- Wine-Commits - list.winehq.org

Francisco Casas : vkd3d-shader/hlsl: Absorb hlsl_ir_constant deref offsets into const_offset.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 74767beaf658296950f3f2ca527192e471d39bf4 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/74767beaf658296950f3f2ca52719… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Thu May 11 15:18:02 2023 -0400 vkd3d-shader/hlsl: Absorb hlsl_ir_constant deref offsets into const_offset. --- libs/vkd3d-shader/hlsl_codegen.c | 29 +++++++++++++++++++++-------- 1 file changed, 21 insertions(+), 8 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 5c3b7eb6..0a9d3f67 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -174,6 +174,24 @@ static bool replace_deref_path_with_offset(struct hlsl_ctx *ctx, struct hlsl_der return true; } +static bool clean_constant_deref_offset_srcs(struct hlsl_ctx *ctx, struct hlsl_deref *deref, + struct hlsl_ir_node *instr) +{ + if (deref->offset.node && deref->offset.node->type == HLSL_IR_CONSTANT) + { + enum hlsl_regset regset = hlsl_deref_get_regset(ctx, deref); + + if (regset == HLSL_REGSET_NUMERIC) + deref->const_offset += 4 * hlsl_ir_constant(deref->offset.node)->value.u[0].u; + else + deref->const_offset += hlsl_ir_constant(deref->offset.node)->value.u[0].u; + hlsl_src_remove(&deref->offset); + return true; + } + return false; +} + + /* Split uniforms into two variables representing the constant and temp * registers, and copy the former to the latter, so that writes to uniforms * work. */ @@ -4465,14 +4483,8 @@ bool hlsl_offset_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref /* We should always have generated a cast to UINT. */ assert(offset_node->data_type->class == HLSL_CLASS_SCALAR && offset_node->data_type->base_type == HLSL_TYPE_UINT); - - if (offset_node->type != HLSL_IR_CONSTANT) - return false; - - if (regset == HLSL_REGSET_NUMERIC) - *offset += 4 * hlsl_ir_constant(offset_node)->value.u[0].u; - else - *offset += hlsl_ir_constant(offset_node)->value.u[0].u; + assert(offset_node->type != HLSL_IR_CONSTANT); + return false; } size = deref->var->data_type->reg_size[regset]; @@ -4789,6 +4801,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry /* TODO: move forward, remove when no longer needed */ transform_derefs(ctx, replace_deref_path_with_offset, body); while (hlsl_transform_ir(ctx, hlsl_fold_constant_exprs, body, NULL)); + transform_derefs(ctx, clean_constant_deref_offset_srcs, body); do compute_liveness(ctx, entry_func);

1 0

Zebediah Figura : vkd3d-shader/d3dbc: Set the interpolation mode for signature elements.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 6b7834d6356a547994b2569e0d47962a7994fb71 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/6b7834d6356a547994b2569e0d479… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Thu Oct 19 21:34:23 2023 -0500 vkd3d-shader/d3dbc: Set the interpolation mode for signature elements. --- libs/vkd3d-shader/d3dbc.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/d3dbc.c b/libs/vkd3d-shader/d3dbc.c index 67fa3271..a41aaaa6 100644 --- a/libs/vkd3d-shader/d3dbc.c +++ b/libs/vkd3d-shader/d3dbc.c @@ -555,9 +555,9 @@ static bool add_signature_element(struct vkd3d_shader_sm1_parser *sm1, bool outp return false; element = &signature->elements[signature->element_count++]; + memset(element, 0, sizeof(*element)); element->semantic_name = name; element->semantic_index = index; - element->stream_index = 0; element->sysval_semantic = sysval; element->component_type = VKD3D_SHADER_COMPONENT_FLOAT; element->register_index = register_index; @@ -565,7 +565,8 @@ static bool add_signature_element(struct vkd3d_shader_sm1_parser *sm1, bool outp element->register_count = 1; element->mask = mask; element->used_mask = is_dcl ? 0 : mask; - element->min_precision = VKD3D_SHADER_MINIMUM_PRECISION_NONE; + if (sm1->p.shader_version.type == VKD3D_SHADER_TYPE_PIXEL && !output) + element->interpolation_mode = VKD3DSIM_LINEAR; return true; }

