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
+ enum hlsl_base_type type = dst_type->base_type;
unsigned int k;
- assert(dst_type->base_type == HLSL_TYPE_BOOL);
- assert(src1->node.data_type->base_type == src2->node.data_type->base_type);
+ assert(type == src1->node.data_type->base_type);
+ assert(type == src2->node.data_type->base_type);
- for (k = 0; k < 4; ++k)
+ for (k = 0; k < dst_type->dimx; ++k)
{
- switch (src1->node.data_type->base_type)
+ switch (type)
{
- case HLSL_TYPE_FLOAT:
- case HLSL_TYPE_HALF:
- dst->u[k].u = src1->value.u[k].f != src2->value.u[k].f;
+ case HLSL_TYPE_INT:
+ case HLSL_TYPE_UINT:
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u | src2->value.u[k].u;
break;
- case HLSL_TYPE_DOUBLE:
- dst->u[k].u = src1->value.u[k].d != src2->value.u[k].d;
- break;
+ default:
+ FIXME("Fold bit/logic or for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ return false;
+ }
+ }
+ return true;
+}
+
+static bool fold_bit_xor(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
+{
+ enum hlsl_base_type type = dst_type->base_type;
+ unsigned int k;
+
+ assert(type == src1->node.data_type->base_type);
+ assert(type == src2->node.data_type->base_type);
+ for (k = 0; k < dst_type->dimx; ++k)
+ {
+ switch (type)
+ {
case HLSL_TYPE_INT:
case HLSL_TYPE_UINT:
- case HLSL_TYPE_BOOL:
- dst->u[k].u = src1->value.u[k].u != src2->value.u[k].u;
+ dst->u[k].u = src1->value.u[k].u ^ src2->value.u[k].u;
break;
default:
- vkd3d_unreachable();
+ FIXME("Fold bit xor for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ return false;
}
-
- dst->u[k].u *= ~0u;
}
return true;
}
@@ -363,45 +417,116 @@ static bool fold_div(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
return true;
}
-static bool fold_mod(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
- const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2,
- const struct vkd3d_shader_location *loc)
+static bool fold_equal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
- enum hlsl_base_type type = dst_type->base_type;
unsigned int k;
- assert(type == src1->node.data_type->base_type);
- assert(type == src2->node.data_type->base_type);
+ assert(dst_type->base_type == HLSL_TYPE_BOOL);
+ assert(src1->node.data_type->base_type == src2->node.data_type->base_type);
for (k = 0; k < dst_type->dimx; ++k)
{
- switch (type)
+ switch (src1->node.data_type->base_type)
{
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].u = src1->value.u[k].f == src2->value.u[k].f;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].u = src1->value.u[k].d == src2->value.u[k].d;
+ break;
+
case HLSL_TYPE_INT:
- if (src2->value.u[k].i == 0)
- {
- hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_DIVISION_BY_ZERO, "Division by zero.");
- return false;
- }
- if (src1->value.u[k].i == INT_MIN && src2->value.u[k].i == -1)
- dst->u[k].i = 0;
- else
- dst->u[k].i = src1->value.u[k].i % src2->value.u[k].i;
+ case HLSL_TYPE_UINT:
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u == src2->value.u[k].u;
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+
+ dst->u[k].u *= ~0u;
+ }
+ return true;
+}
+
+static bool fold_gequal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
+{
+ unsigned int k;
+
+ assert(dst_type->base_type == HLSL_TYPE_BOOL);
+ assert(src1->node.data_type->base_type == src2->node.data_type->base_type);
+
+ for (k = 0; k < dst_type->dimx; ++k)
+ {
+ switch (src1->node.data_type->base_type)
+ {
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].u = src1->value.u[k].f >= src2->value.u[k].f;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].u = src1->value.u[k].d >= src2->value.u[k].d;
+ break;
+
+ case HLSL_TYPE_INT:
+ dst->u[k].u = src1->value.u[k].i >= src2->value.u[k].i;
break;
case HLSL_TYPE_UINT:
- if (src2->value.u[k].u == 0)
- {
- hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_DIVISION_BY_ZERO, "Division by zero.");
- return false;
- }
- dst->u[k].u = src1->value.u[k].u % src2->value.u[k].u;
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u >= src2->value.u[k].u;
break;
default:
- FIXME("Fold modulus for type %s.\n", debug_hlsl_type(ctx, dst_type));
- return false;
+ vkd3d_unreachable();
}
+
+ dst->u[k].u *= ~0u;
+ }
+ return true;
+}
+
+static bool fold_less(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
+{
+ unsigned int k;
+
+ assert(dst_type->base_type == HLSL_TYPE_BOOL);
+ assert(src1->node.data_type->base_type == src2->node.data_type->base_type);
+
+ for (k = 0; k < dst_type->dimx; ++k)
+ {
+ switch (src1->node.data_type->base_type)
+ {
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].u = src1->value.u[k].f < src2->value.u[k].f;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].u = src1->value.u[k].d < src2->value.u[k].d;
+ break;
+
+ case HLSL_TYPE_INT:
+ dst->u[k].u = src1->value.u[k].i < src2->value.u[k].i;
+ break;
+
+ case HLSL_TYPE_UINT:
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u < src2->value.u[k].u;
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+
+ dst->u[k].u *= ~0u;
}
return true;
}
@@ -419,6 +544,15 @@ static bool fold_max(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
{
switch (type)
{
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].f = fmaxf(src1->value.u[k].f, src2->value.u[k].f);
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].d = fmax(src1->value.u[k].d, src2->value.u[k].d);
+ break;
+
case HLSL_TYPE_INT:
dst->u[k].i = max(src1->value.u[k].i, src2->value.u[k].i);
break;
@@ -448,6 +582,15 @@ static bool fold_min(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
{
switch (type)
{
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].f = fminf(src1->value.u[k].f, src2->value.u[k].f);
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].d = fmin(src1->value.u[k].d, src2->value.u[k].d);
+ break;
+
case HLSL_TYPE_INT:
dst->u[k].i = min(src1->value.u[k].i, src2->value.u[k].i);
break;
@@ -464,8 +607,9 @@ static bool fold_min(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, cons
return true;
}
-static bool fold_bit_xor(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
- const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
+static bool fold_mod(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+ const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2,
+ const struct vkd3d_shader_location *loc)
{
enum hlsl_base_type type = dst_type->base_type;
unsigned int k;
@@ -478,19 +622,35 @@ static bool fold_bit_xor(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst,
switch (type)
{
case HLSL_TYPE_INT:
+ if (src2->value.u[k].i == 0)
+ {
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_DIVISION_BY_ZERO, "Division by zero.");
+ return false;
+ }
+ if (src1->value.u[k].i == INT_MIN && src2->value.u[k].i == -1)
+ dst->u[k].i = 0;
+ else
+ dst->u[k].i = src1->value.u[k].i % src2->value.u[k].i;
+ break;
+
case HLSL_TYPE_UINT:
- dst->u[k].u = src1->value.u[k].u ^ src2->value.u[k].u;
+ if (src2->value.u[k].u == 0)
+ {
+ hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_DIVISION_BY_ZERO, "Division by zero.");
+ return false;
+ }
+ dst->u[k].u = src1->value.u[k].u % src2->value.u[k].u;
break;
default:
- FIXME("Fold bit xor for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ FIXME("Fold modulus for type %s.\n", debug_hlsl_type(ctx, dst_type));
return false;
}
}
return true;
}
-static bool fold_bit_and(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+static bool fold_mul(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
enum hlsl_base_type type = dst_type->base_type;
@@ -503,41 +663,60 @@ static bool fold_bit_and(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst,
{
switch (type)
{
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].f = src1->value.u[k].f * src2->value.u[k].f;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].d = src1->value.u[k].d * src2->value.u[k].d;
+ break;
+
case HLSL_TYPE_INT:
case HLSL_TYPE_UINT:
- dst->u[k].u = src1->value.u[k].u & src2->value.u[k].u;
+ dst->u[k].u = src1->value.u[k].u * src2->value.u[k].u;
break;
default:
- FIXME("Fold bit and for type %s.\n", debug_hlsl_type(ctx, dst_type));
+ FIXME("Fold multiplication for type %s.\n", debug_hlsl_type(ctx, dst_type));
return false;
}
}
return true;
}
-static bool fold_bit_or(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
+static bool fold_nequal(struct hlsl_ctx *ctx, struct hlsl_constant_value *dst, const struct hlsl_type *dst_type,
const struct hlsl_ir_constant *src1, const struct hlsl_ir_constant *src2)
{
- enum hlsl_base_type type = dst_type->base_type;
unsigned int k;
- assert(type == src1->node.data_type->base_type);
- assert(type == src2->node.data_type->base_type);
+ assert(dst_type->base_type == HLSL_TYPE_BOOL);
+ assert(src1->node.data_type->base_type == src2->node.data_type->base_type);
for (k = 0; k < dst_type->dimx; ++k)
{
- switch (type)
+ switch (src1->node.data_type->base_type)
{
+ case HLSL_TYPE_FLOAT:
+ case HLSL_TYPE_HALF:
+ dst->u[k].u = src1->value.u[k].f != src2->value.u[k].f;
+ break;
+
+ case HLSL_TYPE_DOUBLE:
+ dst->u[k].u = src1->value.u[k].d != src2->value.u[k].d;
+ break;
+
case HLSL_TYPE_INT:
case HLSL_TYPE_UINT:
- dst->u[k].u = src1->value.u[k].u | src2->value.u[k].u;
+ case HLSL_TYPE_BOOL:
+ dst->u[k].u = src1->value.u[k].u != src2->value.u[k].u;
break;
default:
- FIXME("Fold bit or for type %s.\n", debug_hlsl_type(ctx, dst_type));
- return false;
+ vkd3d_unreachable();
}
+
+ dst->u[k].u *= ~0u;
}
return true;
}
@@ -587,24 +766,42 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr,
success = fold_neg(ctx, &res, instr->data_type, arg1);
break;
+ case HLSL_OP1_RCP:
+ success = fold_rcp(ctx, &res, instr->data_type, arg1, &instr->loc);
+ break;
+
case HLSL_OP2_ADD:
success = fold_add(ctx, &res, instr->data_type, arg1, arg2);
break;
- case HLSL_OP2_MUL:
- success = fold_mul(ctx, &res, instr->data_type, arg1, arg2);
+ case HLSL_OP2_BIT_AND:
+ case HLSL_OP2_LOGIC_AND:
+ success = fold_and(ctx, &res, instr->data_type, arg1, arg2);
break;
- case HLSL_OP2_NEQUAL:
