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
ad13b6a9e1
wine-staging/patches/vkd3d-latest/0001-Updated-vkd3d-to-d3ba810c98b4d2df260a527f74586a0b314.patch
Alistair Leslie-Hughes 12fb6d73ff Updated vkd3d-latest patchset
2024-07-11 09:54:47 +10:00

3889 lines
172 KiB
Diff
Raw Blame History

From bcb85270b8635f3d35b7d559c1800597589c62d1 Mon Sep 17 00:00:00 2001
From: Alistair Leslie-Hughes <leslie_alistair@hotmail.com>
Date: Thu, 7 Mar 2024 10:40:41 +1100
Subject: [PATCH] Updated vkd3d to d3ba810c98b4d2df260a527f74586a0b31408510.
---
libs/vkd3d/include/private/vkd3d_common.h | 2 +-
libs/vkd3d/libs/vkd3d-common/blob.c | 1 +
libs/vkd3d/libs/vkd3d-shader/d3d_asm.c | 25 +-
libs/vkd3d/libs/vkd3d-shader/d3dbc.c | 83 ++++-
libs/vkd3d/libs/vkd3d-shader/dxil.c | 99 ++++--
libs/vkd3d/libs/vkd3d-shader/fx.c | 21 +-
libs/vkd3d/libs/vkd3d-shader/glsl.c | 6 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.c | 53 ++-
libs/vkd3d/libs/vkd3d-shader/hlsl.h | 46 ++-
libs/vkd3d/libs/vkd3d-shader/hlsl.l | 3 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.y | 234 ++++++++++---
libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c | 117 +++++--
libs/vkd3d/libs/vkd3d-shader/ir.c | 144 ++++----
libs/vkd3d/libs/vkd3d-shader/preproc.l | 1 +
libs/vkd3d/libs/vkd3d-shader/spirv.c | 229 +++++++++----
libs/vkd3d/libs/vkd3d-shader/tpf.c | 317 ++++++++++++------
.../libs/vkd3d-shader/vkd3d_shader_main.c | 30 +-
.../libs/vkd3d-shader/vkd3d_shader_private.h | 11 +-
libs/vkd3d/libs/vkd3d/command.c | 136 ++++----
libs/vkd3d/libs/vkd3d/device.c | 15 +-
libs/vkd3d/libs/vkd3d/resource.c | 8 -
libs/vkd3d/libs/vkd3d/vkd3d_private.h | 2 +
22 files changed, 1116 insertions(+), 467 deletions(-)
diff --git a/libs/vkd3d/include/private/vkd3d_common.h b/libs/vkd3d/include/private/vkd3d_common.h
index a9d709d10fe..2d950b4f7aa 100644
--- a/libs/vkd3d/include/private/vkd3d_common.h
+++ b/libs/vkd3d/include/private/vkd3d_common.h
@@ -233,7 +233,7 @@ static inline unsigned int vkd3d_popcount(unsigned int v)
{
#ifdef _MSC_VER
return __popcnt(v);
-#elif defined(__MINGW32__)
+#elif defined(HAVE_BUILTIN_POPCOUNT)
return __builtin_popcount(v);
#else
v -= (v >> 1) & 0x55555555;
diff --git a/libs/vkd3d/libs/vkd3d-common/blob.c b/libs/vkd3d/libs/vkd3d-common/blob.c
index f60ef7db769..c2c6ad67804 100644
--- a/libs/vkd3d/libs/vkd3d-common/blob.c
+++ b/libs/vkd3d/libs/vkd3d-common/blob.c
@@ -20,6 +20,7 @@
#define WIDL_C_INLINE_WRAPPERS
#endif
#define COBJMACROS
+
#define CONST_VTABLE
#include "vkd3d.h"
#include "vkd3d_blob.h"
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
index 9abc2c4db70..8a3eb5a367a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3d_asm.c
@@ -254,6 +254,10 @@ static const char * const shader_opcode_names[] =
[VKD3DSIH_PHASE ] = "phase",
[VKD3DSIH_PHI ] = "phi",
[VKD3DSIH_POW ] = "pow",
+ [VKD3DSIH_QUAD_READ_ACROSS_D ] = "quad_read_across_d",
+ [VKD3DSIH_QUAD_READ_ACROSS_X ] = "quad_read_across_x",
+ [VKD3DSIH_QUAD_READ_ACROSS_Y ] = "quad_read_across_y",
+ [VKD3DSIH_QUAD_READ_LANE_AT ] = "quad_read_lane_at",
[VKD3DSIH_RCP ] = "rcp",
[VKD3DSIH_REP ] = "rep",
[VKD3DSIH_RESINFO ] = "resinfo",
@@ -1199,7 +1203,7 @@ static void shader_print_register(struct vkd3d_d3d_asm_compiler *compiler, const
{
bool untyped = false;
- switch (compiler->current->handler_idx)
+ switch (compiler->current->opcode)
{
case VKD3DSIH_MOV:
case VKD3DSIH_MOVC:
@@ -1755,7 +1759,7 @@ static void shader_dump_instruction_flags(struct vkd3d_d3d_asm_compiler *compile
{
struct vkd3d_string_buffer *buffer = &compiler->buffer;
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_BREAKP:
case VKD3DSIH_CONTINUEP:
@@ -1853,8 +1857,13 @@ static void shader_dump_instruction_flags(struct vkd3d_d3d_asm_compiler *compile
break;
case VKD3DSIH_TEX:
- if (vkd3d_shader_ver_ge(&compiler->shader_version, 2, 0) && (ins->flags & VKD3DSI_TEXLD_PROJECT))
- vkd3d_string_buffer_printf(buffer, "p");
+ if (vkd3d_shader_ver_ge(&compiler->shader_version, 2, 0))
+ {
+ if (ins->flags & VKD3DSI_TEXLD_PROJECT)
+ vkd3d_string_buffer_printf(buffer, "p");
+ else if (ins->flags & VKD3DSI_TEXLD_BIAS)
+ vkd3d_string_buffer_printf(buffer, "b");
+ }
break;
case VKD3DSIH_WAVE_OP_ADD:
@@ -1937,9 +1946,9 @@ static void shader_dump_instruction(struct vkd3d_d3d_asm_compiler *compiler,
if (ins->coissue)
vkd3d_string_buffer_printf(buffer, "+");
- shader_print_opcode(compiler, ins->handler_idx);
+ shader_print_opcode(compiler, ins->opcode);
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_DCL:
case VKD3DSIH_DCL_UAV_TYPED:
@@ -2430,7 +2439,7 @@ enum vkd3d_result d3d_asm_compile(const struct vsir_program *program,
{
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_ELSE:
case VKD3DSIH_ENDIF:
@@ -2459,7 +2468,7 @@ enum vkd3d_result d3d_asm_compile(const struct vsir_program *program,
shader_dump_instruction(&compiler, ins);
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_ELSE:
case VKD3DSIH_IF:
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
index bfd5b52b436..2290385da76 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
@@ -1060,7 +1060,7 @@ static void shader_sm1_read_comment(struct vkd3d_shader_sm1_parser *sm1)
static void shader_sm1_validate_instruction(struct vkd3d_shader_sm1_parser *sm1, struct vkd3d_shader_instruction *ins)
{
- if ((ins->handler_idx == VKD3DSIH_BREAKP || ins->handler_idx == VKD3DSIH_IF) && ins->flags)
+ if ((ins->opcode == VKD3DSIH_BREAKP || ins->opcode == VKD3DSIH_IF) && ins->flags)
{
vkd3d_shader_parser_warning(&sm1->p, VKD3D_SHADER_WARNING_D3DBC_IGNORED_INSTRUCTION_FLAGS,
"Ignoring unexpected instruction flags %#x.", ins->flags);
@@ -1142,23 +1142,23 @@ static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, str
goto fail;
}
- if (ins->handler_idx == VKD3DSIH_DCL)
+ if (ins->opcode == VKD3DSIH_DCL)
{
shader_sm1_read_semantic(sm1, &p, &ins->declaration.semantic);
}
- else if (ins->handler_idx == VKD3DSIH_DEF)
+ else if (ins->opcode == VKD3DSIH_DEF)
{
shader_sm1_read_dst_param(sm1, &p, dst_param);
shader_sm1_read_immconst(sm1, &p, &src_params[0], VSIR_DIMENSION_VEC4, VKD3D_DATA_FLOAT);
shader_sm1_scan_register(sm1, &dst_param->reg, dst_param->write_mask, true);
}
- else if (ins->handler_idx == VKD3DSIH_DEFB)
+ else if (ins->opcode == VKD3DSIH_DEFB)
{
shader_sm1_read_dst_param(sm1, &p, dst_param);
shader_sm1_read_immconst(sm1, &p, &src_params[0], VSIR_DIMENSION_SCALAR, VKD3D_DATA_UINT);
shader_sm1_scan_register(sm1, &dst_param->reg, dst_param->write_mask, true);
}
- else if (ins->handler_idx == VKD3DSIH_DEFI)
+ else if (ins->opcode == VKD3DSIH_DEFI)
{
shader_sm1_read_dst_param(sm1, &p, dst_param);
shader_sm1_read_immconst(sm1, &p, &src_params[0], VSIR_DIMENSION_VEC4, VKD3D_DATA_INT);
@@ -1195,7 +1195,7 @@ static void shader_sm1_read_instruction(struct vkd3d_shader_sm1_parser *sm1, str
return;
fail:
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
*ptr = sm1->end;
}
@@ -1326,7 +1326,7 @@ int d3dbc_parse(const struct vkd3d_shader_compile_info *compile_info, uint64_t c
ins = &instructions->elements[instructions->count];
shader_sm1_read_instruction(&sm1, ins);
- if (ins->handler_idx == VKD3DSIH_INVALID)
+ if (ins->opcode == VKD3DSIH_INVALID)
{
WARN("Encountered unrecognized or invalid instruction.\n");
vsir_program_cleanup(program);
@@ -1739,11 +1739,11 @@ static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffe
}
else
{
- put_u32(buffer, vkd3d_make_u32(D3DXRS_SAMPLER, var->regs[r].id));
+ put_u32(buffer, vkd3d_make_u32(D3DXRS_SAMPLER, var->regs[r].index));
put_u32(buffer, var->bind_count[r]);
}
put_u32(buffer, 0); /* type */
- put_u32(buffer, 0); /* FIXME: default value */
+ put_u32(buffer, 0); /* default value */
}
}
@@ -1767,6 +1767,62 @@ static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffe
write_sm1_type(buffer, var->data_type, ctab_start);
set_u32(buffer, var_offset + 3 * sizeof(uint32_t), var->data_type->bytecode_offset - ctab_start);
+
+ if (var->default_values)
+ {
+ unsigned int reg_size = var->data_type->reg_size[HLSL_REGSET_NUMERIC];
+ unsigned int comp_count = hlsl_type_component_count(var->data_type);
+ unsigned int default_value_offset;
+ unsigned int k;
+
+ default_value_offset = bytecode_reserve_bytes(buffer, reg_size * sizeof(uint32_t));
+ set_u32(buffer, var_offset + 4 * sizeof(uint32_t), default_value_offset - ctab_start);
+
+ for (k = 0; k < comp_count; ++k)
+ {
+ struct hlsl_type *comp_type = hlsl_type_get_component_type(ctx, var->data_type, k);
+ unsigned int comp_offset;
+ enum hlsl_regset regset;
+
+ comp_offset = hlsl_type_get_component_offset(ctx, var->data_type, k, &regset);
+ if (regset == HLSL_REGSET_NUMERIC)
+ {
+ union
+ {
+ uint32_t u;
+ float f;
+ } uni;
+
+ switch (comp_type->e.numeric.type)
+ {
+ case HLSL_TYPE_DOUBLE:
+ hlsl_fixme(ctx, &var->loc, "Write double default values.");
+ uni.u = 0;
+ break;
+
+ case HLSL_TYPE_INT:
+ uni.f = var->default_values[k].value.i;
+ break;
+
+ case HLSL_TYPE_UINT:
+ case HLSL_TYPE_BOOL:
+ uni.f = var->default_values[k].value.u;
+ break;
+
+ case HLSL_TYPE_HALF:
+ case HLSL_TYPE_FLOAT:
+ uni.u = var->default_values[k].value.u;
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+
+ set_u32(buffer, default_value_offset + comp_offset * sizeof(uint32_t), uni.u);
+ }
+ }
+ }
+
++uniform_count;
}
}
@@ -2210,7 +2266,7 @@ static void write_sm1_sampler_dcls(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b
continue;
}
- reg_id = var->regs[HLSL_REGSET_SAMPLERS].id + i;
+ reg_id = var->regs[HLSL_REGSET_SAMPLERS].index + i;
write_sm1_sampler_dcl(ctx, buffer, reg_id, sampler_dim);
}
}
@@ -2515,7 +2571,7 @@ static void write_sm1_resource_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_
struct sm1_instruction sm1_instr;
sampler_offset = hlsl_offset_from_deref_safe(ctx, &load->resource);
- reg_id = load->resource.var->regs[HLSL_REGSET_SAMPLERS].id + sampler_offset;
+ reg_id = load->resource.var->regs[HLSL_REGSET_SAMPLERS].index + sampler_offset;
sm1_instr = (struct sm1_instruction)
{
@@ -2546,6 +2602,11 @@ static void write_sm1_resource_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_
sm1_instr.opcode |= VKD3DSI_TEXLD_PROJECT << VKD3D_SM1_INSTRUCTION_FLAGS_SHIFT;
break;
+ case HLSL_RESOURCE_SAMPLE_LOD_BIAS:
+ sm1_instr.opcode = D3DSIO_TEX;
+ sm1_instr.opcode |= VKD3DSI_TEXLD_BIAS << VKD3D_SM1_INSTRUCTION_FLAGS_SHIFT;
+ break;
+
default:
hlsl_fixme(ctx, &instr->loc, "Resource load type %u.", load->load_type);
return;
diff --git a/libs/vkd3d/libs/vkd3d-shader/dxil.c b/libs/vkd3d/libs/vkd3d-shader/dxil.c
index 73a8d8687c5..2176debc7d2 100644
--- a/libs/vkd3d/libs/vkd3d-shader/dxil.c
+++ b/libs/vkd3d/libs/vkd3d-shader/dxil.c
@@ -458,6 +458,8 @@ enum dx_intrinsic_opcode
DX_WAVE_ACTIVE_OP = 119,
DX_WAVE_ACTIVE_BIT = 120,
DX_WAVE_PREFIX_OP = 121,
+ DX_QUAD_READ_LANE_AT = 122,
+ DX_QUAD_OP = 123,
DX_LEGACY_F32TOF16 = 130,
DX_LEGACY_F16TOF32 = 131,
DX_WAVE_ALL_BIT_COUNT = 135,
@@ -576,6 +578,13 @@ enum dxil_wave_op_kind
WAVE_OP_MAX = 3,
};
+enum dxil_quad_op_kind
+{
+ QUAD_READ_ACROSS_X = 0,
+ QUAD_READ_ACROSS_Y = 1,
+ QUAD_READ_ACROSS_D = 2,
+};
+
struct sm6_pointer_info
{
const struct sm6_type *type;
@@ -3755,21 +3764,21 @@ static enum vkd3d_result sm6_parser_globals_init(struct sm6_parser *sm6)
for (i = 0; i < sm6->p.program->instructions.count; ++i)
{
ins = &sm6->p.program->instructions.elements[i];
- if (ins->handler_idx == VKD3DSIH_DCL_INDEXABLE_TEMP && ins->declaration.indexable_temp.initialiser)
+ if (ins->opcode == VKD3DSIH_DCL_INDEXABLE_TEMP && ins->declaration.indexable_temp.initialiser)
{
ins->declaration.indexable_temp.initialiser = resolve_forward_initialiser(
(uintptr_t)ins->declaration.indexable_temp.initialiser, sm6);
}
- else if (ins->handler_idx == VKD3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER)
+ else if (ins->opcode == VKD3DSIH_DCL_IMMEDIATE_CONSTANT_BUFFER)
{
ins->declaration.icb = resolve_forward_initialiser((uintptr_t)ins->declaration.icb, sm6);
}
- else if (ins->handler_idx == VKD3DSIH_DCL_TGSM_RAW)
+ else if (ins->opcode == VKD3DSIH_DCL_TGSM_RAW)
{
ins->declaration.tgsm_raw.zero_init = resolve_forward_zero_initialiser(ins->flags, sm6);
ins->flags = 0;
}
- else if (ins->handler_idx == VKD3DSIH_DCL_TGSM_STRUCTURED)
+ else if (ins->opcode == VKD3DSIH_DCL_TGSM_STRUCTURED)
{
ins->declaration.tgsm_structured.zero_init = resolve_forward_zero_initialiser(ins->flags, sm6);
ins->flags = 0;
@@ -4402,7 +4411,7 @@ static void sm6_parser_emit_br(struct sm6_parser *sm6, const struct dxil_record
code_block->terminator.false_block = sm6_function_get_block(function, record->operands[1], sm6);
}
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
}
static bool sm6_parser_emit_reg_composite_construct(struct sm6_parser *sm6, const struct vkd3d_shader_register **operand_regs,
@@ -4619,6 +4628,8 @@ static enum vkd3d_shader_opcode map_dx_binary_op(enum dx_intrinsic_opcode op, co
return VKD3DSIH_IMAX;
case DX_IMIN:
return VKD3DSIH_IMIN;
+ case DX_QUAD_READ_LANE_AT:
+ return VKD3DSIH_QUAD_READ_LANE_AT;
case DX_UMAX:
return VKD3DSIH_UMAX;
case DX_UMIN:
@@ -4962,7 +4973,7 @@ static void sm6_parser_emit_dx_create_handle(struct sm6_parser *sm6, enum dx_int
reg->non_uniform = !!sm6_value_get_constant_uint(operands[3]);
/* NOP is used to flag no instruction emitted. */
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
}
static void sm6_parser_emit_dx_stream(struct sm6_parser *sm6, enum dx_intrinsic_opcode op,
@@ -5370,6 +5381,47 @@ static void sm6_parser_emit_dx_primitive_id(struct sm6_parser *sm6, enum dx_intr
sm6_parser_emit_dx_input_register_mov(sm6, state->ins, VKD3DSPR_PRIMID, VKD3D_DATA_UINT);
}
+static enum vkd3d_shader_opcode dx_map_quad_op(enum dxil_quad_op_kind op)
+{
+ switch (op)
+ {
+ case QUAD_READ_ACROSS_X:
+ return VKD3DSIH_QUAD_READ_ACROSS_X;
+ case QUAD_READ_ACROSS_Y:
+ return VKD3DSIH_QUAD_READ_ACROSS_Y;
+ case QUAD_READ_ACROSS_D:
+ return VKD3DSIH_QUAD_READ_ACROSS_D;
+ default:
+ return VKD3DSIH_INVALID;
+ }
+}
+
+static void sm6_parser_emit_dx_quad_op(struct sm6_parser *sm6, enum dx_intrinsic_opcode op,
+ const struct sm6_value **operands, struct function_emission_state *state)
+{
+ struct vkd3d_shader_instruction *ins = state->ins;
+ struct vkd3d_shader_src_param *src_param;
+ enum vkd3d_shader_opcode opcode;
+ enum dxil_quad_op_kind quad_op;
+
+ quad_op = sm6_value_get_constant_uint(operands[1]);
+ if ((opcode = dx_map_quad_op(quad_op)) == VKD3DSIH_INVALID)
+ {
+ FIXME("Unhandled quad op kind %u.\n", quad_op);
+ vkd3d_shader_parser_error(&sm6->p, VKD3D_SHADER_ERROR_DXIL_UNHANDLED_INTRINSIC,
+ "Quad op kind %u is unhandled.", quad_op);
+ return;
+ }
+
+ vsir_instruction_init(ins, &sm6->p.location, opcode);
+
+ if (!(src_param = instruction_src_params_alloc(ins, 1, sm6)))
+ return;
+ src_param_init_from_value(src_param, operands[0]);
+
+ instruction_dst_param_init_ssa_scalar(ins, sm6);
+}
+
static void sm6_parser_emit_dx_raw_buffer_load(struct sm6_parser *sm6, enum dx_intrinsic_opcode op,
const struct sm6_value **operands, struct function_emission_state *state)
{
@@ -6229,6 +6281,8 @@ static const struct sm6_dx_opcode_info sm6_dx_op_table[] =
[DX_MAKE_DOUBLE ] = {"d", "ii", sm6_parser_emit_dx_make_double},
[DX_OUTPUT_CONTROL_POINT_ID ] = {"i", "", sm6_parser_emit_dx_output_control_point_id},
[DX_PRIMITIVE_ID ] = {"i", "", sm6_parser_emit_dx_primitive_id},
+ [DX_QUAD_OP ] = {"n", "Rc", sm6_parser_emit_dx_quad_op},
+ [DX_QUAD_READ_LANE_AT ] = {"n", "Ri", sm6_parser_emit_dx_binary},
[DX_RAW_BUFFER_LOAD ] = {"o", "Hii8i", sm6_parser_emit_dx_raw_buffer_load},
[DX_RAW_BUFFER_STORE ] = {"v", "Hiioooocc", sm6_parser_emit_dx_raw_buffer_store},
[DX_ROUND_NE ] = {"g", "R", sm6_parser_emit_dx_unary},
@@ -6381,7 +6435,7 @@ static void sm6_parser_emit_unhandled(struct sm6_parser *sm6, struct vkd3d_shade
{
const struct sm6_type *type;