1 0

Zebediah Figura : vkd3d-shader/tpf: Set the interpolation mode for signature elements.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 15b69721de4a8a68c16782e23dd531a4defacb15 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/15b69721de4a8a68c16782e23dd53… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Thu Oct 19 21:34:41 2023 -0500 vkd3d-shader/tpf: Set the interpolation mode for signature elements. --- libs/vkd3d-shader/ir.c | 6 ++++++ libs/vkd3d-shader/tpf.c | 21 ++++++++++++++++++--- libs/vkd3d-shader/vkd3d_shader_private.h | 2 ++ 3 files changed, 26 insertions(+), 3 deletions(-) diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 511b0e8f..c8cbf1e4 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -581,6 +581,12 @@ static unsigned int shader_signature_find_element_for_reg(const struct shader_si vkd3d_unreachable(); } +struct signature_element *vsir_signature_find_element_for_reg(const struct shader_signature *signature, + unsigned int reg_idx, unsigned int write_mask) +{ + return &signature->elements[shader_signature_find_element_for_reg(signature, reg_idx, write_mask)]; +} + static unsigned int range_map_get_register_count(uint8_t range_map[][VKD3D_VEC4_SIZE], unsigned int register_idx, unsigned int write_mask) { diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c index 8c5af8e2..a49457dc 100644 --- a/libs/vkd3d-shader/tpf.c +++ b/libs/vkd3d-shader/tpf.c @@ -1070,16 +1070,31 @@ static void shader_sm4_read_declaration_register_semantic(struct vkd3d_shader_in static void shader_sm4_read_dcl_input_ps(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *priv) { + struct vkd3d_shader_dst_param *dst = &ins->declaration.dst; + ins->flags = (opcode_token & VKD3D_SM4_INTERPOLATION_MODE_MASK) >> VKD3D_SM4_INTERPOLATION_MODE_SHIFT; - shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VKD3D_DATA_FLOAT, &ins->declaration.dst); + if (shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VKD3D_DATA_FLOAT, dst)) + { + struct signature_element *e = vsir_signature_find_element_for_reg( + &priv->p.shader_desc.input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); + + e->interpolation_mode = ins->flags; + } } static void shader_sm4_read_dcl_input_ps_siv(struct vkd3d_shader_instruction *ins, uint32_t opcode, uint32_t opcode_token, const uint32_t *tokens, unsigned int token_count, struct vkd3d_shader_sm4_parser *priv) { + struct vkd3d_shader_dst_param *dst = &ins->declaration.register_semantic.reg; + ins->flags = (opcode_token & VKD3D_SM4_INTERPOLATION_MODE_MASK) >> VKD3D_SM4_INTERPOLATION_MODE_SHIFT; - shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VKD3D_DATA_FLOAT, - &ins->declaration.register_semantic.reg); + if (shader_sm4_read_dst_param(priv, &tokens, &tokens[token_count], VKD3D_DATA_FLOAT, dst)) + { + struct signature_element *e = vsir_signature_find_element_for_reg( + &priv->p.shader_desc.input_signature, dst->reg.idx[dst->reg.idx_count - 1].offset, dst->write_mask); + + e->interpolation_mode = ins->flags; + } ins->declaration.register_semantic.sysval_semantic = *tokens; } diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index 93154c0f..76b401fd 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -916,6 +916,8 @@ struct shader_signature unsigned int element_count; }; +struct signature_element *vsir_signature_find_element_for_reg(const struct shader_signature *signature, + unsigned int reg_idx, unsigned int write_mask); void shader_signature_cleanup(struct shader_signature *signature); struct vkd3d_shader_desc

1 0

Francisco Casas : vkd3d-shader/tpf: Declare indexable temps.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: eef2163375f5342c6f032886a1c8f14aac85d20f URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/eef2163375f5342c6f032886a1c8f… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Wed Oct 4 18:03:14 2023 -0300 vkd3d-shader/tpf: Declare indexable temps. If var->indexable, then the variable is given a unique register number, regardless of its lifetime. --- libs/vkd3d-shader/hlsl_codegen.c | 24 ++++++++++++++++++++---- libs/vkd3d-shader/tpf.c | 36 +++++++++++++++++++++++++++++++++++- 2 files changed, 55 insertions(+), 5 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index d957bd0e..51c3d68f 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -3473,6 +3473,10 @@ struct register_allocator unsigned int writemask; unsigned int first_write, last_read; } *allocations; + + /* Indexable temps are allocated separately and always keep their index regardless of their + * lifetime. */ + size_t indexable_count; }; static unsigned int get_available_writemask(const struct register_allocator *allocator, @@ -3719,11 +3723,23 @@ static void allocate_variable_temp_register(struct hlsl_ctx *ctx, if (!var->regs[HLSL_REGSET_NUMERIC].allocated && var->last_read) { - var->regs[HLSL_REGSET_NUMERIC] = allocate_numeric_registers_for_type(ctx, allocator, - var->first_write, var->last_read, var->data_type); + if (var->indexable) + { + var->regs[HLSL_REGSET_NUMERIC].id = allocator->indexable_count++; + var->regs[HLSL_REGSET_NUMERIC].allocation_size = 1; + var->regs[HLSL_REGSET_NUMERIC].writemask = 0; + var->regs[HLSL_REGSET_NUMERIC].allocated = true; - TRACE("Allocated %s to %s (liveness %u-%u).\n", var->name, debug_register('r', - var->regs[HLSL_REGSET_NUMERIC], var->data_type), var->first_write, var->last_read); + TRACE("Allocated %s to x%u[].\n", var->name, var->regs[HLSL_REGSET_NUMERIC].id); + } + else + { + var->regs[HLSL_REGSET_NUMERIC] = allocate_numeric_registers_for_type(ctx, allocator, + var->first_write, var->last_read, var->data_type); + + TRACE("Allocated %s to %s (liveness %u-%u).\n", var->name, debug_register('r', + var->regs[HLSL_REGSET_NUMERIC], var->data_type), var->first_write, var->last_read); + } } } diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c index 0be0ec4f..8c5af8e2 100644 --- a/libs/vkd3d-shader/tpf.c +++ b/libs/vkd3d-shader/tpf.c @@ -3732,7 +3732,7 @@ static void sm4_register_from_deref(struct hlsl_ctx *ctx, struct vkd3d_shader_re struct hlsl_reg hlsl_reg = hlsl_reg_from_deref(ctx, deref); assert(hlsl_reg.allocated); - reg->type = VKD3DSPR_TEMP; + reg->type = deref->var->indexable ? VKD3DSPR_IDXTEMP : VKD3DSPR_TEMP; reg->dimension = VSIR_DIMENSION_VEC4; reg->idx[0].offset = hlsl_reg.id; reg->idx_count = 1; @@ -4251,6 +4251,20 @@ static void write_sm4_dcl_temps(const struct tpf_writer *tpf, uint32_t temp_coun write_sm4_instruction(tpf, &instr); } +static void write_sm4_dcl_indexable_temp(const struct tpf_writer *tpf, uint32_t idx, + uint32_t size, uint32_t comp_count) +{ + struct sm4_instruction instr = + { + .opcode = VKD3D_SM4_OP_DCL_INDEXABLE_TEMP, + + .idx = {idx, size, comp_count}, + .idx_count = 3, + }; + + write_sm4_instruction(tpf, &instr); +} + static void write_sm4_dcl_thread_group(const struct tpf_writer *tpf, const uint32_t thread_count[3]) { struct sm4_instruction instr = @@ -5616,6 +5630,7 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx, struct extern_resource *extern_resources; unsigned int extern_resources_count, i; const struct hlsl_buffer *cbuffer; + const struct hlsl_scope *scope; const struct hlsl_ir_var *var; size_t token_count_position; struct tpf_writer tpf; @@ -5670,6 +5685,25 @@ static void write_sm4_shdr(struct hlsl_ctx *ctx, if (ctx->temp_count) write_sm4_dcl_temps(&tpf, ctx->temp_count); + LIST_FOR_EACH_ENTRY(scope, &ctx->scopes, struct hlsl_scope, entry) + { + LIST_FOR_EACH_ENTRY(var, &scope->vars, struct hlsl_ir_var, scope_entry) + { + if (var->is_uniform || var->is_input_semantic || var->is_output_semantic) + continue; + if (!var->regs[HLSL_REGSET_NUMERIC].allocated) + continue; + + if (var->indexable) + { + unsigned int id = var->regs[HLSL_REGSET_NUMERIC].id; + unsigned int size = align(var->data_type->reg_size[HLSL_REGSET_NUMERIC], 4) / 4; + + write_sm4_dcl_indexable_temp(&tpf, id, size, 4); + } + } + } + write_sm4_block(&tpf, &entry_func->body); write_sm4_ret(&tpf);

