wine/wine-staging
0
0
Fork 0
mirror of https://gitlab.winehq.org/wine/wine-staging.git synced 2024-09-13 09:17:20 -07:00
Code Issues Packages Projects Releases Wiki Activity

Updated vkd3d-latest patchset

Browse Source 
Squash to previous wine-staging release.
This commit is contained in:
Alistair Leslie-Hughes 2024-07-15 10:04:33 +10:00
parent a442564f05
commit 577974c563
7 changed files with 9107 additions and 8577 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8054
patches/vkd3d-latest/0001-Updated-vkd3d-to-c792114a6a58c7c97abf827d154d7ecd22d.patch Normal file
View File

File diff suppressed because it is too large Load Diff

3888
patches/vkd3d-latest/0001-Updated-vkd3d-to-d3ba810c98b4d2df260a527f74586a0b314.patch
View File

File diff suppressed because it is too large Load Diff

1053
patches/vkd3d-latest/0002-Updated-vkd3d-to-5a53b739959db74e8dcce023a7d49356b90.patch Normal file
View File

File diff suppressed because it is too large Load Diff

206
patches/vkd3d-latest/0002-Updated-vkd3d-to-ccb6150aabc7cce9e26a39366c611f5a7da.patch
View File

@ -1,206 +0,0 @@
From f9d3eae7971ce35994e20546183c7147d33c1d4a Mon Sep 17 00:00:00 2001
From: Alistair Leslie-Hughes <leslie_alistair@hotmail.com>
Date: Sun, 23 Jun 2024 15:40:43 +1000
Subject: [PATCH] Updated vkd3d to ccb6150aabc7cce9e26a39366c611f5a7da789e4.
---
libs/vkd3d/libs/vkd3d-shader/d3dbc.c | 2 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.y | 110 ++++++++++++++++++--
libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c | 4 +-
3 files changed, 108 insertions(+), 8 deletions(-)
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
index 2290385da76..3665b99aed7 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
@@ -2582,7 +2582,7 @@ static void write_sm1_resource_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_
.srcs[0].type = D3DSPR_TEMP,
.srcs[0].reg = coords->reg.id,
- .srcs[0].swizzle = hlsl_swizzle_from_writemask(VKD3DSP_WRITEMASK_ALL),
+ .srcs[0].swizzle = hlsl_swizzle_from_writemask(coords->reg.writemask),
.srcs[1].type = D3DSPR_SAMPLER,
.srcs[1].reg = reg_id,
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.y b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
index a2c89b15e4a..9c75c87d36e 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
@@ -1878,12 +1878,57 @@ static bool invert_swizzle(uint32_t *swizzle, unsigned int *writemask, unsigned
return true;
}
+static bool invert_swizzle_matrix(uint32_t *swizzle, unsigned int *writemask, unsigned int *ret_width)
+{
+ /* swizzle is 8 bits per component, each component is (from LSB) 4 bits X, then 4 bits Y.
+ * components are indexed by their sources. i.e. the first component comes from the first
+ * component of the rhs. */
+ unsigned int i, j, bit = 0, inverted = 0, width, new_writemask = 0, new_swizzle = 0;
+
+ /* First, we filter the swizzle to remove components that aren't enabled by writemask. */
+ for (i = 0; i < 4; ++i)
+ {
+ if (*writemask & (1 << i))
+ {
+ unsigned int s = (*swizzle >> (i * 8)) & 0xff;
+ unsigned int x = s & 0xf, y = (s >> 4) & 0xf;
+ unsigned int idx = x + y * 4;
+ new_swizzle |= s << (bit++ * 8);
+ if (new_writemask & (1 << idx))
+ return false;
+ new_writemask |= 1 << idx;
+ }
+ }
+ width = bit;
+
+ /* Then we invert the swizzle. The resulting swizzle has 2 bits per component, because it's for the
+ * incoming vector. */
+ bit = 0;
+ for (i = 0; i < 16; ++i)
+ {
+ for (j = 0; j < width; ++j)
+ {
+ unsigned int s = (new_swizzle >> (j * 8)) & 0xff;
+ unsigned int x = s & 0xf, y = (s >> 4) & 0xf;
+ unsigned int idx = x + y * 4;
+ if (idx == i)
+ inverted |= j << (bit++ * 2);
+ }
+ }
+
+ *swizzle = inverted;
+ *writemask = new_writemask;
+ *ret_width = width;
+ return true;
+}
+
static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct hlsl_block *block, struct hlsl_ir_node *lhs,
enum parse_assign_op assign_op, struct hlsl_ir_node *rhs)
{
struct hlsl_type *lhs_type = lhs->data_type;
struct hlsl_ir_node *copy;