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
if (!dst->type)
return;
@@ -6628,7 +6682,7 @@ static void sm6_parser_emit_cast(struct sm6_parser *sm6, const struct dxil_recor
{
*dst = *value;
dst->type = type;
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
return;
}
@@ -6739,7 +6793,7 @@ static void sm6_parser_emit_cmp2(struct sm6_parser *sm6, const struct dxil_recor
* do not otherwise occur, so deleting these avoids the need for backend support. */
if (sm6_type_is_bool(type_a) && code == ICMP_NE && sm6_value_is_constant_zero(b))
{
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
*dst = *a;
return;
}
@@ -7039,7 +7093,7 @@ static void sm6_parser_emit_gep(struct sm6_parser *sm6, const struct dxil_record
reg->idx_count = 2;
dst->structure_stride = src->structure_stride;
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
}
static void sm6_parser_emit_load(struct sm6_parser *sm6, const struct dxil_record *record,
@@ -7189,7 +7243,7 @@ static void sm6_parser_emit_phi(struct sm6_parser *sm6, const struct dxil_record
incoming[j].block = sm6_function_get_block(function, record->operands[i + 1], sm6);
}
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
qsort(incoming, phi->incoming_count, sizeof(*incoming), phi_incoming_compare);
@@ -7224,7 +7278,7 @@ static void sm6_parser_emit_ret(struct sm6_parser *sm6, const struct dxil_record
code_block->terminator.type = TERMINATOR_RET;
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
}
static void sm6_parser_emit_store(struct sm6_parser *sm6, const struct dxil_record *record,
@@ -7384,7 +7438,7 @@ static void sm6_parser_emit_switch(struct sm6_parser *sm6, const struct dxil_rec
terminator->cases[i / 2u].value = sm6_value_get_constant_uint64(src);
}
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
}
static void sm6_parser_emit_vselect(struct sm6_parser *sm6, const struct dxil_record *record,
@@ -7843,7 +7897,7 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const
}
ins = &code_block->instructions[code_block->instruction_count];
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
dst = sm6_parser_get_current_value(sm6);
fwd_type = dst->type;
@@ -7922,7 +7976,6 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const
if (sm6->p.failed)
return VKD3D_ERROR;
- assert(ins->handler_idx != VKD3DSIH_INVALID);
if (record->attachment)
metadata_attachment_record_apply(record->attachment, record->code, ins, dst, sm6);
@@ -7933,9 +7986,7 @@ static enum vkd3d_result sm6_parser_function_init(struct sm6_parser *sm6, const
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);
+ code_block->instruction_count += ins->opcode != VKD3DSIH_NOP;
if (dst->type && fwd_type && dst->type != fwd_type)
{
@@ -8735,7 +8786,7 @@ static struct vkd3d_shader_resource *sm6_parser_resources_load_common_info(struc
if (!m)
{
- ins->handler_idx = is_uav ? VKD3DSIH_DCL_UAV_RAW : VKD3DSIH_DCL_RESOURCE_RAW;
+ ins->opcode = is_uav ? VKD3DSIH_DCL_UAV_RAW : VKD3DSIH_DCL_RESOURCE_RAW;
ins->declaration.raw_resource.resource.reg.write_mask = 0;
return &ins->declaration.raw_resource.resource;
}
@@ -8760,7 +8811,7 @@ static struct vkd3d_shader_resource *sm6_parser_resources_load_common_info(struc
"A typed resource has no data type.");
}
- ins->handler_idx = is_uav ? VKD3DSIH_DCL_UAV_TYPED : VKD3DSIH_DCL;
+ ins->opcode = is_uav ? VKD3DSIH_DCL_UAV_TYPED : VKD3DSIH_DCL;
for (i = 0; i < VKD3D_VEC4_SIZE; ++i)
ins->declaration.semantic.resource_data_type[i] = resource_values.data_type;
ins->declaration.semantic.resource_type = resource_type;
@@ -8770,14 +8821,14 @@ static struct vkd3d_shader_resource *sm6_parser_resources_load_common_info(struc
}
else if (kind == RESOURCE_KIND_RAWBUFFER)
{
- ins->handler_idx = is_uav ? VKD3DSIH_DCL_UAV_RAW : VKD3DSIH_DCL_RESOURCE_RAW;
+ ins->opcode = is_uav ? VKD3DSIH_DCL_UAV_RAW : VKD3DSIH_DCL_RESOURCE_RAW;
ins->declaration.raw_resource.resource.reg.write_mask = 0;
return &ins->declaration.raw_resource.resource;
}
else if (kind == RESOURCE_KIND_STRUCTUREDBUFFER)
{
- ins->handler_idx = is_uav ? VKD3DSIH_DCL_UAV_STRUCTURED : VKD3DSIH_DCL_RESOURCE_STRUCTURED;
+ ins->opcode = is_uav ? VKD3DSIH_DCL_UAV_STRUCTURED : VKD3DSIH_DCL_RESOURCE_STRUCTURED;
ins->declaration.structured_resource.byte_stride = resource_values.byte_stride;
ins->declaration.structured_resource.resource.reg.write_mask = 0;
@@ -8858,7 +8909,7 @@ static enum vkd3d_result sm6_parser_resources_load_srv(struct sm6_parser *sm6,
d->kind = kind;
d->reg_type = VKD3DSPR_RESOURCE;
d->reg_data_type = (ins->resource_type == VKD3D_SHADER_RESOURCE_BUFFER) ? VKD3D_DATA_UINT : VKD3D_DATA_RESOURCE;
- d->resource_data_type = (ins->handler_idx == VKD3DSIH_DCL)
+ d->resource_data_type = (ins->opcode == VKD3DSIH_DCL)
? ins->declaration.semantic.resource_data_type[0] : VKD3D_DATA_UNUSED;
init_resource_declaration(resource, VKD3DSPR_RESOURCE, d->reg_data_type, d->id, &d->range);
@@ -8932,7 +8983,7 @@ static enum vkd3d_result sm6_parser_resources_load_uav(struct sm6_parser *sm6,
d->kind = values[0];
d->reg_type = VKD3DSPR_UAV;
d->reg_data_type = (ins->resource_type == VKD3D_SHADER_RESOURCE_BUFFER) ? VKD3D_DATA_UINT : VKD3D_DATA_UAV;
- d->resource_data_type = (ins->handler_idx == VKD3DSIH_DCL_UAV_TYPED)
+ d->resource_data_type = (ins->opcode == VKD3DSIH_DCL_UAV_TYPED)
? ins->declaration.semantic.resource_data_type[0] : VKD3D_DATA_UNUSED;
init_resource_declaration(resource, VKD3DSPR_UAV, d->reg_data_type, d->id, &d->range);
diff --git a/libs/vkd3d/libs/vkd3d-shader/fx.c b/libs/vkd3d/libs/vkd3d-shader/fx.c
index 57b4ac24212..3ba0a9ba994 100644
--- a/libs/vkd3d/libs/vkd3d-shader/fx.c
+++ b/libs/vkd3d/libs/vkd3d-shader/fx.c
@@ -494,7 +494,7 @@ static uint32_t write_fx_4_type(const struct hlsl_type *type, struct fx_write_co
uint32_t semantic_offset, field_type_offset;
name_offset = write_string(field->name, fx);
- semantic_offset = write_string(field->semantic.name, fx);
+ semantic_offset = write_string(field->semantic.raw_name, fx);
field_type_offset = write_type(field->type, fx);
put_u32_unaligned(buffer, name_offset);
@@ -683,7 +683,7 @@ static uint32_t write_fx_2_parameter(const struct hlsl_type *type, const char *n
}
name_offset = write_string(name, fx);
- semantic_offset = write_string(semantic->name, fx);
+ semantic_offset = write_string(semantic->raw_name, fx);
offset = put_u32(buffer, hlsl_sm1_base_type(type));
put_u32(buffer, hlsl_sm1_class(type));
@@ -794,6 +794,9 @@ static uint32_t write_fx_2_initial_value(const struct hlsl_ir_var *var, struct f
case HLSL_CLASS_MATRIX:
case HLSL_CLASS_STRUCT:
/* FIXME: write actual initial value */
+ if (var->default_values)
+ hlsl_fixme(fx->ctx, &var->loc, "Write default values.\n");
+
offset = put_u32(buffer, 0);
for (uint32_t i = 1; i < size / sizeof(uint32_t); ++i)
@@ -986,19 +989,18 @@ static void write_fx_4_numeric_variable(struct hlsl_ir_var *var, bool shared, st
};
struct hlsl_ctx *ctx = fx->ctx;
- /* Explicit bind point. */
- if (var->reg_reservation.reg_type)
+ if (var->has_explicit_bind_point)
flags |= HAS_EXPLICIT_BIND_POINT;
type_offset = write_type(var->data_type, fx);
name_offset = write_string(var->name, fx);
- semantic_offset = write_string(var->semantic.name, fx);
+ semantic_offset = write_string(var->semantic.raw_name, fx);
put_u32(buffer, name_offset);
put_u32(buffer, type_offset);
semantic_offset = put_u32(buffer, semantic_offset); /* Semantic */
- put_u32(buffer, var->buffer_offset); /* Offset in the constant buffer */
+ put_u32(buffer, var->buffer_offset * 4); /* Offset in the constant buffer, in bytes. */
value_offset = put_u32(buffer, 0); /* Default value offset */
put_u32(buffer, flags); /* Flags */
@@ -1010,6 +1012,8 @@ static void write_fx_4_numeric_variable(struct hlsl_ir_var *var, bool shared, st
{
/* FIXME: write default value */
set_u32(buffer, value_offset, 0);
+ if (var->default_values)
+ hlsl_fixme(fx->ctx, &var->loc, "Write default values.\n");
put_u32(buffer, 0); /* Annotations count */
if (has_annotations(var))
@@ -1344,7 +1348,7 @@ static void write_fx_4_object_variable(struct hlsl_ir_var *var, struct fx_write_
type_offset = write_type(var->data_type, fx);
name_offset = write_string(var->name, fx);
- semantic_offset = write_string(var->semantic.name, fx);
+ semantic_offset = write_string(var->semantic.raw_name, fx);
put_u32(buffer, name_offset);
put_u32(buffer, type_offset);
@@ -1464,6 +1468,9 @@ static void write_buffers(struct fx_write_context *fx, bool shared)
{
struct hlsl_buffer *buffer;
+ if (shared && !fx->child_effect)
+ return;
+
LIST_FOR_EACH_ENTRY(buffer, &fx->ctx->buffers, struct hlsl_buffer, entry)
{
if (!buffer->size && !fx->include_empty_buffers)
diff --git a/libs/vkd3d/libs/vkd3d-shader/glsl.c b/libs/vkd3d/libs/vkd3d-shader/glsl.c
index 3e482a5fc70..8725724a239 100644
--- a/libs/vkd3d/libs/vkd3d-shader/glsl.c
+++ b/libs/vkd3d/libs/vkd3d-shader/glsl.c
@@ -48,9 +48,9 @@ static void shader_glsl_print_indent(struct vkd3d_string_buffer *buffer, unsigne
static void shader_glsl_unhandled(struct vkd3d_glsl_generator *gen, const struct vkd3d_shader_instruction *ins)
{
shader_glsl_print_indent(&gen->buffer, gen->indent);
- vkd3d_string_buffer_printf(&gen->buffer, "/* <unhandled instruction %#x> */\n", ins->handler_idx);
+ vkd3d_string_buffer_printf(&gen->buffer, "/* <unhandled instruction %#x> */\n", ins->opcode);
vkd3d_glsl_compiler_error(gen, VKD3D_SHADER_ERROR_GLSL_INTERNAL,
- "Internal compiler error: Unhandled instruction %#x.", ins->handler_idx);
+ "Internal compiler error: Unhandled instruction %#x.", ins->opcode);
}
static void shader_glsl_ret(struct vkd3d_glsl_generator *generator,
@@ -74,7 +74,7 @@ static void vkd3d_glsl_handle_instruction(struct vkd3d_glsl_generator *generator
{
generator->location = instruction->location;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_DCL_INPUT:
case VKD3DSIH_DCL_OUTPUT:
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.c b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
index 99214fba6de..c69af4e94bb 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
@@ -167,6 +167,8 @@ void hlsl_free_var(struct hlsl_ir_var *decl)
for (k = 0; k <= HLSL_REGSET_LAST_OBJECT; ++k)
vkd3d_free((void *)decl->objects_usage[k]);
+ vkd3d_free(decl->default_values);
+
for (i = 0; i < decl->state_block_count; ++i)
hlsl_free_state_block(decl->state_blocks[i]);
vkd3d_free(decl->state_blocks);
@@ -556,12 +558,14 @@ unsigned int hlsl_type_get_component_offset(struct hlsl_ctx *ctx, struct hlsl_ty
switch (type->class)
{
- case HLSL_CLASS_SCALAR:
case HLSL_CLASS_VECTOR:
- case HLSL_CLASS_MATRIX:
offset[HLSL_REGSET_NUMERIC] += idx;
break;
+ case HLSL_CLASS_MATRIX:
+ offset[HLSL_REGSET_NUMERIC] += 4 * idx;
+ break;
+
case HLSL_CLASS_STRUCT:
for (r = 0; r <= HLSL_REGSET_LAST; ++r)
offset[r] += type->e.record.fields[idx].reg_offset[r];
@@ -592,6 +596,7 @@ unsigned int hlsl_type_get_component_offset(struct hlsl_ctx *ctx, struct hlsl_ty
case HLSL_CLASS_PASS:
case HLSL_CLASS_TECHNIQUE:
case HLSL_CLASS_VOID:
+ case HLSL_CLASS_SCALAR:
vkd3d_unreachable();
}
type = next_type;
@@ -1247,6 +1252,7 @@ struct hlsl_ir_var *hlsl_new_synthetic_var_named(struct hlsl_ctx *ctx, const cha
list_add_tail(&ctx->dummy_scope->vars, &var->scope_entry);
else
list_add_tail(&ctx->globals->vars, &var->scope_entry);
+ var->is_synthetic = true;
}
return var;
}
@@ -3086,6 +3092,33 @@ void hlsl_dump_function(struct hlsl_ctx *ctx, const struct hlsl_ir_function_decl
vkd3d_string_buffer_cleanup(&buffer);
}
+void hlsl_dump_var_default_values(const struct hlsl_ir_var *var)
+{
+ unsigned int k, component_count = hlsl_type_component_count(var->data_type);
+ struct vkd3d_string_buffer buffer;
+
+ vkd3d_string_buffer_init(&buffer);
+ if (!var->default_values)
+ {
+ vkd3d_string_buffer_printf(&buffer, "var \"%s\" has no default values.\n", var->name);
+ vkd3d_string_buffer_trace(&buffer);
+ vkd3d_string_buffer_cleanup(&buffer);
+ return;
+ }
+
+ vkd3d_string_buffer_printf(&buffer, "var \"%s\" default values:", var->name);
+ for (k = 0; k < component_count; ++k)
+ {
+ if (k % 4 == 0)
+ vkd3d_string_buffer_printf(&buffer, "\n ");
+ vkd3d_string_buffer_printf(&buffer, " 0x%08x", var->default_values[k].value.u);
+ }
+ vkd3d_string_buffer_printf(&buffer, "\n");
+
+ vkd3d_string_buffer_trace(&buffer);
+ vkd3d_string_buffer_cleanup(&buffer);
+}
+
void hlsl_replace_node(struct hlsl_ir_node *old, struct hlsl_ir_node *new)
{
struct hlsl_src *src, *next;
@@ -3319,9 +3352,25 @@ void hlsl_free_attribute(struct hlsl_attribute *attr)
void hlsl_cleanup_semantic(struct hlsl_semantic *semantic)
{
vkd3d_free((void *)semantic->name);
+ vkd3d_free((void *)semantic->raw_name);
memset(semantic, 0, sizeof(*semantic));
}
+bool hlsl_clone_semantic(struct hlsl_ctx *ctx, struct hlsl_semantic *dst, const struct hlsl_semantic *src)
+{
+ *dst = *src;
+ dst->name = dst->raw_name = NULL;
+ if (src->name && !(dst->name = hlsl_strdup(ctx, src->name)))
+ return false;
+ if (src->raw_name && !(dst->raw_name = hlsl_strdup(ctx, src->raw_name)))
+ {
+ hlsl_cleanup_semantic(dst);
+ return false;
+ }
+
+ return true;
+}
+
static void free_function_decl(struct hlsl_ir_function_decl *decl)
{
unsigned int i;
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.h b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
index 27814f3a56f..179cc219e68 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.h
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
@@ -222,6 +222,8 @@ struct hlsl_semantic
const char *name;
uint32_t index;
+ /* Name exactly as it appears in the sources. */
+ const char *raw_name;
/* If the variable or field that stores this hlsl_semantic has already reported that it is missing. */
bool reported_missing;
/* In case the variable or field that stores this semantic has already reported to use a
@@ -259,8 +261,20 @@ struct hlsl_struct_field
* struct. */
struct hlsl_reg
{
- /* Index of the first register allocated. */
+ /* Register number of the first register allocated. */
uint32_t id;
+ /* For descriptors (buffer, texture, sampler, UAV) this is the base binding
+ * index of the descriptor.
+ * For 5.1 and above descriptors have space and may be arrayed, in which
+ * case the array shares a single register ID but has a range of register
+ * indices, and "id" and "index" are as a rule not equal.
+ * For versions below 5.1, the register number for descriptors is the same
+ * as its external binding index, so only "index" is used, and "id" is
+ * ignored.
+ * For numeric registers "index" is not used. */
+ uint32_t index;
+ /* Register space of a descriptor. Not used for numeric registers. */
+ uint32_t space;
/* Number of registers to be allocated.
* Unlike the variable's type's regsize, it is not expressed in register components, but rather
* in whole registers, and may depend on which components are used within the shader. */
@@ -396,6 +410,14 @@ struct hlsl_reg_reservation
unsigned int offset_index;
};
+union hlsl_constant_value_component
+{
+ uint32_t u;
+ int32_t i;
+ float f;
+ double d;
+};
+
struct hlsl_ir_var
{
struct hlsl_type *data_type;
@@ -418,6 +440,15 @@ struct hlsl_ir_var
/* Scope that contains annotations for this variable. */
struct hlsl_scope *annotations;
+ /* Array of default values the variable was initialized with, one for each component.
+ * Only for variables that need it, such as uniforms and variables inside constant buffers.
+ * This pointer is NULL for others. */
+ struct hlsl_default_value
+ {
+ /* Default value, in case the component is a numeric value. */
+ union hlsl_constant_value_component value;
+ } *default_values;
+
/* A dynamic array containing the state block on the variable's declaration, if any.
* An array variable may contain multiple state blocks.
* A technique pass will always contain one.
