[PATCH v4 0/4] MR450: vkd3d-shader/ir: Flatten structured control flow instructions.
-- v4: vkd3d-shader/ir: Store code block names in struct vkd3d_shader_desc. vkd3d-shader/ir: Flatten structured control flow instructions. https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450
From: Conor McCarthy <cmccarthy(a)codeweavers.com> Ensures they are loaded only once per function independent of the control flow graph. --- libs/vkd3d-shader/spirv.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 61d75757..1745d521 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -2370,6 +2370,7 @@ static bool is_in_fork_or_join_phase(const struct spirv_compiler *compiler) } static void spirv_compiler_emit_initial_declarations(struct spirv_compiler *compiler); +static size_t spirv_compiler_get_current_function_location(struct spirv_compiler *compiler); static const char *spirv_compiler_get_entry_point_name(const struct spirv_compiler *compiler) { @@ -3444,11 +3445,13 @@ static uint32_t spirv_compiler_get_descriptor_index(struct spirv_compiler *compi index_ids[0] = compiler->descriptor_offsets_member_id; index_ids[1] = spirv_compiler_get_constant_uint(compiler, push_constant_index); ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassPushConstant, type_id); + vkd3d_spirv_begin_function_stream_insertion(builder, + spirv_compiler_get_current_function_location(compiler)); ptr_id = vkd3d_spirv_build_op_in_bounds_access_chain(builder, ptr_type_id, compiler->push_constants_var_id, index_ids, 2); offset_id = vkd3d_spirv_build_op_load(builder, type_id, ptr_id, SpvMemoryAccessMaskNone); - if (!compiler->control_flow_depth) - compiler->descriptor_offset_ids[push_constant_index] = offset_id; + vkd3d_spirv_end_function_stream_insertion(builder); + compiler->descriptor_offset_ids[push_constant_index] = offset_id; } index_id = vkd3d_spirv_build_op_iadd(builder, type_id, index_id, offset_id); } @@ -6523,6 +6526,9 @@ static void spirv_compiler_emit_hull_shader_main(struct spirv_compiler *compiler struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; uint32_t void_id; + /* If a patch constant function used descriptor indexing the offsets must be reloaded. */ + memset(compiler->descriptor_offset_ids, 0, compiler->offset_info.descriptor_table_count + * sizeof(*compiler->descriptor_offset_ids)); vkd3d_spirv_builder_begin_main_function(builder); void_id = vkd3d_spirv_get_op_type_void(builder); -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450
From: Conor McCarthy <cmccarthy(a)codeweavers.com> --- libs/vkd3d-shader/dxil.c | 2 +- libs/vkd3d-shader/tpf.c | 2 +- libs/vkd3d-shader/vkd3d_shader_main.c | 18 +++++++++--------- libs/vkd3d-shader/vkd3d_shader_private.h | 10 +++++----- 4 files changed, 16 insertions(+), 16 deletions(-) diff --git a/libs/vkd3d-shader/dxil.c b/libs/vkd3d-shader/dxil.c index e0e242cb..a59692dc 100644 --- a/libs/vkd3d-shader/dxil.c +++ b/libs/vkd3d-shader/dxil.c @@ -2307,7 +2307,7 @@ static enum vkd3d_result register_allocate_constant_array(struct vkd3d_shader_re "Out of memory allocating an immediate constant buffer of count %u.", count); return VKD3D_ERROR_OUT_OF_MEMORY; } - if ((reg->idx[0].offset = shader_instruction_array_add_icb(&sm6->p.instructions, icb)) == UINT_MAX) + if ((reg->idx[0].offset = shader_instruction_array_add_opaque_param(&sm6->p.instructions, icb)) == UINT_MAX) { ERR("Failed to store icb object.