1 0

Francisco Casas : vkd3d-shader/hlsl: Mark vars that require non-constant dereferences.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 83c313ecc6d1f16f27edaf1e337e79d23b73fe21 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/83c313ecc6d1f16f27edaf1e337e7… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Wed Oct 4 17:28:02 2023 -0300 vkd3d-shader/hlsl: Mark vars that require non-constant dereferences. --- libs/vkd3d-shader/hlsl.h | 3 +++ libs/vkd3d-shader/hlsl_codegen.c | 15 +++++++++++++++ 2 files changed, 18 insertions(+) diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index 38e605ba..0655bb8e 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -425,6 +425,9 @@ struct hlsl_ir_var * It may be less than the allocation size, e.g. for texture arrays. */ unsigned int bind_count[HLSL_REGSET_LAST_OBJECT + 1]; + /* Whether the shader performs dereferences with non-constant offsets in the variable. */ + bool indexable; + uint32_t is_input_semantic : 1; uint32_t is_output_semantic : 1; uint32_t is_uniform : 1; diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 4ea4f4cb..d957bd0e 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -3210,6 +3210,19 @@ static void dump_function(struct rb_entry *entry, void *context) rb_for_each_entry(&func->overloads, dump_function_decl, ctx); } +static bool mark_indexable_vars(struct hlsl_ctx *ctx, struct hlsl_deref *deref, + struct hlsl_ir_node *instr) +{ + if (!deref->rel_offset.node) + return false; + + assert(deref->var); + assert(deref->rel_offset.node->type != HLSL_IR_CONSTANT); + deref->var->indexable = true; + + return true; +} + static char get_regset_name(enum hlsl_regset regset) { switch (regset) @@ -4812,6 +4825,8 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry if (TRACE_ON()) rb_for_each_entry(&ctx->functions, dump_function, ctx); + transform_derefs(ctx, mark_indexable_vars, body); + calculate_resource_register_counts(ctx); allocate_register_reservations(ctx);