- success = fold_nequal(ctx, &res, instr->data_type, arg1, arg2);
+ case HLSL_OP2_BIT_OR:
+ case HLSL_OP2_LOGIC_OR:
+ success = fold_or(ctx, &res, instr->data_type, arg1, arg2);
+ break;
+
+ case HLSL_OP2_BIT_XOR:
+ success = fold_bit_xor(ctx, &res, instr->data_type, arg1, arg2);
break;
case HLSL_OP2_DIV:
success = fold_div(ctx, &res, instr->data_type, arg1, arg2, &instr->loc);
break;
- case HLSL_OP2_MOD:
- success = fold_mod(ctx, &res, instr->data_type, arg1, arg2, &instr->loc);
+ case HLSL_OP2_EQUAL:
+ success = fold_equal(ctx, &res, instr->data_type, arg1, arg2);
+ break;
+
+ case HLSL_OP2_GEQUAL:
+ success = fold_gequal(ctx, &res, instr->data_type, arg1, arg2);
+ break;
+
+ case HLSL_OP2_LESS:
+ success = fold_less(ctx, &res, instr->data_type, arg1, arg2);
break;
case HLSL_OP2_MAX:
@@ -615,16 +812,16 @@ bool hlsl_fold_constant_exprs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr,
success = fold_min(ctx, &res, instr->data_type, arg1, arg2);
break;
- case HLSL_OP2_BIT_XOR:
- success = fold_bit_xor(ctx, &res, instr->data_type, arg1, arg2);
+ case HLSL_OP2_MOD:
+ success = fold_mod(ctx, &res, instr->data_type, arg1, arg2, &instr->loc);
break;
- case HLSL_OP2_BIT_AND:
- success = fold_bit_and(ctx, &res, instr->data_type, arg1, arg2);
+ case HLSL_OP2_MUL:
+ success = fold_mul(ctx, &res, instr->data_type, arg1, arg2);
break;
- case HLSL_OP2_BIT_OR:
- success = fold_bit_or(ctx, &res, instr->data_type, arg1, arg2);
+ case HLSL_OP2_NEQUAL:
+ success = fold_nequal(ctx, &res, instr->data_type, arg1, arg2);
break;
default:
diff --git a/libs/vkd3d/libs/vkd3d-shader/spirv.c b/libs/vkd3d/libs/vkd3d-shader/spirv.c
index 3542b5fac51..9725a5c7e25 100644
--- a/libs/vkd3d/libs/vkd3d-shader/spirv.c
+++ b/libs/vkd3d/libs/vkd3d-shader/spirv.c
@@ -3217,7 +3217,7 @@ static bool spirv_compiler_get_register_info(const struct spirv_compiler *compil
{
assert(reg->idx[0].offset < compiler->temp_count);
register_info->id = compiler->temp_id + reg->idx[0].offset;
- register_info->storage_class = SpvStorageClassFunction;
+ register_info->storage_class = SpvStorageClassPrivate;
register_info->descriptor_array = NULL;
register_info->member_idx = 0;
register_info->component_type = VKD3D_SHADER_COMPONENT_FLOAT;
@@ -3998,6 +3998,11 @@ static void spirv_compiler_emit_interpolation_decorations(struct spirv_compiler
vkd3d_spirv_enable_capability(builder, SpvCapabilitySampleRateShading);
vkd3d_spirv_build_op_decorate(builder, id, SpvDecorationSample, NULL, 0);
break;
+ case VKD3DSIM_LINEAR_NOPERSPECTIVE_SAMPLE:
+ vkd3d_spirv_build_op_decorate(builder, id, SpvDecorationNoPerspective, NULL, 0);
+ vkd3d_spirv_enable_capability(builder, SpvCapabilitySampleRateShading);
+ vkd3d_spirv_build_op_decorate(builder, id, SpvDecorationSample, NULL, 0);
+ break;
default:
FIXME("Unhandled interpolation mode %#x.\n", mode);
break;
@@ -5258,8 +5263,7 @@ static void spirv_compiler_emit_dcl_global_flags(struct spirv_compiler *compiler
WARN("Unhandled global flags %#x.\n", flags);
}
-static void spirv_compiler_emit_dcl_temps(struct spirv_compiler *compiler,
- const struct vkd3d_shader_instruction *instruction)
+static void spirv_compiler_emit_temps(struct spirv_compiler *compiler, uint32_t count)
{
struct vkd3d_spirv_builder *builder = &compiler->spirv_builder;
size_t function_location;
@@ -5270,11 +5274,11 @@ static void spirv_compiler_emit_dcl_temps(struct spirv_compiler *compiler,
vkd3d_spirv_begin_function_stream_insertion(builder, function_location);
assert(!compiler->temp_count);
- compiler->temp_count = instruction->declaration.count;
+ compiler->temp_count = count;
for (i = 0; i < compiler->temp_count; ++i)
{
- id = spirv_compiler_emit_variable(compiler, &builder->function_stream,
- SpvStorageClassFunction, VKD3D_SHADER_COMPONENT_FLOAT, VKD3D_VEC4_SIZE);
+ id = spirv_compiler_emit_variable(compiler, &builder->global_stream,
+ SpvStorageClassPrivate, VKD3D_SHADER_COMPONENT_FLOAT, VKD3D_VEC4_SIZE);
if (!i)
compiler->temp_id = id;
assert(id == compiler->temp_id + i);
@@ -6236,9 +6240,6 @@ static void spirv_compiler_leave_shader_phase(struct spirv_compiler *compiler)
vkd3d_spirv_build_op_function_end(builder);
- compiler->temp_id = 0;
- compiler->temp_count = 0;
-
if (is_in_control_point_phase(compiler))
{
if (compiler->epilogue_function_id)
@@ -9103,9 +9104,6 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler,
case VKD3DSIH_DCL_GLOBAL_FLAGS:
spirv_compiler_emit_dcl_global_flags(compiler, instruction);
break;
- case VKD3DSIH_DCL_TEMPS:
- spirv_compiler_emit_dcl_temps(compiler, instruction);
- break;
case VKD3DSIH_DCL_INDEXABLE_TEMP:
spirv_compiler_emit_dcl_indexable_temp(compiler, instruction);
break;
@@ -9426,6 +9424,7 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler,
spirv_compiler_emit_cut_stream(compiler, instruction);
break;
case VKD3DSIH_DCL_HS_MAX_TESSFACTOR:
+ case VKD3DSIH_DCL_TEMPS:
case VKD3DSIH_HS_DECLS:
case VKD3DSIH_NOP:
/* nothing to do */
@@ -9448,6 +9447,9 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler,
enum vkd3d_result result = VKD3D_OK;
unsigned int i;
+ if (parser->shader_desc.temp_count)
+ spirv_compiler_emit_temps(compiler, parser->shader_desc.temp_count);
+
compiler->location.column = 0;
compiler->location.line = 1;
diff --git a/libs/vkd3d/libs/vkd3d-shader/tpf.c b/libs/vkd3d/libs/vkd3d-shader/tpf.c
index d066b13ee4e..801c688a297 100644
--- a/libs/vkd3d/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d/libs/vkd3d-shader/tpf.c
@@ -989,6 +989,8 @@ static void shader_sm4_read_declaration_count(struct vkd3d_shader_instruction *i
uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *priv)
{
ins->declaration.count = *tokens;
+ if (opcode == VKD3D_SM4_OP_DCL_TEMPS)
+ priv->p.shader_desc.temp_count = max(priv->p.shader_desc.temp_count, *tokens);
}
static void shader_sm4_read_declaration_dst(struct vkd3d_shader_instruction *ins, uint32_t opcode,
@@ -2442,6 +2444,7 @@ int vkd3d_shader_sm4_parser_create(const struct vkd3d_shader_compile_info *compi
}
shader_desc = &sm4->p.shader_desc;
+ shader_desc->is_dxil = false;
if ((ret = shader_extract_from_dxbc(&compile_info->source,
message_context, compile_info->source_name, shader_desc)) < 0)
{
@@ -2967,47 +2970,154 @@ static D3D_SRV_DIMENSION sm4_rdef_resource_dimension(const struct hlsl_type *typ
}
}
+struct extern_resource
+{
+ /* var is only not NULL if this resource is a whole variable, so it may be responsible for more
+ * than one component. */
+ const struct hlsl_ir_var *var;
+
+ char *name;
+ struct hlsl_type *data_type;
+ bool is_user_packed;
+
+ enum hlsl_regset regset;
+ unsigned int id, bind_count;
+};
+
static int sm4_compare_extern_resources(const void *a, const void *b)
{
- const struct hlsl_ir_var *aa = *(const struct hlsl_ir_var **)a;
- const struct hlsl_ir_var *bb = *(const struct hlsl_ir_var **)b;
- enum hlsl_regset aa_regset, bb_regset;
+ const struct extern_resource *aa = (const struct extern_resource *)a;
+ const struct extern_resource *bb = (const struct extern_resource *)b;
+ int r;
- aa_regset = hlsl_type_get_regset(aa->data_type);
- bb_regset = hlsl_type_get_regset(bb->data_type);
+ if ((r = vkd3d_u32_compare(aa->regset, bb->regset)))
+ return r;
- if (aa_regset != bb_regset)
- return aa_regset - bb_regset;
+ return vkd3d_u32_compare(aa->id, bb->id);
+}
+
+static void sm4_free_extern_resources(struct extern_resource *extern_resources, unsigned int count)
+{
+ unsigned int i;
- return aa->regs[aa_regset].id - bb->regs[bb_regset].id;
+ for (i = 0; i < count; ++i)
+ vkd3d_free(extern_resources[i].name);
+ vkd3d_free(extern_resources);
}
-static const struct hlsl_ir_var **sm4_get_extern_resources(struct hlsl_ctx *ctx, unsigned int *count)
+static const char *string_skip_tag(const char *string)
{
- const struct hlsl_ir_var **extern_resources = NULL;
+ if (!strncmp(string, "<resource>", strlen("<resource>")))
+ return string + strlen("<resource>");
+ return string;
+}
+
+static struct extern_resource *sm4_get_extern_resources(struct hlsl_ctx *ctx, unsigned int *count)
+{
+ bool separate_components = ctx->profile->major_version == 5 && ctx->profile->minor_version == 0;
+ struct extern_resource *extern_resources = NULL;
const struct hlsl_ir_var *var;
enum hlsl_regset regset;
size_t capacity = 0;
+ char *name;
*count = 0;
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
- if (!hlsl_type_is_resource(var->data_type))
- continue;
- regset = hlsl_type_get_regset(var->data_type);
- if (!var->regs[regset].allocated)
- continue;
-
- if (!(hlsl_array_reserve(ctx, (void **)&extern_resources, &capacity, *count + 1,
- sizeof(*extern_resources))))
+ if (separate_components)
{
- *count = 0;
- return NULL;
+ unsigned int component_count = hlsl_type_component_count(var->data_type);
+ unsigned int k, regset_offset;
+
+ for (k = 0; k < component_count; ++k)
+ {
+ struct hlsl_type *component_type = hlsl_type_get_component_type(ctx, var->data_type, k);
+ struct vkd3d_string_buffer *name_buffer;
+
+ if (!hlsl_type_is_resource(component_type))
+ continue;
+
+ regset = hlsl_type_get_regset(component_type);
+ regset_offset = hlsl_type_get_component_offset(ctx, var->data_type, regset, k);
+
+ if (regset_offset > var->regs[regset].bind_count)
+ continue;
+
+ if (var->objects_usage[regset][regset_offset].used)
+ {
+ if (!(hlsl_array_reserve(ctx, (void **)&extern_resources, &capacity, *count + 1,
+ sizeof(*extern_resources))))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ return NULL;
+ }
+
+ if (!(name_buffer = hlsl_component_to_string(ctx, var, k)))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ return NULL;
+ }
+ if (!(name = hlsl_strdup(ctx, string_skip_tag(name_buffer->buffer))))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ hlsl_release_string_buffer(ctx, name_buffer);
+ return NULL;
+ }
+ hlsl_release_string_buffer(ctx, name_buffer);
+
+ extern_resources[*count].var = NULL;
+
+ extern_resources[*count].name = name;
+ extern_resources[*count].data_type = component_type;
+ extern_resources[*count].is_user_packed = false;