@@ -460,6 +491,8 @@ struct hlsl_ir_var
uint32_t is_uniform : 1;
uint32_t is_param : 1;
uint32_t is_separated_resource : 1;
+ uint32_t is_synthetic : 1;
+ uint32_t has_explicit_bind_point : 1;
};
/* This struct is used to represent assignments in state block entries:
@@ -775,13 +808,7 @@ struct hlsl_ir_constant
struct hlsl_ir_node node;
struct hlsl_constant_value
{
- union hlsl_constant_value_component
- {
- uint32_t u;
- int32_t i;
- float f;
- double d;
- } u[4];
+ union hlsl_constant_value_component u[4];
} value;
/* Constant register of type 'c' where the constant value is stored for SM1. */
struct hlsl_reg reg;
@@ -1249,6 +1276,7 @@ void hlsl_block_cleanup(struct hlsl_block *block);
bool hlsl_clone_block(struct hlsl_ctx *ctx, struct hlsl_block *dst_block, const struct hlsl_block *src_block);
void hlsl_dump_function(struct hlsl_ctx *ctx, const struct hlsl_ir_function_decl *func);
+void hlsl_dump_var_default_values(const struct hlsl_ir_var *var);
void hlsl_run_const_passes(struct hlsl_ctx *ctx, struct hlsl_block *body);
int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func,
@@ -1259,7 +1287,9 @@ bool hlsl_init_deref_from_index_chain(struct hlsl_ctx *ctx, struct hlsl_deref *d
bool hlsl_copy_deref(struct hlsl_ctx *ctx, struct hlsl_deref *deref, const struct hlsl_deref *other);
void hlsl_cleanup_deref(struct hlsl_deref *deref);
+
void hlsl_cleanup_semantic(struct hlsl_semantic *semantic);
+bool hlsl_clone_semantic(struct hlsl_ctx *ctx, struct hlsl_semantic *dst, const struct hlsl_semantic *src);
void hlsl_cleanup_ir_switch_cases(struct list *cases);
void hlsl_free_ir_switch_case(struct hlsl_ir_switch_case *c);
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.l b/libs/vkd3d/libs/vkd3d-shader/hlsl.l
index a5923d8bf8e..91418775e1b 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.l
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.l
@@ -49,7 +49,7 @@ static void update_location(struct hlsl_ctx *ctx, YYLTYPE *loc);
RESERVED1 auto|catch|char|class|const_cast|delete|dynamic_cast|enum
RESERVED2 explicit|friend|goto|long|mutable|new|operator|private|protected|public
RESERVED3 reinterpret_cast|short|signed|sizeof|static_cast|template|this|throw|try
-RESERVED4 typename|union|unsigned|using|virtual
+RESERVED4 typename|union|using|virtual
WS [ \t]
NEWLINE (\n)|(\r\n)
@@ -164,6 +164,7 @@ textureCUBE {return KW_TEXTURECUBE; }
TextureCubeArray {return KW_TEXTURECUBEARRAY; }
true {return KW_TRUE; }
typedef {return KW_TYPEDEF; }
+unsigned {return KW_UNSIGNED; }
uniform {return KW_UNIFORM; }
vector {return KW_VECTOR; }
VertexShader {return KW_VERTEXSHADER; }
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.y b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
index 9c1bdef926d..a2c89b15e4a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
@@ -1013,6 +1013,10 @@ static bool gen_struct_fields(struct hlsl_ctx *ctx, struct parse_fields *fields,
field->type = hlsl_new_array_type(ctx, field->type, v->arrays.sizes[k]);
}
}
+
+ if (hlsl_version_ge(ctx, 5, 1) && field->type->class == HLSL_CLASS_ARRAY && hlsl_type_is_resource(field->type))
+ hlsl_fixme(ctx, &v->loc, "Shader model 5.1+ resource array.");
+
vkd3d_free(v->arrays.sizes);
field->loc = v->loc;
field->name = v->name;
@@ -1286,13 +1290,13 @@ static struct hlsl_block *make_block(struct hlsl_ctx *ctx, struct hlsl_ir_node *
return block;
}
-static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, struct hlsl_block *block,
- const struct vkd3d_shader_location *loc)
+static union hlsl_constant_value_component evaluate_static_expression(struct hlsl_ctx *ctx,
+ struct hlsl_block *block, struct hlsl_type *dst_type, const struct vkd3d_shader_location *loc)
{
+ union hlsl_constant_value_component ret = {0};
struct hlsl_ir_constant *constant;
struct hlsl_ir_node *node;
struct hlsl_block expr;
- unsigned int ret = 0;
struct hlsl_src src;
LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry)
@@ -1305,29 +1309,32 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str
case HLSL_IR_LOAD:
case HLSL_IR_INDEX:
continue;
+ case HLSL_IR_STORE:
+ if (hlsl_ir_store(node)->lhs.var->is_synthetic)
+ break;
+ /* fall-through */
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:
case HLSL_IR_SWITCH:
case HLSL_IR_STATEBLOCK_CONSTANT:
hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
"Expected literal expression.");
+ break;
}
}
if (!hlsl_clone_block(ctx, &expr, &ctx->static_initializers))
- return 0;
+ return ret;
hlsl_block_add_block(&expr, block);
- if (!add_implicit_conversion(ctx, &expr, node_from_block(&expr),
- hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), loc))
+ if (!add_implicit_conversion(ctx, &expr, node_from_block(&expr), dst_type, loc))
{
hlsl_block_cleanup(&expr);
- return 0;
+ return ret;
}
/* Wrap the node into a src to allow the reference to survive the multiple const passes. */
@@ -1339,7 +1346,7 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str
if (node->type == HLSL_IR_CONSTANT)
{
constant = hlsl_ir_constant(node);
- ret = constant->value.u[0].u;
+ ret = constant->value.u[0];
}
else
{
@@ -1352,6 +1359,15 @@ static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, str
return ret;
}
+static unsigned int evaluate_static_expression_as_uint(struct hlsl_ctx *ctx, struct hlsl_block *block,
+ const struct vkd3d_shader_location *loc)
+{
+ union hlsl_constant_value_component res;
+
+ res = evaluate_static_expression(ctx, block, hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), loc);
+ return res.u;
+}
+
static bool expr_compatible_data_types(struct hlsl_type *t1, struct hlsl_type *t2)
{
/* Scalar vars can be converted to pretty much everything */
@@ -2067,6 +2083,50 @@ static bool add_increment(struct hlsl_ctx *ctx, struct hlsl_block *block, bool d
return true;
}
+/* For some reason, for matrices, values from default value initializers end up in different
+ * components than from regular initializers. Default value initializers fill the matrix in
+ * vertical reading order (left-to-right top-to-bottom) instead of regular reading order
+ * (top-to-bottom left-to-right), so they have to be adjusted. */
+static unsigned int get_component_index_from_default_initializer_index(struct hlsl_ctx *ctx,
+ struct hlsl_type *type, unsigned int index)
+{
+ unsigned int element_comp_count, element, x, y, i;
+ unsigned int base = 0;
+
+ if (ctx->profile->major_version < 4)
+ return index;
+
+ switch (type->class)
+ {
+ case HLSL_CLASS_MATRIX:
+ x = index / type->dimy;
+ y = index % type->dimy;
+ return y * type->dimx + x;
+
+ case HLSL_CLASS_ARRAY:
+ element_comp_count = hlsl_type_component_count(type->e.array.type);
+ element = index / element_comp_count;
+ base = element * element_comp_count;
+ return base + get_component_index_from_default_initializer_index(ctx, type->e.array.type, index - base);
+
+ case HLSL_CLASS_STRUCT:
+ for (i = 0; i < type->e.record.field_count; ++i)
+ {
+ struct hlsl_type *field_type = type->e.record.fields[i].type;
+
+ element_comp_count = hlsl_type_component_count(field_type);
+ if (index - base < element_comp_count)
+ return base + get_component_index_from_default_initializer_index(ctx, field_type, index - base);
+ base += element_comp_count;
+ }
+ break;
+
+ default:
+ return index;
+ }
+ vkd3d_unreachable();
+}
+
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)
{
@@ -2087,12 +2147,29 @@ static void initialize_var_components(struct hlsl_ctx *ctx, struct hlsl_block *i
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)))
- return;
+ if (dst->default_values)
+ {
+ struct hlsl_default_value default_value = {0};
+ unsigned int dst_index;
- if (!hlsl_new_store_component(ctx, &block, &dst_deref, *store_index, conv))
- return;
- hlsl_block_add_block(instrs, &block);
+ if (!hlsl_clone_block(ctx, &block, instrs))
+ return;
+ default_value.value = evaluate_static_expression(ctx, &block, dst_comp_type, &src->loc);
+
+ dst_index = get_component_index_from_default_initializer_index(ctx, dst->data_type, *store_index);
+ dst->default_values[dst_index] = default_value;
+
+ hlsl_block_cleanup(&block);
+ }
+ else
+ {
+ if (!(conv = add_implicit_conversion(ctx, instrs, load, dst_comp_type, &src->loc)))
+ return;
+
+ if (!hlsl_new_store_component(ctx, &block, &dst_deref, *store_index, conv))
+ return;
+ hlsl_block_add_block(instrs, &block);
+ }
++*store_index;
}
@@ -2246,17 +2323,22 @@ static void declare_var(struct hlsl_ctx *ctx, struct parse_variable_def *v)
}
}
+ if (hlsl_version_ge(ctx, 5, 1) && type->class == HLSL_CLASS_ARRAY && hlsl_type_is_resource(type))
+ {
+ /* SM 5.1/6.x descriptor arrays act differently from previous versions.
+ * Not only are they treated as a single object in reflection, but they
+ * act as a single component for the purposes of assignment and
+ * initialization. */
+ hlsl_fixme(ctx, &v->loc, "Shader model 5.1+ resource array.");
+ }
+
if (!(var_name = vkd3d_strdup(v->name)))
return;
- new_semantic = v->semantic;
- if (v->semantic.name)
+ if (!hlsl_clone_semantic(ctx, &new_semantic, &v->semantic))
{
- if (!(new_semantic.name = vkd3d_strdup(v->semantic.name)))
- {
- vkd3d_free(var_name);
- return;
- }
+ vkd3d_free(var_name);
+ return;
}
if (!(var = hlsl_new_var(ctx, var_name, type, &v->loc, &new_semantic, modifiers, &v->reg_reservation)))
@@ -2348,6 +2430,7 @@ static struct hlsl_block *initialize_vars(struct hlsl_ctx *ctx, struct list *var
{
struct parse_variable_def *v, *v_next;
struct hlsl_block *initializers;
+ unsigned int component_count;
struct hlsl_ir_var *var;
struct hlsl_type *type;
@@ -2371,6 +2454,7 @@ static struct hlsl_block *initialize_vars(struct hlsl_ctx *ctx, struct list *var
}
type = var->data_type;
+ component_count = hlsl_type_component_count(type);
var->state_blocks = v->state_blocks;
var->state_block_count = v->state_block_count;
@@ -2379,51 +2463,70 @@ static struct hlsl_block *initialize_vars(struct hlsl_ctx *ctx, struct list *var
v->state_block_capacity = 0;
v->state_blocks = NULL;
- if (var->state_blocks && hlsl_type_component_count(type) != var->state_block_count)
+ if (var->state_blocks && component_count != var->state_block_count)
{
hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_WRONG_PARAMETER_COUNT,
- "Expected %u state blocks, but got %u.",
- hlsl_type_component_count(type), var->state_block_count);
+ "Expected %u state blocks, but got %u.", component_count, var->state_block_count);
free_parse_variable_def(v);
continue;
}
if (v->initializer.args_count)
{
- if (v->initializer.braces)
- {
- unsigned int size = initializer_size(&v->initializer);
- unsigned int store_index = 0;
- unsigned int k;
+ unsigned int store_index = 0;
+ bool is_default_values_initializer;
+ unsigned int size, k;
+
+ is_default_values_initializer = (ctx->cur_buffer != ctx->globals_buffer)
+ || (var->storage_modifiers & HLSL_STORAGE_UNIFORM);
- if (hlsl_type_component_count(type) != size)
+ if (is_default_values_initializer)
+ {
+ assert(!var->default_values);
+ if (!(var->default_values = hlsl_calloc(ctx, component_count, sizeof(*var->default_values))))
{
- 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_variable_def(v);
continue;
}
+ }
- for (k = 0; k < v->initializer.args_count; ++k)
+ if (!v->initializer.braces)
+ {
+ if (!(add_implicit_conversion(ctx, v->initializer.instrs, v->initializer.args[0], type, &v->loc)))
{
- initialize_var_components(ctx, v->initializer.instrs, var,
- &store_index, v->initializer.args[k]);
+ free_parse_variable_def(v);
+ continue;
}
+
+ v->initializer.args[0] = node_from_block(v->initializer.instrs);
}
- else
+
+ size = initializer_size(&v->initializer);
+ if (component_count != size)
{
- struct hlsl_ir_load *load = hlsl_new_var_load(ctx, var, &var->loc);
+ hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_WRONG_PARAMETER_COUNT,
+ "Expected %u components in initializer, but got %u.", component_count, size);
+ free_parse_variable_def(v);
+ continue;
+ }
- assert(v->initializer.args_count == 1);
- hlsl_block_add_instr(v->initializer.instrs, &load->node);
- add_assignment(ctx, v->initializer.instrs, &load->node, ASSIGN_OP_ASSIGN, v->initializer.args[0]);
+ for (k = 0; k < v->initializer.args_count; ++k)
+ {
+ initialize_var_components(ctx, v->initializer.instrs, var, &store_index, v->initializer.args[k]);
}
- if (var->storage_modifiers & HLSL_STORAGE_STATIC)
+ if (is_default_values_initializer)
+ {
+ hlsl_dump_var_default_values(var);
+ }
+ else if (var->storage_modifiers & HLSL_STORAGE_STATIC)
+ {
hlsl_block_add_block(&ctx->static_initializers, v->initializer.instrs);
+ }
else
+ {
hlsl_block_add_block(initializers, v->initializer.instrs);
+ }
}
else if (var->storage_modifiers & HLSL_STORAGE_STATIC)
{
@@ -4061,11 +4164,15 @@ static bool intrinsic_tex(struct hlsl_ctx *ctx, const struct parse_initializer *
hlsl_release_string_buffer(ctx, string);
}
- if (!strcmp(name, "tex2Dlod"))
+ if (!strcmp(name, "tex2Dbias")
+ || !strcmp(name, "tex2Dlod"))
{
struct hlsl_ir_node *lod, *c;
- load_params.type = HLSL_RESOURCE_SAMPLE_LOD;
+ if (!strcmp(name, "tex2Dlod"))
+ load_params.type = HLSL_RESOURCE_SAMPLE_LOD;
+ else
+ load_params.type = HLSL_RESOURCE_SAMPLE_LOD_BIAS;
if (!(c = hlsl_new_swizzle(ctx, HLSL_SWIZZLE(X, Y, Z, W), hlsl_sampler_dim_count(dim), params->args[1], loc)))
return false;
@@ -4187,6 +4294,12 @@ static bool intrinsic_tex2D(struct hlsl_ctx *ctx,
return intrinsic_tex(ctx, params, loc, "tex2D", HLSL_SAMPLER_DIM_2D);
}
+static bool intrinsic_tex2Dbias(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ return intrinsic_tex(ctx, params, loc, "tex2Dbias", HLSL_SAMPLER_DIM_2D);
+}
+
static bool intrinsic_tex2Dlod(struct hlsl_ctx *ctx,
const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
{
@@ -4407,6 +4520,7 @@ intrinsic_functions[] =
{"tanh", 1, true, intrinsic_tanh},
{"tex1D", -1, false, intrinsic_tex1D},
{"tex2D", -1, false, intrinsic_tex2D},
+ {"tex2Dbias", 2, false, intrinsic_tex2Dbias},
{"tex2Dlod", 2, false, intrinsic_tex2Dlod},
{"tex2Dproj", 2, false, intrinsic_tex2Dproj},
{"tex3D", -1, false, intrinsic_tex3D},
@@ -5566,6 +5680,7 @@ static bool state_block_add_entry(struct hlsl_state_block *state_block, struct h
%token KW_TEXTURECUBEARRAY
%token KW_TRUE
%token KW_TYPEDEF
+%token KW_UNSIGNED
%token KW_UNIFORM
%token KW_VECTOR
%token KW_VERTEXSHADER
@@ -5717,8 +5832,7 @@ hlsl_prog:
| hlsl_prog buffer_declaration buffer_body
| hlsl_prog declaration_statement
{
- if (!list_empty(&$2->instrs))
- hlsl_fixme(ctx, &@2, "Uniform initializer.");
+ hlsl_block_add_block(&ctx->static_initializers, $2);
destroy_block($2);
}
| hlsl_prog preproc_directive
@@ -6315,6 +6429,9 @@ semantic:
{
char *p;
+ if (!($$.raw_name = hlsl_strdup(ctx, $2)))
+ YYABORT;
+
for (p = $2 + strlen($2); p > $2 && isdigit(p[-1]); --p)
;
$$.name = $2;
@@ -6451,6 +6568,9 @@ parameter:
}
$$.type = type;
+ if (hlsl_version_ge(ctx, 5, 1) && type->class == HLSL_CLASS_ARRAY && hlsl_type_is_resource(type))
+ hlsl_fixme(ctx, &@2, "Shader model 5.1+ resource array.");
+
$$.name = $3;
$$.semantic = $5.semantic;
$$.reg_reservation = $5.reg_reservation;
@@ -6713,6 +6833,26 @@ type_no_void:
}
vkd3d_free($1);
}
+ | KW_UNSIGNED TYPE_IDENTIFIER
+ {
+ struct hlsl_type *type = hlsl_get_type(ctx->cur_scope, $2, true, true);
+
+ if (hlsl_is_numeric_type(type) && type->e.numeric.type == HLSL_TYPE_INT)
+ {
+ if (!(type = hlsl_type_clone(ctx, type, 0, 0)))
+ YYABORT;
+ vkd3d_free((void *)type->name);
+ type->name = NULL;
+ type->e.numeric.type = HLSL_TYPE_UINT;
+ }
+ else
+ {
+ hlsl_error(ctx, &@2, VKD3D_SHADER_ERROR_HLSL_INVALID_TYPE,
+ "The 'unsigned' keyword can't be used with type %s.", $2);
+ }
+
+ $$ = type;
+ }
| KW_STRUCT TYPE_IDENTIFIER
{
$$ = hlsl_get_type(ctx->cur_scope, $2, true, true);
@@ -7156,6 +7296,8 @@ var_modifiers:
}
| var_identifier var_modifiers
{
+ $$ = $2;
+
if (!strcmp($1, "precise"))
$$ = add_modifiers(ctx, $2, HLSL_MODIFIER_PRECISE, &@1);
else if (!strcmp($1, "single"))
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
index bdb72a1fab9..09066a6191a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
@@ -218,6 +218,14 @@ static void prepend_uniform_copy(struct hlsl_ctx *ctx, struct hlsl_block *block,
uniform->is_uniform = 1;
uniform->is_param = temp->is_param;
uniform->buffer = temp->buffer;
+ if (temp->default_values)
+ {
+ /* Transfer default values from the temp to the uniform. */
+ assert(!uniform->default_values);
+ assert(hlsl_type_component_count(temp->data_type) == hlsl_type_component_count(uniform->data_type));
+ uniform->default_values = temp->default_values;
+ temp->default_values = NULL;
+ }
if (!(new_name = hlsl_sprintf_alloc(ctx, "<temp-%s>", temp->name)))
return;
@@ -312,7 +320,7 @@ static struct hlsl_ir_var *add_semantic_var(struct hlsl_ctx *ctx, struct hlsl_ir
}
}
- if (!(new_semantic.name = hlsl_strdup(ctx, semantic->name)))
+ if (!(hlsl_clone_semantic(ctx, &new_semantic, semantic)))
{
vkd3d_free(new_name);
return NULL;
@@ -3815,15 +3823,16 @@ static void allocate_register_reservations(struct hlsl_ctx *ctx)
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
+ const struct hlsl_reg_reservation *reservation = &var->reg_reservation;
unsigned int r;
- if (var->reg_reservation.reg_type)
+ if (reservation->reg_type)
{
for (r = 0; r <= HLSL_REGSET_LAST_OBJECT; ++r)
{
if (var->regs[r].allocation_size > 0)
{
- if (var->reg_reservation.reg_type != get_regset_name(r))
+ if (reservation->reg_type != get_regset_name(r))
{
struct vkd3d_string_buffer *type_string;
@@ -3839,10 +3848,8 @@ static void allocate_register_reservations(struct hlsl_ctx *ctx)
else
{
var->regs[r].allocated = true;
- var->regs[r].id = var->reg_reservation.reg_index;
- TRACE("Allocated reserved %s to %c%u-%c%u.\n", var->name, var->reg_reservation.reg_type,
- var->reg_reservation.reg_index, var->reg_reservation.reg_type,
- var->reg_reservation.reg_index + var->regs[r].allocation_size);
+ var->regs[r].space = reservation->reg_space;
+ var->regs[r].index = reservation->reg_index;
}
}
}
@@ -4762,13 +4769,14 @@ static void allocate_semantic_registers(struct hlsl_ctx *ctx)
}
}
-static const struct hlsl_buffer *get_reserved_buffer(struct hlsl_ctx *ctx, uint32_t index)
+static const struct hlsl_buffer *get_reserved_buffer(struct hlsl_ctx *ctx, uint32_t space, uint32_t index)
{
const struct hlsl_buffer *buffer;
LIST_FOR_EACH_ENTRY(buffer, &ctx->buffers, const struct hlsl_buffer, entry)
{
- if (buffer->used_size && buffer->reservation.reg_type == 'b' && buffer->reservation.reg_index == index)
+ if (buffer->used_size && buffer->reservation.reg_type == 'b'
+ && buffer->reservation.reg_space == space && buffer->reservation.reg_index == index)
return buffer;
}
return NULL;
@@ -4783,6 +4791,7 @@ static void hlsl_calculate_buffer_offset(struct hlsl_ctx *ctx, struct hlsl_ir_va
if (register_reservation)