\n"); vkd3d_free(icb); diff --git a/libs/vkd3d-shader/tpf.c b/libs/vkd3d-shader/tpf.c index a9d89bd7..381e1ccf 100644 --- a/libs/vkd3d-shader/tpf.c +++ b/libs/vkd3d-shader/tpf.c @@ -796,7 +796,7 @@ static void shader_sm4_read_shader_data(struct vkd3d_shader_instruction *ins, ui icb->component_count = VKD3D_VEC4_SIZE; icb->element_count = icb_size / VKD3D_VEC4_SIZE; memcpy(icb->data, tokens, sizeof(*tokens) * icb_size); - shader_instruction_array_add_icb(&priv->p.instructions, icb); + shader_instruction_array_add_opaque_param(&priv->p.instructions, icb); ins->declaration.icb = icb; } diff --git a/libs/vkd3d-shader/vkd3d_shader_main.c b/libs/vkd3d-shader/vkd3d_shader_main.c index 28d647b3..55940da8 100644 --- a/libs/vkd3d-shader/vkd3d_shader_main.c +++ b/libs/vkd3d-shader/vkd3d_shader_main.c @@ -1864,14 +1864,14 @@ bool shader_instruction_array_reserve(struct vkd3d_shader_instruction_array *ins return true; } -unsigned int shader_instruction_array_add_icb(struct vkd3d_shader_instruction_array *instructions, - struct vkd3d_shader_immediate_constant_buffer *icb) +unsigned int shader_instruction_array_add_opaque_param(struct vkd3d_shader_instruction_array *instructions, + void *param) { - if (!vkd3d_array_reserve((void **)&instructions->icbs, &instructions->icb_capacity, instructions->icb_count + 1, - sizeof(*instructions->icbs))) + if (!vkd3d_array_reserve((void **)&instructions->opaque_params, &instructions->opaque_param_capacity, + instructions->opaque_param_count + 1, sizeof(*instructions->opaque_params))) return UINT_MAX; - instructions->icbs[instructions->icb_count] = icb; - return instructions->icb_count++; + instructions->opaque_params[instructions->opaque_param_count] = param; + return instructions->opaque_param_count++; } static struct vkd3d_shader_src_param *shader_instruction_array_clone_src_params( @@ -1959,9 +1959,9 @@ void shader_instruction_array_destroy(struct vkd3d_shader_instruction_array *ins vkd3d_free(instructions->elements); shader_param_allocator_destroy(&instructions->dst_params); shader_param_allocator_destroy(&instructions->src_params); - for (i = 0; i < instructions->icb_count; ++i) - vkd3d_free(instructions->icbs[i]); - vkd3d_free(instructions->icbs); + for (i = 0; i < instructions->opaque_param_count; ++i) + vkd3d_free(instructions->opaque_params[i]); + vkd3d_free(instructions->opaque_params); } void vkd3d_shader_build_varying_map(const struct vkd3d_shader_signature *output_signature, diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index 0871f8b1..fd812a6c 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -1188,15 +1188,15 @@ struct vkd3d_shader_instruction_array struct vkd3d_shader_param_allocator src_params; struct vkd3d_shader_param_allocator dst_params; - struct vkd3d_shader_immediate_constant_buffer **icbs; - size_t icb_capacity; - size_t icb_count; + void **opaque_params; + size_t opaque_param_capacity; + size_t opaque_param_count; }; bool shader_instruction_array_init(struct vkd3d_shader_instruction_array *instructions, unsigned int reserve); bool shader_instruction_array_reserve(struct vkd3d_shader_instruction_array *instructions, unsigned int reserve); -unsigned int shader_instruction_array_add_icb(struct vkd3d_shader_instruction_array *instructions, - struct vkd3d_shader_immediate_constant_buffer *icb); +unsigned int shader_instruction_array_add_opaque_param(struct vkd3d_shader_instruction_array *instructions, + void *param); bool shader_instruction_array_clone_instruction(struct vkd3d_shader_instruction_array *instructions, unsigned int dst, unsigned int src); void shader_instruction_array_destroy(struct vkd3d_shader_instruction_array *instructions); -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450