1 0

Francisco Casas : vkd3d-shader/hlsl: Rename hlsl_deref.offset to hlsl_deref.rel_offset.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 313df300ad317b49b45b38148755095988a974ed URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/313df300ad317b49b45b381487550… Author: Francisco Casas <fcasas(a)codeweavers.com> Date: Fri Oct 6 13:56:24 2023 -0300 vkd3d-shader/hlsl: Rename hlsl_deref.offset to hlsl_deref.rel_offset. This field is now analogous to vkd3d_shader_register_index.rel_addr. Also, it makes sense to rename it now because all the constant part of the offset is now handled to hlsl_deref.const_offset. Consequently, it may also be NULL now. --- libs/vkd3d-shader/hlsl.c | 12 ++++++------ libs/vkd3d-shader/hlsl.h | 4 ++-- libs/vkd3d-shader/hlsl_codegen.c | 36 ++++++++++++++++++------------------ 3 files changed, 26 insertions(+), 26 deletions(-) diff --git a/libs/vkd3d-shader/hlsl.c b/libs/vkd3d-shader/hlsl.c index b26bde31..f56ecc90 100644 --- a/libs/vkd3d-shader/hlsl.c +++ b/libs/vkd3d-shader/hlsl.c @@ -512,7 +512,7 @@ static bool init_deref(struct hlsl_ctx *ctx, struct hlsl_deref *deref, struct hl { deref->var = var; deref->path_len = path_len; - deref->offset.node = NULL; + deref->rel_offset.node = NULL; deref->const_offset = 0; deref->data_type = NULL; @@ -541,7 +541,7 @@ bool hlsl_init_deref_from_index_chain(struct hlsl_ctx *ctx, struct hlsl_deref *d deref->path = NULL; deref->path_len = 0; - deref->offset.node = NULL; + deref->rel_offset.node = NULL; deref->const_offset = 0; assert(chain); @@ -1138,7 +1138,7 @@ void hlsl_cleanup_deref(struct hlsl_deref *deref) deref->path = NULL; deref->path_len = 0; - hlsl_src_remove(&deref->offset); + hlsl_src_remove(&deref->rel_offset); deref->const_offset = 0; } @@ -2432,12 +2432,12 @@ static void dump_deref(struct vkd3d_string_buffer *buffer, const struct hlsl_der { bool show_rel, show_const; - show_rel = deref->offset.node; + show_rel = deref->rel_offset.node; show_const = deref->const_offset != 0 || !show_rel; vkd3d_string_buffer_printf(buffer, "["); if (show_rel) - dump_src(buffer, &deref->offset); + dump_src(buffer, &deref->rel_offset); if (show_rel && show_const) vkd3d_string_buffer_printf(buffer, " + "); if (show_const) @@ -3011,7 +3011,7 @@ static void free_ir_resource_load(struct hlsl_ir_resource_load *load) static void free_ir_resource_store(struct hlsl_ir_resource_store *store) { - hlsl_src_remove(&store->resource.offset); + hlsl_src_remove(&store->resource.rel_offset); hlsl_src_remove(&store->coords); hlsl_src_remove(&store->value); vkd3d_free(store); diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index 39b57d2f..38e605ba 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -642,11 +642,11 @@ struct hlsl_deref * regset) from the start of the variable, to the part of the variable that is referenced. * This offset is stored using two fields, one for a variable part and other for a constant * part, which are added together: - * - offset: An offset given by an instruction node, in whole registers. + * - rel_offset: An offset given by an instruction node, in whole registers. * - const_offset: A constant number of register components. * Since the type information cannot longer be retrieved from the offset alone, the type is * stored in the data_type field, which remains NULL if the deref hasn't been lowered yet. */ - struct hlsl_src offset; + struct hlsl_src rel_offset; unsigned int const_offset; struct hlsl_type *data_type; }; diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 0a9d3f67..4ea4f4cb 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -168,7 +168,7 @@ static bool replace_deref_path_with_offset(struct hlsl_ctx *ctx, struct hlsl_der list_move_before(&instr->entry, &block.instrs); hlsl_cleanup_deref(deref); - hlsl_src_from_node(&deref->offset, offset); + hlsl_src_from_node(&deref->rel_offset, offset); deref->const_offset = offset_component; return true; @@ -177,15 +177,15 @@ static bool replace_deref_path_with_offset(struct hlsl_ctx *ctx, struct hlsl_der static bool clean_constant_deref_offset_srcs(struct hlsl_ctx *ctx, struct hlsl_deref *deref, struct hlsl_ir_node *instr) { - if (deref->offset.node && deref->offset.node->type == HLSL_IR_CONSTANT) + if (deref->rel_offset.node && deref->rel_offset.node->type == HLSL_IR_CONSTANT) { enum hlsl_regset regset = hlsl_deref_get_regset(ctx, deref); if (regset == HLSL_REGSET_NUMERIC) - deref->const_offset += 4 * hlsl_ir_constant(deref->offset.node)->value.u[0].u; + deref->const_offset += 4 * hlsl_ir_constant(deref->rel_offset.node)->value.u[0].u; else - deref->const_offset += hlsl_ir_constant(deref->offset.node)->value.u[0].u; - hlsl_src_remove(&deref->offset); + deref->const_offset += hlsl_ir_constant(deref->rel_offset.node)->value.u[0].u; + hlsl_src_remove(&deref->rel_offset); return true; } return false; @@ -3298,8 +3298,8 @@ static void compute_liveness_recurse(struct hlsl_block *block, unsigned int loop if (!var->first_write) var->first_write = loop_first ? min(instr->index, loop_first) : instr->index; store->rhs.node->last_read = last_read; - if (store->lhs.offset.node) - store->lhs.offset.node->last_read = last_read; + if (store->lhs.rel_offset.node) + store->lhs.rel_offset.node->last_read = last_read; break; } case HLSL_IR_EXPR: @@ -3326,8 +3326,8 @@ static void compute_liveness_recurse(struct hlsl_block *block, unsigned int loop var = load->src.var; var->last_read = max(var->last_read, last_read); - if (load->src.offset.node) - load->src.offset.node->last_read = last_read; + if (load->src.rel_offset.node) + load->src.rel_offset.node->last_read = last_read; break; } case HLSL_IR_LOOP: @@ -3344,14 +3344,14 @@ static void compute_liveness_recurse(struct hlsl_block *block, unsigned int loop var = load->resource.var; var->last_read = max(var->last_read, last_read); - if (load->resource.offset.node) - load->resource.offset.node->last_read = last_read; + if (load->resource.rel_offset.node) + load->resource.rel_offset.node->last_read = last_read; if ((var = load->sampler.var)) { var->last_read = max(var->last_read, last_read); - if (load->sampler.offset.node) - load->sampler.offset.node->last_read = last_read; + if (load->sampler.rel_offset.node) + load->sampler.rel_offset.node->last_read = last_read; } if (load->coords.node) @@ -3376,8 +3376,8 @@ static void compute_liveness_recurse(struct hlsl_block *block, unsigned int loop var = store->resource.var; var->last_read = max(var->last_read, last_read); - if (store->resource.offset.node) - store->resource.offset.node->last_read = last_read; + if (store->resource.rel_offset.node) + store->resource.rel_offset.node->last_read = last_read; store->coords.node->last_read = last_read; store->value.node->last_read = last_read; break; @@ -4473,7 +4473,7 @@ bool hlsl_regset_index_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref bool hlsl_offset_from_deref(struct hlsl_ctx *ctx, const struct hlsl_deref *deref, unsigned int *offset) { enum hlsl_regset regset = hlsl_deref_get_regset(ctx, deref); - struct hlsl_ir_node *offset_node = deref->offset.node; + struct hlsl_ir_node *offset_node = deref->rel_offset.node; unsigned int size; *offset = deref->const_offset; @@ -4505,8 +4505,8 @@ unsigned int hlsl_offset_from_deref_safe(struct hlsl_ctx *ctx, const struct hlsl if (hlsl_offset_from_deref(ctx, deref, &offset)) return offset; - hlsl_fixme(ctx, &deref->offset.node->loc, "Dereference with non-constant offset of type %s.", - hlsl_node_type_to_string(deref->offset.node->type)); + hlsl_fixme(ctx, &deref->rel_offset.node->loc, "Dereference with non-constant offset of type %s.", + hlsl_node_type_to_string(deref->rel_offset.node->type)); return 0; }