- unsigned int writemask = 0;
+ unsigned int writemask = 0, width = 0;
+ bool matrix_writemask = false;
if (assign_op == ASSIGN_OP_SUB)
{
@@ -1901,7 +1946,10 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct hlsl_blo
}
if (hlsl_is_numeric_type(lhs_type))
+ {
writemask = (1 << lhs_type->dimx) - 1;
+ width = lhs_type->dimx;
+ }
if (!(rhs = add_implicit_conversion(ctx, block, rhs, lhs_type, &rhs->loc)))
return NULL;
@@ -1918,12 +1966,24 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct hlsl_blo
struct hlsl_ir_swizzle *swizzle = hlsl_ir_swizzle(lhs);
struct hlsl_ir_node *new_swizzle;
uint32_t s = swizzle->swizzle;
- unsigned int width;
- if (lhs->data_type->class == HLSL_CLASS_MATRIX)
- hlsl_fixme(ctx, &lhs->loc, "Matrix assignment with a writemask.");
+ assert(!matrix_writemask);
- if (!invert_swizzle(&s, &writemask, &width))
+ if (swizzle->val.node->data_type->class == HLSL_CLASS_MATRIX)
+ {
+ if (swizzle->val.node->type != HLSL_IR_LOAD && swizzle->val.node->type != HLSL_IR_INDEX)
+ {
+ hlsl_fixme(ctx, &lhs->loc, "Unhandled source of matrix swizzle.");
+ return NULL;
+ }
+ if (!invert_swizzle_matrix(&s, &writemask, &width))
+ {
+ hlsl_error(ctx, &lhs->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_WRITEMASK, "Invalid writemask for matrix.");
+ return NULL;
+ }
+ matrix_writemask = true;
+ }
+ else if (!invert_swizzle(&s, &writemask, &width))
{
hlsl_error(ctx, &lhs->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_WRITEMASK, "Invalid writemask.");
return NULL;
@@ -1971,7 +2031,7 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct hlsl_blo
dim_count = hlsl_sampler_dim_count(resource_type->sampler_dim);
- if (writemask != ((1u << resource_type->e.resource.format->dimx) - 1))
+ if (width != resource_type->e.resource.format->dimx * resource_type->e.resource.format->dimy)
hlsl_error(ctx, &lhs->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_WRITEMASK,
"Resource store expressions must write to all components.");
@@ -1987,12 +2047,50 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct hlsl_blo
hlsl_block_add_instr(block, store);
hlsl_cleanup_deref(&resource_deref);
}
+ else if (matrix_writemask)
+ {
+ struct hlsl_deref deref;
+ unsigned int i, j, k = 0;
+
+ hlsl_init_deref_from_index_chain(ctx, &deref, lhs);
+
+ for (i = 0; i < lhs->data_type->dimy; ++i)
+ {
+ for (j = 0; j < lhs->data_type->dimx; ++j)
+ {
+ struct hlsl_ir_node *load;
+ struct hlsl_block store_block;
+ const unsigned int idx = i * 4 + j;
+ const unsigned int component = i * lhs->data_type->dimx + j;
+
+ if (!(writemask & (1 << idx)))
+ continue;
+
+ if (!(load = hlsl_add_load_component(ctx, block, rhs, k++, &rhs->loc)))
+ {
+ hlsl_cleanup_deref(&deref);
+ return NULL;
+ }
+
+ if (!hlsl_new_store_component(ctx, &store_block, &deref, component, load))
+ {
+ hlsl_cleanup_deref(&deref);
+ return NULL;
+ }
+ hlsl_block_add_block(block, &store_block);
+ }
+ }
+
+ hlsl_cleanup_deref(&deref);
+ }
else if (lhs->type == HLSL_IR_INDEX && hlsl_index_is_noncontiguous(hlsl_ir_index(lhs)))
{
struct hlsl_ir_index *row = hlsl_ir_index(lhs);
struct hlsl_ir_node *mat = row->val.node;
unsigned int i, k = 0;
+ assert(!matrix_writemask);
+
for (i = 0; i < mat->data_type->dimx; ++i)
{
struct hlsl_ir_node *cell, *load, *store, *c;
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
index 09066a6191a..36270b159a5 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
@@ -4188,8 +4188,10 @@ static struct hlsl_reg allocate_numeric_registers_for_type(struct hlsl_ctx *ctx,
{
unsigned int reg_size = type->reg_size[HLSL_REGSET_NUMERIC];
+ /* FIXME: We could potentially pack structs or arrays more efficiently... */
+
if (type->class <= HLSL_CLASS_VECTOR)
- return allocate_register(ctx, allocator, first_write, last_read, reg_size, type->dimx);
+ return allocate_register(ctx, allocator, first_write, last_read, type->dimx, type->dimx);
else
return allocate_range(ctx, allocator, first_write, last_read, reg_size);
}
--
2.43.0