+
+ extern_resources[*count].regset = regset;
+ extern_resources[*count].id = var->regs[regset].id + regset_offset;
+ extern_resources[*count].bind_count = 1;
+
+ ++*count;
+ }
+ }
}
+ else
+ {
+ if (!hlsl_type_is_resource(var->data_type))
+ continue;
+ regset = hlsl_type_get_regset(var->data_type);
+ if (!var->regs[regset].allocated)
+ continue;
+
+ if (!(hlsl_array_reserve(ctx, (void **)&extern_resources, &capacity, *count + 1,
+ sizeof(*extern_resources))))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ return NULL;
+ }
+
+ if (!(name = hlsl_strdup(ctx, string_skip_tag(var->name))))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ return NULL;
+ }
+
+ extern_resources[*count].var = var;
- extern_resources[*count] = var;
- ++*count;
+ extern_resources[*count].name = name;
+ extern_resources[*count].data_type = var->data_type;
+ extern_resources[*count].is_user_packed = !!var->reg_reservation.reg_type;
+
+ extern_resources[*count].regset = regset;
+ extern_resources[*count].id = var->regs[regset].id;
+ extern_resources[*count].bind_count = var->regs[regset].bind_count;
+
+ ++*count;
+ }
}
qsort(extern_resources, *count, sizeof(*extern_resources), sm4_compare_extern_resources);
@@ -3020,8 +3130,8 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
size_t cbuffers_offset, resources_offset, creator_offset, string_offset;
size_t cbuffer_position, resource_position, creator_position;
const struct hlsl_profile_info *profile = ctx->profile;
- const struct hlsl_ir_var **extern_resources;
struct vkd3d_bytecode_buffer buffer = {0};
+ struct extern_resource *extern_resources;
const struct hlsl_buffer *cbuffer;
const struct hlsl_ir_var *var;
@@ -3075,18 +3185,15 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
for (i = 0; i < extern_resources_count; ++i)
{
- enum hlsl_regset regset;
+ const struct extern_resource *resource = &extern_resources[i];
uint32_t flags = 0;
- var = extern_resources[i];
- regset = hlsl_type_get_regset(var->data_type);
-
- if (var->reg_reservation.reg_type)
+ if (resource->is_user_packed)
flags |= D3D_SIF_USERPACKED;
put_u32(&buffer, 0); /* name */
- put_u32(&buffer, sm4_resource_type(var->data_type));
- if (regset == HLSL_REGSET_SAMPLERS)
+ put_u32(&buffer, sm4_resource_type(resource->data_type));
+ if (resource->regset == HLSL_REGSET_SAMPLERS)
{
put_u32(&buffer, 0);
put_u32(&buffer, 0);
@@ -3094,15 +3201,15 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
}
else
{
- unsigned int dimx = hlsl_type_get_component_type(ctx, var->data_type, 0)->e.resource_format->dimx;
+ unsigned int dimx = hlsl_type_get_component_type(ctx, resource->data_type, 0)->e.resource_format->dimx;
- put_u32(&buffer, sm4_resource_format(var->data_type));
- put_u32(&buffer, sm4_rdef_resource_dimension(var->data_type));
+ put_u32(&buffer, sm4_resource_format(resource->data_type));
+ put_u32(&buffer, sm4_rdef_resource_dimension(resource->data_type));
put_u32(&buffer, ~0u); /* FIXME: multisample count */
flags |= (dimx - 1) << VKD3D_SM4_SIF_TEXTURE_COMPONENTS_SHIFT;
}
- put_u32(&buffer, var->regs[regset].id);
- put_u32(&buffer, var->regs[regset].bind_count);
+ put_u32(&buffer, resource->id);
+ put_u32(&buffer, resource->bind_count);
put_u32(&buffer, flags);
}
@@ -3128,9 +3235,9 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
for (i = 0; i < extern_resources_count; ++i)
{
- var = extern_resources[i];
+ const struct extern_resource *resource = &extern_resources[i];
- string_offset = put_string(&buffer, var->name);
+ string_offset = put_string(&buffer, resource->name);
set_u32(&buffer, resources_offset + i * 8 * sizeof(uint32_t), string_offset);
}
@@ -3236,7 +3343,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
add_section(dxbc, TAG_RDEF, &buffer);
- vkd3d_free(extern_resources);
+ sm4_free_extern_resources(extern_resources, extern_resources_count);
}
static enum vkd3d_sm4_resource_type sm4_resource_dimension(const struct hlsl_type *type)
@@ -3346,8 +3453,9 @@ struct sm4_instruction
static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *reg,
unsigned int *writemask, enum vkd3d_sm4_swizzle_type *swizzle_type,
- const struct hlsl_deref *deref, const struct hlsl_type *data_type)
+ const struct hlsl_deref *deref)
{
+ const struct hlsl_type *data_type = hlsl_deref_get_type(ctx, deref);
const struct hlsl_ir_var *var = deref->var;
if (var->is_uniform)
@@ -3362,7 +3470,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
*swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
reg->idx[0] = var->regs[HLSL_REGSET_TEXTURES].id;
reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
- assert(deref->offset_regset == HLSL_REGSET_TEXTURES);
+ assert(regset == HLSL_REGSET_TEXTURES);
reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
@@ -3374,7 +3482,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
*swizzle_type = VKD3D_SM4_SWIZZLE_VEC4;
reg->idx[0] = var->regs[HLSL_REGSET_UAVS].id;
reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
- assert(deref->offset_regset == HLSL_REGSET_UAVS);
+ assert(regset == HLSL_REGSET_UAVS);
reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
@@ -3386,7 +3494,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
*swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
reg->idx[0] = var->regs[HLSL_REGSET_SAMPLERS].id;
reg->idx[0] += hlsl_offset_from_deref_safe(ctx, deref);
- assert(deref->offset_regset == HLSL_REGSET_SAMPLERS);
+ assert(regset == HLSL_REGSET_SAMPLERS);
reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
@@ -3484,11 +3592,11 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct sm4_register *r
}
static void sm4_src_from_deref(struct hlsl_ctx *ctx, struct sm4_src_register *src,
- const struct hlsl_deref *deref, const struct hlsl_type *data_type, unsigned int map_writemask)
+ const struct hlsl_deref *deref, unsigned int map_writemask)
{
unsigned int writemask;
- sm4_register_from_deref(ctx, &src->reg, &writemask, &src->swizzle_type, deref, data_type);
+ sm4_register_from_deref(ctx, &src->reg, &writemask, &src->swizzle_type, deref);
if (src->swizzle_type == VKD3D_SM4_SWIZZLE_VEC4)
src->swizzle = hlsl_map_swizzle(hlsl_swizzle_from_writemask(writemask), map_writemask);
}
@@ -3689,9 +3797,11 @@ static void write_sm4_dcl_constant_buffer(struct vkd3d_bytecode_buffer *buffer,
write_sm4_instruction(buffer, &instr);
}
-static void write_sm4_dcl_samplers(struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_var *var)
+static void write_sm4_dcl_samplers(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
+ const struct extern_resource *resource)
{
- unsigned int i, count = var->data_type->reg_size[HLSL_REGSET_SAMPLERS];
+ struct hlsl_type *component_type;
+ unsigned int i;
struct sm4_instruction instr =
{
.opcode = VKD3D_SM4_OP_DCL_SAMPLER,
@@ -3701,38 +3811,44 @@ static void write_sm4_dcl_samplers(struct vkd3d_bytecode_buffer *buffer, const s
.dst_count = 1,
};
- if (var->data_type->sampler_dim == HLSL_SAMPLER_DIM_COMPARISON)
+ component_type = hlsl_type_get_component_type(ctx, resource->data_type, 0);
+
+ if (component_type->sampler_dim == HLSL_SAMPLER_DIM_COMPARISON)
instr.opcode |= VKD3D_SM4_SAMPLER_COMPARISON << VKD3D_SM4_SAMPLER_MODE_SHIFT;
- for (i = 0; i < count; ++i)
+ assert(resource->regset == HLSL_REGSET_SAMPLERS);
+
+ for (i = 0; i < resource->bind_count; ++i)
{
- if (!var->objects_usage[HLSL_REGSET_SAMPLERS][i].used)
+ if (resource->var && !resource->var->objects_usage[HLSL_REGSET_SAMPLERS][i].used)
continue;
- instr.dsts[0].reg.idx[0] = var->regs[HLSL_REGSET_SAMPLERS].id + i;
+ instr.dsts[0].reg.idx[0] = resource->id + i;
write_sm4_instruction(buffer, &instr);
}
}
static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
- const struct hlsl_ir_var *var, bool uav)
+ const struct extern_resource *resource, bool uav)
{
enum hlsl_regset regset = uav ? HLSL_REGSET_UAVS : HLSL_REGSET_TEXTURES;
- unsigned int i, count = var->data_type->reg_size[regset];
struct hlsl_type *component_type;
struct sm4_instruction instr;
+ unsigned int i;
- component_type = hlsl_type_get_component_type(ctx, var->data_type, 0);
+ assert(resource->regset == regset);
- for (i = 0; i < count; ++i)
+ component_type = hlsl_type_get_component_type(ctx, resource->data_type, 0);
+
+ for (i = 0; i < resource->bind_count; ++i)
{
- if (!var->objects_usage[regset][i].used)
+ if (resource->var && !resource->var->objects_usage[regset][i].used)
continue;
instr = (struct sm4_instruction)
{
.dsts[0].reg.type = uav ? VKD3D_SM5_RT_UAV : VKD3D_SM4_RT_RESOURCE,
- .dsts[0].reg.idx = {var->regs[regset].id + i},
+ .dsts[0].reg.idx = {resource->id + i},
.dsts[0].reg.idx_count = 1,
.dst_count = 1,
@@ -3742,11 +3858,11 @@ static void write_sm4_dcl_textures(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
if (uav)
{
- switch (var->data_type->sampler_dim)
+ switch (resource->data_type->sampler_dim)
{
case HLSL_SAMPLER_DIM_STRUCTURED_BUFFER:
instr.opcode = VKD3D_SM5_OP_DCL_UAV_STRUCTURED;
- instr.byte_stride = var->data_type->e.resource_format->reg_size[HLSL_REGSET_NUMERIC] * 4;
+ instr.byte_stride = resource->data_type->e.resource_format->reg_size[HLSL_REGSET_NUMERIC] * 4;
break;
default:
instr.opcode = VKD3D_SM5_OP_DCL_UAV_TYPED;
@@ -4008,11 +4124,11 @@ static void write_sm4_binary_op_with_two_destinations(struct vkd3d_bytecode_buff
}
static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
- const struct hlsl_type *resource_type, const struct hlsl_ir_node *dst,
- const struct hlsl_deref *resource, const struct hlsl_ir_node *coords,
- const struct hlsl_ir_node *sample_index, const struct hlsl_ir_node *texel_offset,
- enum hlsl_sampler_dim dim)
+ const struct hlsl_ir_node *dst, const struct hlsl_deref *resource,
+ const struct hlsl_ir_node *coords, const struct hlsl_ir_node *sample_index,
+ const struct hlsl_ir_node *texel_offset, enum hlsl_sampler_dim dim)
{
+ const struct hlsl_type *resource_type = hlsl_deref_get_type(ctx, resource);
bool multisampled = resource_type->base_type == HLSL_TYPE_TEXTURE