1 0

Nikolay Sivov : tests: Remove unused parameter from mul() tests functions.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: ea554b91d7733f7650731d5139e5dc8701b34ea2 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/ea554b91d7733f7650731d5139e5d… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Fri Oct 27 22:21:29 2023 +0200 tests: Remove unused parameter from mul() tests functions. --- tests/hlsl/mul.shader_test | 37 ++++++++++++++++++------------------- 1 file changed, 18 insertions(+), 19 deletions(-) diff --git a/tests/hlsl/mul.shader_test b/tests/hlsl/mul.shader_test index cb104a9e..7ddd50b6 100644 --- a/tests/hlsl/mul.shader_test +++ b/tests/hlsl/mul.shader_test @@ -1,6 +1,5 @@ - [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -16,7 +15,7 @@ draw quad probe all rgba (30.0, 70.0, 110.0, 150.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4 x = float4(1.0, 2.0, 3.0, 4.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -32,7 +31,7 @@ draw quad probe all rgba (90.0, 100.0, 110.0, 120.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -48,7 +47,7 @@ draw quad probe all rgba (14.0, 38.0, 62.0, 86.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float3 x = float3(1.0, 2.0, 3.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -64,7 +63,7 @@ draw quad probe all rgba (38.0, 44.0, 50.0, 56.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float3x3 x = float3x3(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, @@ -79,7 +78,7 @@ draw quad probe all rgba (14.0, 32.0, 50.0, 0.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4 x = float4(1.0, 2.0, 3.0, 4.0); float3x3 y = float3x3(1.0, 2.0, 3.0, @@ -94,7 +93,7 @@ draw quad probe all rgba (30.0, 36.0, 42.0, 0.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float x = 10.0; float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -110,7 +109,7 @@ draw quad probe all rgba (50.0, 60.0, 70.0, 80.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -126,7 +125,7 @@ draw quad probe all rgba (50.0, 60.0, 70.0, 80.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float1 x = float1(10.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -142,7 +141,7 @@ draw quad probe all rgba (10.0, 20.0, 30.0, 40.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -158,7 +157,7 @@ draw quad probe all rgba (10.0, 50.0, 90.0, 130.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float1x1 x = float1x1(10.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -174,7 +173,7 @@ draw quad probe all rgba (10.0, 20.0, 30.0, 40.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -190,7 +189,7 @@ draw quad probe all rgba (10.0, 50.0, 90.0, 130.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float1x4 x = float1x4(1.0, 2.0, 3.0, 4.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -206,7 +205,7 @@ draw quad probe all rgba (90.0, 100.0, 110.0, 120.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -222,7 +221,7 @@ draw quad probe all rgba (5.0, 10.0, 15.0, 20.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x1 x = float4x1(1.0, 2.0, 3.0, 4.0); float4x4 y = float4x4(1.0, 2.0, 3.0, 4.0, @@ -238,7 +237,7 @@ draw quad probe all rgba (2.0, 4.0, 6.0, 8.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, @@ -254,7 +253,7 @@ draw quad probe all rgba (30.0, 70.0, 110.0, 150.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float3x3 x = float3x3(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, @@ -272,7 +271,7 @@ draw quad probe all rgba (83.0, 98.0, 113.0, 128.0) [pixel shader] -float4 main(float4 pos : sv_position) : sv_target +float4 main() : sv_target { float4x4 x = float4x4(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0,