1883
patches/vkd3d-latest/0003-Updated-vkd3d-to-3dc43e8945f68c42268b8d5e43525b9e108.patch
View File

File diff suppressed because it is too large Load Diff

1640
patches/vkd3d-latest/0004-Updated-vkd3d-to-746222b349e9c009ed270fb5ca400497dfb.patch
View File

File diff suppressed because it is too large Load Diff

960
patches/vkd3d-latest/0005-Updated-vkd3d-to-c792114a6a58c7c97abf827d154d7ecd22d.patch
View File

@ -1,960 +0,0 @@
From e940aca803c12bbd55ebe3fb26920373a56a0fab Mon Sep 17 00:00:00 2001
From: Alistair Leslie-Hughes <leslie_alistair@hotmail.com>
Date: Thu, 11 Jul 2024 09:52:56 +1000
Subject: [PATCH] Updated vkd3d to c792114a6a58c7c97abf827d154d7ecd22d81536.
---
libs/vkd3d/libs/vkd3d-shader/d3dbc.c | 4 +-
libs/vkd3d/libs/vkd3d-shader/fx.c | 2 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.c | 13 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.h | 12 +-
libs/vkd3d/libs/vkd3d-shader/hlsl.y | 197 +++++----
libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c | 390 +++++++++++++++---
libs/vkd3d/libs/vkd3d-shader/tpf.c | 29 ++
.../libs/vkd3d-shader/vkd3d_shader_private.h | 1 +
libs/vkd3d/libs/vkd3d/resource.c | 6 +-
9 files changed, 510 insertions(+), 144 deletions(-)
diff --git a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
index ea9fe77532d..4522d56c5c9 100644
--- a/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
+++ b/libs/vkd3d/libs/vkd3d-shader/d3dbc.c
@@ -2670,11 +2670,11 @@ static void d3dbc_write_resource_load(struct d3dbc_compiler *d3dbc, const struct
case HLSL_RESOURCE_SAMPLE_GRAD:
sm1_instr.opcode = D3DSIO_TEXLDD;
- sm1_instr.srcs[2].type = D3DSPR_TEMP;
+ sm1_instr.srcs[2].type = VKD3DSPR_TEMP;
sm1_instr.srcs[2].reg = ddx->reg.id;
sm1_instr.srcs[2].swizzle = hlsl_swizzle_from_writemask(ddx->reg.writemask);
- sm1_instr.srcs[3].type = D3DSPR_TEMP;
+ sm1_instr.srcs[3].type = VKD3DSPR_TEMP;
sm1_instr.srcs[3].reg = ddy->reg.id;
sm1_instr.srcs[3].swizzle = hlsl_swizzle_from_writemask(ddy->reg.writemask);
diff --git a/libs/vkd3d/libs/vkd3d-shader/fx.c b/libs/vkd3d/libs/vkd3d-shader/fx.c
index 0857ebb65d5..bd2ad1290cd 100644
--- a/libs/vkd3d/libs/vkd3d-shader/fx.c
+++ b/libs/vkd3d/libs/vkd3d-shader/fx.c
@@ -830,7 +830,7 @@ 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->raw_name, fx);
+ semantic_offset = semantic->raw_name ? write_string(semantic->raw_name, fx) : 0;
offset = put_u32(buffer, hlsl_sm1_base_type(type));
put_u32(buffer, hlsl_sm1_class(type));
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.c b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
index a157590c97a..acf50869a40 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.c
@@ -1810,7 +1810,8 @@ struct hlsl_ir_node *hlsl_new_jump(struct hlsl_ctx *ctx, enum hlsl_ir_jump_type
}
struct hlsl_ir_node *hlsl_new_loop(struct hlsl_ctx *ctx,
- struct hlsl_block *block, const struct vkd3d_shader_location *loc)
+ struct hlsl_block *block, enum hlsl_ir_loop_unroll_type unroll_type,
+ unsigned int unroll_limit, const struct vkd3d_shader_location *loc)
{
struct hlsl_ir_loop *loop;
@@ -1819,6 +1820,9 @@ struct hlsl_ir_node *hlsl_new_loop(struct hlsl_ctx *ctx,
init_node(&loop->node, HLSL_IR_LOOP, NULL, loc);
hlsl_block_init(&loop->body);
hlsl_block_add_block(&loop->body, block);
+
+ loop->unroll_type = unroll_type;
+ loop->unroll_limit = unroll_limit;
return &loop->node;
}
@@ -1881,9 +1885,7 @@ static struct hlsl_ir_node *map_instr(const struct clone_instr_map *map, struct
return map->instrs[i].dst;
}
- /* The block passed to hlsl_clone_block() should have been free of external
- * references. */
- vkd3d_unreachable();
+ return src;
}
static bool clone_deref(struct hlsl_ctx *ctx, struct clone_instr_map *map,
@@ -1980,7 +1982,7 @@ static struct hlsl_ir_node *clone_loop(struct hlsl_ctx *ctx, struct clone_instr_
if (!clone_block(ctx, &body, &src->body, map))
return NULL;
- if (!(dst = hlsl_new_loop(ctx, &body, &src->node.loc)))
+ if (!(dst = hlsl_new_loop(ctx, &body, src->unroll_type, src->unroll_limit, &src->node.loc)))
{
hlsl_block_cleanup(&body);
return NULL;
@@ -2791,6 +2793,7 @@ const char *debug_hlsl_expr_op(enum hlsl_ir_expr_op op)
static const char *const op_names[] =
{
[HLSL_OP0_VOID] = "void",
+ [HLSL_OP0_RASTERIZER_SAMPLE_COUNT] = "GetRenderTargetSampleCount",
[HLSL_OP1_ABS] = "abs",
[HLSL_OP1_BIT_NOT] = "~",
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.h b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
index a79d2b064cf..5832958712a 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.h
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.h
@@ -602,12 +602,21 @@ struct hlsl_ir_if
struct hlsl_block else_block;
};
+enum hlsl_ir_loop_unroll_type
+{
+ HLSL_IR_LOOP_UNROLL,
+ HLSL_IR_LOOP_FORCE_UNROLL,
+ HLSL_IR_LOOP_FORCE_LOOP
+};
+
struct hlsl_ir_loop
{
struct hlsl_ir_node node;
/* loop condition is stored in the body (as "if (!condition) break;") */
struct hlsl_block body;
unsigned int next_index; /* liveness index of the end of the loop */
+ unsigned int unroll_limit;
+ enum hlsl_ir_loop_unroll_type unroll_type;
};
struct hlsl_ir_switch_case
@@ -629,6 +638,7 @@ struct hlsl_ir_switch
enum hlsl_ir_expr_op
{
HLSL_OP0_VOID,