From: Conor McCarthy <cmccarthy(a)codeweavers.com> --- libs/vkd3d-shader/d3d_asm.c | 40 +- libs/vkd3d-shader/ir.c | 560 +++++++++++++++++++++++ libs/vkd3d-shader/spirv.c | 548 +++++----------------- libs/vkd3d-shader/vkd3d_shader_private.h | 26 ++ 4 files changed, 741 insertions(+), 433 deletions(-) diff --git a/libs/vkd3d-shader/d3d_asm.c b/libs/vkd3d-shader/d3d_asm.c index 5cea0c0c..ae9efdb3 100644 --- a/libs/vkd3d-shader/d3d_asm.c +++ b/libs/vkd3d-shader/d3d_asm.c @@ -44,6 +44,7 @@ static const char * const shader_opcode_names[] = [VKD3DSIH_BEM ] = "bem", [VKD3DSIH_BFI ] = "bfi", [VKD3DSIH_BFREV ] = "bfrev", + [VKD3DSIH_BRANCH ] = "branch", [VKD3DSIH_BREAK ] = "break", [VKD3DSIH_BREAKC ] = "breakc", [VKD3DSIH_BREAKP ] = "breakp", @@ -1516,6 +1517,7 @@ static void shader_dump_instruction_flags(struct vkd3d_d3d_asm_compiler *compile switch (ins->handler_idx) { + case VKD3DSIH_BRANCH: case VKD3DSIH_BREAKP: case VKD3DSIH_CONTINUEP: case VKD3DSIH_DISCARD: @@ -1610,6 +1612,27 @@ static void shader_dump_instruction(struct vkd3d_d3d_asm_compiler *compiler, switch (ins->handler_idx) { + case VKD3DSIH_BRANCH: + { + const struct vkd3d_shader_branch *branch = &ins->declaration.branch; + if (ins->src_count) + { + shader_dump_instruction_flags(compiler, ins); + vkd3d_string_buffer_printf(buffer, " "); + shader_dump_src_param(compiler, ins->src); + vkd3d_string_buffer_printf(buffer, " ? l%u : l%u", branch->true_id, branch->false_id); + } + else + { + vkd3d_string_buffer_printf(buffer, " l%u", branch->true_id); + } + if (branch->merge_block_id) + vkd3d_string_buffer_printf(buffer, ", merge l%u", branch->merge_block_id); + if (branch->continue_block_id) + vkd3d_string_buffer_printf(buffer, ", continue l%u", branch->continue_block_id); + break; + } + case VKD3DSIH_DCL: case VKD3DSIH_DCL_UAV_TYPED: vkd3d_string_buffer_printf(buffer, "%s", compiler->colours.opcode); @@ -1825,6 +1848,19 @@ static void shader_dump_instruction(struct vkd3d_d3d_asm_compiler *compiler, shader_print_bool_literal(compiler, " = ", ins->src[0].reg.u.immconst_uint[0], ""); break; + case VKD3DSIH_SWITCH: + if (ins->declaration.switch_.case_count) + { + /* New monolithic switch variant. */ + const struct vkd3d_shader_switch *switch_ = &ins->declaration.switch_; + vkd3d_string_buffer_printf(buffer, " "); + shader_dump_src_param(compiler, ins->src); + for (i = 0; i < switch_->case_count; ++i) + vkd3d_string_buffer_printf(buffer, ", %u: l%u", switch_->cases[i].value, switch_->cases[i].block_id); + vkd3d_string_buffer_printf(buffer, ", default: l%u", switch_->default_id); + break; + } + /* fall through */ default: shader_dump_instruction_flags(compiler, ins); @@ -1972,9 +2008,11 @@ enum vkd3d_result vkd3d_dxbc_binary_to_text(const struct vkd3d_shader_instructio case VKD3DSIH_ELSE: case VKD3DSIH_IF: case VKD3DSIH_LOOP: - case VKD3DSIH_SWITCH: ++indent; break; + case VKD3DSIH_SWITCH: + indent += !ins->declaration.switch_.case_count; + break; default: break; diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 16e5948f..3c535d56 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -1424,6 +1424,544 @@ static enum vkd3d_result normalise_combined_samplers(struct vkd3d_shader_parser return VKD3D_OK; } +struct vkd3d_if_cf_info +{ + size_t ins_location; + unsigned int id; + uint32_t merge_block_id; + unsigned int else_block_id; +}; + +struct vkd3d_loop_cf_info +{ + unsigned int header_block_id; + unsigned int continue_block_id; + uint32_t merge_block_id; +}; + +struct vkd3d_switch_cf_info +{ + size_t ins_location; + unsigned int id; + unsigned int merge_block_id; + unsigned int default_block_id; + struct vkd3d_shader_switch_case *cases; + size_t cases_size; + unsigned int cases_count; +}; + +struct vkd3d_control_flow_info +{ + union + { + struct vkd3d_if_cf_info if_; + struct vkd3d_loop_cf_info loop; + struct vkd3d_switch_cf_info switch_; + } u; + + enum + { + VKD3D_BLOCK_IF, + VKD3D_BLOCK_LOOP, + VKD3D_BLOCK_SWITCH, + } current_block; + bool inside_block; +}; + +struct cf_flattener +{ + struct vkd3d_shader_parser *parser; + + struct vkd3d_shader_location location; + bool allocation_failed; + + struct vkd3d_shader_instruction *instructions; + size_t instruction_capacity; + size_t instruction_count; + + unsigned int block_id; + + unsigned int branch_id; + unsigned int loop_id; + unsigned int switch_id; + unsigned int control_flow_depth; + struct vkd3d_control_flow_info *control_flow_info; + size_t control_flow_info_size; +}; + +static struct vkd3d_shader_instruction *cf_flattener_require_space(struct cf_flattener *flattener, size_t count) +{ + if (!vkd3d_array_reserve((void **)&flattener->instructions, &flattener->instruction_capacity, + flattener->instruction_count + count, sizeof(*flattener->instructions))) + { + ERR("Failed to allocate instructions.\n"); + flattener->allocation_failed = true; + return NULL; + } + return &flattener->instructions[flattener->instruction_count]; +} + +static bool cf_flattener_copy_instruction(struct cf_flattener *flattener, + const struct vkd3d_shader_instruction *instruction) +{ + struct vkd3d_shader_instruction *dst_ins; + + if (instruction->handler_idx == VKD3DSIH_NOP) + return true; + + if (!(dst_ins = cf_flattener_require_space(flattener, 1))) + return false; + + *dst_ins = *instruction; + ++flattener->instruction_count; + return true; +} + +static unsigned int cf_flattener_alloc_block_id(struct cf_flattener *flattener) +{ + return ++flattener->block_id; +} + +static void cf_flattener_emit_label(struct cf_flattener *flattener, unsigned int label_id) +{ + struct vkd3d_shader_src_param *src_param; + struct vkd3d_shader_instruction *ins; + + if (!(src_param = shader_parser_get_src_params(flattener->parser, 1))) + { + flattener->allocation_failed = true; + return; + } + if (!(ins = cf_flattener_require_space(flattener, 1))) + return; + + vsir_register_init(&src_param->reg, VKD3DSPR_LABEL, VKD3D_DATA_UINT, 1); + src_param->reg.idx[0].offset = label_id; + src_param->swizzle = 0; + src_param->modifiers = 0; + + vsir_instruction_init(ins, &flattener->location, VKD3DSIH_LABEL); + ins->src = src_param; + ins->src_count = 1; + ++flattener->instruction_count; +} + +static void cf_flattener_emit_branch(struct cf_flattener *flattener, + unsigned int merge_block_id, unsigned int continue_block_id, + const struct vkd3d_shader_src_param *condition, unsigned int true_id, unsigned int false_id, + unsigned int flags) +{ + struct vkd3d_shader_instruction *ins; + + if (!