1 0

Zebediah Figura : vkd3d-shader: Handle a NULL descriptor info in vkd3d_shader_scan_add_descriptor().
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: a668f1ce1c72813dac4626f1ff7c1a3de10f9448 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/a668f1ce1c72813dac4626f1ff7c1… Author: Zebediah Figura <zfigura(a)codeweavers.com> Date: Fri Oct 27 11:48:24 2023 -0500 vkd3d-shader: Handle a NULL descriptor info in vkd3d_shader_scan_add_descriptor(). --- libs/vkd3d-shader/vkd3d_shader_main.c | 12 +++--------- 1 file changed, 3 insertions(+), 9 deletions(-) diff --git a/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d-shader/vkd3d_shader_main.c index 7aaec49f..dcd40137 100644 --- a/libs/vkd3d-shader/vkd3d_shader_main.c +++ b/libs/vkd3d-shader/vkd3d_shader_main.c @@ -784,6 +784,9 @@ static struct vkd3d_shader_descriptor_info1 *vkd3d_shader_scan_add_descriptor(st struct vkd3d_shader_scan_descriptor_info1 *info = context->scan_descriptor_info; struct vkd3d_shader_descriptor_info1 *d; + if (!info) + return NULL; + if (!vkd3d_array_reserve((void **)&info->descriptors, &context->descriptors_size, info->descriptor_count + 1, sizeof(*info->descriptors))) { @@ -811,9 +814,6 @@ static void vkd3d_shader_scan_constant_buffer_declaration(struct vkd3d_shader_sc const struct vkd3d_shader_constant_buffer *cb = &instruction->declaration.cb; struct vkd3d_shader_descriptor_info1 *d; - if (!context->scan_descriptor_info) - return; - if (!(d = vkd3d_shader_scan_add_descriptor(context, VKD3D_SHADER_DESCRIPTOR_TYPE_CBV, &cb->src.reg, &cb->range, VKD3D_SHADER_RESOURCE_BUFFER, VKD3D_SHADER_RESOURCE_DATA_UINT))) return; @@ -826,9 +826,6 @@ static void vkd3d_shader_scan_sampler_declaration(struct vkd3d_shader_scan_conte const struct vkd3d_shader_sampler *sampler = &instruction->declaration.sampler; struct vkd3d_shader_descriptor_info1 *d; - if (!context->scan_descriptor_info) - return; - if (!(d = vkd3d_shader_scan_add_descriptor(context, VKD3D_SHADER_DESCRIPTOR_TYPE_SAMPLER, &sampler->src.reg, &sampler->range, VKD3D_SHADER_RESOURCE_NONE, VKD3D_SHADER_RESOURCE_DATA_UINT))) return; @@ -854,9 +851,6 @@ static void vkd3d_shader_scan_resource_declaration(struct vkd3d_shader_scan_cont struct vkd3d_shader_descriptor_info1 *d; enum vkd3d_shader_descriptor_type type; - if (!context->scan_descriptor_info) - return; - if (resource->reg.reg.type == VKD3DSPR_UAV) type = VKD3D_SHADER_DESCRIPTOR_TYPE_UAV; else