+ HLSL_OP0_RASTERIZER_SAMPLE_COUNT,
HLSL_OP1_ABS,
HLSL_OP1_BIT_NOT,
@@ -1390,7 +1400,7 @@ bool hlsl_index_chain_has_resource_access(struct hlsl_ir_index *index);
struct hlsl_ir_node *hlsl_new_index(struct hlsl_ctx *ctx, struct hlsl_ir_node *val,
struct hlsl_ir_node *idx, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_loop(struct hlsl_ctx *ctx,
- struct hlsl_block *block, const struct vkd3d_shader_location *loc);
+ struct hlsl_block *block, enum hlsl_ir_loop_unroll_type unroll_type, unsigned int unroll_limit, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_resource_load(struct hlsl_ctx *ctx,
const struct hlsl_resource_load_params *params, const struct vkd3d_shader_location *loc);
struct hlsl_ir_node *hlsl_new_resource_store(struct hlsl_ctx *ctx, const struct hlsl_deref *resource,
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl.y b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
index ed6b41bf403..7b058a65bc1 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl.y
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl.y
@@ -573,12 +573,91 @@ static void check_loop_attributes(struct hlsl_ctx *ctx, const struct parse_attri
hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, "Unroll attribute can't be used with 'fastopt' attribute.");
}
+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;
+ struct hlsl_src src;
+
+ LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry)
+ {
+ switch (node->type)
+ {
+ case HLSL_IR_CONSTANT:
+ case HLSL_IR_EXPR:
+ case HLSL_IR_SWIZZLE:
+ case HLSL_IR_LOAD:
+ case HLSL_IR_INDEX:
+ continue;
+ case HLSL_IR_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_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 ret;
+ hlsl_block_add_block(&expr, block);
+
+ if (!add_implicit_conversion(ctx, &expr, node_from_block(&expr), dst_type, loc))
+ {
+ hlsl_block_cleanup(&expr);
+ return ret;
+ }
+
+ /* Wrap the node into a src to allow the reference to survive the multiple const passes. */
+ hlsl_src_from_node(&src, node_from_block(&expr));
+ hlsl_run_const_passes(ctx, &expr);
+ node = src.node;
+ hlsl_src_remove(&src);
+
+ if (node->type == HLSL_IR_CONSTANT)
+ {
+ constant = hlsl_ir_constant(node);
+ ret = constant->value.u[0];
+ }
+ else
+ {
+ hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
+ "Failed to evaluate constant expression.");
+ }
+
+ hlsl_block_cleanup(&expr);
+
+ 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 struct hlsl_block *create_loop(struct hlsl_ctx *ctx, enum loop_type type,
const struct parse_attribute_list *attributes, struct hlsl_block *init, struct hlsl_block *cond,
struct hlsl_block *iter, struct hlsl_block *body, const struct vkd3d_shader_location *loc)
{
+ enum hlsl_ir_loop_unroll_type unroll_type = HLSL_IR_LOOP_UNROLL;
+ unsigned int i, unroll_limit = 0;
struct hlsl_ir_node *loop;
- unsigned int i;
if (attribute_list_has_duplicates(attributes))
hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, "Found duplicate attribute.");
@@ -591,18 +670,29 @@ static struct hlsl_block *create_loop(struct hlsl_ctx *ctx, enum loop_type type,
const struct hlsl_attribute *attr = attributes->attrs[i];
if (!strcmp(attr->name, "unroll"))
{
- if (attr->args_count)
+ if (attr->args_count > 1)
{
- hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED, "Unroll attribute with iteration count.");
+ hlsl_warning(ctx, &attr->loc, VKD3D_SHADER_WARNING_HLSL_IGNORED_ATTRIBUTE,
+ "Ignoring 'unroll' attribute with more than 1 argument.");
+ continue;
}
- else
+
+ if (attr->args_count == 1)
{
- hlsl_warning(ctx, loc, VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED, "Loop unrolling is not implemented.");
+ struct hlsl_block expr;
+ hlsl_block_init(&expr);
+ if (!hlsl_clone_block(ctx, &expr, &attr->instrs))
+ return NULL;
+
+ unroll_limit = evaluate_static_expression_as_uint(ctx, &expr, loc);
+ hlsl_block_cleanup(&expr);
}
+
+ unroll_type = HLSL_IR_LOOP_FORCE_UNROLL;
}
else if (!strcmp(attr->name, "loop"))
{
- /* TODO: this attribute will be used to disable unrolling, once it's implememented. */
+ unroll_type = HLSL_IR_LOOP_FORCE_LOOP;
}
else if (!strcmp(attr->name, "fastopt")
|| !strcmp(attr->name, "allow_uav_condition"))
@@ -631,7 +721,7 @@ static struct hlsl_block *create_loop(struct hlsl_ctx *ctx, enum loop_type type,
else
list_move_head(&body->instrs, &cond->instrs);
- if (!(loop = hlsl_new_loop(ctx, body, loc)))
+ if (!(loop = hlsl_new_loop(ctx, body, unroll_type, unroll_limit, loc)))
goto oom;
hlsl_block_add_instr(init, loop);
@@ -1320,84 +1410,6 @@ static struct hlsl_block *make_block(struct hlsl_ctx *ctx, struct hlsl_ir_node *
return block;
}
-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;
- struct hlsl_src src;
-
- LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry)
- {
- switch (node->type)
- {
- case HLSL_IR_CONSTANT:
- case HLSL_IR_EXPR:
- case HLSL_IR_SWIZZLE:
- case HLSL_IR_LOAD:
- case HLSL_IR_INDEX:
- continue;
- case HLSL_IR_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_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 ret;
- hlsl_block_add_block(&expr, block);
-
- if (!add_implicit_conversion(ctx, &expr, node_from_block(&expr), dst_type, loc))
- {
- hlsl_block_cleanup(&expr);