(ins = cf_flattener_require_space(flattener, 1))) + return; + vsir_instruction_init(ins, &flattener->location, VKD3DSIH_BRANCH); + ins->src = condition; + ins->src_count = !!condition; + ins->declaration.branch.merge_block_id = merge_block_id; + ins->declaration.branch.continue_block_id = continue_block_id; + ins->declaration.branch.true_id = true_id; + ins->declaration.branch.false_id = false_id; + ins->flags = flags; + ++flattener->instruction_count; +} + +static void cf_flattener_emit_conditional_branch_and_merge(struct cf_flattener *flattener, + const struct vkd3d_shader_src_param *condition, unsigned int true_id, unsigned int flags) +{ + unsigned int merge_block_id; + + merge_block_id = cf_flattener_alloc_block_id(flattener); + cf_flattener_emit_branch(flattener, merge_block_id, 0, condition, true_id, merge_block_id, flags); + cf_flattener_emit_label(flattener, merge_block_id); +} + +static void cf_flattener_emit_unconditional_branch(struct cf_flattener *flattener, unsigned int target_block_id) +{ + cf_flattener_emit_branch(flattener, 0, 0, NULL, target_block_id, 0, 0); +} + +static struct vkd3d_control_flow_info *cf_flattener_push_control_flow_level( + struct cf_flattener *flattener) +{ + if (!vkd3d_array_reserve((void **)&flattener->control_flow_info, &flattener->control_flow_info_size, + flattener->control_flow_depth + 1, sizeof(*flattener->control_flow_info))) + { + ERR("Failed to allocate control flow info structure.\n"); + flattener->allocation_failed = true; + return NULL; + } + + return &flattener->control_flow_info[flattener->control_flow_depth++]; +} + +static void cf_flattener_pop_control_flow_level(struct cf_flattener *flattener) +{ + struct vkd3d_control_flow_info *cf_info; + + assert(flattener->control_flow_depth); + + cf_info = &flattener->control_flow_info[--flattener->control_flow_depth]; + memset(cf_info, 0, sizeof(*cf_info)); +} + +static struct vkd3d_control_flow_info *cf_flattener_find_innermost_loop( + struct cf_flattener *flattener) +{ + int depth; + + for (depth = flattener->control_flow_depth - 1; depth >= 0; --depth) + { + if (flattener->control_flow_info[depth].current_block == VKD3D_BLOCK_LOOP) + return &flattener->control_flow_info[depth]; + } + + return NULL; +} + +static struct vkd3d_control_flow_info *cf_flattener_find_innermost_breakable_cf_construct( + struct cf_flattener *flattener) +{ + int depth; + + for (depth = flattener->control_flow_depth - 1; depth >= 0; --depth) + { + if (flattener->control_flow_info[depth].current_block == VKD3D_BLOCK_LOOP + || flattener->control_flow_info[depth].current_block == VKD3D_BLOCK_SWITCH) + return &flattener->control_flow_info[depth]; + } + + return NULL; +} + +static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_parser *parser, + struct cf_flattener *flattener) +{ + struct vkd3d_shader_instruction_array *instructions = &parser->instructions; + struct vkd3d_shader_instruction *dst_ins; + bool main_block_open; + size_t i; + + flattener->parser = parser; + main_block_open = parser->shader_version.type != VKD3D_SHADER_TYPE_HULL; + + if (!cf_flattener_require_space(flattener, instructions->count)) + return VKD3D_ERROR_OUT_OF_MEMORY; + + for (i = 0; i < instructions->count; ++i) + { + unsigned int loop_header_block_id, loop_body_block_id, continue_block_id, merge_block_id, true_block_id; + const struct vkd3d_shader_instruction *instruction = &instructions->elements[i]; + const struct vkd3d_shader_src_param *src = instruction->src; + struct vkd3d_control_flow_info *cf_info; + + flattener->location = instruction->location; + + cf_info = flattener->control_flow_depth + ? &flattener->control_flow_info[flattener->control_flow_depth - 1] : NULL; + + switch (instruction->handler_idx) + { + case VKD3DSIH_IF: + if (!