1 0

Nikolay Sivov : vkd3d-shader/tpf: Write out 'switch' statements.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: bc2a4ee66a082d6892d7c01ec6a22a7e85a7ee91 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/bc2a4ee66a082d6892d7c01ec6a22… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Thu Oct 26 22:26:33 2023 +0200 vkd3d-shader/tpf: Write out 'switch' statements. Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- libs/vkd3d-shader/tpf.c | 44 +++++++++++++++++++++++++ tests/hlsl/switch.shader_test | 74 +++++++++++++++++++++---------------------- 2 files changed, 81 insertions(+), 37 deletions(-) diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c index 0ea5a682..0be0ec4f 100644 --- a/libs/vkd3d-shader/tpf.c +++ b/libs/vkd3d-shader/tpf.c @@ -5467,6 +5467,46 @@ static void write_sm4_store(const struct tpf_writer *tpf, const struct hlsl_ir_s write_sm4_instruction(tpf, &instr); } +static void write_sm4_switch(const struct tpf_writer *tpf, const struct hlsl_ir_switch *s) +{ + const struct hlsl_ir_node *selector = s->selector.node; + struct hlsl_ir_switch_case *c; + struct sm4_instruction instr; + + memset(&instr, 0, sizeof(instr)); + instr.opcode = VKD3D_SM4_OP_SWITCH; + + sm4_src_from_node(tpf, &instr.srcs[0], selector, VKD3DSP_WRITEMASK_ALL); + instr.src_count = 1; + + write_sm4_instruction(tpf, &instr); + + LIST_FOR_EACH_ENTRY(c, &s->cases, struct hlsl_ir_switch_case, entry) + { + memset(&instr, 0, sizeof(instr)); + if (c->is_default) + { + instr.opcode = VKD3D_SM4_OP_DEFAULT; + } + else + { + struct hlsl_constant_value value = { .u[0].u = c->value }; + + instr.opcode = VKD3D_SM4_OP_CASE; + sm4_src_from_constant_value(&instr.srcs[0], &value, 1, VKD3DSP_WRITEMASK_ALL); + instr.src_count = 1; + } + + write_sm4_instruction(tpf, &instr); + write_sm4_block(tpf, &c->body); + } + + memset(&instr, 0, sizeof(instr)); + instr.opcode = VKD3D_SM4_OP_ENDSWITCH; + + write_sm4_instruction(tpf, &instr); +} + static void write_sm4_swizzle(const struct tpf_writer *tpf, const struct hlsl_ir_swizzle *swizzle) { unsigned int hlsl_swizzle; @@ -5554,6 +5594,10 @@ static void write_sm4_block(const struct tpf_writer *tpf, const struct hlsl_bloc write_sm4_store(tpf, hlsl_ir_store(instr)); break; + case HLSL_IR_SWITCH: + write_sm4_switch(tpf, hlsl_ir_switch(instr)); + break; + case HLSL_IR_SWIZZLE: write_sm4_swizzle(tpf, hlsl_ir_swizzle(instr)); break; diff --git a/tests/hlsl/switch.shader_test b/tests/hlsl/switch.shader_test index aab12485..243b0b11 100644 --- a/tests/hlsl/switch.shader_test +++ b/tests/hlsl/switch.shader_test @@ -1,7 +1,7 @@ [require] shader model >= 4.0 -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -19,17 +19,17 @@ float4 main() : sv_target [test] uniform 0 uint4 3 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (5.0, 5.0, 5.0, 5.0) uniform 0 uint4 1 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (4.0, 4.0, 4.0, 4.0) uniform 0 uint4 0 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (3.0, 3.0, 3.0, 3.0) % falling through is only supported for empty case statements -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -49,17 +49,17 @@ float4 main() : sv_target [test] uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) uniform 0 uint4 1 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.0, 3.0, 4.0) uniform 0 uint4 0 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.0, 3.0, 4.0) % case value evaluation -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -81,14 +81,14 @@ float4 main() : sv_target [test] uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.1, 3.1, 4.1) uniform 0 uint4 1 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) % floats are accepted -[pixel shader todo fail(sm>=6)] +[pixel shader fail(sm>=6)] uint4 v; float4 main() : sv_target @@ -110,13 +110,13 @@ float4 main() : sv_target [test] uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.1, 3.1, 4.1) uniform 0 uint4 1 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) -[pixel shader todo fail(sm>=6)] +[pixel shader fail(sm>=6)] float4 v; float4 main() : sv_target @@ -138,10 +138,10 @@ float4 main() : sv_target [test] uniform 0 float4 2.0 0.0 0.0 0.0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.1, 3.1, 4.1) uniform 0 float4 1.0 0.0 0.0 0.0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) [pixel shader fail] @@ -312,7 +312,7 @@ float4 main() : sv_target } % more complicated breaks -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -339,17 +339,17 @@ float4 main() : sv_target [test] uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.1, 2.1, 3.1, 4.1) uniform 0 uint4 1 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.2, 2.2, 3.2, 4.2) uniform 0 uint4 0 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) % switch breaks within a loop -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -377,11 +377,11 @@ float4 main() : sv_target [test] uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (5.0, 6.0, 7.0, 8.0) % default case placement -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -408,16 +408,16 @@ float4 main() : sv_target [test] uniform 0 uint4 0 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (4.0, 5.0, 6.0, 7.0) uniform 0 uint4 2 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (2.0, 3.0, 4.0, 5.0) uniform 0 uint4 3 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (4.0, 5.0, 6.0, 7.0) -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -445,13 +445,13 @@ float4 main() : sv_target [test] uniform 0 uint4 3 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) uniform 0 uint4 0 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (4.0, 5.0, 6.0, 7.0) uniform 0 uint4 5 0 0 0 -todo draw quad +todo(sm>=6) draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) % 'continue' is not supported in switches @@ -480,7 +480,7 @@ float4 main() : sv_target return c; } -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -511,14 +511,14 @@ float4 main() : sv_target [test] uniform 0 uint4 0 0 3 1 -todo draw quad +todo(sm>=6) draw quad probe all rgba (10.0, 11.0, 12.0, 13.0) uniform 0 uint4 1 0 3 1 -todo draw quad +todo(sm>=6) draw quad probe all rgba (7.0, 8.0, 9.0, 10.0) % return from a switch nested in a loop -[pixel shader todo] +[pixel shader] uint4 v; float4 main() : sv_target @@ -545,8 +545,8 @@ float4 main() : sv_target [test] uniform 0 uint4 0 0 3 1 -todo draw quad +todo(sm>=6) draw quad probe all rgba (304.0, 305.0, 306.0, 307.0) uniform 0 uint4 1 0 3 1 -todo draw quad +todo(sm>=6) draw quad probe all rgba (3.0, 4.0, 5.0, 6.0)

1 0

Nikolay Sivov : vkd3d-shader/hlsl: Add a pass to normalize switch cases blocks.
by Alexandre Julliard Oct. 31, 2023