- return ret;
- }
-
- /* Wrap the node into a src to allow the reference to survive the multiple const passes. */
- hlsl_src_from_node(&src, node_from_block(&expr));
- hlsl_run_const_passes(ctx, &expr);
- node = src.node;
- hlsl_src_remove(&src);
-
- if (node->type == HLSL_IR_CONSTANT)
- {
- constant = hlsl_ir_constant(node);
- ret = constant->value.u[0];
- }
- else
- {
- hlsl_error(ctx, &node->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX,
- "Failed to evaluate constant expression.");
- }
-
- hlsl_block_cleanup(&expr);
-
- 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 */
@@ -4676,6 +4688,20 @@ static bool intrinsic_d3dcolor_to_ubyte4(struct hlsl_ctx *ctx,
return true;
}
+static bool intrinsic_GetRenderTargetSampleCount(struct hlsl_ctx *ctx,
+ const struct parse_initializer *params, const struct vkd3d_shader_location *loc)
+{
+ struct hlsl_ir_node *operands[HLSL_MAX_OPERANDS] = {0};
+ struct hlsl_ir_node *expr;
+
+ if (!(expr = hlsl_new_expr(ctx, HLSL_OP0_RASTERIZER_SAMPLE_COUNT,
+ operands, hlsl_get_scalar_type(ctx, HLSL_TYPE_UINT), loc)))
+ return false;
+ hlsl_block_add_instr(params->instrs, expr);
+
+ return true;
+}
+
static const struct intrinsic_function
{
const char *name;
@@ -4688,6 +4714,7 @@ intrinsic_functions[] =
{
/* Note: these entries should be kept in alphabetical order. */
{"D3DCOLORtoUBYTE4", 1, true, intrinsic_d3dcolor_to_ubyte4},
+ {"GetRenderTargetSampleCount", 0, true, intrinsic_GetRenderTargetSampleCount},
{"abs", 1, true, intrinsic_abs},
{"acos", 1, true, intrinsic_acos},
{"all", 1, true, intrinsic_all},
diff --git a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
index f9f5c8ed58a..7e4f168675e 100644
--- a/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
+++ b/libs/vkd3d/libs/vkd3d-shader/hlsl_codegen.c
@@ -1981,6 +1981,76 @@ bool hlsl_copy_propagation_execute(struct hlsl_ctx *ctx, struct hlsl_block *bloc
return progress;
}
+enum validation_result
+{
+ DEREF_VALIDATION_OK,
+ DEREF_VALIDATION_OUT_OF_BOUNDS,
+ DEREF_VALIDATION_NOT_CONSTANT,
+};
+
+static enum validation_result validate_component_index_range_from_deref(struct hlsl_ctx *ctx,
+ const struct hlsl_deref *deref)
+{
+ struct hlsl_type *type = deref->var->data_type;
+ unsigned int i;
+
+ for (i = 0; i < deref->path_len; ++i)
+ {
+ struct hlsl_ir_node *path_node = deref->path[i].node;
+ unsigned int idx = 0;
+
+ assert(path_node);
+ if (path_node->type != HLSL_IR_CONSTANT)
+ return DEREF_VALIDATION_NOT_CONSTANT;
+
+ /* We should always have generated a cast to UINT. */
+ assert(path_node->data_type->class == HLSL_CLASS_SCALAR
+ && path_node->data_type->e.numeric.type == HLSL_TYPE_UINT);
+
+ idx = hlsl_ir_constant(path_node)->value.u[0].u;
+
+ switch (type->class)
+ {
+ case HLSL_CLASS_VECTOR:
+ if (idx >= type->dimx)
+ {
+ hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
+ "Vector index is out of bounds. %u/%u", idx, type->dimx);
+ return DEREF_VALIDATION_OUT_OF_BOUNDS;
+ }
+ break;
+
+ case HLSL_CLASS_MATRIX:
+ if (idx >= hlsl_type_major_size(type))
+ {
+ hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
+ "Matrix index is out of bounds. %u/%u", idx, hlsl_type_major_size(type));
+ return DEREF_VALIDATION_OUT_OF_BOUNDS;
+ }
+ break;
+
+ case HLSL_CLASS_ARRAY:
+ if (idx >= type->e.array.elements_count)
+ {
+ hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
+ "Array index is out of bounds. %u/%u", idx, type->e.array.elements_count);
+ return DEREF_VALIDATION_OUT_OF_BOUNDS;
+ }
+ break;
+
+ case HLSL_CLASS_STRUCT:
+ break;
+
+ default:
+ vkd3d_unreachable();
+ }
+
+ type = hlsl_get_element_type_from_path_index(ctx, type, path_node);
+ }
+
+ return DEREF_VALIDATION_OK;
+}
+
static void note_non_static_deref_expressions(struct hlsl_ctx *ctx, const struct hlsl_deref *deref,
const char *usage)
{
@@ -1998,60 +2068,77 @@ static void note_non_static_deref_expressions(struct hlsl_ctx *ctx, const struct
}
}
-static bool validate_static_object_references(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr,
+static bool validate_dereferences(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr,
void *context)
{
- unsigned int start, count;
-
- if (instr->type == HLSL_IR_RESOURCE_LOAD)
+ switch (instr->type)
{
- struct hlsl_ir_resource_load *load = hlsl_ir_resource_load(instr);
-
- if (!load->resource.var->is_uniform)
+ case HLSL_IR_RESOURCE_LOAD:
{
- hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Loaded resource must have a single uniform source.");
+ struct hlsl_ir_resource_load *load = hlsl_ir_resource_load(instr);
+
+ if (!load->resource.var->is_uniform)
+ {
+ hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
+ "Loaded resource must have a single uniform source.");
+ }
+ else if (validate_component_index_range_from_deref(ctx, &load->resource) == DEREF_VALIDATION_NOT_CONSTANT)
+ {
+ hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
+ "Loaded resource from \"%s\" must be determinable at compile time.",
+ load->resource.var->name);
+ note_non_static_deref_expressions(ctx, &load->resource, "loaded resource");
+ }
+
+ if (load->sampler.var)
+ {
+ if (!load->sampler.var->is_uniform)
+ {
+ hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