(cf_info = cf_flattener_push_control_flow_level(flattener))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + true_block_id = cf_flattener_alloc_block_id(flattener); + merge_block_id = cf_flattener_alloc_block_id(flattener); + cf_info->u.if_.ins_location = flattener->instruction_count; + cf_flattener_emit_branch(flattener, merge_block_id, 0, + src, true_block_id, merge_block_id, instruction->flags); + + cf_flattener_emit_label(flattener, true_block_id); + + cf_info->u.if_.id = flattener->branch_id; + cf_info->u.if_.merge_block_id = merge_block_id; + cf_info->u.if_.else_block_id = 0; + cf_info->inside_block = true; + cf_info->current_block = VKD3D_BLOCK_IF; + + ++flattener->branch_id; + break; + + case VKD3DSIH_ELSE: + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_IF); + + if (cf_info->inside_block) + cf_flattener_emit_unconditional_branch(flattener, cf_info->u.if_.merge_block_id); + + cf_info->u.if_.else_block_id = cf_flattener_alloc_block_id(flattener); + dst_ins = &flattener->instructions[cf_info->u.if_.ins_location]; + dst_ins->declaration.branch.false_id = cf_info->u.if_.else_block_id; + + cf_flattener_emit_label(flattener, cf_info->u.if_.else_block_id); + + cf_info->inside_block = true; + break; + + case VKD3DSIH_ENDIF: + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_IF); + + if (cf_info->inside_block) + cf_flattener_emit_unconditional_branch(flattener, cf_info->u.if_.merge_block_id); + + cf_flattener_emit_label(flattener, cf_info->u.if_.merge_block_id); + + cf_flattener_pop_control_flow_level(flattener); + break; + + case VKD3DSIH_LOOP: + if (!(cf_info = cf_flattener_push_control_flow_level(flattener))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + loop_header_block_id = cf_flattener_alloc_block_id(flattener); + loop_body_block_id = cf_flattener_alloc_block_id(flattener); + continue_block_id = cf_flattener_alloc_block_id(flattener); + merge_block_id = cf_flattener_alloc_block_id(flattener); + + cf_flattener_emit_unconditional_branch(flattener, loop_header_block_id); + cf_flattener_emit_label(flattener, loop_header_block_id); + cf_flattener_emit_branch(flattener, merge_block_id, continue_block_id, + NULL, loop_body_block_id, 0, 0); + + cf_flattener_emit_label(flattener, loop_body_block_id); + + cf_info->u.loop.header_block_id = loop_header_block_id; + cf_info->u.loop.continue_block_id = continue_block_id; + cf_info->u.loop.merge_block_id = merge_block_id; + cf_info->current_block = VKD3D_BLOCK_LOOP; + cf_info->inside_block = true; + + ++flattener->loop_id; + break; + + case VKD3DSIH_ENDLOOP: + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_LOOP); + + if (cf_info->inside_block) + cf_flattener_emit_unconditional_branch(flattener, cf_info->u.loop.continue_block_id); + + cf_flattener_emit_label(flattener, cf_info->u.loop.continue_block_id); + cf_flattener_emit_unconditional_branch(flattener, cf_info->u.loop.header_block_id); + cf_flattener_emit_label(flattener, cf_info->u.loop.merge_block_id); + + cf_flattener_pop_control_flow_level(flattener); + break; + + case VKD3DSIH_SWITCH: + if (!