&& (resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMS || resource_type->sampler_dim == HLSL_SAMPLER_DIM_2DMSARRAY);
bool uav = (hlsl_type_get_regset(resource_type) == HLSL_REGSET_UAVS);
@@ -4052,7 +4168,7 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
sm4_src_from_node(&instr.srcs[0], coords, coords_writemask);
- sm4_src_from_deref(ctx, &instr.srcs[1], resource, resource_type, instr.dsts[0].writemask);
+ sm4_src_from_deref(ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);
instr.src_count = 2;
@@ -4089,7 +4205,6 @@ static void write_sm4_ld(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buf
static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
const struct hlsl_ir_resource_load *load)
{
- const struct hlsl_type *resource_type = load->resource.var->data_type;
const struct hlsl_ir_node *texel_offset = load->texel_offset.node;
const struct hlsl_ir_node *coords = load->coords.node;
const struct hlsl_deref *resource = &load->resource;
@@ -4142,8 +4257,8 @@ static void write_sm4_sample(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
instr.dst_count = 1;
sm4_src_from_node(&instr.srcs[0], coords, VKD3DSP_WRITEMASK_ALL);
- sm4_src_from_deref(ctx, &instr.srcs[1], resource, resource_type, instr.dsts[0].writemask);
- sm4_src_from_deref(ctx, &instr.srcs[2], sampler, sampler->var->data_type, VKD3DSP_WRITEMASK_ALL);
+ sm4_src_from_deref(ctx, &instr.srcs[1], resource, instr.dsts[0].writemask);
+ sm4_src_from_deref(ctx, &instr.srcs[2], sampler, VKD3DSP_WRITEMASK_ALL);
instr.src_count = 3;
if (load->load_type == HLSL_RESOURCE_SAMPLE_LOD
@@ -4313,7 +4428,7 @@ static void write_sm4_store_uav_typed(struct hlsl_ctx *ctx, struct vkd3d_bytecod
memset(&instr, 0, sizeof(instr));
instr.opcode = VKD3D_SM5_OP_STORE_UAV_TYPED;
- sm4_register_from_deref(ctx, &instr.dsts[0].reg, &instr.dsts[0].writemask, NULL, dst, dst->var->data_type);
+ sm4_register_from_deref(ctx, &instr.dsts[0].reg, &instr.dsts[0].writemask, NULL, dst);
instr.dst_count = 1;
sm4_src_from_node(&instr.srcs[0], coords, VKD3DSP_WRITEMASK_ALL);
@@ -4369,11 +4484,31 @@ static void write_sm4_expr(struct hlsl_ctx *ctx,
write_sm4_unary_op(buffer, VKD3D_SM4_OP_DERIV_RTX, &expr->node, arg1, 0);
break;
+ case HLSL_OP1_DSX_COARSE:
+ assert(type_is_float(dst_type));
+ write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTX_COARSE, &expr->node, arg1, 0);
+ break;
+
+ case HLSL_OP1_DSX_FINE:
+ assert(type_is_float(dst_type));
+ write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTX_FINE, &expr->node, arg1, 0);
+ break;
+
case HLSL_OP1_DSY:
assert(type_is_float(dst_type));
write_sm4_unary_op(buffer, VKD3D_SM4_OP_DERIV_RTY, &expr->node, arg1, 0);
break;
+ case HLSL_OP1_DSY_COARSE:
+ assert(type_is_float(dst_type));
+ write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTY_COARSE, &expr->node, arg1, 0);
+ break;
+
+ case HLSL_OP1_DSY_FINE:
+ assert(type_is_float(dst_type));
+ write_sm4_unary_op(buffer, VKD3D_SM5_OP_DERIV_RTY_FINE, &expr->node, arg1, 0);
+ break;
+
case HLSL_OP1_EXP2:
assert(type_is_float(dst_type));
write_sm4_unary_op(buffer, VKD3D_SM4_OP_EXP, &expr->node, arg1, 0);
@@ -4780,19 +4915,13 @@ static void write_sm4_jump(struct hlsl_ctx *ctx,
instr.opcode = VKD3D_SM4_OP_BREAK;
break;
- case HLSL_IR_JUMP_DISCARD:
+ case HLSL_IR_JUMP_DISCARD_NZ:
{
- struct sm4_register *reg = &instr.srcs[0].reg;
-
instr.opcode = VKD3D_SM4_OP_DISCARD | VKD3D_SM4_CONDITIONAL_NZ;
memset(&instr.srcs[0], 0, sizeof(*instr.srcs));
- instr.srcs[0].swizzle_type = VKD3D_SM4_SWIZZLE_NONE;
instr.src_count = 1;
- reg->type = VKD3D_SM4_RT_IMMCONST;
- reg->dim = VKD3D_SM4_DIMENSION_SCALAR;
- reg->immconst_uint[0] = ~0u;
-
+ sm4_src_from_node(&instr.srcs[0], jump->condition.node, VKD3DSP_WRITEMASK_ALL);
break;
}
@@ -4800,7 +4929,7 @@ static void write_sm4_jump(struct hlsl_ctx *ctx,
vkd3d_unreachable();
default:
- hlsl_fixme(ctx, &jump->node.loc, "Jump type %s.\n", hlsl_jump_type_to_string(jump->type));
+ hlsl_fixme(ctx, &jump->node.loc, "Jump type %s.", hlsl_jump_type_to_string(jump->type));
return;
}
@@ -4839,7 +4968,7 @@ static void write_sm4_load(struct hlsl_ctx *ctx,
instr.opcode = VKD3D_SM4_OP_MOVC;
- sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, type, instr.dsts[0].writemask);
+ sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);
memset(&value, 0xff, sizeof(value));
sm4_src_from_constant_value(&instr.srcs[1], &value, type->dimx, instr.dsts[0].writemask);
@@ -4851,7 +4980,7 @@ static void write_sm4_load(struct hlsl_ctx *ctx,
{
instr.opcode = VKD3D_SM4_OP_MOV;
- sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, type, instr.dsts[0].writemask);
+ sm4_src_from_deref(ctx, &instr.srcs[0], &load->src, instr.dsts[0].writemask);
instr.src_count = 1;
}
@@ -4875,8 +5004,7 @@ static void write_sm4_loop(struct hlsl_ctx *ctx,
}
static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer,
- const struct hlsl_type *resource_type, const struct hlsl_ir_node *dst,
- const struct hlsl_deref *resource, const struct hlsl_deref *sampler,
+ const struct hlsl_ir_node *dst, const struct hlsl_deref *resource, const struct hlsl_deref *sampler,
const struct hlsl_ir_node *coords, unsigned int swizzle, const struct hlsl_ir_node *texel_offset)
{
struct sm4_src_register *src;
@@ -4906,10 +5034,10 @@ static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
}
}
- sm4_src_from_deref(ctx, &instr.srcs[instr.src_count++], resource, resource_type, instr.dsts[0].writemask);
+ sm4_src_from_deref(ctx, &instr.srcs[instr.src_count++], resource, instr.dsts[0].writemask);
src = &instr.srcs[instr.src_count++];
- sm4_src_from_deref(ctx, src, sampler, sampler->var->data_type, VKD3DSP_WRITEMASK_ALL);
+ sm4_src_from_deref(ctx, src, sampler, VKD3DSP_WRITEMASK_ALL);
src->reg.dim = VKD3D_SM4_DIMENSION_VEC4;
src->swizzle_type = VKD3D_SM4_SWIZZLE_SCALAR;
src->swizzle = swizzle;
@@ -4920,34 +5048,16 @@ static void write_sm4_gather(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer
static void write_sm4_resource_load(struct hlsl_ctx *ctx,
struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_resource_load *load)
{
- const struct hlsl_type *resource_type = load->resource.var->data_type;
const struct hlsl_ir_node *texel_offset = load->texel_offset.node;
const struct hlsl_ir_node *sample_index = load->sample_index.node;
const struct hlsl_ir_node *coords = load->coords.node;
- if (!hlsl_type_is_resource(resource_type))
+ if (load->sampler.var && !load->sampler.var->is_uniform)
{
- hlsl_fixme(ctx, &load->node.loc, "Separate object fields as new variables.");
+ hlsl_fixme(ctx, &load->node.loc, "Sample using non-uniform sampler variable.");
return;
}
- if (load->sampler.var)
- {
- const struct hlsl_type *sampler_type = load->sampler.var->data_type;
-
- if (!hlsl_type_is_resource(sampler_type))
- {
- hlsl_fixme(ctx, &load->node.loc, "Separate object fields as new variables.");
- return;
- }
-
- if (!load->sampler.var->is_uniform)
- {
- hlsl_fixme(ctx, &load->node.loc, "Sample using non-uniform sampler variable.");
- return;
- }
- }
-
if (!load->resource.var->is_uniform)
{
hlsl_fixme(ctx, &load->node.loc, "Load from non-uniform resource variable.");
@@ -4957,7 +5067,7 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx,
switch (load->load_type)
{
case HLSL_RESOURCE_LOAD:
- write_sm4_ld(ctx, buffer, resource_type, &load->node, &load->resource,
+ write_sm4_ld(ctx, buffer, &load->node, &load->resource,
coords, sample_index, texel_offset, load->sampling_dim);
break;
@@ -4967,32 +5077,29 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx,
case HLSL_RESOURCE_SAMPLE_LOD:
case HLSL_RESOURCE_SAMPLE_LOD_BIAS:
case HLSL_RESOURCE_SAMPLE_GRAD:
- if (!load->sampler.var)
- {
- hlsl_fixme(ctx, &load->node.loc, "SM4 combined sample expression.");
- return;
- }
+ /* Combined sample expressions were lowered. */
+ assert(load->sampler.var);
write_sm4_sample(ctx, buffer, load);
break;
case HLSL_RESOURCE_GATHER_RED:
- write_sm4_gather(ctx, buffer, resource_type, &load->node, &load->resource,
- &load->sampler, coords, HLSL_SWIZZLE(X, X, X, X), texel_offset);
+ write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+ HLSL_SWIZZLE(X, X, X, X), texel_offset);
break;
case HLSL_RESOURCE_GATHER_GREEN:
- write_sm4_gather(ctx, buffer, resource_type, &load->node, &load->resource,
- &load->sampler, coords, HLSL_SWIZZLE(Y, Y, Y, Y), texel_offset);
+ write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+ HLSL_SWIZZLE(Y, Y, Y, Y), texel_offset);
break;
case HLSL_RESOURCE_GATHER_BLUE:
- write_sm4_gather(ctx, buffer, resource_type, &load->node, &load->resource,
- &load->sampler, coords, HLSL_SWIZZLE(Z, Z, Z, Z), texel_offset);
+ write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+ HLSL_SWIZZLE(Z, Z, Z, Z), texel_offset);
break;
case HLSL_RESOURCE_GATHER_ALPHA:
- write_sm4_gather(ctx, buffer, resource_type, &load->node, &load->resource,
- &load->sampler, coords, HLSL_SWIZZLE(W, W, W, W), texel_offset);
+ write_sm4_gather(ctx, buffer, &load->node, &load->resource, &load->sampler, coords,
+ HLSL_SWIZZLE(W, W, W, W), texel_offset);
break;
}
}
@@ -5000,13 +5107,7 @@ static void write_sm4_resource_load(struct hlsl_ctx *ctx,
static void write_sm4_resource_store(struct hlsl_ctx *ctx,
struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_resource_store *store)
{
- const struct hlsl_type *resource_type = store->resource.var->data_type;
-
- if (!hlsl_type_is_resource(resource_type))
- {
- hlsl_fixme(ctx, &store->node.loc, "Separate object fields as new variables.");
- return;
- }
+ struct hlsl_type *resource_type = hlsl_deref_get_type(ctx, &store->resource);
if (!store->resource.var->is_uniform)
{
@@ -5016,7 +5117,7 @@ static void write_sm4_resource_store(struct hlsl_ctx *ctx,
if (resource_type->sampler_dim == HLSL_SAMPLER_DIM_STRUCTURED_BUFFER)
{
- hlsl_fixme(ctx, &store->node.loc, "Structured buffers store is not implemented.\n");
+ hlsl_fixme(ctx, &store->node.loc, "Structured buffers store is not implemented.");
return;
}
@@ -5033,7 +5134,7 @@ static void write_sm4_store(struct hlsl_ctx *ctx,
memset(&instr, 0, sizeof(instr));