+ "Resource load sampler must have a single uniform source.");
+ }
+ else if (validate_component_index_range_from_deref(ctx, &load->sampler) == DEREF_VALIDATION_NOT_CONSTANT)
+ {
+ hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
+ "Resource load sampler from \"%s\" must be determinable at compile time.",
+ load->sampler.var->name);
+ note_non_static_deref_expressions(ctx, &load->sampler, "resource load sampler");
+ }
+ }
+ break;
}
- else if (!hlsl_component_index_range_from_deref(ctx, &load->resource, &start, &count))
+ case HLSL_IR_RESOURCE_STORE:
{
- hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Loaded resource from \"%s\" must be determinable at compile time.",
- load->resource.var->name);
- note_non_static_deref_expressions(ctx, &load->resource, "loaded resource");
- }
+ struct hlsl_ir_resource_store *store = hlsl_ir_resource_store(instr);
- if (load->sampler.var)
- {
- if (!load->sampler.var->is_uniform)
+ if (!store->resource.var->is_uniform)
{
hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Resource load sampler must have a single uniform source.");
+ "Accessed resource must have a single uniform source.");
}
- else if (!hlsl_component_index_range_from_deref(ctx, &load->sampler, &start, &count))
+ else if (validate_component_index_range_from_deref(ctx, &store->resource) == DEREF_VALIDATION_NOT_CONSTANT)
{
hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Resource load sampler from \"%s\" must be determinable at compile time.",
- load->sampler.var->name);
- note_non_static_deref_expressions(ctx, &load->sampler, "resource load sampler");
+ "Accessed resource from \"%s\" must be determinable at compile time.",
+ store->resource.var->name);
+ note_non_static_deref_expressions(ctx, &store->resource, "accessed resource");
}
+ break;
}
- }
- else if (instr->type == HLSL_IR_RESOURCE_STORE)
- {
- struct hlsl_ir_resource_store *store = hlsl_ir_resource_store(instr);
-
- if (!store->resource.var->is_uniform)
+ case HLSL_IR_LOAD:
{
- hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Accessed resource must have a single uniform source.");
+ struct hlsl_ir_load *load = hlsl_ir_load(instr);
+ validate_component_index_range_from_deref(ctx, &load->src);
+ break;
}
- else if (!hlsl_component_index_range_from_deref(ctx, &store->resource, &start, &count))
+ case HLSL_IR_STORE:
{
- hlsl_error(ctx, &instr->loc, VKD3D_SHADER_ERROR_HLSL_NON_STATIC_OBJECT_REF,
- "Accessed resource from \"%s\" must be determinable at compile time.",
- store->resource.var->name);
- note_non_static_deref_expressions(ctx, &store->resource, "accessed resource");
+ struct hlsl_ir_store *store = hlsl_ir_store(instr);
+ validate_component_index_range_from_deref(ctx, &store->lhs);
+ break;
}
+ default:
+ break;
}
return false;
@@ -5203,21 +5290,13 @@ bool hlsl_component_index_range_from_deref(struct hlsl_ctx *ctx, const struct hl
{
case HLSL_CLASS_VECTOR:
if (idx >= type->dimx)
- {
- hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
- "Vector index is out of bounds. %u/%u", idx, type->dimx);
return false;
- }
*start += idx;
break;
case HLSL_CLASS_MATRIX:
if (idx >= hlsl_type_major_size(type))
- {
- hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
- "Matrix index is out of bounds. %u/%u", idx, hlsl_type_major_size(type));
return false;
- }
if (hlsl_type_is_row_major(type))
*start += idx * type->dimx;
else
@@ -5226,11 +5305,7 @@ bool hlsl_component_index_range_from_deref(struct hlsl_ctx *ctx, const struct hl
case HLSL_CLASS_ARRAY:
if (idx >= type->e.array.elements_count)
- {
- hlsl_error(ctx, &path_node->loc, VKD3D_SHADER_ERROR_HLSL_OFFSET_OUT_OF_BOUNDS,
- "Array index is out of bounds. %u/%u", idx, type->e.array.elements_count);
return false;
- }
*start += idx * hlsl_type_component_count(type->e.array.type);
break;
@@ -5635,6 +5710,222 @@ static void sm1_generate_vsir(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl
sm1_generate_vsir_signature(ctx, program);
}
+static struct hlsl_ir_jump *loop_unrolling_find_jump(struct hlsl_block *block, struct hlsl_ir_node *stop_point,
+ struct hlsl_block **found_block)
+{
+ struct hlsl_ir_node *node;
+
+ LIST_FOR_EACH_ENTRY(node, &block->instrs, struct hlsl_ir_node, entry)
+ {
+ if (node == stop_point)
+ return NULL;
+
+ if (node->type == HLSL_IR_IF)
+ {
+ struct hlsl_ir_if *iff = hlsl_ir_if(node);
+ struct hlsl_ir_jump *jump = NULL;
+
+ if ((jump = loop_unrolling_find_jump(&iff->then_block, stop_point, found_block)))
+ return jump;
+ if ((jump = loop_unrolling_find_jump(&iff->else_block, stop_point, found_block)))
+ return jump;
+ }
+ else if (node->type == HLSL_IR_JUMP)
+ {
+ struct hlsl_ir_jump *jump = hlsl_ir_jump(node);
+
+ if (jump->type == HLSL_IR_JUMP_BREAK || jump->type == HLSL_IR_JUMP_CONTINUE)
+ {
+ *found_block = block;
+ return jump;
+ }
+ }
+ }
+
+ return NULL;
+}
+
+static unsigned int loop_unrolling_get_max_iterations(struct hlsl_ctx *ctx, struct hlsl_ir_loop *loop)
+{
+ /* Always use the explicit limit if it has been passed. */
+ if (loop->unroll_limit)
+ return loop->unroll_limit;
+
+ /* All SMs will default to 1024 if [unroll] has been specified without an explicit limit. */