{
var->buffer_offset = 4 * var->reg_reservation.reg_index;
+ var->has_explicit_bind_point = 1;
}
else
{
@@ -4815,6 +4824,7 @@ static void hlsl_calculate_buffer_offset(struct hlsl_ctx *ctx, struct hlsl_ir_va
}
}
var->buffer_offset = var->reg_reservation.offset_index;
+ var->has_explicit_bind_point = 1;
}
else
{
@@ -4916,8 +4926,8 @@ void hlsl_calculate_buffer_offsets(struct hlsl_ctx *ctx)
static void allocate_buffers(struct hlsl_ctx *ctx)
{
struct hlsl_buffer *buffer;
+ uint32_t index = 0, id = 0;
struct hlsl_ir_var *var;
- uint32_t index = 0;
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
@@ -4938,32 +4948,48 @@ static void allocate_buffers(struct hlsl_ctx *ctx)
if (buffer->type == HLSL_BUFFER_CONSTANT)
{
- if (buffer->reservation.reg_type == 'b')
+ const struct hlsl_reg_reservation *reservation = &buffer->reservation;
+
+ if (reservation->reg_type == 'b')
{
- const struct hlsl_buffer *reserved_buffer = get_reserved_buffer(ctx, buffer->reservation.reg_index);
+ const struct hlsl_buffer *reserved_buffer = get_reserved_buffer(ctx,
+ reservation->reg_space, reservation->reg_index);
if (reserved_buffer && reserved_buffer != buffer)
{
hlsl_error(ctx, &buffer->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS,
- "Multiple buffers bound to cb%u.", buffer->reservation.reg_index);
+ "Multiple buffers bound to space %u, index %u.",
+ reservation->reg_space, reservation->reg_index);
hlsl_note(ctx, &reserved_buffer->loc, VKD3D_SHADER_LOG_ERROR,
- "Buffer %s is already bound to cb%u.", reserved_buffer->name, buffer->reservation.reg_index);
+ "Buffer %s is already bound to space %u, index %u.",
+ reserved_buffer->name, reservation->reg_space, reservation->reg_index);
}
- buffer->reg.id = buffer->reservation.reg_index;
+ buffer->reg.space = reservation->reg_space;
+ buffer->reg.index = reservation->reg_index;
+ if (hlsl_version_ge(ctx, 5, 1))
+ buffer->reg.id = id++;
+ else
+ buffer->reg.id = buffer->reg.index;
buffer->reg.allocation_size = 1;
buffer->reg.allocated = true;
- TRACE("Allocated reserved %s to cb%u.\n", buffer->name, index);
+ TRACE("Allocated reserved %s to space %u, index %u, id %u.\n",
+ buffer->name, buffer->reg.space, buffer->reg.index, buffer->reg.id);
}
- else if (!buffer->reservation.reg_type)
+ else if (!reservation->reg_type)
{
- while (get_reserved_buffer(ctx, index))
+ while (get_reserved_buffer(ctx, 0, index))
++index;
- buffer->reg.id = index;
+ buffer->reg.space = 0;
+ buffer->reg.index = index;
+ if (hlsl_version_ge(ctx, 5, 1))
+ buffer->reg.id = id++;
+ else
+ buffer->reg.id = buffer->reg.index;
buffer->reg.allocation_size = 1;
buffer->reg.allocated = true;
- TRACE("Allocated %s to cb%u.\n", buffer->name, index);
+ TRACE("Allocated %s to space 0, index %u, id %u.\n", buffer->name, buffer->reg.index, buffer->reg.id);
++index;
}
else
@@ -4980,7 +5006,7 @@ 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, bool allocated_only)
+ uint32_t space, uint32_t index, bool allocated_only)
{
const struct hlsl_ir_var *var;
unsigned int start, count;
@@ -4995,12 +5021,18 @@ static const struct hlsl_ir_var *get_allocated_object(struct hlsl_ctx *ctx, enum
start = var->reg_reservation.reg_index;
count = var->data_type->reg_size[regset];
+ if (var->reg_reservation.reg_space != space)
+ continue;
+
if (!var->regs[regset].allocated && allocated_only)
continue;
}
else if (var->regs[regset].allocated)
{
- start = var->regs[regset].id;
+ if (var->regs[regset].space != space)
+ continue;
+
+ start = var->regs[regset].index;
count = var->regs[regset].allocation_size;
}
else
@@ -5017,8 +5049,8 @@ static const struct hlsl_ir_var *get_allocated_object(struct hlsl_ctx *ctx, enum
static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
{
char regset_name = get_regset_name(regset);
+ uint32_t min_index = 0, id = 0;
struct hlsl_ir_var *var;
- uint32_t min_index = 0;
if (regset == HLSL_REGSET_UAVS)
{
@@ -5041,35 +5073,44 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
if (var->regs[regset].allocated)
{
const struct hlsl_ir_var *reserved_object, *last_reported = NULL;
- unsigned int index, i;
+ unsigned int i;
- if (var->regs[regset].id < min_index)
+ if (var->regs[regset].index < min_index)
{
assert(regset == HLSL_REGSET_UAVS);
hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS,
"UAV index (%u) must be higher than the maximum render target index (%u).",
- var->regs[regset].id, min_index - 1);
+ var->regs[regset].index, min_index - 1);
continue;
}
for (i = 0; i < count; ++i)
{
- index = var->regs[regset].id + i;
+ unsigned int space = var->regs[regset].space;
+ unsigned int index = var->regs[regset].index + i;
/* 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);
+ reserved_object = get_allocated_object(ctx, regset, space, index, true);
if (reserved_object && reserved_object != var && reserved_object != last_reported)
{
hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS,
- "Multiple variables bound to %c%u.", regset_name, index);
+ "Multiple variables bound to space %u, %c%u.", regset_name, space, index);
hlsl_note(ctx, &reserved_object->loc, VKD3D_SHADER_LOG_ERROR,
- "Variable '%s' is already bound to %c%u.", reserved_object->name,
- regset_name, index);
+ "Variable '%s' is already bound to space %u, %c%u.",
+ reserved_object->name, regset_name, space, index);
last_reported = reserved_object;
}
}
+
+ if (hlsl_version_ge(ctx, 5, 1))
+ var->regs[regset].id = id++;
+ else
+ var->regs[regset].id = var->regs[regset].index;
+ TRACE("Allocated reserved variable %s to space %u, indices %c%u-%c%u, id %u.\n",
+ var->name, var->regs[regset].space, regset_name, var->regs[regset].index,
+ regset_name, var->regs[regset].index + count, var->regs[regset].id);
}
else
{
@@ -5078,7 +5119,7 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
while (available < count)
{
- if (get_allocated_object(ctx, regset, index, false))
+ if (get_allocated_object(ctx, regset, 0, index, false))
available = 0;
else
++available;
@@ -5086,10 +5127,15 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset)
}
index -= count;
- var->regs[regset].id = index;
+ var->regs[regset].space = 0;
+ var->regs[regset].index = index;
+ if (hlsl_version_ge(ctx, 5, 1))
+ var->regs[regset].id = id++;
+ else
+ var->regs[regset].id = var->regs[regset].index;
var->regs[regset].allocated = true;
- TRACE("Allocated variable %s to %c%u-%c%u.\n", var->name, regset_name, index, regset_name,
- index + count);
+ TRACE("Allocated variable %s to space 0, indices %c%u-%c%u, id %u.\n", var->name,
+ regset_name, index, regset_name, index + count, var->regs[regset].id);
++index;
}
}
@@ -5295,6 +5341,7 @@ struct hlsl_reg hlsl_reg_from_deref(struct hlsl_ctx *ctx, const struct hlsl_dere
assert(deref->data_type);
assert(hlsl_is_numeric_type(deref->data_type));
+ ret.index += offset / 4;
ret.id += offset / 4;
ret.writemask = 0xf & (0xf << (offset % 4));
diff --git a/libs/vkd3d/libs/vkd3d-shader/ir.c b/libs/vkd3d/libs/vkd3d-shader/ir.c
index b3b745fc1b2..5f99be632f2 100644
--- a/libs/vkd3d/libs/vkd3d-shader/ir.c
+++ b/libs/vkd3d/libs/vkd3d-shader/ir.c
@@ -46,9 +46,9 @@ static inline bool shader_register_is_phase_instance_id(const struct vkd3d_shade
static bool vsir_instruction_is_dcl(const struct vkd3d_shader_instruction *instruction)
{
- enum vkd3d_shader_opcode handler_idx = instruction->handler_idx;
- return (VKD3DSIH_DCL <= handler_idx && handler_idx <= VKD3DSIH_DCL_VERTICES_OUT)
- || handler_idx == VKD3DSIH_HS_DECLS;
+ enum vkd3d_shader_opcode opcode = instruction->opcode;
+ return (VKD3DSIH_DCL <= opcode && opcode <= VKD3DSIH_DCL_VERTICES_OUT)
+ || opcode == VKD3DSIH_HS_DECLS;
}
static void vkd3d_shader_instruction_make_nop(struct vkd3d_shader_instruction *ins)
@@ -60,9 +60,9 @@ static void vkd3d_shader_instruction_make_nop(struct vkd3d_shader_instruction *i
static bool vsir_instruction_init_with_params(struct vsir_program *program,
struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location,
- enum vkd3d_shader_opcode handler_idx, unsigned int dst_count, unsigned int src_count)
+ enum vkd3d_shader_opcode opcode, unsigned int dst_count, unsigned int src_count)
{
- vsir_instruction_init(ins, location, handler_idx);
+ vsir_instruction_init(ins, location, opcode);
ins->dst_count = dst_count;
ins->src_count = src_count;
@@ -287,7 +287,7 @@ static enum vkd3d_result vsir_program_lower_precise_mad(struct vsir_program *pro
mul_ins = &instructions->elements[pos];
add_ins = &instructions->elements[pos + 1];
- mul_ins->handler_idx = VKD3DSIH_MUL;
+ mul_ins->opcode = VKD3DSIH_MUL;
mul_ins->src_count = 2;
if (!(vsir_instruction_init_with_params(program, add_ins, &mul_ins->location, VKD3DSIH_ADD, 1, 2)))
@@ -322,7 +322,7 @@ static enum vkd3d_result vsir_program_lower_instructions(struct vsir_program *pr
{
struct vkd3d_shader_instruction *ins = &instructions->elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_IFC:
if ((ret = vsir_program_lower_ifc(program, ins, &tmp_idx, message_context)) < 0)
@@ -492,26 +492,26 @@ static void flattener_eliminate_phase_related_dcls(struct hull_flattener *normal
struct shader_phase_location *loc;
bool b;
- if (ins->handler_idx == VKD3DSIH_HS_FORK_PHASE || ins->handler_idx == VKD3DSIH_HS_JOIN_PHASE)
+ if (ins->opcode == VKD3DSIH_HS_FORK_PHASE || ins->opcode == VKD3DSIH_HS_JOIN_PHASE)
{
b = flattener_is_in_fork_or_join_phase(normaliser);
/* Reset the phase info. */
normaliser->phase_body_idx = ~0u;
- normaliser->phase = ins->handler_idx;
+ normaliser->phase = ins->opcode;
normaliser->instance_count = 1;
/* Leave the first occurrence and delete the rest. */
if (b)
vkd3d_shader_instruction_make_nop(ins);
return;
}
- else if (ins->handler_idx == VKD3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT
- || ins->handler_idx == VKD3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT)
+ else if (ins->opcode == VKD3DSIH_DCL_HS_FORK_PHASE_INSTANCE_COUNT
+ || ins->opcode == VKD3DSIH_DCL_HS_JOIN_PHASE_INSTANCE_COUNT)
{
normaliser->instance_count = ins->declaration.count + !ins->declaration.count;
vkd3d_shader_instruction_make_nop(ins);
return;
}
- else if (ins->handler_idx == VKD3DSIH_DCL_INPUT && shader_register_is_phase_instance_id(
+ else if (ins->opcode == VKD3DSIH_DCL_INPUT && shader_register_is_phase_instance_id(
&ins->declaration.dst.reg))
{
vkd3d_shader_instruction_make_nop(ins);
@@ -524,7 +524,7 @@ static void flattener_eliminate_phase_related_dcls(struct hull_flattener *normal
if (normaliser->phase_body_idx == ~0u)
normaliser->phase_body_idx = index;
- if (ins->handler_idx == VKD3DSIH_RET)
+ if (ins->opcode == VKD3DSIH_RET)
{
normaliser->last_ret_location = ins->location;
vkd3d_shader_instruction_make_nop(ins);
@@ -679,11 +679,11 @@ static void src_param_init_const_uint(struct vkd3d_shader_src_param *src, uint32
}
void vsir_instruction_init(struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location,
- enum vkd3d_shader_opcode handler_idx)
+ enum vkd3d_shader_opcode opcode)
{
memset(ins, 0, sizeof(*ins));
ins->location = *location;
- ins->handler_idx = handler_idx;
+ ins->opcode = opcode;
}
static bool vsir_instruction_init_label(struct vkd3d_shader_instruction *ins,
@@ -865,12 +865,12 @@ static enum vkd3d_result instruction_array_normalise_hull_shader_control_point_i
{
ins = &instructions->elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_HS_CONTROL_POINT_PHASE:
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
- normaliser.phase = ins->handler_idx;
+ normaliser.phase = ins->opcode;
break;
default:
if (vsir_instruction_is_dcl(ins))
@@ -888,7 +888,7 @@ static enum vkd3d_result instruction_array_normalise_hull_shader_control_point_i
{
ins = &instructions->elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_DCL_INPUT_CONTROL_POINT_COUNT:
input_control_point_count = ins->declaration.count;
@@ -1526,7 +1526,7 @@ static void shader_instruction_normalise_io_params(struct vkd3d_shader_instructi
struct vkd3d_shader_register *reg;
unsigned int i;
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_DCL_INPUT:
if (normaliser->shader_type == VKD3D_SHADER_TYPE_HULL)
@@ -1560,7 +1560,7 @@ static void shader_instruction_normalise_io_params(struct vkd3d_shader_instructi
case VKD3DSIH_HS_CONTROL_POINT_PHASE:
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
- normaliser->phase = ins->handler_idx;
+ normaliser->phase = ins->opcode;
memset(normaliser->input_dcl_params, 0, sizeof(normaliser->input_dcl_params));
memset(normaliser->output_dcl_params, 0, sizeof(normaliser->output_dcl_params));
memset(normaliser->pc_dcl_params, 0, sizeof(normaliser->pc_dcl_params));
@@ -1594,7 +1594,7 @@ static enum vkd3d_result vsir_program_normalise_io_registers(struct vsir_program
{
ins = &program->instructions.elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_DCL_OUTPUT_CONTROL_POINT_COUNT:
normaliser.output_control_point_count = ins->declaration.count;
@@ -1608,7 +1608,7 @@ static enum vkd3d_result vsir_program_normalise_io_registers(struct vsir_program
/* fall through */
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
- normaliser.phase = ins->handler_idx;
+ normaliser.phase = ins->opcode;
break;
default:
break;
@@ -1740,7 +1740,7 @@ static enum vkd3d_result instruction_array_normalise_flat_constants(struct vsir_
{
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
- if (ins->handler_idx == VKD3DSIH_DEF || ins->handler_idx == VKD3DSIH_DEFI || ins->handler_idx == VKD3DSIH_DEFB)
+ if (ins->opcode == VKD3DSIH_DEF || ins->opcode == VKD3DSIH_DEFI || ins->opcode == VKD3DSIH_DEFB)
{
struct flat_constant_def *def;
@@ -1779,7 +1779,7 @@ static void remove_dead_code(struct vsir_program *program)
{
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_IF:
case VKD3DSIH_LOOP:
@@ -1799,7 +1799,7 @@ static void remove_dead_code(struct vsir_program *program)
{
if (depth > 0)
{
- if (ins->handler_idx != VKD3DSIH_ELSE)
+ if (ins->opcode != VKD3DSIH_ELSE)
--depth;
vkd3d_shader_instruction_make_nop(ins);
}
@@ -1870,14 +1870,14 @@ static enum vkd3d_result vsir_program_normalise_combined_samplers(struct vsir_pr
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
struct vkd3d_shader_src_param *srcs;
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_TEX:
if (!(srcs = shader_src_param_allocator_get(&program->instructions.src_params, 3)))
return VKD3D_ERROR_OUT_OF_MEMORY;
memset(srcs, 0, sizeof(*srcs) * 3);
- ins->handler_idx = VKD3DSIH_SAMPLE;
+ ins->opcode = VKD3DSIH_SAMPLE;
srcs[0] = ins->src[0];
@@ -1919,7 +1919,7 @@ static enum vkd3d_result vsir_program_normalise_combined_samplers(struct vsir_pr
case VKD3DSIH_TEXREG2RGB:
vkd3d_shader_error(message_context, &ins->location, VKD3D_SHADER_ERROR_VSIR_NOT_IMPLEMENTED,
"Aborting due to not yet implemented feature: "
- "Combined sampler instruction %#x.", ins->handler_idx);
+ "Combined sampler instruction %#x.", ins->opcode);
return VKD3D_ERROR_NOT_IMPLEMENTED;
default:
@@ -2030,7 +2030,7 @@ static bool cf_flattener_copy_instruction(struct cf_flattener *flattener,
{
struct vkd3d_shader_instruction *dst_ins;
- if (instruction->handler_idx == VKD3DSIH_NOP)
+ if (instruction->opcode == VKD3DSIH_NOP)
return true;
if (!(dst_ins = cf_flattener_require_space(flattener, 1)))
@@ -2245,9 +2245,9 @@ static enum vkd3d_result cf_flattener_iterate_instruction_array(struct cf_flatte
* phase instruction, and in all other shader types begins with the first label instruction.
* Declaring an indexable temp with function scope is not considered a declaration,
* because it needs to live inside a function. */
- if (!after_declarations_section && instruction->handler_idx != VKD3DSIH_NOP)
+ if (!after_declarations_section && instruction->opcode != VKD3DSIH_NOP)
{
- bool is_function_indexable = instruction->handler_idx == VKD3DSIH_DCL_INDEXABLE_TEMP
+ bool is_function_indexable = instruction->opcode == VKD3DSIH_DCL_INDEXABLE_TEMP
&& instruction->declaration.indexable_temp.has_function_scope;
if (!vsir_instruction_is_dcl(instruction) || is_function_indexable)
@@ -2260,14 +2260,14 @@ static enum vkd3d_result cf_flattener_iterate_instruction_array(struct cf_flatte
cf_info = flattener->control_flow_depth
? &flattener->control_flow_info[flattener->control_flow_depth - 1] : NULL;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_HS_CONTROL_POINT_PHASE:
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
if (!cf_flattener_copy_instruction(flattener, instruction))
return VKD3D_ERROR_OUT_OF_MEMORY;
- if (instruction->handler_idx != VKD3DSIH_HS_CONTROL_POINT_PHASE || !instruction->flags)
+ if (instruction->opcode != VKD3DSIH_HS_CONTROL_POINT_PHASE || !instruction->flags)
after_declarations_section = false;
break;
@@ -2662,7 +2662,7 @@ static enum vkd3d_result lower_switch_to_if_ladder(struct vsir_program *program)
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
unsigned int case_count, j, default_label;
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_LABEL:
current_label = label_from_src_param(&ins->src[0]);
@@ -2858,7 +2858,7 @@ static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_
/* Only phi src/dst SSA values need be converted here. Structurisation may
* introduce new cases of undominated SSA use, which will be handled later. */
- if (ins->handler_idx != VKD3DSIH_PHI)
+ if (ins->opcode != VKD3DSIH_PHI)
continue;
++phi_count;
@@ -2907,7 +2907,7 @@ static enum vkd3d_result vsir_program_materialise_phi_ssas_to_temps(struct vsir_
for (j = 0; j < ins->src_count; ++j)
materialize_ssas_to_temps_process_reg(program, &alloc, &ins->src[j].reg);
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_LABEL:
current_label = label_from_src_param(&ins->src[0]);
@@ -3336,7 +3336,7 @@ static void vsir_cfg_dump_dot(struct vsir_cfg *cfg)
if (block->label == 0)
continue;
- switch (block->end->handler_idx)
+ switch (block->end->opcode)
{
case VKD3DSIH_RET:
shape = "trapezium";
@@ -3478,7 +3478,7 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program
struct vkd3d_shader_instruction *instruction = &program->instructions.elements[i];
bool finish = false;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_PHI:
case VKD3DSIH_SWITCH_MONOLITHIC:
@@ -3533,7 +3533,7 @@ static enum vkd3d_result vsir_cfg_init(struct vsir_cfg *cfg, struct vsir_program
if (block->label == 0)
continue;
- switch (block->end->handler_idx)
+ switch (block->end->opcode)
{
case VKD3DSIH_RET:
break;
@@ -4192,7 +4192,7 @@ static enum vkd3d_result vsir_cfg_build_structured_program(struct vsir_cfg *cfg)
structure->u.block = block;
/* Generate between zero and two jump instructions. */
- switch (block->end->handler_idx)
+ switch (block->end->opcode)
{
case VKD3DSIH_BRANCH:
{
@@ -5049,7 +5049,7 @@ static enum vkd3d_result vsir_program_structurize(struct vsir_program *program,
{
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_LABEL:
assert(program->shader_version.type != VKD3D_SHADER_TYPE_HULL);
@@ -5064,7 +5064,7 @@ static enum vkd3d_result vsir_program_structurize(struct vsir_program *program,
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
assert(program->shader_version.type == VKD3D_SHADER_TYPE_HULL);
- TRACE("Structurizing phase %u of a hull shader.\n", ins->handler_idx);
+ TRACE("Structurizing phase %u of a hull shader.\n", ins->opcode);
target.instructions[target.ins_count++] = *ins;
++i;
if ((ret = vsir_program_structurize_function(program, message_context,
@@ -5222,7 +5222,7 @@ static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(stru
{
struct vkd3d_shader_instruction *ins = &program->instructions.elements[i];
- switch (ins->handler_idx)
+ switch (ins->opcode)
{
case VKD3DSIH_LABEL:
assert(program->shader_version.type != VKD3D_SHADER_TYPE_HULL);
@@ -5237,7 +5237,7 @@ static enum vkd3d_result vsir_program_materialize_undominated_ssas_to_temps(stru
case VKD3DSIH_HS_FORK_PHASE:
case VKD3DSIH_HS_JOIN_PHASE:
assert(program->shader_version.type == VKD3D_SHADER_TYPE_HULL);
- TRACE("Materializing undominated SSAs in phase %u of a hull shader.\n", ins->handler_idx);
+ TRACE("Materializing undominated SSAs in phase %u of a hull shader.\n", ins->opcode);
++i;
if ((ret = vsir_program_materialize_undominated_ssas_to_temps_in_function(
program, message_context, &i)) < 0)
@@ -5641,7 +5641,7 @@ static void vsir_validate_dst_count(struct validation_context *ctx,
if (instruction->dst_count != count)
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_DEST_COUNT,
"Invalid destination count %u for an instruction of type %#x, expected %u.",
- instruction->dst_count, instruction->handler_idx, count);
+ instruction->dst_count, instruction->opcode, count);
}
static void vsir_validate_src_count(struct validation_context *ctx,
@@ -5650,7 +5650,7 @@ static void vsir_validate_src_count(struct validation_context *ctx,
if (instruction->src_count != count)
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SOURCE_COUNT,
"Invalid source count %u for an instruction of type %#x, expected %u.",
- instruction->src_count, instruction->handler_idx, count);
+ instruction->src_count, instruction->opcode, count);
}
static bool vsir_validate_src_min_count(struct validation_context *ctx,
@@ -5660,7 +5660,7 @@ static bool vsir_validate_src_min_count(struct validation_context *ctx,
{
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SOURCE_COUNT,
"Invalid source count %u for an instruction of type %#x, expected at least %u.",
- instruction->src_count, instruction->handler_idx, count);
+ instruction->src_count, instruction->opcode, count);
return false;
}
@@ -5674,7 +5674,7 @@ static bool vsir_validate_src_max_count(struct validation_context *ctx,
{
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_SOURCE_COUNT,
"Invalid source count %u for an instruction of type %#x, expected at most %u.",
- instruction->src_count, instruction->handler_idx, count);
+ instruction->src_count, instruction->opcode, count);
return false;
}
@@ -5701,7 +5701,7 @@ static void vsir_validate_cf_type(struct validation_context *ctx,
assert(expected_type != CF_TYPE_UNKNOWN);
if (ctx->cf_type != expected_type)
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "Invalid instruction %#x in %s shader.",
- instruction->handler_idx, name_from_cf_type(ctx->cf_type));
+ instruction->opcode, name_from_cf_type(ctx->cf_type));
}
static void vsir_validate_instruction(struct validation_context *ctx)
@@ -5718,13 +5718,13 @@ static void vsir_validate_instruction(struct validation_context *ctx)
for (i = 0; i < instruction->src_count; ++i)
vsir_validate_src_param(ctx, &instruction->src[i]);
- if (instruction->handler_idx >= VKD3DSIH_INVALID)
+ if (instruction->opcode >= VKD3DSIH_INVALID)
{
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER, "Invalid instruction handler %#x.",
- instruction->handler_idx);
+ instruction->opcode);
}
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_HS_DECLS:
case VKD3DSIH_HS_CONTROL_POINT_PHASE:
@@ -5733,12 +5733,14 @@ static void vsir_validate_instruction(struct validation_context *ctx)
vsir_validate_dst_count(ctx, instruction, 0);
vsir_validate_src_count(ctx, instruction, 0);
if (version->type != VKD3D_SHADER_TYPE_HULL)
- validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER, "Phase instruction %#x is only valid in a hull shader.",
- instruction->handler_idx);
+ validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER,
+ "Phase instruction %#x is only valid in a hull shader.",
+ instruction->opcode);
if (ctx->depth != 0)
- validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "Phase instruction %#x must appear to top level.",
- instruction->handler_idx);
- ctx->phase = instruction->handler_idx;
+ validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW,
+ "Phase instruction %#x must appear to top level.",
+ instruction->opcode);
+ ctx->phase = instruction->opcode;
ctx->dcl_temps_found = false;
return;
@@ -5812,7 +5814,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
&& ctx->phase == VKD3DSIH_INVALID)
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_HANDLER,
"Instruction %#x appear before any phase instruction in a hull shader.",
- instruction->handler_idx);
+ instruction->opcode);
/* We support two different control flow types in shaders:
* block-based, like DXIL and SPIR-V, and structured, like D3DBC
@@ -5824,7 +5826,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
* block, but need for that hasn't arisen yet, so we don't. */
if (ctx->cf_type == CF_TYPE_UNKNOWN && !vsir_instruction_is_dcl(instruction))
{
- if (instruction->handler_idx == VKD3DSIH_LABEL)
+ if (instruction->opcode == VKD3DSIH_LABEL)
ctx->cf_type = CF_TYPE_BLOCKS;
else
ctx->cf_type = CF_TYPE_STRUCTURED;
@@ -5832,7 +5834,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
if (ctx->cf_type == CF_TYPE_BLOCKS && !vsir_instruction_is_dcl(instruction))
{
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_LABEL:
if (ctx->inside_block)
@@ -5844,20 +5846,22 @@ static void vsir_validate_instruction(struct validation_context *ctx)
case VKD3DSIH_BRANCH:
case VKD3DSIH_SWITCH_MONOLITHIC:
if (!ctx->inside_block)
- validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "Invalid instruction %#x outside any block.",
- instruction->handler_idx);
+ validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW,
+ "Invalid instruction %#x outside any block.",
+ instruction->opcode);
ctx->inside_block = false;
break;
default:
if (!ctx->inside_block)
- validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "Invalid instruction %#x outside any block.",
- instruction->handler_idx);
+ validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW,
+ "Invalid instruction %#x outside any block.",
+ instruction->opcode);
break;
}
}
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_DCL_TEMPS:
vsir_validate_dst_count(ctx, instruction, 0);
@@ -5877,7 +5881,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
vsir_validate_src_count(ctx, instruction, 1);
if (!vkd3d_array_reserve((void **)&ctx->blocks, &ctx->blocks_capacity, ctx->depth + 1, sizeof(*ctx->blocks)))
return;
- ctx->blocks[ctx->depth++] = instruction->handler_idx;
+ ctx->blocks[ctx->depth++] = instruction->opcode;
break;
case VKD3DSIH_IFC:
@@ -5896,7 +5900,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
if (ctx->depth == 0 || ctx->blocks[ctx->depth - 1] != VKD3DSIH_IF)
validator_error(ctx, VKD3D_SHADER_ERROR_VSIR_INVALID_CONTROL_FLOW, "ELSE instruction doesn't terminate IF block.");
else
- ctx->blocks[ctx->depth - 1] = instruction->handler_idx;
+ ctx->blocks[ctx->depth - 1] = instruction->opcode;
break;
case VKD3DSIH_ENDIF:
@@ -5915,7 +5919,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
vsir_validate_src_count(ctx, instruction, version->major <= 3 ? 2 : 0);
if (!vkd3d_array_reserve((void **)&ctx->blocks, &ctx->blocks_capacity, ctx->depth + 1, sizeof(*ctx->blocks)))
return;
- ctx->blocks[ctx->depth++] = instruction->handler_idx;
+ ctx->blocks[ctx->depth++] = instruction->opcode;
break;
case VKD3DSIH_ENDLOOP:
@@ -5934,7 +5938,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
vsir_validate_src_count(ctx, instruction, 1);
if (!vkd3d_array_reserve((void **)&ctx->blocks, &ctx->blocks_capacity, ctx->depth + 1, sizeof(*ctx->blocks)))
return;
- ctx->blocks[ctx->depth++] = instruction->handler_idx;
+ ctx->blocks[ctx->depth++] = instruction->opcode;
break;
case VKD3DSIH_ENDREP:
@@ -5953,7 +5957,7 @@ static void vsir_validate_instruction(struct validation_context *ctx)
vsir_validate_src_count(ctx, instruction, 1);
if (!vkd3d_array_reserve((void **)&ctx->blocks, &ctx->blocks_capacity, ctx->depth + 1, sizeof(*ctx->blocks)))
return;
- ctx->blocks[ctx->depth++] = instruction->handler_idx;
+ ctx->blocks[ctx->depth++] = instruction->opcode;
break;
case VKD3DSIH_ENDSWITCH:
diff --git a/libs/vkd3d/libs/vkd3d-shader/preproc.l b/libs/vkd3d/libs/vkd3d-shader/preproc.l
index be50d3b9020..a3cdbe559a7 100644
--- a/libs/vkd3d/libs/vkd3d-shader/preproc.l
+++ b/libs/vkd3d/libs/vkd3d-shader/preproc.l
@@ -20,6 +20,7 @@
%{
+#include "preproc.h"
#include "preproc.tab.h"
#undef ERROR /* defined in wingdi.h */
diff --git a/libs/vkd3d/libs/vkd3d-shader/spirv.c b/libs/vkd3d/libs/vkd3d-shader/spirv.c
index 984a4f894f6..524fb8e9b1f 100644
--- a/libs/vkd3d/libs/vkd3d-shader/spirv.c
+++ b/libs/vkd3d/libs/vkd3d-shader/spirv.c
@@ -1752,6 +1752,22 @@ static uint32_t vkd3d_spirv_get_op_scope_subgroup(struct vkd3d_spirv_builder *bu
return vkd3d_spirv_build_once(builder, &builder->scope_subgroup_id, vkd3d_spirv_build_op_scope_subgroup);
}
+static uint32_t vkd3d_spirv_build_op_group_nonuniform_quad_swap(struct vkd3d_spirv_builder *builder,
+ uint32_t result_type, uint32_t val_id, uint32_t op_id)
+{
+ vkd3d_spirv_enable_capability(builder, SpvCapabilityGroupNonUniformQuad);
+ return vkd3d_spirv_build_op_tr3(builder, &builder->function_stream, SpvOpGroupNonUniformQuadSwap, result_type,
+ vkd3d_spirv_get_op_scope_subgroup(builder), val_id, op_id);
+}
+
+static uint32_t vkd3d_spirv_build_op_group_nonuniform_quad_broadcast(struct vkd3d_spirv_builder *builder,
+ uint32_t result_type, uint32_t val_id, uint32_t index_id)
+{
+ vkd3d_spirv_enable_capability(builder, SpvCapabilityGroupNonUniformQuad);
+ return vkd3d_spirv_build_op_tr3(builder, &builder->function_stream, SpvOpGroupNonUniformQuadBroadcast, result_type,
+ vkd3d_spirv_get_op_scope_subgroup(builder), val_id, index_id);
+}
+
static uint32_t vkd3d_spirv_build_op_group_nonuniform_ballot(struct vkd3d_spirv_builder *builder,
uint32_t result_type, uint32_t val_id)
{
@@ -6831,7 +6847,7 @@ static void spirv_compiler_enter_shader_phase(struct spirv_compiler *compiler,
uint32_t function_id, void_id, function_type_id;
struct vkd3d_shader_phase *phase;
- assert(compiler->phase != instruction->handler_idx);
+ assert(compiler->phase != instruction->opcode);
if (!is_in_default_phase(compiler))
spirv_compiler_leave_shader_phase(compiler);
@@ -6843,16 +6859,16 @@ static void spirv_compiler_enter_shader_phase(struct spirv_compiler *compiler,
vkd3d_spirv_build_op_function(builder, void_id, function_id,
SpvFunctionControlMaskNone, function_type_id);
- compiler->phase = instruction->handler_idx;
+ compiler->phase = instruction->opcode;
spirv_compiler_emit_shader_phase_name(compiler, function_id, NULL);
- phase = (instruction->handler_idx == VKD3DSIH_HS_CONTROL_POINT_PHASE)
+ phase = (instruction->opcode == VKD3DSIH_HS_CONTROL_POINT_PHASE)
? &compiler->control_point_phase : &compiler->patch_constant_phase;
phase->function_id = function_id;
/* The insertion location must be set after the label is emitted. */
phase->function_location = 0;
- if (instruction->handler_idx == VKD3DSIH_HS_CONTROL_POINT_PHASE)
+ if (instruction->opcode == VKD3DSIH_HS_CONTROL_POINT_PHASE)
compiler->emit_default_control_point_phase = instruction->flags;
}
@@ -7016,7 +7032,7 @@ static SpvOp spirv_compiler_map_alu_instruction(const struct vkd3d_shader_instru
{
static const struct
{
- enum vkd3d_shader_opcode handler_idx;
+ enum vkd3d_shader_opcode opcode;
SpvOp spirv_op;
}
alu_ops[] =
@@ -7056,7 +7072,7 @@ static SpvOp spirv_compiler_map_alu_instruction(const struct vkd3d_shader_instru
for (i = 0; i < ARRAY_SIZE(alu_ops); ++i)
{
- if (alu_ops[i].handler_idx == instruction->handler_idx)
+ if (alu_ops[i].opcode == instruction->opcode)
return alu_ops[i].spirv_op;
}
@@ -7065,7 +7081,7 @@ static SpvOp spirv_compiler_map_alu_instruction(const struct vkd3d_shader_instru
static SpvOp spirv_compiler_map_logical_instruction(const struct vkd3d_shader_instruction *instruction)
{
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_AND:
return SpvOpLogicalAnd;
@@ -7090,20 +7106,20 @@ static void spirv_compiler_emit_bool_cast(struct spirv_compiler *compiler,
val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask);
if (dst->reg.data_type == VKD3D_DATA_HALF || dst->reg.data_type == VKD3D_DATA_FLOAT)
{
- val_id = spirv_compiler_emit_bool_to_float(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOF);
+ val_id = spirv_compiler_emit_bool_to_float(compiler, 1, val_id, instruction->opcode == VKD3DSIH_ITOF);
}
else if (dst->reg.data_type == VKD3D_DATA_DOUBLE)
{
/* ITOD is not supported. Frontends which emit bool casts must use ITOF for double. */
- val_id = spirv_compiler_emit_bool_to_double(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOF);
+ val_id = spirv_compiler_emit_bool_to_double(compiler, 1, val_id, instruction->opcode == VKD3DSIH_ITOF);
}
else if (dst->reg.data_type == VKD3D_DATA_UINT16 || dst->reg.data_type == VKD3D_DATA_UINT)
{
- val_id = spirv_compiler_emit_bool_to_int(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOI);
+ val_id = spirv_compiler_emit_bool_to_int(compiler, 1, val_id, instruction->opcode == VKD3DSIH_ITOI);
}
else if (dst->reg.data_type == VKD3D_DATA_UINT64)
{
- val_id = spirv_compiler_emit_bool_to_int64(compiler, 1, val_id, instruction->handler_idx == VKD3DSIH_ITOI);
+ val_id = spirv_compiler_emit_bool_to_int64(compiler, 1, val_id, instruction->opcode == VKD3DSIH_ITOI);
}
else
{
@@ -7126,7 +7142,7 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil
SpvOp op = SpvOpMax;
unsigned int i;
- if (src->reg.data_type == VKD3D_DATA_UINT64 && instruction->handler_idx == VKD3DSIH_COUNTBITS)
+ if (src->reg.data_type == VKD3D_DATA_UINT64 && instruction->opcode == VKD3DSIH_COUNTBITS)
{
/* At least some drivers support this anyway, but if validation is enabled it will fail. */
FIXME("Unsupported 64-bit source for bit count.\n");
@@ -7142,8 +7158,8 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil
/* VSIR supports logic ops AND/OR/XOR on bool values. */
op = spirv_compiler_map_logical_instruction(instruction);
}
- else if (instruction->handler_idx == VKD3DSIH_ITOF || instruction->handler_idx == VKD3DSIH_UTOF
- || instruction->handler_idx == VKD3DSIH_ITOI || instruction->handler_idx == VKD3DSIH_UTOU)
+ else if (instruction->opcode == VKD3DSIH_ITOF || instruction->opcode == VKD3DSIH_UTOF
+ || instruction->opcode == VKD3DSIH_ITOI || instruction->opcode == VKD3DSIH_UTOU)
{
/* VSIR supports cast from bool to signed/unsigned integer types and floating point types,
* where bool is treated as a 1-bit integer and a signed 'true' value converts to -1. */
@@ -7158,9 +7174,9 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil
if (op == SpvOpMax)
{
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_HANDLER,
- "Encountered invalid/unhandled instruction handler %#x.", instruction->handler_idx);
+ "Encountered invalid/unhandled instruction handler %#x.", instruction->opcode);
return VKD3D_ERROR_INVALID_SHADER;
}
@@ -7179,8 +7195,8 @@ static enum vkd3d_result spirv_compiler_emit_alu_instruction(struct spirv_compil
* Microsoft fxc will compile immediate constants larger than 5 bits.
* Fixing up the constants would be more elegant, but the simplest way is
* to let this handle constants too. */
- if (!(instruction->flags & VKD3DSI_SHIFT_UNMASKED) && (instruction->handler_idx == VKD3DSIH_ISHL
- || instruction->handler_idx == VKD3DSIH_ISHR || instruction->handler_idx == VKD3DSIH_USHR))
+ if (!(instruction->flags & VKD3DSI_SHIFT_UNMASKED) && (instruction->opcode == VKD3DSIH_ISHL
+ || instruction->opcode == VKD3DSIH_ISHR || instruction->opcode == VKD3DSIH_USHR))
{
uint32_t mask_id = spirv_compiler_get_constant_vector(compiler,
VKD3D_SHADER_COMPONENT_UINT, vsir_write_mask_component_count(dst->write_mask), 0x1f);
@@ -7218,7 +7234,7 @@ static enum GLSLstd450 spirv_compiler_map_ext_glsl_instruction(
{
static const struct
{
- enum vkd3d_shader_opcode handler_idx;
+ enum vkd3d_shader_opcode opcode;
enum GLSLstd450 glsl_inst;
}
glsl_insts[] =
@@ -7258,7 +7274,7 @@ static enum GLSLstd450 spirv_compiler_map_ext_glsl_instruction(
for (i = 0; i < ARRAY_SIZE(glsl_insts); ++i)
{
- if (glsl_insts[i].handler_idx == instruction->handler_idx)
+ if (glsl_insts[i].opcode == instruction->opcode)
return glsl_insts[i].glsl_inst;
}
@@ -7276,20 +7292,20 @@ static void spirv_compiler_emit_ext_glsl_instruction(struct spirv_compiler *comp
unsigned int i, component_count;
enum GLSLstd450 glsl_inst;
- if (src[0].reg.data_type == VKD3D_DATA_UINT64 && (instruction->handler_idx == VKD3DSIH_FIRSTBIT_HI
- || instruction->handler_idx == VKD3DSIH_FIRSTBIT_LO || instruction->handler_idx == VKD3DSIH_FIRSTBIT_SHI))
+ if (src[0].reg.data_type == VKD3D_DATA_UINT64 && (instruction->opcode == VKD3DSIH_FIRSTBIT_HI
+ || instruction->opcode == VKD3DSIH_FIRSTBIT_LO || instruction->opcode == VKD3DSIH_FIRSTBIT_SHI))
{
/* At least some drivers support this anyway, but if validation is enabled it will fail. */
- FIXME("Unsupported 64-bit source for handler %#x.\n", instruction->handler_idx);
+ FIXME("Unsupported 64-bit source for handler %#x.\n", instruction->opcode);
spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED,
- "64-bit source for handler %#x is not supported.", instruction->handler_idx);
+ "64-bit source for handler %#x is not supported.", instruction->opcode);
return;
}
glsl_inst = spirv_compiler_map_ext_glsl_instruction(instruction);
if (glsl_inst == GLSLstd450Bad)
{
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -7306,8 +7322,8 @@ static void spirv_compiler_emit_ext_glsl_instruction(struct spirv_compiler *comp
val_id = vkd3d_spirv_build_op_ext_inst(builder, type_id,
instr_set_id, glsl_inst, src_id, instruction->src_count);
- if (instruction->handler_idx == VKD3DSIH_FIRSTBIT_HI
- || instruction->handler_idx == VKD3DSIH_FIRSTBIT_SHI)
+ if (instruction->opcode == VKD3DSIH_FIRSTBIT_HI
+ || instruction->opcode == VKD3DSIH_FIRSTBIT_SHI)
{
/* In D3D bits are numbered from the most significant bit. */
component_count = vsir_write_mask_component_count(dst->write_mask);