instr.opcode = VKD3D_SM4_OP_MOV;
- sm4_register_from_deref(ctx, &instr.dsts[0].reg, &writemask, NULL, &store->lhs, rhs->data_type);
+ sm4_register_from_deref(ctx, &instr.dsts[0].reg, &writemask, NULL, &store->lhs);
instr.dsts[0].writemask = hlsl_combine_writemasks(writemask, store->writemask);
instr.dst_count = 1;
@@ -5144,8 +5245,8 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
const struct hlsl_ir_function_decl *entry_func, struct dxbc_writer *dxbc)
{
const struct hlsl_profile_info *profile = ctx->profile;
- const struct hlsl_ir_var **extern_resources;
struct vkd3d_bytecode_buffer buffer = {0};
+ struct extern_resource *extern_resources;
unsigned int extern_resources_count, i;
const struct hlsl_buffer *cbuffer;
const struct hlsl_ir_var *var;
@@ -5177,17 +5278,14 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
for (i = 0; i < extern_resources_count; ++i)
{
- enum hlsl_regset regset;
+ const struct extern_resource *resource = &extern_resources[i];
- var = extern_resources[i];
- regset = hlsl_type_get_regset(var->data_type);
-
- if (regset == HLSL_REGSET_SAMPLERS)
- write_sm4_dcl_samplers(&buffer, var);
- else if (regset == HLSL_REGSET_TEXTURES)
- write_sm4_dcl_textures(ctx, &buffer, var, false);
- else if (regset == HLSL_REGSET_UAVS)
- write_sm4_dcl_textures(ctx, &buffer, var, true);
+ if (resource->regset == HLSL_REGSET_SAMPLERS)
+ write_sm4_dcl_samplers(ctx, &buffer, resource);
+ else if (resource->regset == HLSL_REGSET_TEXTURES)
+ write_sm4_dcl_textures(ctx, &buffer, resource, false);
+ else if (resource->regset == HLSL_REGSET_UAVS)
+ write_sm4_dcl_textures(ctx, &buffer, resource, true);
}
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
@@ -5210,7 +5308,7 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
add_section(dxbc, TAG_SHDR, &buffer);
- vkd3d_free(extern_resources);
+ sm4_free_extern_resources(extern_resources, extern_resources_count);
}
int hlsl_sm4_write(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, struct vkd3d_shader_code *out)
diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
index 343fdb2252e..33d8c60e59a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
+++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
@@ -415,6 +415,8 @@ static const char *shader_get_source_type_suffix(enum vkd3d_shader_source_type t
return "hlsl";
case VKD3D_SHADER_SOURCE_D3D_BYTECODE:
return "d3dbc";
+ case VKD3D_SHADER_SOURCE_DXBC_DXIL:
+ return "dxil";
default:
FIXME("Unhandled source type %#x.\n", type);
return "bin";
@@ -440,6 +442,18 @@ void vkd3d_shader_dump_shader(enum vkd3d_shader_source_type source_type,
shader_get_source_type_suffix(source_type), shader->code, shader->size);
}
+static void init_scan_signature_info(const struct vkd3d_shader_compile_info *info)
+{
+ struct vkd3d_shader_scan_signature_info *signature_info;
+
+ if ((signature_info = vkd3d_find_struct(info->next, SCAN_SIGNATURE_INFO)))
+ {
+ memset(&signature_info->input, 0, sizeof(signature_info->input));
+ memset(&signature_info->output, 0, sizeof(signature_info->output));
+ memset(&signature_info->patch_constant, 0, sizeof(signature_info->patch_constant));
+ }
+}
+
bool vkd3d_shader_parser_init(struct vkd3d_shader_parser *parser,
struct vkd3d_shader_message_context *message_context, const char *source_name,
const struct vkd3d_shader_version *version, const struct vkd3d_shader_parser_ops *ops,
@@ -526,6 +540,43 @@ void vkd3d_shader_free_messages(char *messages)
vkd3d_free(messages);
}
+static bool vkd3d_shader_signature_from_shader_signature(struct vkd3d_shader_signature *signature,
+ const struct shader_signature *src)
+{
+ unsigned int i;
+
+ signature->element_count = src->element_count;
+ if (!src->elements)
+ {
+ assert(!signature->element_count);
+ signature->elements = NULL;
+ return true;
+ }
+
+ if (!(signature->elements = vkd3d_calloc(signature->element_count, sizeof(*signature->elements))))
+ return false;
+
+ for (i = 0; i < signature->element_count; ++i)
+ {
+ struct vkd3d_shader_signature_element *d = &signature->elements[i];
+ struct signature_element *e = &src->elements[i];
+
+ d->semantic_name = e->semantic_name;
+ d->semantic_index = e->semantic_index;
+ d->stream_index = e->stream_index;
+ d->sysval_semantic = e->sysval_semantic;
+ d->component_type = e->component_type;
+ d->register_index = e->register_index;
+ if (e->register_count > 1)
+ FIXME("Arrayed elements are not supported yet.\n");
+ d->mask = e->mask;
+ d->used_mask = e->used_mask;
+ d->min_precision = e->min_precision;
+ }
+
+ return true;
+}
+
struct vkd3d_shader_scan_context
{
struct vkd3d_shader_scan_descriptor_info *scan_descriptor_info;
@@ -1070,6 +1121,7 @@ static int scan_with_parser(const struct vkd3d_shader_compile_info *compile_info
struct vkd3d_shader_message_context *message_context, struct vkd3d_shader_parser *parser)
{
struct vkd3d_shader_scan_descriptor_info *scan_descriptor_info;
+ struct vkd3d_shader_scan_signature_info *signature_info;
struct vkd3d_shader_instruction *instruction;
struct vkd3d_shader_scan_context context;
int ret = VKD3D_OK;
@@ -1080,6 +1132,7 @@ static int scan_with_parser(const struct vkd3d_shader_compile_info *compile_info
scan_descriptor_info->descriptors = NULL;
scan_descriptor_info->descriptor_count = 0;
}
+ signature_info = vkd3d_find_struct(compile_info->next, SCAN_SIGNATURE_INFO);
vkd3d_shader_scan_context_init(&context, compile_info, scan_descriptor_info, message_context);
@@ -1099,6 +1152,21 @@ static int scan_with_parser(const struct vkd3d_shader_compile_info *compile_info
}
}
+ if (!ret && signature_info)
+ {
+ if (!vkd3d_shader_signature_from_shader_signature(&signature_info->input, &parser->shader_desc.input_signature)
+ || !vkd3d_shader_signature_from_shader_signature(&signature_info->output,
+ &parser->shader_desc.output_signature)
+ || !vkd3d_shader_signature_from_shader_signature(&signature_info->patch_constant,
+ &parser->shader_desc.patch_constant_signature))
+ {
+ vkd3d_shader_free_scan_signature_info(signature_info);
+ if (scan_descriptor_info)
+ vkd3d_shader_free_scan_descriptor_info(scan_descriptor_info);
+ ret = VKD3D_ERROR_OUT_OF_MEMORY;
+ }
+ }
+
vkd3d_shader_scan_context_cleanup(&context);
return ret;
}
@@ -1139,6 +1207,24 @@ static int scan_d3dbc(const struct vkd3d_shader_compile_info *compile_info,
return ret;
}
+static int scan_dxil(const struct vkd3d_shader_compile_info *compile_info,
+ struct vkd3d_shader_message_context *message_context)
+{
+ struct vkd3d_shader_parser *parser;
+ int ret;
+
+ if ((ret = vkd3d_shader_sm6_parser_create(compile_info, message_context, &parser)) < 0)
+ {
+ WARN("Failed to initialise shader parser.\n");
+ return ret;
+ }
+
+ ret = scan_with_parser(compile_info, message_context, parser);
+ vkd3d_shader_parser_destroy(parser);
+
+ return ret;
+}
+
int vkd3d_shader_scan(const struct vkd3d_shader_compile_info *compile_info, char **messages)
{
struct vkd3d_shader_message_context message_context;
@@ -1152,6 +1238,8 @@ int vkd3d_shader_scan(const struct vkd3d_shader_compile_info *compile_info, char
if ((ret = vkd3d_shader_validate_compile_info(compile_info, false)) < 0)
return ret;
+ init_scan_signature_info(compile_info);
+
vkd3d_shader_message_context_init(&message_context, compile_info->log_level);
switch (compile_info->source_type)
@@ -1169,6 +1257,10 @@ int vkd3d_shader_scan(const struct vkd3d_shader_compile_info *compile_info, char
ret = scan_d3dbc(compile_info, &message_context);
break;
+ case VKD3D_SHADER_SOURCE_DXBC_DXIL:
+ ret = scan_dxil(compile_info, &message_context);
+ break;
+
default:
ERR("Unsupported source type %#x.\n", compile_info->source_type);
ret = VKD3D_ERROR_INVALID_ARGUMENT;
@@ -1291,6 +1383,24 @@ static int compile_d3d_bytecode(const struct vkd3d_shader_compile_info *compile_
return VKD3D_ERROR;
}
+static int compile_dxbc_dxil(const struct vkd3d_shader_compile_info *compile_info,
+ struct vkd3d_shader_code *out, struct vkd3d_shader_message_context *message_context)
+{
+ struct vkd3d_shader_parser *parser;
+ int ret;
+
+ if ((ret = vkd3d_shader_sm6_parser_create(compile_info, message_context, &parser)) < 0)
+ {
+ WARN("Failed to initialise shader parser.\n");
+ return ret;
+ }
+
+ ret = vkd3d_shader_parser_compile(parser, compile_info, out, message_context);
+
+ vkd3d_shader_parser_destroy(parser);
+ return ret;
+}
+
int vkd3d_shader_compile(const struct vkd3d_shader_compile_info *compile_info,
struct vkd3d_shader_code *out, char **messages)
{
@@ -1305,6 +1415,8 @@ int vkd3d_shader_compile(const struct vkd3d_shader_compile_info *compile_info,
if ((ret = vkd3d_shader_validate_compile_info(compile_info, true)) < 0)
return ret;
+ init_scan_signature_info(compile_info);
+
vkd3d_shader_message_context_init(&message_context, compile_info->log_level);
switch (compile_info->source_type)
@@ -1321,6 +1433,10 @@ int vkd3d_shader_compile(const struct vkd3d_shader_compile_info *compile_info,
ret = compile_d3d_bytecode(compile_info, out, &message_context);
break;
+ case VKD3D_SHADER_SOURCE_DXBC_DXIL:
+ ret = compile_dxbc_dxil(compile_info, out, &message_context);
+ break;
+
default:
vkd3d_unreachable();
}
@@ -1339,6 +1455,15 @@ void vkd3d_shader_free_scan_descriptor_info(struct vkd3d_shader_scan_descriptor_
vkd3d_free(scan_descriptor_info->descriptors);
}
+void vkd3d_shader_free_scan_signature_info(struct vkd3d_shader_scan_signature_info *info)
+{
+ TRACE("info %p.\n", info);
+
+ vkd3d_shader_free_shader_signature(&info->input);
+ vkd3d_shader_free_shader_signature(&info->output);
+ vkd3d_shader_free_shader_signature(&info->patch_constant);
+}
+
void vkd3d_shader_free_shader_code(struct vkd3d_shader_code *shader_code)
{
TRACE("shader_code %p.\n", shader_code);
@@ -1401,43 +1526,6 @@ void vkd3d_shader_free_root_signature(struct vkd3d_shader_versioned_root_signatu
desc->version = 0;
}
-static bool vkd3d_shader_signature_from_shader_signature(struct vkd3d_shader_signature *signature,
- const struct shader_signature *src)
-{
- unsigned int i;
-
- signature->element_count = src->element_count;