+ if (loop->unroll_type == HLSL_IR_LOOP_FORCE_UNROLL)
+ return 1024;
+
+ /* SM4 limits implicit unrolling to 254 iterations. */
+ if (hlsl_version_ge(ctx, 4, 0))
+ return 254;
+
+ /* SM<3 implicitly unrolls up to 1024 iterations. */
+ return 1024;
+}
+
+static bool loop_unrolling_unroll_loop(struct hlsl_ctx *ctx, struct hlsl_block *block,
+ struct hlsl_block *loop_parent, struct hlsl_ir_loop *loop)
+{
+ unsigned int max_iterations, i;
+
+ max_iterations = loop_unrolling_get_max_iterations(ctx, loop);
+
+ for (i = 0; i < max_iterations; ++i)
+ {
+ struct hlsl_block tmp_dst, *jump_block;
+ struct hlsl_ir_jump *jump = NULL;
+
+ if (!hlsl_clone_block(ctx, &tmp_dst, &loop->body))
+ return false;
+ list_move_before(&loop->node.entry, &tmp_dst.instrs);
+ hlsl_block_cleanup(&tmp_dst);
+
+ hlsl_run_const_passes(ctx, block);
+
+ if ((jump = loop_unrolling_find_jump(loop_parent, &loop->node, &jump_block)))
+ {
+ enum hlsl_ir_jump_type type = jump->type;
+
+ if (jump_block != loop_parent)
+ {
+ if (loop->unroll_type == HLSL_IR_LOOP_FORCE_UNROLL)
+ hlsl_error(ctx, &jump->node.loc, VKD3D_SHADER_ERROR_HLSL_FAILED_FORCED_UNROLL,
+ "Unable to unroll loop, unrolling loops with conditional jumps is currently not supported.");
+ return false;
+ }
+
+ list_move_slice_tail(&tmp_dst.instrs, &jump->node.entry, list_prev(&loop_parent->instrs, &loop->node.entry));
+ hlsl_block_cleanup(&tmp_dst);
+
+ if (type == HLSL_IR_JUMP_BREAK)
+ break;
+ }
+ }
+
+ /* Native will not emit an error if max_iterations has been reached with an
+ * explicit limit. It also will not insert a loop if there are iterations left
+ * i.e [unroll(4)] for (i = 0; i < 8; ++i)) */
+ if (!loop->unroll_limit && i == max_iterations)
+ {
+ if (loop->unroll_type == HLSL_IR_LOOP_FORCE_UNROLL)
+ hlsl_error(ctx, &loop->node.loc, VKD3D_SHADER_ERROR_HLSL_FAILED_FORCED_UNROLL,
+ "Unable to unroll loop, maximum iterations reached (%u).", max_iterations);
+ return false;
+ }
+
+ list_remove(&loop->node.entry);
+ hlsl_free_instr(&loop->node);
+
+ return true;
+}
+
+/*
+ * loop_unrolling_find_unrollable_loop() is not the normal way to do things;
+ * normal passes simply iterate over the whole block and apply a transformation
+ * to every relevant instruction. However, loop unrolling can fail, and we want
+ * to leave the loop in its previous state in that case. That isn't a problem by
+ * itself, except that loop unrolling needs copy-prop in order to work properly,
+ * and copy-prop state at the time of the loop depends on the rest of the program
+ * up to that point. This means we need to clone the whole program, and at that
+ * point we have to search it again anyway to find the clone of the loop we were
+ * going to unroll.
+ *
+ * FIXME: Ideally we wouldn't clone the whole program; instead we would run copyprop
+ * up until the loop instruction, clone just that loop, then use copyprop again
+ * with the saved state after unrolling. However, copyprop currently isn't built
+ * for that yet [notably, it still relies on indices]. Note also this still doesn't
+ * really let us use transform_ir() anyway [since we don't have a good way to say
+ * "copyprop from the beginning of the program up to the instruction we're
+ * currently processing" from the callback]; we'd have to use a dedicated
+ * recursive function instead. */
+static struct hlsl_ir_loop *loop_unrolling_find_unrollable_loop(struct hlsl_ctx *ctx, struct hlsl_block *block,
+ struct hlsl_block **containing_block)
+{
+ struct hlsl_ir_node *instr;
+
+ LIST_FOR_EACH_ENTRY(instr, &block->instrs, struct hlsl_ir_node, entry)
+ {
+ switch (instr->type)
+ {
+ case HLSL_IR_LOOP:
+ {
+ struct hlsl_ir_loop *nested_loop;
+ struct hlsl_ir_loop *loop = hlsl_ir_loop(instr);
+
+ if ((nested_loop = loop_unrolling_find_unrollable_loop(ctx, &loop->body, containing_block)))
+ return nested_loop;
+
+ if (loop->unroll_type == HLSL_IR_LOOP_UNROLL || loop->unroll_type == HLSL_IR_LOOP_FORCE_UNROLL)
+ {
+ *containing_block = block;
+ return loop;
+ }
+
+ break;
+ }
+ case HLSL_IR_IF:
+ {
+ struct hlsl_ir_loop *loop;
+ struct hlsl_ir_if *iff = hlsl_ir_if(instr);
+
+ if ((loop = loop_unrolling_find_unrollable_loop(ctx, &iff->then_block, containing_block)))
+ return loop;
+ if ((loop = loop_unrolling_find_unrollable_loop(ctx, &iff->else_block, containing_block)))
+ return loop;
+
+ break;
+ }
+ case HLSL_IR_SWITCH:
+ {
+ struct hlsl_ir_switch *s = hlsl_ir_switch(instr);
+ struct hlsl_ir_switch_case *c;
+ struct hlsl_ir_loop *loop;
+
+ LIST_FOR_EACH_ENTRY(c, &s->cases, struct hlsl_ir_switch_case, entry)
+ {
+ if ((loop = loop_unrolling_find_unrollable_loop(ctx, &c->body, containing_block)))
+ return loop;
+ }
+
+ break;
+ }
+ default:
+ break;
+ }
+ }
+
+ return NULL;
+}
+
+static void transform_unroll_loops(struct hlsl_ctx *ctx, struct hlsl_block *block)
+{
+ while (true)
+ {
+ struct hlsl_block clone, *containing_block;
+ struct hlsl_ir_loop *loop, *cloned_loop;
+
+ if (!(loop = loop_unrolling_find_unrollable_loop(ctx, block, &containing_block)))
+ return;
+
+ if (!hlsl_clone_block(ctx, &clone, block))
+ return;
+
+ cloned_loop = loop_unrolling_find_unrollable_loop(ctx, &clone, &containing_block);
+ assert(cloned_loop);