@@ -7415,7 +7431,7 @@ static void spirv_compiler_emit_movc(struct spirv_compiler *compiler,
if (src[0].reg.data_type != VKD3D_DATA_BOOL)
{
- if (instruction->handler_idx == VKD3DSIH_CMP)
+ if (instruction->opcode == VKD3DSIH_CMP)
condition_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, SpvOpFOrdGreaterThanEqual,
vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_BOOL, component_count), condition_id,
spirv_compiler_get_constant_float_vector(compiler, 0.0f, component_count));
@@ -7469,9 +7485,9 @@ static void spirv_compiler_emit_dot(struct spirv_compiler *compiler,
component_count = vsir_write_mask_component_count(dst->write_mask);
component_type = vkd3d_component_type_from_data_type(dst->reg.data_type);
- if (instruction->handler_idx == VKD3DSIH_DP4)
+ if (instruction->opcode == VKD3DSIH_DP4)
write_mask = VKD3DSP_WRITEMASK_ALL;
- else if (instruction->handler_idx == VKD3DSIH_DP3)
+ else if (instruction->opcode == VKD3DSIH_DP3)
write_mask = VKD3DSP_WRITEMASK_0 | VKD3DSP_WRITEMASK_1 | VKD3DSP_WRITEMASK_2;
else
write_mask = VKD3DSP_WRITEMASK_0 | VKD3DSP_WRITEMASK_1;
@@ -7606,8 +7622,8 @@ static void spirv_compiler_emit_int_div(struct spirv_compiler *compiler,
unsigned int component_count = 0;
SpvOp div_op, mod_op;
- div_op = instruction->handler_idx == VKD3DSIH_IDIV ? SpvOpSDiv : SpvOpUDiv;
- mod_op = instruction->handler_idx == VKD3DSIH_IDIV ? SpvOpSRem : SpvOpUMod;
+ div_op = instruction->opcode == VKD3DSIH_IDIV ? SpvOpSDiv : SpvOpUDiv;
+ mod_op = instruction->opcode == VKD3DSIH_IDIV ? SpvOpSRem : SpvOpUMod;
if (dst[0].reg.type != VKD3DSPR_NULL)
{
@@ -7778,13 +7794,13 @@ static void spirv_compiler_emit_bitfield_instruction(struct spirv_compiler *comp
mask_id = spirv_compiler_get_constant_uint(compiler, size - 1);
size_id = spirv_compiler_get_constant_uint(compiler, size);
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_BFI: op = SpvOpBitFieldInsert; break;
case VKD3DSIH_IBFE: op = SpvOpBitFieldSExtract; break;
case VKD3DSIH_UBFE: op = SpvOpBitFieldUExtract; break;
default:
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -7895,7 +7911,7 @@ static void spirv_compiler_emit_comparison_instruction(struct spirv_compiler *co
unsigned int component_count;
SpvOp op;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_DEQO:
case VKD3DSIH_EQO: op = SpvOpFOrdEqual; break;
@@ -7916,7 +7932,7 @@ static void spirv_compiler_emit_comparison_instruction(struct spirv_compiler *co
case VKD3DSIH_UGE: op = SpvOpUGreaterThanEqual; break;
case VKD3DSIH_ULT: op = SpvOpULessThan; break;
default:
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -7949,7 +7965,7 @@ static void spirv_compiler_emit_orderedness_instruction(struct spirv_compiler *c
src0_id = vkd3d_spirv_build_op_is_nan(builder, type_id, src0_id);
src1_id = vkd3d_spirv_build_op_is_nan(builder, type_id, src1_id);
val_id = vkd3d_spirv_build_op_logical_or(builder, type_id, src0_id, src1_id);
- if (instruction->handler_idx == VKD3DSIH_ORD)
+ if (instruction->opcode == VKD3DSIH_ORD)
val_id = vkd3d_spirv_build_op_logical_not(builder, type_id, val_id);
spirv_compiler_emit_store_dst(compiler, dst, val_id);
}
@@ -7964,7 +7980,7 @@ static void spirv_compiler_emit_float_comparison_instruction(struct spirv_compil
unsigned int component_count;
SpvOp op;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_SLT: op = SpvOpFOrdLessThan; break;
case VKD3DSIH_SGE: op = SpvOpFOrdGreaterThanEqual; break;
@@ -8262,7 +8278,7 @@ static void spirv_compiler_emit_deriv_instruction(struct spirv_compiler *compile
static const struct instruction_info
{
- enum vkd3d_shader_opcode handler_idx;
+ enum vkd3d_shader_opcode opcode;
SpvOp op;
bool needs_derivative_control;
}
@@ -8279,7 +8295,7 @@ static void spirv_compiler_emit_deriv_instruction(struct spirv_compiler *compile
info = NULL;
for (i = 0; i < ARRAY_SIZE(deriv_instructions); ++i)
{
- if (deriv_instructions[i].handler_idx == instruction->handler_idx)
+ if (deriv_instructions[i].opcode == instruction->opcode)
{
info = &deriv_instructions[i];
break;
@@ -8287,7 +8303,7 @@ static void spirv_compiler_emit_deriv_instruction(struct spirv_compiler *compile
}
if (!info)
{
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -8497,7 +8513,7 @@ static void spirv_compiler_emit_ld(struct spirv_compiler *compiler,
uint32_t coordinate_mask;
bool multisample;
- multisample = instruction->handler_idx == VKD3DSIH_LD2DMS;
+ multisample = instruction->opcode == VKD3DSIH_LD2DMS;
spirv_compiler_prepare_image(compiler, &image, &src[1].reg, NULL, VKD3D_IMAGE_FLAG_NONE);
@@ -8576,7 +8592,7 @@ static void spirv_compiler_emit_sample(struct spirv_compiler *compiler,
spirv_compiler_prepare_image(compiler, &image,
&resource->reg, &sampler->reg, VKD3D_IMAGE_FLAG_SAMPLED);
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_SAMPLE:
op = SpvOpImageSampleImplicitLod;
@@ -8603,7 +8619,7 @@ static void spirv_compiler_emit_sample(struct spirv_compiler *compiler,
&src[3], VKD3DSP_WRITEMASK_0);
break;
default:
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -8637,7 +8653,7 @@ static void spirv_compiler_emit_sample_c(struct spirv_compiler *compiler,
uint32_t image_operands[2];
SpvOp op;
- if (instruction->handler_idx == VKD3DSIH_SAMPLE_C_LZ)
+ if (instruction->opcode == VKD3DSIH_SAMPLE_C_LZ)
{
op = SpvOpImageSampleDrefExplicitLod;
operands_mask |= SpvImageOperandsLodMask;
@@ -8687,12 +8703,12 @@ static void spirv_compiler_emit_gather4(struct spirv_compiler *compiler,
uint32_t coordinate_mask;
bool extended_offset;
- if (instruction->handler_idx == VKD3DSIH_GATHER4_C
- || instruction->handler_idx == VKD3DSIH_GATHER4_PO_C)
+ if (instruction->opcode == VKD3DSIH_GATHER4_C
+ || instruction->opcode == VKD3DSIH_GATHER4_PO_C)
image_flags |= VKD3D_IMAGE_FLAG_DEPTH;
- extended_offset = instruction->handler_idx == VKD3DSIH_GATHER4_PO
- || instruction->handler_idx == VKD3DSIH_GATHER4_PO_C;
+ extended_offset = instruction->opcode == VKD3DSIH_GATHER4_PO
+ || instruction->opcode == VKD3DSIH_GATHER4_PO_C;
addr = &src[0];
offset = extended_offset ? &src[1] : NULL;
@@ -8963,7 +8979,6 @@ static void spirv_compiler_emit_store_uav_raw_structured(struct spirv_compiler *
{
type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1);
spirv_compiler_prepare_image(compiler, &image, &dst->reg, NULL, VKD3D_IMAGE_FLAG_NONE);
- assert((instruction->handler_idx == VKD3DSIH_STORE_STRUCTURED) != !image.structure_stride);
base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler,
type_id, image.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0);
@@ -9007,7 +9022,6 @@ static void spirv_compiler_emit_store_tgsm(struct spirv_compiler *compiler,
type_id = vkd3d_spirv_get_type_id(builder, VKD3D_SHADER_COMPONENT_UINT, 1);
ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, reg_info.storage_class, type_id);
- assert((instruction->handler_idx == VKD3DSIH_STORE_STRUCTURED) != !reg_info.structure_stride);
base_coordinate_id = spirv_compiler_emit_raw_structured_addressing(compiler,
type_id, reg_info.structure_stride, &src[0], VKD3DSP_WRITEMASK_0, &src[1], VKD3DSP_WRITEMASK_0);
@@ -9145,7 +9159,7 @@ static void spirv_compiler_emit_uav_counter_instruction(struct spirv_compiler *c
uint32_t operands[3];
SpvOp op;
- op = instruction->handler_idx == VKD3DSIH_IMM_ATOMIC_ALLOC
+ op = instruction->opcode == VKD3DSIH_IMM_ATOMIC_ALLOC
? SpvOpAtomicIIncrement : SpvOpAtomicIDecrement;
resource_symbol = spirv_compiler_find_resource(compiler, &src->reg);
@@ -9211,7 +9225,7 @@ static SpvOp spirv_compiler_map_atomic_instruction(const struct vkd3d_shader_ins
{
static const struct
{
- enum vkd3d_shader_opcode handler_idx;
+ enum vkd3d_shader_opcode opcode;
SpvOp spirv_op;
}
atomic_ops[] =
@@ -9240,16 +9254,16 @@ static SpvOp spirv_compiler_map_atomic_instruction(const struct vkd3d_shader_ins
for (i = 0; i < ARRAY_SIZE(atomic_ops); ++i)
{
- if (atomic_ops[i].handler_idx == instruction->handler_idx)
+ if (atomic_ops[i].opcode == instruction->opcode)
return atomic_ops[i].spirv_op;
}
return SpvOpMax;
}
-static bool is_imm_atomic_instruction(enum vkd3d_shader_opcode handler_idx)
+static bool is_imm_atomic_instruction(enum vkd3d_shader_opcode opcode)
{
- return VKD3DSIH_IMM_ATOMIC_ALLOC <= handler_idx && handler_idx <= VKD3DSIH_IMM_ATOMIC_XOR;
+ return VKD3DSIH_IMM_ATOMIC_ALLOC <= opcode && opcode <= VKD3DSIH_IMM_ATOMIC_XOR;
}
static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compiler,
@@ -9274,12 +9288,12 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil
bool raw;
SpvOp op;
- resource = is_imm_atomic_instruction(instruction->handler_idx) ? &dst[1] : &dst[0];
+ resource = is_imm_atomic_instruction(instruction->opcode) ? &dst[1] : &dst[0];
op = spirv_compiler_map_atomic_instruction(instruction);
if (op == SpvOpMax)
{
- ERR("Unexpected instruction %#x.\n", instruction->handler_idx);
+ ERR("Unexpected instruction %#x.\n", instruction->opcode);
return;
}
@@ -9360,7 +9374,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil
{
WARN("Ignoring 'volatile' attribute.\n");
spirv_compiler_warning(compiler, VKD3D_SHADER_WARNING_SPV_IGNORING_FLAG,
- "Ignoring the 'volatile' attribute flag for atomic instruction %#x.", instruction->handler_idx);
+ "Ignoring the 'volatile' attribute flag for atomic instruction %#x.", instruction->opcode);
}
memory_semantic = (instruction->flags & VKD3DARF_SEQ_CST)
@@ -9379,7 +9393,7 @@ static void spirv_compiler_emit_atomic_instruction(struct spirv_compiler *compil
result_id = vkd3d_spirv_build_op_trv(builder, &builder->function_stream,
op, type_id, operands, i);
- if (is_imm_atomic_instruction(instruction->handler_idx))
+ if (is_imm_atomic_instruction(instruction->opcode))
spirv_compiler_emit_store_dst(compiler, dst, result_id);
}
@@ -9684,13 +9698,13 @@ static void spirv_compiler_emit_eval_attrib(struct spirv_compiler *compiler,
src_ids[src_count++] = register_info.id;
- if (instruction->handler_idx == VKD3DSIH_EVAL_CENTROID)
+ if (instruction->opcode == VKD3DSIH_EVAL_CENTROID)
{
op = GLSLstd450InterpolateAtCentroid;
}
else
{
- assert(instruction->handler_idx == VKD3DSIH_EVAL_SAMPLE_INDEX);
+ assert(instruction->opcode == VKD3DSIH_EVAL_SAMPLE_INDEX);
op = GLSLstd450InterpolateAtSample;
src_ids[src_count++] = spirv_compiler_emit_load_src(compiler, &src[1], VKD3DSP_WRITEMASK_0);
}
@@ -9772,7 +9786,7 @@ static void spirv_compiler_emit_emit_stream(struct spirv_compiler *compiler,
struct vkd3d_spirv_builder *builder = &compiler->spirv_builder;
unsigned int stream_idx;
- if (instruction->handler_idx == VKD3DSIH_EMIT_STREAM)
+ if (instruction->opcode == VKD3DSIH_EMIT_STREAM)
stream_idx = instruction->src[0].reg.idx[0].offset;
else
stream_idx = 0;
@@ -9793,7 +9807,7 @@ static void spirv_compiler_emit_cut_stream(struct spirv_compiler *compiler,
struct vkd3d_spirv_builder *builder = &compiler->spirv_builder;
unsigned int stream_idx;
- if (instruction->handler_idx == VKD3DSIH_CUT_STREAM)
+ if (instruction->opcode == VKD3DSIH_CUT_STREAM)
stream_idx = instruction->src[0].reg.idx[0].offset;
else
stream_idx = 0;
@@ -9807,9 +9821,68 @@ static void spirv_compiler_emit_cut_stream(struct spirv_compiler *compiler,
vkd3d_spirv_build_op_end_primitive(builder);
}
-static SpvOp map_wave_bool_op(enum vkd3d_shader_opcode handler_idx)
+static uint32_t map_quad_read_across_direction(enum vkd3d_shader_opcode opcode)
{
- switch (handler_idx)
+ switch (opcode)
+ {
+ case VKD3DSIH_QUAD_READ_ACROSS_X:
+ return 0;
+ case VKD3DSIH_QUAD_READ_ACROSS_Y:
+ return 1;
+ case VKD3DSIH_QUAD_READ_ACROSS_D:
+ return 2;
+ default:
+ vkd3d_unreachable();
+ }
+}
+
+static void spirv_compiler_emit_quad_read_across(struct spirv_compiler *compiler,
+ const struct vkd3d_shader_instruction *instruction)
+{
+ struct vkd3d_spirv_builder *builder = &compiler->spirv_builder;
+ const struct vkd3d_shader_dst_param *dst = instruction->dst;
+ const struct vkd3d_shader_src_param *src = instruction->src;
+ uint32_t type_id, direction_type_id, direction_id, val_id;
+
+ type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type,
+ vsir_write_mask_component_count(dst->write_mask));
+ direction_type_id = vkd3d_spirv_get_type_id_for_data_type(builder, VKD3D_DATA_UINT, 1);
+ val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask);
+ direction_id = map_quad_read_across_direction(instruction->opcode);
+ direction_id = vkd3d_spirv_get_op_constant(builder, direction_type_id, direction_id);
+ val_id = vkd3d_spirv_build_op_group_nonuniform_quad_swap(builder, type_id, val_id, direction_id);
+
+ spirv_compiler_emit_store_dst(compiler, dst, val_id);
+}
+
+static void spirv_compiler_emit_quad_read_lane_at(struct spirv_compiler *compiler,
+ const struct vkd3d_shader_instruction *instruction)
+{
+ struct vkd3d_spirv_builder *builder = &compiler->spirv_builder;
+ const struct vkd3d_shader_dst_param *dst = instruction->dst;
+ const struct vkd3d_shader_src_param *src = instruction->src;
+ uint32_t type_id, val_id, lane_id;
+
+ if (!register_is_constant_or_undef(&src[1].reg))
+ {
+ FIXME("Unsupported non-constant quad read lane index.\n");
+ spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_NOT_IMPLEMENTED,
+ "Non-constant quad read lane indices are not supported.");
+ return;
+ }
+
+ type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type,
+ vsir_write_mask_component_count(dst->write_mask));
+ val_id = spirv_compiler_emit_load_src(compiler, &src[0], dst->write_mask);
+ lane_id = spirv_compiler_emit_load_src(compiler, &src[1], VKD3DSP_WRITEMASK_0);
+ val_id = vkd3d_spirv_build_op_group_nonuniform_quad_broadcast(builder, type_id, val_id, lane_id);
+
+ spirv_compiler_emit_store_dst(compiler, dst, val_id);
+}
+
+static SpvOp map_wave_bool_op(enum vkd3d_shader_opcode opcode)
+{
+ switch (opcode)
{
case VKD3DSIH_WAVE_ACTIVE_ALL_EQUAL:
return SpvOpGroupNonUniformAllEqual;
@@ -9833,7 +9906,7 @@ static void spirv_compiler_emit_wave_bool_op(struct spirv_compiler *compiler,
vkd3d_spirv_enable_capability(builder, SpvCapabilityGroupNonUniformVote);
- op = map_wave_bool_op(instruction->handler_idx);
+ op = map_wave_bool_op(instruction->opcode);
type_id = vkd3d_spirv_get_op_type_bool(builder);
val_id = spirv_compiler_emit_load_src(compiler, src, dst->write_mask);
val_id = vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, op,
@@ -9865,9 +9938,9 @@ static void spirv_compiler_emit_wave_active_ballot(struct spirv_compiler *compil
spirv_compiler_emit_store_dst(compiler, dst, val_id);
}
-static SpvOp map_wave_alu_op(enum vkd3d_shader_opcode handler_idx, bool is_float)
+static SpvOp map_wave_alu_op(enum vkd3d_shader_opcode opcode, bool is_float)
{
- switch (handler_idx)
+ switch (opcode)
{
case VKD3DSIH_WAVE_ACTIVE_BIT_AND:
return SpvOpGroupNonUniformBitwiseAnd;
@@ -9905,7 +9978,7 @@ static void spirv_compiler_emit_wave_alu_op(struct spirv_compiler *compiler,
uint32_t type_id, val_id;
SpvOp op;
- op = map_wave_alu_op(instruction->handler_idx, data_type_is_floating_point(src->reg.data_type));
+ op = map_wave_alu_op(instruction->opcode, data_type_is_floating_point(src->reg.data_type));
type_id = vkd3d_spirv_get_type_id_for_data_type(builder, dst->reg.data_type,
vsir_write_mask_component_count(dst->write_mask));
@@ -9928,7 +10001,7 @@ static void spirv_compiler_emit_wave_bit_count(struct spirv_compiler *compiler,
SpvGroupOperation group_op;
uint32_t type_id, val_id;
- group_op = (instruction->handler_idx == VKD3DSIH_WAVE_PREFIX_BIT_COUNT) ? SpvGroupOperationExclusiveScan
+ group_op = (instruction->opcode == VKD3DSIH_WAVE_PREFIX_BIT_COUNT) ? SpvGroupOperationExclusiveScan
: SpvGroupOperationReduce;
val_id = spirv_compiler_emit_group_nonuniform_ballot(compiler, instruction->src);
@@ -10014,7 +10087,7 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler,
compiler->location = instruction->location;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_DCL_GLOBAL_FLAGS:
spirv_compiler_emit_dcl_global_flags(compiler, instruction);
@@ -10337,6 +10410,14 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler,
case VKD3DSIH_CUT_STREAM:
spirv_compiler_emit_cut_stream(compiler, instruction);
break;
+ case VKD3DSIH_QUAD_READ_ACROSS_D:
+ case VKD3DSIH_QUAD_READ_ACROSS_X:
+ case VKD3DSIH_QUAD_READ_ACROSS_Y:
+ spirv_compiler_emit_quad_read_across(compiler, instruction);
+ break;
+ case VKD3DSIH_QUAD_READ_LANE_AT:
+ spirv_compiler_emit_quad_read_lane_at(compiler, instruction);
+ break;
case VKD3DSIH_WAVE_ACTIVE_ALL_EQUAL:
case VKD3DSIH_WAVE_ALL_TRUE:
case VKD3DSIH_WAVE_ANY_TRUE:
@@ -10390,9 +10471,9 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler,
/* nothing to do */
break;
default:
- FIXME("Unhandled instruction %#x.\n", instruction->handler_idx);
+ FIXME("Unhandled instruction %#x.\n", instruction->opcode);
spirv_compiler_error(compiler, VKD3D_SHADER_ERROR_SPV_INVALID_HANDLER,
- "Encountered invalid/unhandled instruction handler %#x.", instruction->handler_idx);
+ "Encountered invalid/unhandled instruction handler %#x.", instruction->opcode);
break;
}
diff --git a/libs/vkd3d/libs/vkd3d-shader/tpf.c b/libs/vkd3d/libs/vkd3d-shader/tpf.c
index b562e815a81..24206ae9a4d 100644
--- a/libs/vkd3d/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d/libs/vkd3d-shader/tpf.c
@@ -780,7 +780,7 @@ static void shader_sm4_read_shader_data(struct vkd3d_shader_instruction *ins, ui
if (type != VKD3D_SM4_SHADER_DATA_IMMEDIATE_CONSTANT_BUFFER)
{
FIXME("Ignoring shader data type %#x.\n", type);
- ins->handler_idx = VKD3DSIH_NOP;
+ ins->opcode = VKD3DSIH_NOP;
return;
}
@@ -789,7 +789,7 @@ static void shader_sm4_read_shader_data(struct vkd3d_shader_instruction *ins, ui
if (icb_size % 4)
{
FIXME("Unexpected immediate constant buffer size %u.\n", icb_size);
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
@@ -797,7 +797,7 @@ static void shader_sm4_read_shader_data(struct vkd3d_shader_instruction *ins, ui
{
ERR("Failed to allocate immediate constant buffer, size %u.\n", icb_size);
vkd3d_shader_parser_error(&priv->p, VKD3D_SHADER_ERROR_TPF_OUT_OF_MEMORY, "Out of memory.");
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
icb->register_idx = 0;
@@ -2395,16 +2395,16 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
if (!(opcode_info = get_info_from_sm4_opcode(&sm4->lookup, opcode)))
{
FIXME("Unrecognized opcode %#x, opcode_token 0x%08x.\n", opcode, opcode_token);
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
*ptr += len;
return;
}
vsir_instruction_init(ins, &sm4->p.location, opcode_info->handler_idx);
- if (ins->handler_idx == VKD3DSIH_HS_CONTROL_POINT_PHASE || ins->handler_idx == VKD3DSIH_HS_FORK_PHASE
- || ins->handler_idx == VKD3DSIH_HS_JOIN_PHASE)
- sm4->phase = ins->handler_idx;
- sm4->has_control_point_phase |= ins->handler_idx == VKD3DSIH_HS_CONTROL_POINT_PHASE;
+ if (ins->opcode == VKD3DSIH_HS_CONTROL_POINT_PHASE || ins->opcode == VKD3DSIH_HS_FORK_PHASE