- if (!src->elements)
- {
- assert(!signature->element_count);
- signature->elements = NULL;
- return true;
- }
-
- if (!(signature->elements = vkd3d_calloc(signature->element_count, sizeof(*signature->elements))))
- return false;
-
- for (i = 0; i < signature->element_count; ++i)
- {
- struct vkd3d_shader_signature_element *d = &signature->elements[i];
- struct signature_element *e = &src->elements[i];
-
- d->semantic_name = e->semantic_name;
- d->semantic_index = e->semantic_index;
- d->stream_index = e->stream_index;
- d->sysval_semantic = e->sysval_semantic;
- d->component_type = e->component_type;
- d->register_index = e->register_index;
- if (e->register_count > 1)
- FIXME("Arrayed elements are not supported yet.\n");
- d->mask = e->mask;
- d->used_mask = e->used_mask;
- d->min_precision = e->min_precision;
- }
-
- return true;
-}
-
void shader_signature_cleanup(struct shader_signature *signature)
{
vkd3d_free(signature->elements);
@@ -1526,6 +1614,7 @@ const enum vkd3d_shader_source_type *vkd3d_shader_get_supported_source_types(uns
VKD3D_SHADER_SOURCE_DXBC_TPF,
VKD3D_SHADER_SOURCE_HLSL,
VKD3D_SHADER_SOURCE_D3D_BYTECODE,
+ VKD3D_SHADER_SOURCE_DXBC_DXIL,
};
TRACE("count %p.\n", count);
@@ -1564,6 +1653,7 @@ const enum vkd3d_shader_target_type *vkd3d_shader_get_supported_target_types(
switch (source_type)
{
+ case VKD3D_SHADER_SOURCE_DXBC_DXIL:
case VKD3D_SHADER_SOURCE_DXBC_TPF:
*count = ARRAY_SIZE(dxbc_tpf_types);
return dxbc_tpf_types;
diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
index 406d53a3391..0e93f3a556a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
+++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
@@ -145,8 +145,29 @@ enum vkd3d_shader_error
VKD3D_SHADER_ERROR_D3DBC_INVALID_OPCODE = 7002,
VKD3D_SHADER_ERROR_D3DBC_INVALID_RESOURCE_TYPE = 7003,
VKD3D_SHADER_ERROR_D3DBC_OUT_OF_MEMORY = 7004,
+ VKD3D_SHADER_ERROR_D3DBC_INVALID_REGISTER_INDEX = 7005,
+ VKD3D_SHADER_ERROR_D3DBC_UNDECLARED_SEMANTIC = 7006,
VKD3D_SHADER_WARNING_D3DBC_IGNORED_INSTRUCTION_FLAGS= 7300,
+
+ VKD3D_SHADER_ERROR_DXIL_OUT_OF_MEMORY = 8000,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_SIZE = 8001,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_CHUNK_OFFSET = 8002,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_CHUNK_SIZE = 8003,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_BITCODE = 8004,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_OPERAND_COUNT = 8005,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_TABLE = 8006,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_VALUE_SYMTAB = 8007,
+ VKD3D_SHADER_ERROR_DXIL_UNSUPPORTED_BITCODE_FORMAT = 8008,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_FUNCTION_DCL = 8009,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_TYPE_ID = 8010,
+ VKD3D_SHADER_ERROR_DXIL_INVALID_MODULE = 8011,
+
+ VKD3D_SHADER_WARNING_DXIL_UNKNOWN_MAGIC_NUMBER = 8300,
+ VKD3D_SHADER_WARNING_DXIL_UNKNOWN_SHADER_TYPE = 8301,
+ VKD3D_SHADER_WARNING_DXIL_INVALID_BLOCK_LENGTH = 8302,
+ VKD3D_SHADER_WARNING_DXIL_INVALID_MODULE_LENGTH = 8303,
+ VKD3D_SHADER_WARNING_DXIL_IGNORING_OPERANDS = 8304,
};
enum vkd3d_shader_opcode
@@ -516,6 +537,7 @@ enum vkd3d_data_type
VKD3D_DATA_DOUBLE,
VKD3D_DATA_CONTINUED,
VKD3D_DATA_UNUSED,
+ VKD3D_DATA_UINT8,
};
enum vkd3d_immconst_type
@@ -802,6 +824,7 @@ struct signature_element
struct shader_signature
{
struct signature_element *elements;
+ size_t elements_capacity;
unsigned int element_count;
};
@@ -811,9 +834,12 @@ struct vkd3d_shader_desc
{
const uint32_t *byte_code;
size_t byte_code_size;
+ bool is_dxil;
struct shader_signature input_signature;
struct shader_signature output_signature;
struct shader_signature patch_constant_signature;
+
+ uint32_t temp_count;
};
struct vkd3d_shader_register_semantic
@@ -1167,6 +1193,8 @@ int vkd3d_shader_sm1_parser_create(const struct vkd3d_shader_compile_info *compi
struct vkd3d_shader_message_context *message_context, struct vkd3d_shader_parser **parser);
int vkd3d_shader_sm4_parser_create(const struct vkd3d_shader_compile_info *compile_info,
struct vkd3d_shader_message_context *message_context, struct vkd3d_shader_parser **parser);
+int vkd3d_shader_sm6_parser_create(const struct vkd3d_shader_compile_info *compile_info,
+ struct vkd3d_shader_message_context *message_context, struct vkd3d_shader_parser **parser);
void free_shader_desc(struct vkd3d_shader_desc *desc);
@@ -1339,6 +1367,7 @@ static inline void *vkd3d_find_struct_(const struct vkd3d_struct *chain,
}
#define VKD3D_DXBC_HEADER_SIZE (8 * sizeof(uint32_t))
+#define VKD3D_DXBC_CHUNK_ALIGNMENT sizeof(uint32_t)
#define TAG_AON9 VKD3D_MAKE_TAG('A', 'o', 'n', '9')
#define TAG_DXBC VKD3D_MAKE_TAG('D', 'X', 'B', 'C')
diff --git a/libs/vkd3d/libs/vkd3d/command.c b/libs/vkd3d/libs/vkd3d/command.c
index 32439eec7eb..53cb5d9582c 100644
--- a/libs/vkd3d/libs/vkd3d/command.c
+++ b/libs/vkd3d/libs/vkd3d/command.c
@@ -26,6 +26,7 @@ static HRESULT d3d12_fence_signal(struct d3d12_fence *fence, uint64_t value, VkF
static void d3d12_fence_signal_timeline_semaphore(struct d3d12_fence *fence, uint64_t timeline_value);
static HRESULT d3d12_command_queue_signal(struct d3d12_command_queue *command_queue,
struct d3d12_fence *fence, uint64_t value);
+static void d3d12_command_queue_submit_locked(struct d3d12_command_queue *queue);
static HRESULT d3d12_command_queue_flush_ops(struct d3d12_command_queue *queue, bool *flushed_any);
static HRESULT d3d12_command_queue_flush_ops_locked(struct d3d12_command_queue *queue, bool *flushed_any);
@@ -3624,7 +3625,7 @@ static HRESULT d3d12_command_list_allocate_transfer_buffer(struct d3d12_command_
static void d3d12_command_list_copy_incompatible_texture_region(struct d3d12_command_list *list,
struct d3d12_resource *dst_resource, unsigned int dst_sub_resource_idx,
const struct vkd3d_format *dst_format, struct d3d12_resource *src_resource,
- unsigned int src_sub_resource_idx, const struct vkd3d_format *src_format)
+ unsigned int src_sub_resource_idx, const struct vkd3d_format *src_format, unsigned int layer_count)
{
const struct vkd3d_vk_device_procs *vk_procs = &list->device->vk_procs;
const D3D12_RESOURCE_DESC *dst_desc = &dst_resource->desc;
@@ -3651,6 +3652,7 @@ static void d3d12_command_list_copy_incompatible_texture_region(struct d3d12_com
buffer_image_copy.bufferImageHeight = 0;
vk_image_subresource_layers_from_d3d12(&buffer_image_copy.imageSubresource,
src_format, src_sub_resource_idx, src_desc->MipLevels);
+ buffer_image_copy.imageSubresource.layerCount = layer_count;
src_miplevel_idx = buffer_image_copy.imageSubresource.mipLevel;
buffer_image_copy.imageOffset.x = 0;
buffer_image_copy.imageOffset.y = 0;
@@ -3658,7 +3660,7 @@ static void d3d12_command_list_copy_incompatible_texture_region(struct d3d12_com
vk_extent_3d_from_d3d12_miplevel(&buffer_image_copy.imageExtent, src_desc, src_miplevel_idx);
buffer_size = src_format->byte_count * buffer_image_copy.imageExtent.width *
- buffer_image_copy.imageExtent.height * buffer_image_copy.imageExtent.depth;
+ buffer_image_copy.imageExtent.height * buffer_image_copy.imageExtent.depth * layer_count;
if (FAILED(hr = d3d12_command_list_allocate_transfer_buffer(list, buffer_size, &transfer_buffer)))
{
ERR("Failed to allocate transfer buffer, hr %#x.\n", hr);
@@ -3684,6 +3686,7 @@ static void d3d12_command_list_copy_incompatible_texture_region(struct d3d12_com
vk_image_subresource_layers_from_d3d12(&buffer_image_copy.imageSubresource,
dst_format, dst_sub_resource_idx, dst_desc->MipLevels);
+ buffer_image_copy.imageSubresource.layerCount = layer_count;
dst_miplevel_idx = buffer_image_copy.imageSubresource.mipLevel;
assert(d3d12_resource_desc_get_width(src_desc, src_miplevel_idx) ==
@@ -3813,7 +3816,7 @@ static void STDMETHODCALLTYPE d3d12_command_list_CopyTextureRegion(ID3D12Graphic
{
d3d12_command_list_copy_incompatible_texture_region(list,
dst_resource, dst->u.SubresourceIndex, dst_format,
- src_resource, src->u.SubresourceIndex, src_format);
+ src_resource, src->u.SubresourceIndex, src_format, 1);
return;
}
@@ -3835,6 +3838,7 @@ static void STDMETHODCALLTYPE d3d12_command_list_CopyResource(ID3D12GraphicsComm
{
struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList2(iface);
struct d3d12_resource *dst_resource, *src_resource;
+ const struct vkd3d_format *dst_format, *src_format;
const struct vkd3d_vk_device_procs *vk_procs;
VkBufferCopy vk_buffer_copy;
VkImageCopy vk_image_copy;
@@ -3867,16 +3871,29 @@ static void STDMETHODCALLTYPE d3d12_command_list_CopyResource(ID3D12GraphicsComm
else
{
layer_count = d3d12_resource_desc_get_layer_count(&dst_resource->desc);
+ dst_format = dst_resource->format;
+ src_format = src_resource->format;
assert(d3d12_resource_is_texture(dst_resource));
assert(d3d12_resource_is_texture(src_resource));
assert(dst_resource->desc.MipLevels == src_resource->desc.MipLevels);
assert(layer_count == d3d12_resource_desc_get_layer_count(&src_resource->desc));
+ if (src_format->vk_aspect_mask != dst_format->vk_aspect_mask)
+ {
+ for (i = 0; i < dst_resource->desc.MipLevels; ++i)
+ {
+ d3d12_command_list_copy_incompatible_texture_region(list,
+ dst_resource, i, dst_format,
+ src_resource, i, src_format, layer_count);
+ }
+ return;
+ }
+
for (i = 0; i < dst_resource->desc.MipLevels; ++i)
{
vk_image_copy_from_d3d12(&vk_image_copy, i, i, &src_resource->desc, &dst_resource->desc,
- src_resource->format, dst_resource->format, NULL, 0, 0, 0);
+ src_format, dst_format, NULL, 0, 0, 0);
vk_image_copy.dstSubresource.layerCount = layer_count;
vk_image_copy.srcSubresource.layerCount = layer_count;
VK_CALL(vkCmdCopyImage(list->vk_command_buffer, src_resource->u.vk_image,
@@ -6063,8 +6080,35 @@ static ULONG STDMETHODCALLTYPE d3d12_command_queue_AddRef(ID3D12CommandQueue *if
return refcount;
}
+static void d3d12_command_queue_destroy_op(struct vkd3d_cs_op_data *op)
+{
+ switch (op->opcode)
+ {
+ case VKD3D_CS_OP_WAIT:
+ d3d12_fence_decref(op->u.wait.fence);
+ break;
+
+ case VKD3D_CS_OP_SIGNAL:
+ d3d12_fence_decref(op->u.signal.fence);
+ break;
+
+ case VKD3D_CS_OP_EXECUTE:
+ vkd3d_free(op->u.execute.buffers);
+ break;
+
+ case VKD3D_CS_OP_UPDATE_MAPPINGS:
+ case VKD3D_CS_OP_COPY_MAPPINGS:
+ break;
+ }
+}
+
static void d3d12_command_queue_op_array_destroy(struct d3d12_command_queue_op_array *array)
{
+ unsigned int i;
+
+ for (i = 0; i < array->count; ++i)
+ d3d12_command_queue_destroy_op(&array->ops[i]);
+
vkd3d_free(array->ops);
}
@@ -6162,17 +6206,131 @@ static struct vkd3d_cs_op_data *d3d12_command_queue_op_array_require_space(struc
return &array->ops[array->count++];
}
+static bool clone_array_parameter(void **dst, const void *src, size_t elem_size, unsigned int count)
+{
+ void *buffer;
+
+ *dst = NULL;
+ if (src)
+ {
+ if (!(buffer = vkd3d_calloc(count, elem_size)))
+ return false;
+ memcpy(buffer, src, count * elem_size);
+ *dst = buffer;
+ }
+ return true;
+}
+
+static void update_mappings_cleanup(struct vkd3d_cs_update_mappings *update_mappings)
+{
+ vkd3d_free(update_mappings->region_start_coordinates);
+ vkd3d_free(update_mappings->region_sizes);
+ vkd3d_free(update_mappings->range_flags);
+ vkd3d_free(update_mappings->heap_range_offsets);
+ vkd3d_free(update_mappings->range_tile_counts);
+}
+
static void STDMETHODCALLTYPE d3d12_command_queue_UpdateTileMappings(ID3D12CommandQueue *iface,
ID3D12Resource *resource, UINT region_count,
const D3D12_TILED_RESOURCE_COORDINATE *region_start_coordinates, const D3D12_TILE_REGION_SIZE *region_sizes,
ID3D12Heap *heap, UINT range_count, const D3D12_TILE_RANGE_FLAGS *range_flags,
- UINT *heap_range_offsets, UINT *range_tile_counts, D3D12_TILE_MAPPING_FLAGS flags)
+ const UINT *heap_range_offsets, const UINT *range_tile_counts, D3D12_TILE_MAPPING_FLAGS flags)
{
- FIXME("iface %p, resource %p, region_count %u, region_start_coordinates %p, "
+ struct d3d12_resource *resource_impl = unsafe_impl_from_ID3D12Resource(resource);
+ struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface);
+ struct d3d12_heap *heap_impl = unsafe_impl_from_ID3D12Heap(heap);
+ struct vkd3d_cs_update_mappings update_mappings = {0};
+ struct vkd3d_cs_op_data *op;
+
+ TRACE("iface %p, resource %p, region_count %u, region_start_coordinates %p, "
"region_sizes %p, heap %p, range_count %u, range_flags %p, heap_range_offsets %p, "
- "range_tile_counts %p, flags %#x stub!\n",
+ "range_tile_counts %p, flags %#x.\n",
iface, resource, region_count, region_start_coordinates, region_sizes, heap, range_count,
range_flags, heap_range_offsets, range_tile_counts, flags);
+
+ if (!region_count || !range_count)
+ return;
+
+ if (!command_queue->supports_sparse_binding)
+ {
+ FIXME("Command queue %p does not support sparse binding.\n", command_queue);
+ return;
+ }
+
+ if (!resource_impl->tiles.subresource_count)
+ {
+ WARN("Resource %p is not a tiled resource.\n", resource_impl);
+ return;
+ }
+
+ if (region_count > 1 && !region_start_coordinates)
+ {
+ WARN("Region start coordinates must not be NULL when region count is > 1.\n");
+ return;
+ }
+
+ if (range_count > 1 && !range_tile_counts)
+ {
+ WARN("Range tile counts must not be NULL when range count is > 1.\n");
+ return;
+ }
+
+ update_mappings.resource = resource_impl;
+ update_mappings.heap = heap_impl;
+ if (!clone_array_parameter((void **)&update_mappings.region_start_coordinates,
+ region_start_coordinates, sizeof(*region_start_coordinates), region_count))
+ {
+ ERR("Failed to allocate region start coordinates.\n");
+ return;
+ }
+ if (!clone_array_parameter((void **)&update_mappings.region_sizes,
+ region_sizes, sizeof(*region_sizes), region_count))
+ {
+ ERR("Failed to allocate region sizes.\n");
+ goto free_clones;
+ }
+ if (!clone_array_parameter((void **)&update_mappings.range_flags,
+ range_flags, sizeof(*range_flags), range_count))
+ {
+ ERR("Failed to allocate range flags.\n");
+ goto free_clones;
+ }
+ if (!clone_array_parameter((void **)&update_mappings.heap_range_offsets,
+ heap_range_offsets, sizeof(*heap_range_offsets), range_count))
+ {
+ ERR("Failed to allocate heap range offsets.\n");
+ goto free_clones;
+ }
+ if (!clone_array_parameter((void **)&update_mappings.range_tile_counts,
+ range_tile_counts, sizeof(*range_tile_counts), range_count))
+ {
+ ERR("Failed to allocate range tile counts.\n");
+ goto free_clones;
+ }
+ update_mappings.region_count = region_count;
+ update_mappings.range_count = range_count;
+ update_mappings.flags = flags;
+
+ vkd3d_mutex_lock(&command_queue->op_mutex);
+
+ if (!(op = d3d12_command_queue_op_array_require_space(&command_queue->op_queue)))
+ {
+ ERR("Failed to add op.\n");
+ goto unlock_mutex;
+ }
+
+ op->opcode = VKD3D_CS_OP_UPDATE_MAPPINGS;
+ op->u.update_mappings = update_mappings;
+
+ d3d12_command_queue_submit_locked(command_queue);
+
+ vkd3d_mutex_unlock(&command_queue->op_mutex);
+ return;
+
+unlock_mutex:
+ vkd3d_mutex_unlock(&command_queue->op_mutex);
+free_clones:
+ update_mappings_cleanup(&update_mappings);
}
static void STDMETHODCALLTYPE d3d12_command_queue_CopyTileMappings(ID3D12CommandQueue *iface,
@@ -6183,10 +6341,34 @@ static void STDMETHODCALLTYPE d3d12_command_queue_CopyTileMappings(ID3D12Command
const D3D12_TILE_REGION_SIZE *region_size,
D3D12_TILE_MAPPING_FLAGS flags)
{
- FIXME("iface %p, dst_resource %p, dst_region_start_coordinate %p, "
- "src_resource %p, src_region_start_coordinate %p, region_size %p, flags %#x stub!\n",
+ struct d3d12_resource *dst_resource_impl = impl_from_ID3D12Resource(dst_resource);
+ struct d3d12_resource *src_resource_impl = impl_from_ID3D12Resource(src_resource);
+ struct d3d12_command_queue *command_queue = impl_from_ID3D12CommandQueue(iface);
+ struct vkd3d_cs_op_data *op;
+
+ TRACE("iface %p, dst_resource %p, dst_region_start_coordinate %p, "
+ "src_resource %p, src_region_start_coordinate %p, region_size %p, flags %#x.\n",
iface, dst_resource, dst_region_start_coordinate, src_resource,
src_region_start_coordinate, region_size, flags);
+
+ vkd3d_mutex_lock(&command_queue->op_mutex);
+
+ if (!(op = d3d12_command_queue_op_array_require_space(&command_queue->op_queue)))
+ {
+ ERR("Failed to add op.\n");
+ return;
+ }
+ op->opcode = VKD3D_CS_OP_COPY_MAPPINGS;
+ op->u.copy_mappings.dst_resource = dst_resource_impl;
+ op->u.copy_mappings.src_resource = src_resource_impl;
+ op->u.copy_mappings.dst_region_start_coordinate = *dst_region_start_coordinate;
+ op->u.copy_mappings.src_region_start_coordinate = *src_region_start_coordinate;
+ op->u.copy_mappings.region_size = *region_size;
+ op->u.copy_mappings.flags = flags;
+
+ d3d12_command_queue_submit_locked(command_queue);
+
+ vkd3d_mutex_unlock(&command_queue->op_mutex);
}
static void d3d12_command_queue_execute(struct d3d12_command_queue *command_queue,
@@ -6214,8 +6396,6 @@ static void d3d12_command_queue_execute(struct d3d12_command_queue *command_queu
ERR("Failed to submit queue(s), vr %d.\n", vr);
vkd3d_queue_release(vkd3d_queue);
-
- vkd3d_free(buffers);
}
static void d3d12_command_queue_submit_locked(struct d3d12_command_queue *queue)
@@ -6273,7 +6453,7 @@ static void STDMETHODCALLTYPE d3d12_command_queue_ExecuteCommandLists(ID3D12Comm
if (!(op = d3d12_command_queue_op_array_require_space(&command_queue->op_queue)))
{
ERR("Failed to add op.\n");
- return;
+ goto done;
}
op->opcode = VKD3D_CS_OP_EXECUTE;
op->u.execute.buffers = buffers;
@@ -6281,6 +6461,7 @@ static void STDMETHODCALLTYPE d3d12_command_queue_ExecuteCommandLists(ID3D12Comm
d3d12_command_queue_submit_locked(command_queue);
+done:
vkd3d_mutex_unlock(&command_queue->op_mutex);
return;
}
@@ -6348,6 +6529,7 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Signal(ID3D12CommandQueue *
if (!(op = d3d12_command_queue_op_array_require_space(&command_queue->op_queue)))
{
+ ERR("Failed to add op.\n");
hr = E_OUTOFMEMORY;
goto done;
}
@@ -6686,6 +6868,7 @@ static HRESULT STDMETHODCALLTYPE d3d12_command_queue_Wait(ID3D12CommandQueue *if
if (!(op = d3d12_command_queue_op_array_require_space(&command_queue->op_queue)))
{
+ ERR("Failed to add op.\n");
hr = E_OUTOFMEMORY;
goto done;
}
@@ -6922,22 +7105,31 @@ static HRESULT d3d12_command_queue_flush_ops_locked(struct d3d12_command_queue *
return d3d12_command_queue_fixup_after_flush_locked(queue);
}
d3d12_command_queue_wait_locked(queue, fence, op->u.wait.value);
- d3d12_fence_decref(fence);
break;
case VKD3D_CS_OP_SIGNAL:
d3d12_command_queue_signal(queue, op->u.signal.fence, op->u.signal.value);
- d3d12_fence_decref(op->u.signal.fence);
break;
case VKD3D_CS_OP_EXECUTE:
d3d12_command_queue_execute(queue, op->u.execute.buffers, op->u.execute.buffer_count);
break;
+ case VKD3D_CS_OP_UPDATE_MAPPINGS:
+ FIXME("Tiled resource binding is not supported yet.\n");
+ update_mappings_cleanup(&op->u.update_mappings);
+ break;
+
+ case VKD3D_CS_OP_COPY_MAPPINGS:
+ FIXME("Tiled resource mapping copying is not supported yet.\n");
+ break;
+
default:
vkd3d_unreachable();
}
+ d3d12_command_queue_destroy_op(op);
+
*flushed_any |= true;
}
@@ -7000,6 +7192,8 @@ static HRESULT d3d12_command_queue_init(struct d3d12_command_queue *queue,
if (FAILED(hr = vkd3d_fence_worker_start(&queue->fence_worker, queue->vkd3d_queue, device)))
goto fail_destroy_op_mutex;
+ queue->supports_sparse_binding = !!(queue->vkd3d_queue->vk_queue_flags & VK_QUEUE_SPARSE_BINDING_BIT);
+
d3d12_device_add_ref(queue->device = device);
return S_OK;
diff --git a/libs/vkd3d/libs/vkd3d/device.c b/libs/vkd3d/libs/vkd3d/device.c
index 4263dcf4184..b9a8943cc08 100644
--- a/libs/vkd3d/libs/vkd3d/device.c
+++ b/libs/vkd3d/libs/vkd3d/device.c
@@ -1464,6 +1464,8 @@ static HRESULT vkd3d_init_device_caps(struct d3d12_device *device,
vulkan_info->device_limits = physical_device_info->properties2.properties.limits;
vulkan_info->sparse_properties = physical_device_info->properties2.properties.sparseProperties;
+ vulkan_info->sparse_binding = features->sparseBinding;
+ vulkan_info->sparse_residency_3d = features->sparseResidencyImage3D;
vulkan_info->rasterization_stream = physical_device_info->xfb_properties.transformFeedbackRasterizationStreamSelect;