+
+ if (!loop_unrolling_unroll_loop(ctx, &clone, containing_block, cloned_loop))
+ {
+ hlsl_block_cleanup(&clone);
+ loop->unroll_type = HLSL_IR_LOOP_FORCE_LOOP;
+ continue;
+ }
+
+ hlsl_block_cleanup(block);
+ hlsl_block_init(block);
+ hlsl_block_add_block(block, &clone);
+ }
+}
+
int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func,
enum vkd3d_shader_target_type target_type, struct vkd3d_shader_code *out)
{
@@ -5721,6 +6012,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
hlsl_transform_ir(ctx, lower_discard_neg, body, NULL);
}
+ transform_unroll_loops(ctx, body);
hlsl_run_const_passes(ctx, body);
remove_unreachable_code(ctx, body);
@@ -5730,7 +6022,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry
lower_ir(ctx, lower_casts_to_bool, body);
lower_ir(ctx, lower_int_dot, body);
- hlsl_transform_ir(ctx, validate_static_object_references, body, NULL);
+ hlsl_transform_ir(ctx, validate_dereferences, body, NULL);
hlsl_transform_ir(ctx, track_object_components_sampler_dim, body, NULL);
if (profile->major_version >= 4)
hlsl_transform_ir(ctx, lower_combined_samples, body, NULL);
diff --git a/libs/vkd3d/libs/vkd3d-shader/tpf.c b/libs/vkd3d/libs/vkd3d-shader/tpf.c
index ca7cdfd5217..a7c37215e5e 100644
--- a/libs/vkd3d/libs/vkd3d-shader/tpf.c
+++ b/libs/vkd3d/libs/vkd3d-shader/tpf.c
@@ -2762,6 +2762,7 @@ bool hlsl_sm4_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_sem
{"depth", true, VKD3D_SHADER_TYPE_PIXEL, VKD3DSPR_DEPTHOUT, false},
{"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, VKD3DSPR_DEPTHOUT, false},
{"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, VKD3DSPR_OUTPUT, true},
+ {"sv_coverage", true, VKD3D_SHADER_TYPE_PIXEL, VKD3DSPR_SAMPLEMASK, false},
};
for (i = 0; i < ARRAY_SIZE(register_table); ++i)
@@ -2817,6 +2818,7 @@ bool hlsl_sm4_usage_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semant
{"depth", true, VKD3D_SHADER_TYPE_PIXEL, D3D_NAME_DEPTH},
{"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, D3D_NAME_TARGET},
{"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, D3D_NAME_DEPTH},
+ {"sv_coverage", true, VKD3D_SHADER_TYPE_PIXEL, D3D_NAME_COVERAGE},
{"sv_position", false, VKD3D_SHADER_TYPE_VERTEX, D3D_NAME_UNDEFINED},
{"sv_vertexid", false, VKD3D_SHADER_TYPE_VERTEX, D3D_NAME_VERTEX_ID},
@@ -5042,6 +5044,25 @@ static void write_sm4_store_uav_typed(const struct tpf_writer *tpf, const struct
write_sm4_instruction(tpf, &instr);
}
+static void write_sm4_rasterizer_sample_count(const struct tpf_writer *tpf, const struct hlsl_ir_node *dst)
+{
+ struct sm4_instruction instr;
+
+ memset(&instr, 0, sizeof(instr));
+ instr.opcode = VKD3D_SM4_OP_SAMPLE_INFO;
+ instr.extra_bits |= VKD3DSI_SAMPLE_INFO_UINT << VKD3D_SM4_INSTRUCTION_FLAGS_SHIFT;
+
+ sm4_dst_from_node(&instr.dsts[0], dst);
+ instr.dst_count = 1;
+
+ instr.srcs[0].reg.type = VKD3DSPR_RASTERIZER;
+ instr.srcs[0].reg.dimension = VSIR_DIMENSION_VEC4;
+ instr.srcs[0].swizzle = VKD3D_SHADER_SWIZZLE(X, X, X, X);
+ instr.src_count = 1;
+
+ write_sm4_instruction(tpf, &instr);
+}
+
static void write_sm4_expr(const struct tpf_writer *tpf, const struct hlsl_ir_expr *expr)
{
const struct hlsl_ir_node *arg1 = expr->operands[0].node;
@@ -5057,6 +5078,14 @@ static void write_sm4_expr(const struct tpf_writer *tpf, const struct hlsl_ir_ex
switch (expr->op)
{
+ case HLSL_OP0_RASTERIZER_SAMPLE_COUNT:
+ if (tpf->ctx->profile->type == VKD3D_SHADER_TYPE_PIXEL && hlsl_version_ge(tpf->ctx, 4, 1))
+ write_sm4_rasterizer_sample_count(tpf, &expr->node);
+ else
+ hlsl_error(tpf->ctx, &expr->node.loc, VKD3D_SHADER_ERROR_HLSL_INCOMPATIBLE_PROFILE,
+ "GetRenderTargetSampleCount() can only be used from a pixel shader using version 4.1 or higher.");
+ break;
+
case HLSL_OP1_ABS:
switch (dst_type->e.numeric.type)
{
diff --git a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
index c79f845b675..96e613669a6 100644
--- a/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
+++ b/libs/vkd3d/libs/vkd3d-shader/vkd3d_shader_private.h
@@ -152,6 +152,7 @@ enum vkd3d_shader_error
VKD3D_SHADER_ERROR_HLSL_MISSING_TECHNIQUE = 5029,
VKD3D_SHADER_ERROR_HLSL_UNKNOWN_MODIFIER = 5030,
VKD3D_SHADER_ERROR_HLSL_INVALID_STATE_BLOCK_ENTRY = 5031,
+ VKD3D_SHADER_ERROR_HLSL_FAILED_FORCED_UNROLL = 5032,
VKD3D_SHADER_WARNING_HLSL_IMPLICIT_TRUNCATION = 5300,
VKD3D_SHADER_WARNING_HLSL_DIVISION_BY_ZERO = 5301,
diff --git a/libs/vkd3d/libs/vkd3d/resource.c b/libs/vkd3d/libs/vkd3d/resource.c
index db2d87428bb..7d7f40c0953 100644
--- a/libs/vkd3d/libs/vkd3d/resource.c
+++ b/libs/vkd3d/libs/vkd3d/resource.c
@@ -4349,7 +4349,11 @@ static HRESULT d3d12_descriptor_heap_init(struct d3d12_descriptor_heap *descript
return hr;
descriptor_heap->use_vk_heaps = device->use_vk_heaps && (desc->Flags & D3D12_DESCRIPTOR_HEAP_FLAG_SHADER_VISIBLE);
- d3d12_descriptor_heap_vk_descriptor_sets_init(descriptor_heap, device, desc);
+ if (FAILED(hr = d3d12_descriptor_heap_vk_descriptor_sets_init(descriptor_heap, device, desc)))
+ {
+ vkd3d_private_store_destroy(&descriptor_heap->private_store);
+ return hr;
+ }
vkd3d_mutex_init(&descriptor_heap->vk_sets_mutex);
d3d12_device_add_ref(descriptor_heap->device = device);
--
2.43.0