+ || ins->opcode == VKD3DSIH_HS_JOIN_PHASE)
+ sm4->phase = ins->opcode;
+ sm4->has_control_point_phase |= ins->opcode == VKD3DSIH_HS_CONTROL_POINT_PHASE;
ins->flags = 0;
ins->coissue = false;
ins->raw = false;
@@ -2417,7 +2417,7 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
{
ERR("Failed to allocate src parameters.\n");
vkd3d_shader_parser_error(&sm4->p, VKD3D_SHADER_ERROR_TPF_OUT_OF_MEMORY, "Out of memory.");
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
ins->resource_type = VKD3D_SHADER_RESOURCE_NONE;
@@ -2459,7 +2459,7 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
{
ERR("Failed to allocate dst parameters.\n");
vkd3d_shader_parser_error(&sm4->p, VKD3D_SHADER_ERROR_TPF_OUT_OF_MEMORY, "Out of memory.");
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
for (i = 0; i < ins->dst_count; ++i)
@@ -2467,7 +2467,7 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
if (!(shader_sm4_read_dst_param(sm4, &p, *ptr, map_data_type(opcode_info->dst_info[i]),
&dst_params[i])))
{
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
dst_params[i].modifiers |= instruction_dst_modifier;
@@ -2478,7 +2478,7 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
if (!(shader_sm4_read_src_param(sm4, &p, *ptr, map_data_type(opcode_info->src_info[i]),
&src_params[i])))
{
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
}
@@ -2488,7 +2488,7 @@ static void shader_sm4_read_instruction(struct vkd3d_shader_sm4_parser *sm4, str
fail:
*ptr = sm4->end;
- ins->handler_idx = VKD3DSIH_INVALID;
+ ins->opcode = VKD3DSIH_INVALID;
return;
}
@@ -2693,7 +2693,7 @@ int tpf_parse(const struct vkd3d_shader_compile_info *compile_info, uint64_t con
ins = &instructions->elements[instructions->count];
shader_sm4_read_instruction(&sm4, ins);
- if (ins->handler_idx == VKD3DSIH_INVALID)
+ if (ins->opcode == VKD3DSIH_INVALID)
{
WARN("Encountered unrecognized or invalid instruction.\n");
vsir_program_cleanup(program);
@@ -3178,13 +3178,14 @@ 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;
+ const struct hlsl_buffer *buffer;
char *name;
struct hlsl_type *data_type;
bool is_user_packed;
enum hlsl_regset regset;
- unsigned int id, bind_count;
+ unsigned int id, space, index, bind_count;
};
static int sm4_compare_extern_resources(const void *a, const void *b)
@@ -3196,7 +3197,10 @@ static int sm4_compare_extern_resources(const void *a, const void *b)
if ((r = vkd3d_u32_compare(aa->regset, bb->regset)))
return r;
- return vkd3d_u32_compare(aa->id, bb->id);
+ if ((r = vkd3d_u32_compare(aa->space, bb->space)))
+ return r;
+
+ return vkd3d_u32_compare(aa->index, bb->index);
}
static void sm4_free_extern_resources(struct extern_resource *extern_resources, unsigned int count)
@@ -3220,6 +3224,7 @@ static struct extern_resource *sm4_get_extern_resources(struct hlsl_ctx *ctx, un
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;
+ struct hlsl_buffer *buffer;
enum hlsl_regset regset;
size_t capacity = 0;
char *name;
@@ -3272,13 +3277,16 @@ static struct extern_resource *sm4_get_extern_resources(struct hlsl_ctx *ctx, un
hlsl_release_string_buffer(ctx, name_buffer);
extern_resources[*count].var = NULL;
+ extern_resources[*count].buffer = NULL;
extern_resources[*count].name = name;
extern_resources[*count].data_type = component_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 + regset_offset;
+ extern_resources[*count].id = var->regs[regset].id;
+ extern_resources[*count].space = var->regs[regset].space;
+ extern_resources[*count].index = var->regs[regset].index + regset_offset;
extern_resources[*count].bind_count = 1;
++*count;
@@ -3313,13 +3321,19 @@ static struct extern_resource *sm4_get_extern_resources(struct hlsl_ctx *ctx, un
}
extern_resources[*count].var = var;
+ extern_resources[*count].buffer = NULL;
extern_resources[*count].name = name;
extern_resources[*count].data_type = var->data_type;
- extern_resources[*count].is_user_packed = !!var->reg_reservation.reg_type;
+ /* For some reason 5.1 resources aren't marked as
+ * user-packed, but cbuffers still are. */
+ extern_resources[*count].is_user_packed = hlsl_version_lt(ctx, 5, 1)
+ && !!var->reg_reservation.reg_type;
extern_resources[*count].regset = r;
extern_resources[*count].id = var->regs[r].id;
+ extern_resources[*count].space = var->regs[r].space;
+ extern_resources[*count].index = var->regs[r].index;
extern_resources[*count].bind_count = var->bind_count[r];
++*count;
@@ -3327,14 +3341,51 @@ static struct extern_resource *sm4_get_extern_resources(struct hlsl_ctx *ctx, un
}
}
+ LIST_FOR_EACH_ENTRY(buffer, &ctx->buffers, struct hlsl_buffer, entry)
+ {
+ if (!buffer->reg.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, buffer->name)))
+ {
+ sm4_free_extern_resources(extern_resources, *count);
+ *count = 0;
+ return NULL;
+ }
+
+ extern_resources[*count].var = NULL;
+ extern_resources[*count].buffer = buffer;
+
+ extern_resources[*count].name = name;
+ extern_resources[*count].data_type = NULL;
+ extern_resources[*count].is_user_packed = !!buffer->reservation.reg_type;
+
+ extern_resources[*count].regset = HLSL_REGSET_NUMERIC;
+ extern_resources[*count].id = buffer->reg.id;
+ extern_resources[*count].space = buffer->reg.space;
+ extern_resources[*count].index = buffer->reg.index;
+ extern_resources[*count].bind_count = 1;
+
+ ++*count;
+ }
+
qsort(extern_resources, *count, sizeof(*extern_resources), sm4_compare_extern_resources);
return extern_resources;
}
static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
{
- unsigned int cbuffer_count = 0, resource_count = 0, extern_resources_count, i, j;
+ uint32_t binding_desc_size = (hlsl_version_ge(ctx, 5, 1) ? 10 : 8) * sizeof(uint32_t);
size_t cbuffers_offset, resources_offset, creator_offset, string_offset;
+ unsigned int cbuffer_count = 0, extern_resources_count, i, j;
size_t cbuffer_position, resource_position, creator_position;
const struct hlsl_profile_info *profile = ctx->profile;
struct vkd3d_bytecode_buffer buffer = {0};
@@ -3354,19 +3405,15 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
extern_resources = sm4_get_extern_resources(ctx, &extern_resources_count);
- resource_count += extern_resources_count;
LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry)
{
if (cbuffer->reg.allocated)
- {
++cbuffer_count;
- ++resource_count;
- }
}
put_u32(&buffer, cbuffer_count);
cbuffer_position = put_u32(&buffer, 0);
- put_u32(&buffer, resource_count);
+ put_u32(&buffer, extern_resources_count);
resource_position = put_u32(&buffer, 0);
put_u32(&buffer, vkd3d_make_u32(vkd3d_make_u16(profile->minor_version, profile->major_version),
target_types[profile->type]));
@@ -3378,7 +3425,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
put_u32(&buffer, hlsl_version_ge(ctx, 5, 1) ? TAG_RD11_REVERSE : TAG_RD11);
put_u32(&buffer, 15 * sizeof(uint32_t)); /* size of RDEF header including this header */
put_u32(&buffer, 6 * sizeof(uint32_t)); /* size of buffer desc */
- put_u32(&buffer, (hlsl_version_ge(ctx, 5, 1) ? 10 : 8) * sizeof(uint32_t)); /* size of binding desc */
+ put_u32(&buffer, binding_desc_size); /* size of binding desc */
put_u32(&buffer, 10 * sizeof(uint32_t)); /* size of variable desc */
put_u32(&buffer, 9 * sizeof(uint32_t)); /* size of type desc */
put_u32(&buffer, 3 * sizeof(uint32_t)); /* size of member desc */
@@ -3395,21 +3442,15 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
const struct extern_resource *resource = &extern_resources[i];
uint32_t flags = 0;
- if (hlsl_version_ge(ctx, 5, 1))
- hlsl_fixme(ctx, &resource->var->loc, "Shader model 5.1 resource reflection.");
-
if (resource->is_user_packed)
flags |= D3D_SIF_USERPACKED;
put_u32(&buffer, 0); /* name */
- put_u32(&buffer, sm4_resource_type(resource->data_type));
- if (resource->regset == HLSL_REGSET_SAMPLERS)
- {
- put_u32(&buffer, 0);
- put_u32(&buffer, 0);
- put_u32(&buffer, 0);
- }
+ if (resource->buffer)
+ put_u32(&buffer, resource->buffer->type == HLSL_BUFFER_CONSTANT ? D3D_SIT_CBUFFER : D3D_SIT_TBUFFER);
else
+ put_u32(&buffer, sm4_resource_type(resource->data_type));
+ if (resource->regset == HLSL_REGSET_TEXTURES || resource->regset == HLSL_REGSET_UAVS)
{
unsigned int dimx = hlsl_type_get_component_type(ctx, resource->data_type, 0)->e.resource.format->dimx;
@@ -3418,32 +3459,21 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
put_u32(&buffer, ~0u); /* FIXME: multisample count */
flags |= (dimx - 1) << VKD3D_SM4_SIF_TEXTURE_COMPONENTS_SHIFT;
}
- put_u32(&buffer, resource->id);
+ else
+ {
+ put_u32(&buffer, 0);
+ put_u32(&buffer, 0);
+ put_u32(&buffer, 0);
+ }
+ put_u32(&buffer, resource->index);
put_u32(&buffer, resource->bind_count);
put_u32(&buffer, flags);
- }
-
- LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry)
- {
- uint32_t flags = 0;
-
- if (!cbuffer->reg.allocated)
- continue;
if (hlsl_version_ge(ctx, 5, 1))
- hlsl_fixme(ctx, &cbuffer->loc, "Shader model 5.1 resource reflection.");
-
- if (cbuffer->reservation.reg_type)
- flags |= D3D_SIF_USERPACKED;
-
- put_u32(&buffer, 0); /* name */
- put_u32(&buffer, cbuffer->type == HLSL_BUFFER_CONSTANT ? D3D_SIT_CBUFFER : D3D_SIT_TBUFFER);
- put_u32(&buffer, 0); /* return type */
- put_u32(&buffer, 0); /* dimension */
- put_u32(&buffer, 0); /* multisample count */
- put_u32(&buffer, cbuffer->reg.id); /* bind point */
- put_u32(&buffer, 1); /* bind count */
- put_u32(&buffer, flags); /* flags */
+ {
+ put_u32(&buffer, resource->space);
+ put_u32(&buffer, resource->id);
+ }
}
for (i = 0; i < extern_resources_count; ++i)
@@ -3451,16 +3481,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
const struct extern_resource *resource = &extern_resources[i];
string_offset = put_string(&buffer, resource->name);
- set_u32(&buffer, resources_offset + i * 8 * sizeof(uint32_t), string_offset);
- }
-
- LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry)
- {
- if (!cbuffer->reg.allocated)
- continue;
-
- string_offset = put_string(&buffer, cbuffer->name);
- set_u32(&buffer, resources_offset + i++ * 8 * sizeof(uint32_t), string_offset);
+ set_u32(&buffer, resources_offset + i * binding_desc_size, string_offset);
}
/* Buffers. */
@@ -3522,7 +3543,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
put_u32(&buffer, var->data_type->reg_size[HLSL_REGSET_NUMERIC] * sizeof(float));
put_u32(&buffer, flags);
put_u32(&buffer, 0); /* type */
- put_u32(&buffer, 0); /* FIXME: default value */
+ put_u32(&buffer, 0); /* default value */
if (profile->major_version >= 5)
{
@@ -3546,6 +3567,34 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
set_u32(&buffer, var_offset, string_offset);
write_sm4_type(ctx, &buffer, var->data_type);
set_u32(&buffer, var_offset + 4 * sizeof(uint32_t), var->data_type->bytecode_offset);
+
+ if (var->default_values)
+ {
+ unsigned int reg_size = var->data_type->reg_size[HLSL_REGSET_NUMERIC];
+ unsigned int comp_count = hlsl_type_component_count(var->data_type);
+ unsigned int default_value_offset;
+ unsigned int k;
+
+ default_value_offset = bytecode_reserve_bytes(&buffer, reg_size * sizeof(uint32_t));
+ set_u32(&buffer, var_offset + 5 * sizeof(uint32_t), default_value_offset);
+
+ for (k = 0; k < comp_count; ++k)
+ {
+ struct hlsl_type *comp_type = hlsl_type_get_component_type(ctx, var->data_type, k);
+ unsigned int comp_offset;
+ enum hlsl_regset regset;
+
+ comp_offset = hlsl_type_get_component_offset(ctx, var->data_type, k, &regset);
+ if (regset == HLSL_REGSET_NUMERIC)
+ {
+ if (comp_type->e.numeric.type == HLSL_TYPE_DOUBLE)
+ hlsl_fixme(ctx, &var->loc, "Write double default values.");
+
+ set_u32(&buffer, default_value_offset + comp_offset * sizeof(uint32_t),
+ var->default_values[k].value.u);
+ }
+ }
+ }
++j;
}
}
@@ -3720,30 +3769,57 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct vkd3d_shader_re
{
reg->type = VKD3DSPR_RESOURCE;
reg->dimension = VSIR_DIMENSION_VEC4;
- reg->idx[0].offset = var->regs[HLSL_REGSET_TEXTURES].id;
- reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ if (hlsl_version_ge(ctx, 5, 1))
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_TEXTURES].id;
+ reg->idx[1].offset = var->regs[HLSL_REGSET_TEXTURES].index; /* FIXME: array index */
+ reg->idx_count = 2;
+ }
+ else
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_TEXTURES].index;
+ reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ reg->idx_count = 1;
+ }
assert(regset == HLSL_REGSET_TEXTURES);
- reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
else if (regset == HLSL_REGSET_UAVS)
{
reg->type = VKD3DSPR_UAV;
reg->dimension = VSIR_DIMENSION_VEC4;
- reg->idx[0].offset = var->regs[HLSL_REGSET_UAVS].id;
- reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ if (hlsl_version_ge(ctx, 5, 1))
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_UAVS].id;
+ reg->idx[1].offset = var->regs[HLSL_REGSET_UAVS].index; /* FIXME: array index */
+ reg->idx_count = 2;
+ }
+ else
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_UAVS].index;
+ reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ reg->idx_count = 1;
+ }
assert(regset == HLSL_REGSET_UAVS);
- reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
else if (regset == HLSL_REGSET_SAMPLERS)
{
reg->type = VKD3DSPR_SAMPLER;
reg->dimension = VSIR_DIMENSION_NONE;
- reg->idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].id;
- reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ if (hlsl_version_ge(ctx, 5, 1))
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].id;
+ reg->idx[1].offset = var->regs[HLSL_REGSET_SAMPLERS].index; /* FIXME: array index */
+ reg->idx_count = 2;
+ }
+ else
+ {
+ reg->idx[0].offset = var->regs[HLSL_REGSET_SAMPLERS].index;
+ reg->idx[0].offset += hlsl_offset_from_deref_safe(ctx, deref);
+ reg->idx_count = 1;
+ }
assert(regset == HLSL_REGSET_SAMPLERS);
- reg->idx_count = 1;
*writemask = VKD3DSP_WRITEMASK_ALL;
}
else
@@ -3753,9 +3829,19 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct vkd3d_shader_re
assert(data_type->class <= HLSL_CLASS_VECTOR);
reg->type = VKD3DSPR_CONSTBUFFER;
reg->dimension = VSIR_DIMENSION_VEC4;
- reg->idx[0].offset = var->buffer->reg.id;
- reg->idx[1].offset = offset / 4;
- reg->idx_count = 2;
+ if (hlsl_version_ge(ctx, 5, 1))
+ {
+ reg->idx[0].offset = var->buffer->reg.id;
+ reg->idx[1].offset = var->buffer->reg.index; /* FIXME: array index */
+ reg->idx[2].offset = offset / 4;
+ reg->idx_count = 3;
+ }
+ else
+ {
+ reg->idx[0].offset = var->buffer->reg.index;
+ reg->idx[1].offset = offset / 4;
+ reg->idx_count = 2;
+ }
*writemask = ((1u << data_type->dimx) - 1) << (offset & 3);
}
}
@@ -4139,18 +4225,36 @@ static bool encode_texel_offset_as_aoffimmi(struct sm4_instruction *instr,
static void write_sm4_dcl_constant_buffer(const struct tpf_writer *tpf, const struct hlsl_buffer *cbuffer)
{
- const struct sm4_instruction instr =
+ size_t size = (cbuffer->used_size + 3) / 4;
+
+ struct sm4_instruction instr =
{
.opcode = VKD3D_SM4_OP_DCL_CONSTANT_BUFFER,
.srcs[0].reg.dimension = VSIR_DIMENSION_VEC4,
.srcs[0].reg.type = VKD3DSPR_CONSTBUFFER,
- .srcs[0].reg.idx[0].offset = cbuffer->reg.id,
- .srcs[0].reg.idx[1].offset = (cbuffer->used_size + 3) / 4,
- .srcs[0].reg.idx_count = 2,
.srcs[0].swizzle = VKD3D_SHADER_NO_SWIZZLE,
.src_count = 1,
};
+
+ if (hlsl_version_ge(tpf->ctx, 5, 1))
+ {
+ instr.srcs[0].reg.idx[0].offset = cbuffer->reg.id;
+ instr.srcs[0].reg.idx[1].offset = cbuffer->reg.index;
+ instr.srcs[0].reg.idx[2].offset = cbuffer->reg.index; /* FIXME: array end */
+ instr.srcs[0].reg.idx_count = 3;
+
+ instr.idx[0] = size;
+ instr.idx[1] = cbuffer->reg.space;
+ instr.idx_count = 2;
+ }
+ else
+ {
+ instr.srcs[0].reg.idx[0].offset = cbuffer->reg.index;
+ instr.srcs[0].reg.idx[1].offset = size;
+ instr.srcs[0].reg.idx_count = 2;
+ }
+
write_sm4_instruction(tpf, &instr);
}
@@ -4163,7 +4267,6 @@ static void write_sm4_dcl_samplers(const struct tpf_writer *tpf, const struct ex
.opcode = VKD3D_SM4_OP_DCL_SAMPLER,
.dsts[0].reg.type = VKD3DSPR_SAMPLER,
- .dsts[0].reg.idx_count = 1,
.dst_count = 1,
};
@@ -4179,7 +4282,22 @@ static void write_sm4_dcl_samplers(const struct tpf_writer *tpf, const struct ex
if (resource->var && !resource->var->objects_usage[HLSL_REGSET_SAMPLERS][i].used)
continue;
- instr.dsts[0].reg.idx[0].offset = resource->id + i;
+ if (hlsl_version_ge(tpf->ctx, 5, 1))
+ {
+ assert(!i);
+ instr.dsts[0].reg.idx[0].offset = resource->id;
+ instr.dsts[0].reg.idx[1].offset = resource->index;
+ instr.dsts[0].reg.idx[2].offset = resource->index; /* FIXME: array end */
+ instr.dsts[0].reg.idx_count = 3;
+
+ instr.idx[0] = resource->space;
+ instr.idx_count = 1;
+ }
+ else
+ {
+ instr.dsts[0].reg.idx[0].offset = resource->index + i;
+ instr.dsts[0].reg.idx_count = 1;
+ }
write_sm4_instruction(tpf, &instr);
}
}
@@ -4212,6 +4330,23 @@ static void write_sm4_dcl_textures(const struct tpf_writer *tpf, const struct ex
.idx_count = 1,
};
+ if (hlsl_version_ge(tpf->ctx, 5, 1))
+ {
+ assert(!i);
+ instr.dsts[0].reg.idx[0].offset = resource->id;
+ instr.dsts[0].reg.idx[1].offset = resource->index;
+ instr.dsts[0].reg.idx[2].offset = resource->index; /* FIXME: array end */
+ instr.dsts[0].reg.idx_count = 3;
+
+ instr.idx[1] = resource->space;
+ instr.idx_count = 2;
+ }
+ else
+ {
+ instr.dsts[0].reg.idx[0].offset = resource->index + i;
+ instr.dsts[0].reg.idx_count = 1;
+ }
+
if (uav)
{
switch (resource->data_type->sampler_dim)
@@ -5799,21 +5934,13 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx,
LIST_FOR_EACH_ENTRY(cbuffer, &ctx->buffers, struct hlsl_buffer, entry)
{
if (cbuffer->reg.allocated)
- {
- if (hlsl_version_ge(ctx, 5, 1))
- hlsl_fixme(ctx, &cbuffer->loc, "Shader model 5.1 resource definition.");
-
write_sm4_dcl_constant_buffer(&tpf, cbuffer);
- }
}
for (i = 0; i < extern_resources_count; ++i)
{
const struct extern_resource *resource = &extern_resources[i];
- if (hlsl_version_ge(ctx, 5, 1))
- hlsl_fixme(ctx, &resource->var->loc, "Shader model 5.1 resource declaration.");
-
if (resource->regset == HLSL_REGSET_SAMPLERS)
write_sm4_dcl_samplers(&tpf, resource);
else if (resource->regset == HLSL_REGSET_TEXTURES)
@@ -5875,7 +6002,7 @@ static void write_sm4_sfi0(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc)
extern_resources = sm4_get_extern_resources(ctx, &extern_resources_count);
for (unsigned int i = 0; i < extern_resources_count; ++i)
{
- if (extern_resources[i].data_type->e.resource.rasteriser_ordered)
+ if (extern_resources[i].data_type && extern_resources[i].data_type->e.resource.rasteriser_ordered)
*flags |= VKD3D_SM4_REQUIRES_ROVS;
}
sm4_free_extern_resources(extern_resources, extern_resources_count);
diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
index 14a3fa778e5..fdbde019111 100644
--- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
+++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_main.c