vulkan_info->transform_feedback_queries = physical_device_info->xfb_properties.transformFeedbackQueries;
vulkan_info->uav_read_without_format = features->shaderStorageImageReadWithoutFormat;
diff --git a/libs/vkd3d/libs/vkd3d/resource.c b/libs/vkd3d/libs/vkd3d/resource.c
index ea7b6859cc1..4c07d326504 100644
--- a/libs/vkd3d/libs/vkd3d/resource.c
+++ b/libs/vkd3d/libs/vkd3d/resource.c
@@ -779,6 +779,7 @@ static HRESULT vkd3d_create_image(struct d3d12_device *device,
VkImageFormatListCreateInfoKHR format_list;
const struct vkd3d_format *format;
VkImageCreateInfo image_info;
+ uint32_t count;
VkResult vr;
if (resource)
@@ -914,6 +915,20 @@ static HRESULT vkd3d_create_image(struct d3d12_device *device,
if (resource && image_info.tiling == VK_IMAGE_TILING_LINEAR)
resource->flags |= VKD3D_RESOURCE_LINEAR_TILING;
+ if (sparse_resource)
+ {
+ count = 0;
+ VK_CALL(vkGetPhysicalDeviceSparseImageFormatProperties(device->vk_physical_device, image_info.format,
+ image_info.imageType, image_info.samples, image_info.usage, image_info.tiling, &count, NULL));
+
+ if (!count)
+ {
+ FIXME("Sparse images are not supported with format %u, type %u, samples %u, usage %#x.\n",
+ image_info.format, image_info.imageType, image_info.samples, image_info.usage);
+ return E_INVALIDARG;
+ }
+ }
+
if ((vr = VK_CALL(vkCreateImage(device->vk_device, &image_info, NULL, vk_image))) < 0)
WARN("Failed to create Vulkan image, vr %d.\n", vr);
@@ -928,6 +943,7 @@ HRESULT vkd3d_get_image_allocation_info(struct d3d12_device *device,
D3D12_RESOURCE_DESC validated_desc;
VkMemoryRequirements requirements;
VkImage vk_image;
+ bool tiled;
HRESULT hr;
assert(desc->Dimension != D3D12_RESOURCE_DIMENSION_BUFFER);
@@ -940,8 +956,10 @@ HRESULT vkd3d_get_image_allocation_info(struct d3d12_device *device,
desc = &validated_desc;
}
+ tiled = desc->Layout == D3D12_TEXTURE_LAYOUT_64KB_UNDEFINED_SWIZZLE;
+
/* XXX: We have to create an image to get its memory requirements. */
- if (SUCCEEDED(hr = vkd3d_create_image(device, &heap_properties, 0, desc, NULL, &vk_image)))
+ if (SUCCEEDED(hr = vkd3d_create_image(device, tiled ? NULL : &heap_properties, 0, desc, NULL, &vk_image)))
{
VK_CALL(vkGetImageMemoryRequirements(device->vk_device, vk_image, &requirements));
VK_CALL(vkDestroyImage(device->vk_device, vk_image, NULL));
@@ -1039,12 +1057,12 @@ static void d3d12_resource_get_level_box(const struct d3d12_resource *resource,
box->back = d3d12_resource_desc_get_depth(&resource->desc, level);
}
-/* ID3D12Resource */
-static inline struct d3d12_resource *impl_from_ID3D12Resource(ID3D12Resource *iface)
+static void d3d12_resource_init_tiles(struct d3d12_resource *resource)
{
- return CONTAINING_RECORD(iface, struct d3d12_resource, ID3D12Resource_iface);
+ resource->tiles.subresource_count = d3d12_resource_desc_get_sub_resource_count(&resource->desc);
}
+/* ID3D12Resource */
static HRESULT STDMETHODCALLTYPE d3d12_resource_QueryInterface(ID3D12Resource *iface,
REFIID riid, void **object)
{
@@ -1661,6 +1679,21 @@ HRESULT d3d12_resource_validate_desc(const D3D12_RESOURCE_DESC *desc, struct d3d
return E_INVALIDARG;
}
+ if (desc->Layout == D3D12_TEXTURE_LAYOUT_64KB_UNDEFINED_SWIZZLE)
+ {
+ if (desc->Dimension == D3D12_RESOURCE_DIMENSION_TEXTURE3D && !device->vk_info.sparse_residency_3d)
+ {
+ WARN("The device does not support tiled 3D images.\n");
+ return E_INVALIDARG;
+ }
+ if (format->plane_count > 1)
+ {
+ WARN("Invalid format %#x. D3D12 does not support multiplanar formats for tiled resources.\n",
+ format->dxgi_format);
+ return E_INVALIDARG;
+ }
+ }
+
if (!d3d12_resource_validate_texture_format(desc, format)
|| !d3d12_resource_validate_texture_alignment(desc, format))
return E_INVALIDARG;
@@ -1722,6 +1755,12 @@ static HRESULT d3d12_resource_init(struct d3d12_resource *resource, struct d3d12
resource->desc = *desc;
+ if (!heap_properties && !device->vk_info.sparse_binding)
+ {
+ WARN("The device does not support tiled images.\n");
+ return E_INVALIDARG;
+ }
+
if (heap_properties && !d3d12_resource_validate_heap_properties(resource, heap_properties, initial_state))
return E_INVALIDARG;
@@ -1787,6 +1826,8 @@ static HRESULT d3d12_resource_init(struct d3d12_resource *resource, struct d3d12
resource->heap = NULL;
resource->heap_offset = 0;
+ memset(&resource->tiles, 0, sizeof(resource->tiles));
+
if (FAILED(hr = vkd3d_private_store_init(&resource->private_store)))
{
d3d12_resource_destroy(resource, device);
@@ -1972,6 +2013,8 @@ HRESULT d3d12_reserved_resource_create(struct d3d12_device *device,
desc, initial_state, optimized_clear_value, &object)))
return hr;
+ d3d12_resource_init_tiles(object);
+
TRACE("Created reserved resource %p.\n", object);
*resource = object;
diff --git a/libs/vkd3d/libs/vkd3d/state.c b/libs/vkd3d/libs/vkd3d/state.c
index 5e46b467252..2d8138245d8 100644
--- a/libs/vkd3d/libs/vkd3d/state.c
+++ b/libs/vkd3d/libs/vkd3d/state.c
@@ -20,6 +20,7 @@
#include "vkd3d_private.h"
#include "vkd3d_shaders.h"
+#include "vkd3d_shader_utils.h"
/* ID3D12RootSignature */
static inline struct d3d12_root_signature *impl_from_ID3D12RootSignature(ID3D12RootSignature *iface)
@@ -1978,14 +1979,14 @@ static HRESULT create_shader_stage(struct d3d12_device *device,
compile_info.next = shader_interface;
compile_info.source.code = code->pShaderBytecode;
compile_info.source.size = code->BytecodeLength;
- compile_info.source_type = VKD3D_SHADER_SOURCE_DXBC_TPF;
compile_info.target_type = VKD3D_SHADER_TARGET_SPIRV_BINARY;
compile_info.options = options;
compile_info.option_count = ARRAY_SIZE(options);
compile_info.log_level = VKD3D_SHADER_LOG_NONE;
compile_info.source_name = NULL;
- if ((ret = vkd3d_shader_compile(&compile_info, &spirv, NULL)) < 0)
+ if ((ret = vkd3d_shader_parse_dxbc_source_type(&compile_info.source, &compile_info.source_type, NULL)) < 0
+ || (ret = vkd3d_shader_compile(&compile_info, &spirv, NULL)) < 0)
{
WARN("Failed to compile shader, vkd3d result %d.\n", ret);
return hresult_from_vkd3d_result(ret);
@@ -2008,6 +2009,7 @@ static int vkd3d_scan_dxbc(const struct d3d12_device *device, const D3D12_SHADER
struct vkd3d_shader_scan_descriptor_info *descriptor_info)
{
struct vkd3d_shader_compile_info compile_info;
+ enum vkd3d_result ret;
const struct vkd3d_shader_compile_option options[] =
{
@@ -2019,13 +2021,15 @@ static int vkd3d_scan_dxbc(const struct d3d12_device *device, const D3D12_SHADER
compile_info.next = descriptor_info;
compile_info.source.code = code->pShaderBytecode;
compile_info.source.size = code->BytecodeLength;
- compile_info.source_type = VKD3D_SHADER_SOURCE_DXBC_TPF;
compile_info.target_type = VKD3D_SHADER_TARGET_SPIRV_BINARY;
compile_info.options = options;
compile_info.option_count = ARRAY_SIZE(options);
compile_info.log_level = VKD3D_SHADER_LOG_NONE;
compile_info.source_name = NULL;
+ if ((ret = vkd3d_shader_parse_dxbc_source_type(&compile_info.source, &compile_info.source_type, NULL)) < 0)
+ return ret;
+
return vkd3d_shader_scan(&compile_info, NULL);
}
diff --git a/libs/vkd3d/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/libs/vkd3d/vkd3d_private.h
index b0150754434..c5259420acf 100644
--- a/libs/vkd3d/libs/vkd3d/vkd3d_private.h
+++ b/libs/vkd3d/libs/vkd3d/vkd3d_private.h
@@ -147,9 +147,12 @@ struct vkd3d_vulkan_info
unsigned int max_vertex_attrib_divisor;
VkPhysicalDeviceLimits device_limits;
- VkPhysicalDeviceSparseProperties sparse_properties;
struct vkd3d_device_descriptor_limits descriptor_limits;
+ VkPhysicalDeviceSparseProperties sparse_properties;
+ bool sparse_binding;
+ bool sparse_residency_3d;
+
VkPhysicalDeviceTexelBufferAlignmentPropertiesEXT texel_buffer_alignment_properties;
unsigned int shader_extension_count;
@@ -670,6 +673,11 @@ struct d3d12_heap *unsafe_impl_from_ID3D12Heap(ID3D12Heap *iface);
#define VKD3D_RESOURCE_DEDICATED_HEAP 0x00000008
#define VKD3D_RESOURCE_LINEAR_TILING 0x00000010
+struct d3d12_resource_tile_info
+{
+ unsigned int subresource_count;
+};
+
/* ID3D12Resource */
struct d3d12_resource
{
@@ -698,9 +706,16 @@ struct d3d12_resource
struct d3d12_device *device;
+ struct d3d12_resource_tile_info tiles;
+
struct vkd3d_private_store private_store;
};
+static inline struct d3d12_resource *impl_from_ID3D12Resource(ID3D12Resource *iface)
+{
+ return CONTAINING_RECORD(iface, struct d3d12_resource, ID3D12Resource_iface);
+}
+
static inline bool d3d12_resource_is_buffer(const struct d3d12_resource *resource)
{
return resource->desc.Dimension == D3D12_RESOURCE_DIMENSION_BUFFER;
@@ -1454,6 +1469,8 @@ enum vkd3d_cs_op
VKD3D_CS_OP_WAIT,
VKD3D_CS_OP_SIGNAL,
VKD3D_CS_OP_EXECUTE,
+ VKD3D_CS_OP_UPDATE_MAPPINGS,
+ VKD3D_CS_OP_COPY_MAPPINGS,
};
struct vkd3d_cs_wait
@@ -1474,6 +1491,30 @@ struct vkd3d_cs_execute
unsigned int buffer_count;
};
+struct vkd3d_cs_update_mappings
+{
+ struct d3d12_resource *resource;
+ struct d3d12_heap *heap;
+ D3D12_TILED_RESOURCE_COORDINATE *region_start_coordinates;
+ D3D12_TILE_REGION_SIZE *region_sizes;
+ D3D12_TILE_RANGE_FLAGS *range_flags;
+ UINT *heap_range_offsets;
+ UINT *range_tile_counts;
+ UINT region_count;
+ UINT range_count;
+ D3D12_TILE_MAPPING_FLAGS flags;
+};
+
+struct vkd3d_cs_copy_mappings
+{
+ struct d3d12_resource *dst_resource;
+ struct d3d12_resource *src_resource;
+ D3D12_TILED_RESOURCE_COORDINATE dst_region_start_coordinate;
+ D3D12_TILED_RESOURCE_COORDINATE src_region_start_coordinate;
+ D3D12_TILE_REGION_SIZE region_size;
+ D3D12_TILE_MAPPING_FLAGS flags;
+};
+
struct vkd3d_cs_op_data
{
enum vkd3d_cs_op opcode;
@@ -1482,6 +1523,8 @@ struct vkd3d_cs_op_data
struct vkd3d_cs_wait wait;
struct vkd3d_cs_signal signal;
struct vkd3d_cs_execute execute;
+ struct vkd3d_cs_update_mappings update_mappings;
+ struct vkd3d_cs_copy_mappings copy_mappings;
} u;
};
@@ -1519,6 +1562,8 @@ struct d3d12_command_queue
* set, aux_op_queue.count must be zero. */
struct d3d12_command_queue_op_array aux_op_queue;
+ bool supports_sparse_binding;
+
struct vkd3d_private_store private_store;
};
--
2.40.1
Reference in New Issue View Git Blame Copy Permalink