Oct. 31, 2023

Module: vkd3d Branch: master Commit: 68c14079a6bd59f1ec2e15e917b0df0592262cd8 URL: https://gitlab.winehq.org/wine/vkd3d/-/commit/68c14079a6bd59f1ec2e15e917b0d… Author: Nikolay Sivov <nsivov(a)codeweavers.com> Date: Wed Oct 11 14:39:01 2023 +0200 vkd3d-shader/hlsl: Add a pass to normalize switch cases blocks. Signed-off-by: Nikolay Sivov <nsivov(a)codeweavers.com> --- libs/vkd3d-shader/hlsl_codegen.c | 113 +++++++++++++++++++++++++++++++++++++++ tests/hlsl/switch.shader_test | 12 ++--- 2 files changed, 119 insertions(+), 6 deletions(-) diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 013667e2..541a45ac 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -2169,6 +2169,118 @@ static bool remove_trivial_conditional_branches(struct hlsl_ctx *ctx, struct hls return true; } +static bool normalize_switch_cases(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, void *context) +{ + struct hlsl_ir_switch_case *c, *def = NULL; + bool missing_terminal_break = false; + struct hlsl_ir_node *node; + struct hlsl_ir_jump *jump; + struct hlsl_ir_switch *s; + + if (instr->type != HLSL_IR_SWITCH) + return false; + s = hlsl_ir_switch(instr); + + LIST_FOR_EACH_ENTRY(c, &s->cases, struct hlsl_ir_switch_case, entry) + { + bool terminal_break = false; + + if (list_empty(&c->body.instrs)) + { + terminal_break = !!list_next(&s->cases, &c->entry); + } + else + { + node = LIST_ENTRY(list_tail(&c->body.instrs), struct hlsl_ir_node, entry); + if (node->type == HLSL_IR_JUMP) + { + jump = hlsl_ir_jump(node); + terminal_break = jump->type == HLSL_IR_JUMP_BREAK; + } + } + + missing_terminal_break |= !terminal_break; + + if (!terminal_break) + { + if (c->is_default) + { + hlsl_error(ctx, &c->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, + "The 'default' case block is not terminated with 'break' or 'return'."); + } + else + { + hlsl_error(ctx, &c->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, + "Switch case block '%u' is not terminated with 'break' or 'return'.", c->value); + } + } + } + + if (missing_terminal_break) + return true; + + LIST_FOR_EACH_ENTRY(c, &s->cases, struct hlsl_ir_switch_case, entry) + { + if (c->is_default) + { + def = c; + + /* Remove preceding empty cases. */ + while (list_prev(&s->cases, &def->entry)) + { + c = LIST_ENTRY(list_prev(&s->cases, &def->entry), struct hlsl_ir_switch_case, entry); + if (!list_empty(&c->body.instrs)) + break; + hlsl_free_ir_switch_case(c); + } + + if (list_empty(&def->body.instrs)) + { + /* Remove following empty cases. */ + while (list_next(&s->cases, &def->entry)) + { + c = LIST_ENTRY(list_next(&s->cases, &def->entry), struct hlsl_ir_switch_case, entry); + if (!list_empty(&c->body.instrs)) + break; + hlsl_free_ir_switch_case(c); + } + + /* Merge with the next case. */ + if (list_next(&s->cases, &def->entry)) + { + c = LIST_ENTRY(list_next(&s->cases, &def->entry), struct hlsl_ir_switch_case, entry); + c->is_default = true; + hlsl_free_ir_switch_case(def); + def = c; + } + } + + break; + } + } + + if (def) + { + list_remove(&def->entry); + } + else + { + struct hlsl_ir_node *jump; + + if (!(def = hlsl_new_switch_case(ctx, 0, true, NULL, &s->node.loc))) + return true; + if (!(jump = hlsl_new_jump(ctx, HLSL_IR_JUMP_BREAK, NULL, &s->node.loc))) + { + hlsl_free_ir_switch_case(def); + return true; + } + hlsl_block_add_instr(&def->body, jump); + } + list_add_tail(&s->cases, &def->entry); + + return true; +} + static bool lower_nonconstant_vector_derefs(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, struct hlsl_block *block) { struct hlsl_ir_node *idx; @@ -4629,6 +4741,7 @@ int hlsl_emit_bytecode(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry } while (progress); remove_unreachable_code(ctx, body); + hlsl_transform_ir(ctx, normalize_switch_cases, body, NULL); lower_ir(ctx, lower_nonconstant_vector_derefs, body); lower_ir(ctx, lower_casts_to_bool, body); diff --git a/tests/hlsl/switch.shader_test b/tests/hlsl/switch.shader_test index 720672a7..aab12485 100644 --- a/tests/hlsl/switch.shader_test +++ b/tests/hlsl/switch.shader_test @@ -211,7 +211,7 @@ float4 main() : sv_target } % unterminated cases -[pixel shader fail(sm<6) todo] +[pixel shader fail(sm<6)] uint4 v; float4 main() : sv_target @@ -230,7 +230,7 @@ float4 main() : sv_target return c; } -[pixel shader fail todo] +[pixel shader fail] uint4 v; float4 main() : sv_target @@ -246,7 +246,7 @@ float4 main() : sv_target return 0.0; } -[pixel shader fail todo] +[pixel shader fail] uint4 v; float4 main() : sv_target @@ -262,7 +262,7 @@ float4 main() : sv_target return 0.0; } -[pixel shader fail(sm<6) todo] +[pixel shader fail(sm<6)] uint4 v; float4 main() : sv_target @@ -279,7 +279,7 @@ float4 main() : sv_target return 0.0; } -[pixel shader fail(sm<6) todo] +[pixel shader fail(sm<6)] uint4 v; float4 main() : sv_target @@ -296,7 +296,7 @@ float4 main() : sv_target return 0.0; } -[pixel shader fail(sm<6) todo] +[pixel shader fail(sm<6)] uint4 v; float4 main() : sv_target

1 0