@@ -23,6 +23,8 @@
#include <stdio.h>
#include <math.h>
+/* VKD3D_DEBUG_ENV_NAME("VKD3D_SHADER_DEBUG"); */
+
static inline int char_to_int(char c)
{
if ('0' <= c && c <= '9')
@@ -847,12 +849,13 @@ static void vkd3d_shader_scan_add_uav_flag(const struct vkd3d_shader_scan_contex
static bool vkd3d_shader_instruction_is_uav_read(const struct vkd3d_shader_instruction *instruction)
{
- enum vkd3d_shader_opcode handler_idx = instruction->handler_idx;
- return (VKD3DSIH_ATOMIC_AND <= handler_idx && handler_idx <= VKD3DSIH_ATOMIC_XOR)
- || (VKD3DSIH_IMM_ATOMIC_ALLOC <= handler_idx && handler_idx <= VKD3DSIH_IMM_ATOMIC_XOR)
- || handler_idx == VKD3DSIH_LD_UAV_TYPED
- || (handler_idx == VKD3DSIH_LD_RAW && instruction->src[1].reg.type == VKD3DSPR_UAV)
- || (handler_idx == VKD3DSIH_LD_STRUCTURED && instruction->src[2].reg.type == VKD3DSPR_UAV);
+ enum vkd3d_shader_opcode opcode = instruction->opcode;
+
+ return (VKD3DSIH_ATOMIC_AND <= opcode && opcode <= VKD3DSIH_ATOMIC_XOR)
+ || (VKD3DSIH_IMM_ATOMIC_ALLOC <= opcode && opcode <= VKD3DSIH_IMM_ATOMIC_XOR)
+ || opcode == VKD3DSIH_LD_UAV_TYPED
+ || (opcode == VKD3DSIH_LD_RAW && instruction->src[1].reg.type == VKD3DSPR_UAV)
+ || (opcode == VKD3DSIH_LD_STRUCTURED && instruction->src[2].reg.type == VKD3DSPR_UAV);
}
static void vkd3d_shader_scan_record_uav_read(struct vkd3d_shader_scan_context *context,
@@ -863,9 +866,9 @@ static void vkd3d_shader_scan_record_uav_read(struct vkd3d_shader_scan_context *
static bool vkd3d_shader_instruction_is_uav_counter(const struct vkd3d_shader_instruction *instruction)
{
- enum vkd3d_shader_opcode handler_idx = instruction->handler_idx;
- return handler_idx == VKD3DSIH_IMM_ATOMIC_ALLOC
- || handler_idx == VKD3DSIH_IMM_ATOMIC_CONSUME;
+ enum vkd3d_shader_opcode opcode = instruction->opcode;
+
+ return opcode == VKD3DSIH_IMM_ATOMIC_ALLOC || opcode == VKD3DSIH_IMM_ATOMIC_CONSUME;
}
static void vkd3d_shader_scan_record_uav_counter(struct vkd3d_shader_scan_context *context,
@@ -876,9 +879,10 @@ static void vkd3d_shader_scan_record_uav_counter(struct vkd3d_shader_scan_contex
static bool vkd3d_shader_instruction_is_uav_atomic_op(const struct vkd3d_shader_instruction *instruction)
{
- enum vkd3d_shader_opcode handler_idx = instruction->handler_idx;
- return (VKD3DSIH_ATOMIC_AND <= handler_idx && handler_idx <= VKD3DSIH_ATOMIC_XOR)
- || (VKD3DSIH_IMM_ATOMIC_ALLOC <= handler_idx && handler_idx <= VKD3DSIH_IMM_ATOMIC_XOR);
+ enum vkd3d_shader_opcode opcode = instruction->opcode;
+
+ return (VKD3DSIH_ATOMIC_AND <= opcode && opcode <= VKD3DSIH_ATOMIC_XOR)
+ || (VKD3DSIH_IMM_ATOMIC_ALLOC <= opcode && opcode <= VKD3DSIH_IMM_ATOMIC_XOR);
}
static void vkd3d_shader_scan_record_uav_atomic_op(struct vkd3d_shader_scan_context *context,
@@ -1130,7 +1134,7 @@ static int vkd3d_shader_scan_instruction(struct vkd3d_shader_scan_context *conte
context->location = instruction->location;
- switch (instruction->handler_idx)
+ switch (instruction->opcode)
{
case VKD3DSIH_DCL_CONSTANT_BUFFER:
vkd3d_shader_scan_constant_buffer_declaration(context, instruction);
diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
index 29b8d6ad022..37e3d31c995 100644
--- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
+++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
@@ -455,6 +455,10 @@ enum vkd3d_shader_opcode
VKD3DSIH_PHASE,
VKD3DSIH_PHI,
VKD3DSIH_POW,
+ VKD3DSIH_QUAD_READ_ACROSS_D,
+ VKD3DSIH_QUAD_READ_ACROSS_X,
+ VKD3DSIH_QUAD_READ_ACROSS_Y,
+ VKD3DSIH_QUAD_READ_LANE_AT,
VKD3DSIH_RCP,
VKD3DSIH_REP,
VKD3DSIH_RESINFO,
@@ -805,6 +809,7 @@ enum vkd3d_tessellator_domain
#define VKD3DSI_NONE 0x0
#define VKD3DSI_TEXLD_PROJECT 0x1
+#define VKD3DSI_TEXLD_BIAS 0x2
#define VKD3DSI_INDEXED_DYNAMIC 0x4
#define VKD3DSI_RESINFO_RCP_FLOAT 0x1
#define VKD3DSI_RESINFO_UINT 0x2
@@ -1189,7 +1194,7 @@ struct vkd3d_shader_location
struct vkd3d_shader_instruction
{
struct vkd3d_shader_location location;
- enum vkd3d_shader_opcode handler_idx;
+ enum vkd3d_shader_opcode opcode;
uint32_t flags;
unsigned int dst_count;
unsigned int src_count;
@@ -1238,8 +1243,8 @@ static inline bool vkd3d_shader_ver_le(const struct vkd3d_shader_version *v, uns
return v->major < major || (v->major == major && v->minor <= minor);
}
-void vsir_instruction_init(struct vkd3d_shader_instruction *ins, const struct vkd3d_shader_location *location,
- enum vkd3d_shader_opcode handler_idx);
+void vsir_instruction_init(struct vkd3d_shader_instruction *ins,
+ const struct vkd3d_shader_location *location, enum vkd3d_shader_opcode opcode);
static inline bool vkd3d_shader_instruction_has_texel_offset(const struct vkd3d_shader_instruction *ins)
{
diff --git a/libs/vkd3d/libs/vkd3d/command.c b/libs/vkd3d/libs/vkd3d/command.c
index 95366d3441b..6c463be8d60 100644
--- a/libs/vkd3d/libs/vkd3d/command.c
+++ b/libs/vkd3d/libs/vkd3d/command.c
@@ -2025,7 +2025,8 @@ static void d3d12_command_list_invalidate_root_parameters(struct d3d12_command_l
static bool vk_barrier_parameters_from_d3d12_resource_state(unsigned int state, unsigned int stencil_state,
const struct d3d12_resource *resource, VkQueueFlags vk_queue_flags, const struct vkd3d_vulkan_info *vk_info,
- VkAccessFlags *access_mask, VkPipelineStageFlags *stage_flags, VkImageLayout *image_layout)
+ VkAccessFlags *access_mask, VkPipelineStageFlags *stage_flags, VkImageLayout *image_layout,
+ struct d3d12_device *device)
{
bool is_swapchain_image = resource && (resource->flags & VKD3D_RESOURCE_PRESENT_STATE_TRANSITION);
VkPipelineStageFlags queue_shader_stages = 0;
@@ -2035,8 +2036,9 @@ static bool vk_barrier_parameters_from_d3d12_resource_state(unsigned int state,
queue_shader_stages |= VK_PIPELINE_STAGE_VERTEX_SHADER_BIT
| VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT
| VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT
- | VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT
| VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT;
+ if (device->vk_info.geometry_shaders)
+ queue_shader_stages |= VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT;
}
if (vk_queue_flags & VK_QUEUE_COMPUTE_BIT)
queue_shader_stages |= VK_PIPELINE_STAGE_COMPUTE_SHADER_BIT;
@@ -2054,7 +2056,7 @@ static bool vk_barrier_parameters_from_d3d12_resource_state(unsigned int state,
{
if (resource->present_state != D3D12_RESOURCE_STATE_PRESENT)
return vk_barrier_parameters_from_d3d12_resource_state(resource->present_state, 0,
- resource, vk_queue_flags, vk_info, access_mask, stage_flags, image_layout);
+ resource, vk_queue_flags, vk_info, access_mask, stage_flags, image_layout, device);
*access_mask = VK_ACCESS_MEMORY_READ_BIT;
*stage_flags = VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT;
@@ -2251,7 +2253,8 @@ static void d3d12_command_list_transition_resource_to_initial_state(struct d3d12
VK_IMAGE_LAYOUT_PREINITIALIZED : VK_IMAGE_LAYOUT_UNDEFINED;
if (!vk_barrier_parameters_from_d3d12_resource_state(resource->initial_state, 0,
- resource, list->vk_queue_flags, vk_info, &barrier.dstAccessMask, &dst_stage_mask, &barrier.newLayout))
+ resource, list->vk_queue_flags, vk_info, &barrier.dstAccessMask,
+ &dst_stage_mask, &barrier.newLayout, list->device))
{
FIXME("Unhandled state %#x.\n", resource->initial_state);
return;
@@ -4277,13 +4280,15 @@ static void STDMETHODCALLTYPE d3d12_command_list_ResourceBarrier(ID3D12GraphicsC
}
if (!vk_barrier_parameters_from_d3d12_resource_state(state_before, stencil_state_before,
- resource, list->vk_queue_flags, vk_info, &src_access_mask, &src_stage_mask, &layout_before))
+ resource, list->vk_queue_flags, vk_info, &src_access_mask,
+ &src_stage_mask, &layout_before, list->device))
{
FIXME("Unhandled state %#x.\n", state_before);
continue;
}
if (!vk_barrier_parameters_from_d3d12_resource_state(state_after, stencil_state_after,
- resource, list->vk_queue_flags, vk_info, &dst_access_mask, &dst_stage_mask, &layout_after))
+ resource, list->vk_queue_flags, vk_info, &dst_access_mask,
+ &dst_stage_mask, &layout_after, list->device))
{
FIXME("Unhandled state %#x.\n", state_after);
continue;
@@ -4303,7 +4308,8 @@ static void STDMETHODCALLTYPE d3d12_command_list_ResourceBarrier(ID3D12GraphicsC
resource = unsafe_impl_from_ID3D12Resource(uav->pResource);
vk_barrier_parameters_from_d3d12_resource_state(D3D12_RESOURCE_STATE_UNORDERED_ACCESS, 0,
- resource, list->vk_queue_flags, vk_info, &access_mask, &stage_mask, &image_layout);
+ resource, list->vk_queue_flags, vk_info, &access_mask,
+ &stage_mask, &image_layout, list->device);
src_access_mask = dst_access_mask = access_mask;
src_stage_mask = dst_stage_mask = stage_mask;
layout_before = layout_after = image_layout;
@@ -4814,15 +4820,9 @@ static void STDMETHODCALLTYPE d3d12_command_list_IASetVertexBuffers(ID3D12Graphi
return;
}
- if (!views)
- {
- WARN("NULL \"views\" pointer specified.\n");
- return;
- }
-
for (i = 0; i < view_count; ++i)
{
- if (views[i].BufferLocation)
+ if (views && views[i].BufferLocation)
{
resource = vkd3d_gpu_va_allocator_dereference(gpu_va_allocator, views[i].BufferLocation);
buffers[i] = resource->u.vk_buffer;
@@ -5434,6 +5434,52 @@ static const struct vkd3d_format *vkd3d_fixup_clear_uav_uint_colour(struct d3d12
}
}
+static struct vkd3d_view *create_uint_view(struct d3d12_device *device, const struct vkd3d_resource_view *view,
+ struct d3d12_resource *resource, VkClearColorValue *colour)
+{
+ struct vkd3d_texture_view_desc view_desc;
+ const struct vkd3d_format *uint_format;
+ struct vkd3d_view *uint_view;
+
+ if (!(uint_format = vkd3d_find_uint_format(device, view->format->dxgi_format))
+ && !(uint_format = vkd3d_fixup_clear_uav_uint_colour(device, view->format->dxgi_format, colour)))
+ {
+ ERR("Unhandled format %#x.\n", view->format->dxgi_format);
+ return NULL;
+ }
+
+ if (d3d12_resource_is_buffer(resource))
+ {
+ if (!vkd3d_create_buffer_view(device, VKD3D_DESCRIPTOR_MAGIC_UAV, resource->u.vk_buffer,
+ uint_format, view->info.buffer.offset, view->info.buffer.size, &uint_view))
+ {
+ ERR("Failed to create buffer view.\n");
+ return NULL;
+ }
+
+ return uint_view;
+ }
+
+ memset(&view_desc, 0, sizeof(view_desc));
+ view_desc.view_type = view->info.texture.vk_view_type;
+ view_desc.format = uint_format;
+ view_desc.miplevel_idx = view->info.texture.miplevel_idx;
+ view_desc.miplevel_count = 1;
+ view_desc.layer_idx = view->info.texture.layer_idx;
+ view_desc.layer_count = view->info.texture.layer_count;
+ view_desc.vk_image_aspect = VK_IMAGE_ASPECT_COLOR_BIT;
+ view_desc.usage = VK_IMAGE_USAGE_STORAGE_BIT;
+
+ if (!vkd3d_create_texture_view(device, VKD3D_DESCRIPTOR_MAGIC_UAV,
+ resource->u.vk_image, &view_desc, &uint_view))
+ {
+ ERR("Failed to create image view.\n");
+ return NULL;
+ }
+
+ return uint_view;
+}
+
static void STDMETHODCALLTYPE d3d12_command_list_ClearUnorderedAccessViewUint(ID3D12GraphicsCommandList5 *iface,
D3D12_GPU_DESCRIPTOR_HANDLE gpu_handle, D3D12_CPU_DESCRIPTOR_HANDLE cpu_handle, ID3D12Resource *resource,
const UINT values[4], UINT rect_count, const D3D12_RECT *rects)
@@ -5441,8 +5487,6 @@ static void STDMETHODCALLTYPE d3d12_command_list_ClearUnorderedAccessViewUint(ID
struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList5(iface);
struct vkd3d_view *descriptor, *uint_view = NULL;
struct d3d12_device *device = list->device;
- struct vkd3d_texture_view_desc view_desc;
- const struct vkd3d_format *uint_format;
const struct vkd3d_resource_view *view;
struct d3d12_resource *resource_impl;
VkClearColorValue colour;
@@ -5456,44 +5500,11 @@ static void STDMETHODCALLTYPE d3d12_command_list_ClearUnorderedAccessViewUint(ID
view = &descriptor->v;
memcpy(colour.uint32, values, sizeof(colour.uint32));
- if (view->format->type != VKD3D_FORMAT_TYPE_UINT)
+ if (view->format->type != VKD3D_FORMAT_TYPE_UINT
+ && !(descriptor = uint_view = create_uint_view(device, view, resource_impl, &colour)))
{
- if (!(uint_format = vkd3d_find_uint_format(device, view->format->dxgi_format))
- && !(uint_format = vkd3d_fixup_clear_uav_uint_colour(device, view->format->dxgi_format, &colour)))
- {
- ERR("Unhandled format %#x.\n", view->format->dxgi_format);
- return;
- }
-
- if (d3d12_resource_is_buffer(resource_impl))
- {
- if (!vkd3d_create_buffer_view(device, VKD3D_DESCRIPTOR_MAGIC_UAV, resource_impl->u.vk_buffer,
- uint_format, view->info.buffer.offset, view->info.buffer.size, &uint_view))
- {
- ERR("Failed to create buffer view.\n");
- return;
- }
- }
- else
- {
- memset(&view_desc, 0, sizeof(view_desc));
- view_desc.view_type = view->info.texture.vk_view_type;
- view_desc.format = uint_format;
- view_desc.miplevel_idx = view->info.texture.miplevel_idx;
- view_desc.miplevel_count = 1;
- view_desc.layer_idx = view->info.texture.layer_idx;
- view_desc.layer_count = view->info.texture.layer_count;
- view_desc.vk_image_aspect = VK_IMAGE_ASPECT_COLOR_BIT;
- view_desc.usage = VK_IMAGE_USAGE_STORAGE_BIT;
-
- if (!vkd3d_create_texture_view(device, VKD3D_DESCRIPTOR_MAGIC_UAV, resource_impl->u.vk_image, &view_desc,
- &uint_view))
- {
- ERR("Failed to create image view.\n");
- return;
- }
- }
- descriptor = uint_view;
+ ERR("Failed to create UINT view.\n");
+ return;
}
d3d12_command_list_clear_uav(list, resource_impl, descriptor, &colour, rect_count, rects);
@@ -5507,19 +5518,32 @@ static void STDMETHODCALLTYPE d3d12_command_list_ClearUnorderedAccessViewFloat(I
const float values[4], UINT rect_count, const D3D12_RECT *rects)
{
struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList5(iface);
+ struct vkd3d_view *descriptor, *uint_view = NULL;
+ struct d3d12_device *device = list->device;
+ const struct vkd3d_resource_view *view;
struct d3d12_resource *resource_impl;
VkClearColorValue colour;
- struct vkd3d_view *view;
TRACE("iface %p, gpu_handle %s, cpu_handle %s, resource %p, values %p, rect_count %u, rects %p.\n",
iface, debug_gpu_handle(gpu_handle), debug_cpu_handle(cpu_handle), resource, values, rect_count, rects);
resource_impl = unsafe_impl_from_ID3D12Resource(resource);
- if (!(view = d3d12_desc_from_cpu_handle(cpu_handle)->s.u.view))
+ if (!(descriptor = d3d12_desc_from_cpu_handle(cpu_handle)->s.u.view))
return;
+ view = &descriptor->v;
memcpy(colour.float32, values, sizeof(colour.float32));
- d3d12_command_list_clear_uav(list, resource_impl, view, &colour, rect_count, rects);
+ if (view->format->type == VKD3D_FORMAT_TYPE_SINT
+ && !(descriptor = uint_view = create_uint_view(device, view, resource_impl, &colour)))
+ {
+ ERR("Failed to create UINT view.\n");
+ return;
+ }
+
+ d3d12_command_list_clear_uav(list, resource_impl, descriptor, &colour, rect_count, rects);
+
+ if (uint_view)
+ vkd3d_view_decref(uint_view, device);
}
static void STDMETHODCALLTYPE d3d12_command_list_DiscardResource(ID3D12GraphicsCommandList5 *iface,
diff --git a/libs/vkd3d/libs/vkd3d/device.c b/libs/vkd3d/libs/vkd3d/device.c
index cfc9c5f5ed3..674e46fe5c5 100644
--- a/libs/vkd3d/libs/vkd3d/device.c
+++ b/libs/vkd3d/libs/vkd3d/device.c
@@ -76,6 +76,14 @@ static const char * const required_device_extensions[] =
VK_KHR_SHADER_DRAW_PARAMETERS_EXTENSION_NAME,
};
+/* In general we don't want to enable Vulkan beta extensions, but make an
+ * exception for VK_KHR_portability_subset because we draw no real feature from
+ * it, but it's still useful to be able to develop for MoltenVK without being
+ * spammed with validation errors. */
+#ifndef VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME
+#define VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME "VK_KHR_portability_subset"
+#endif
+
static const struct vkd3d_optional_extension_info optional_device_extensions[] =
{
/* KHR extensions */
@@ -85,6 +93,7 @@ static const struct vkd3d_optional_extension_info optional_device_extensions[] =
VK_EXTENSION(KHR_IMAGE_FORMAT_LIST, KHR_image_format_list),
VK_EXTENSION(KHR_MAINTENANCE2, KHR_maintenance2),
VK_EXTENSION(KHR_MAINTENANCE3, KHR_maintenance3),
+ VK_EXTENSION(KHR_PORTABILITY_SUBSET, KHR_portability_subset),
VK_EXTENSION(KHR_PUSH_DESCRIPTOR, KHR_push_descriptor),
VK_EXTENSION(KHR_SAMPLER_MIRROR_CLAMP_TO_EDGE, KHR_sampler_mirror_clamp_to_edge),
VK_EXTENSION(KHR_TIMELINE_SEMAPHORE, KHR_timeline_semaphore),
@@ -92,7 +101,7 @@ static const struct vkd3d_optional_extension_info optional_device_extensions[] =
VK_EXTENSION(EXT_4444_FORMATS, EXT_4444_formats),
VK_EXTENSION(EXT_CALIBRATED_TIMESTAMPS, EXT_calibrated_timestamps),
VK_EXTENSION(EXT_CONDITIONAL_RENDERING, EXT_conditional_rendering),
- VK_EXTENSION(EXT_DEBUG_MARKER, EXT_debug_marker),
+ VK_DEBUG_EXTENSION(EXT_DEBUG_MARKER, EXT_debug_marker),
VK_EXTENSION(EXT_DEPTH_CLIP_ENABLE, EXT_depth_clip_enable),
VK_EXTENSION(EXT_DESCRIPTOR_INDEXING, EXT_descriptor_indexing),
VK_EXTENSION(EXT_FRAGMENT_SHADER_INTERLOCK, EXT_fragment_shader_interlock),
@@ -1634,6 +1643,7 @@ 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->geometry_shaders = physical_device_info->features2.features.geometryShader;
vulkan_info->sparse_binding = features->sparseBinding;
vulkan_info->sparse_residency_3d = features->sparseResidencyImage3D;
vulkan_info->rasterization_stream = physical_device_info->xfb_properties.transformFeedbackRasterizationStreamSelect;
@@ -3806,7 +3816,8 @@ static HRESULT STDMETHODCALLTYPE d3d12_device_CheckFeatureSupport(ID3D12Device9
return E_INVALIDARG;
}
- data->UnalignedBlockTexturesSupported = FALSE;
+ /* Vulkan does not restrict block texture alignment. */
+ data->UnalignedBlockTexturesSupported = TRUE;
TRACE("Unaligned block texture support %#x.\n", data->UnalignedBlockTexturesSupported);
return S_OK;
diff --git a/libs/vkd3d/libs/vkd3d/resource.c b/libs/vkd3d/libs/vkd3d/resource.c
index c897d9f2c5a..db2d87428bb 100644
--- a/libs/vkd3d/libs/vkd3d/resource.c
+++ b/libs/vkd3d/libs/vkd3d/resource.c
@@ -1809,14 +1809,6 @@ static bool d3d12_resource_validate_texture_format(const D3D12_RESOURCE_DESC1 *d
return false;
}
- if (align(desc->Width, format->block_width) != desc->Width
- || align(desc->Height, format->block_height) != desc->Height)
- {
- WARN("Invalid size %"PRIu64"x%u for block compressed format %#x.\n",
- desc->Width, desc->Height, desc->Format);
- return false;
- }
-
return true;
}
diff --git a/libs/vkd3d/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/libs/vkd3d/vkd3d_private.h
index d1fa866d9e3..a55e967cdfc 100644
--- a/libs/vkd3d/libs/vkd3d/vkd3d_private.h
+++ b/libs/vkd3d/libs/vkd3d/vkd3d_private.h
@@ -123,6 +123,7 @@ struct vkd3d_vulkan_info
bool KHR_image_format_list;
bool KHR_maintenance2;
bool KHR_maintenance3;
+ bool KHR_portability_subset;
bool KHR_push_descriptor;
bool KHR_sampler_mirror_clamp_to_edge;
bool KHR_timeline_semaphore;
@@ -145,6 +146,7 @@ struct vkd3d_vulkan_info
bool rasterization_stream;
bool transform_feedback_queries;
+ bool geometry_shaders;
bool uav_read_without_format;
--
2.43.0
Reference in New Issue View Git Blame Copy Permalink