(cf_info = cf_flattener_push_control_flow_level(flattener))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + merge_block_id = cf_flattener_alloc_block_id(flattener); + + assert(src->reg.data_type == VKD3D_DATA_INT); + + cf_info->u.if_.ins_location = flattener->instruction_count; + + if (!(dst_ins = cf_flattener_require_space(flattener, 1))) + return VKD3D_ERROR_OUT_OF_MEMORY; + vsir_instruction_init(dst_ins, &instruction->location, VKD3DSIH_SWITCH); + dst_ins->src = src; + dst_ins->src_count = 1; + dst_ins->declaration.switch_.merge_block_id = merge_block_id; + dst_ins->declaration.switch_.cases = NULL; + dst_ins->declaration.switch_.case_count = 0; + dst_ins->declaration.switch_.default_id = 0; + ++flattener->instruction_count; + + cf_info->u.switch_.id = flattener->switch_id; + cf_info->u.switch_.merge_block_id = merge_block_id; + cf_info->u.switch_.cases = NULL; + cf_info->u.switch_.cases_size = 0; + cf_info->u.switch_.cases_count = 0; + cf_info->u.switch_.default_block_id = 0; + cf_info->inside_block = false; + cf_info->current_block = VKD3D_BLOCK_SWITCH; + + ++flattener->switch_id; + + if (!vkd3d_array_reserve((void **)&cf_info->u.switch_.cases, &cf_info->u.switch_.cases_size, + 10, sizeof(*cf_info->u.switch_.cases))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + break; + + case VKD3DSIH_ENDSWITCH: + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); + assert(!cf_info->inside_block); + + if (!cf_info->u.switch_.default_block_id) + cf_info->u.switch_.default_block_id = cf_info->u.switch_.merge_block_id; + + cf_flattener_emit_label(flattener, cf_info->u.switch_.merge_block_id); + + /* The SWITCH instruction is completed when the endswitch + * instruction is processed because we do not know the number + * of case statements or the default block id in advance.*/ + dst_ins = &flattener->instructions[cf_info->u.switch_.ins_location]; + dst_ins->declaration.switch_.cases = cf_info->u.switch_.cases; + dst_ins->declaration.switch_.case_count = cf_info->u.switch_.cases_count; + dst_ins->declaration.switch_.default_id = cf_info->u.switch_.default_block_id; + shader_instruction_array_add_opaque_param(instructions, cf_info->u.switch_.cases); + + cf_flattener_pop_control_flow_level(flattener); + break; + + case VKD3DSIH_CASE: + { + unsigned int label_id, value; + + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); + + if (src->swizzle != VKD3D_SHADER_SWIZZLE(X, X, X, X)) + { + WARN("Unexpected src swizzle %#x.\n", src->swizzle); + vkd3d_shader_parser_error(parser, VKD3D_SHADER_ERROR_VSIR_INVALID_SWIZZLE, + "The swizzle for a switch case value is not scalar X."); + } + assert(src->reg.type == VKD3DSPR_IMMCONST); + value = *src->reg.u.immconst_uint; + + if (!vkd3d_array_reserve((void **)&cf_info->u.switch_.cases, &cf_info->u.switch_.cases_size, + cf_info->u.switch_.cases_count + 1, sizeof(*cf_info->u.switch_.cases))) + return VKD3D_ERROR_OUT_OF_MEMORY; + + label_id = cf_flattener_alloc_block_id(flattener); + if (cf_info->inside_block) /* fall-through */ + cf_flattener_emit_unconditional_branch(flattener, label_id); + + cf_info->u.switch_.cases[cf_info->u.switch_.cases_count].value = value; + cf_info->u.switch_.cases[cf_info->u.switch_.cases_count].block_id = label_id; + ++cf_info->u.switch_.cases_count; + + cf_flattener_emit_label(flattener, label_id); + cf_info->inside_block = true; + break; + } + + case VKD3DSIH_DEFAULT: + assert(flattener->control_flow_depth); + assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); + assert(!cf_info->u.switch_.default_block_id); + + cf_info->u.switch_.default_block_id = cf_flattener_alloc_block_id(flattener); + if (cf_info->inside_block) /* fall-through */ + cf_flattener_emit_unconditional_branch(flattener, cf_info->u.switch_.default_block_id); + + cf_flattener_emit_label(flattener, cf_info->u.switch_.default_block_id); + cf_info->inside_block = true; + break; + + case VKD3DSIH_BREAK: + { + struct vkd3d_control_flow_info *breakable_cf_info; + + assert(flattener->control_flow_depth); + + if (!(breakable_cf_info = cf_flattener_find_innermost_breakable_cf_construct(flattener))) + { + FIXME("Unhandled break instruction.\n"); + return VKD3D_ERROR_INVALID_SHADER; + } + + if (breakable_cf_info->current_block == VKD3D_BLOCK_LOOP) + { + cf_flattener_emit_unconditional_branch(flattener, breakable_cf_info->u.loop.merge_block_id); + } + else if (breakable_cf_info->current_block == VKD3D_BLOCK_SWITCH) + { + /* The current case block may have already been ended by an + * unconditional continue instruction. */ + if (breakable_cf_info->inside_block) + cf_flattener_emit_unconditional_branch(flattener, breakable_cf_info->u.switch_.merge_block_id); + } + + cf_info->inside_block = false; + break; + } + + case VKD3DSIH_BREAKP: + { + struct vkd3d_control_flow_info *loop_cf_info; + + assert(flattener->control_flow_depth); + + if (!(loop_cf_info = cf_flattener_find_innermost_loop(flattener))) + { + ERR("Invalid 'breakc' instruction outside loop.\n"); + return VKD3D_ERROR_INVALID_SHADER; + } + + cf_flattener_emit_conditional_branch_and_merge(flattener, + src, loop_cf_info->u.loop.merge_block_id, instruction->flags); + break; + } + + case VKD3DSIH_CONTINUE: + { + struct vkd3d_control_flow_info *loop_cf_info; + + assert(flattener->control_flow_depth); + + if (!(loop_cf_info = cf_flattener_find_innermost_loop(flattener))) + { + ERR("Invalid 'continue' instruction outside loop.\n"); + return VKD3D_ERROR_INVALID_SHADER; + } + + cf_flattener_emit_unconditional_branch(flattener, loop_cf_info->u.loop.continue_block_id); + + cf_info->inside_block = false; + break; + } + + case VKD3DSIH_CONTINUEP: + { + struct vkd3d_control_flow_info *loop_cf_info; + + if (!(loop_cf_info = cf_flattener_find_innermost_loop(flattener))) + { + ERR("Invalid 'continuec' instruction outside loop.\n"); + return VKD3D_ERROR_INVALID_SHADER; + } + + cf_flattener_emit_conditional_branch_and_merge(flattener, + src, loop_cf_info->u.loop.continue_block_id, instruction->flags); + break; + } + + case VKD3DSIH_RET: + if (!cf_flattener_copy_instruction(flattener, instruction)) + return VKD3D_ERROR_OUT_OF_MEMORY; + + if (cf_info) + cf_info->inside_block = false; + else + main_block_open = false; + break; + + default: + if (!cf_flattener_copy_instruction(flattener, instruction)) + return VKD3D_ERROR_OUT_OF_MEMORY; + break; + } + } + + if (main_block_open) + { + if (!(dst_ins = cf_flattener_require_space(flattener, 1))) + return VKD3D_ERROR_OUT_OF_MEMORY; + vsir_instruction_init(dst_ins, &flattener->location, VKD3DSIH_RET); + ++flattener->instruction_count; + } + + return flattener->allocation_failed ? VKD3D_ERROR_OUT_OF_MEMORY : VKD3D_OK; +} + enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser, const struct vkd3d_shader_compile_info *compile_info) { @@ -1452,6 +1990,25 @@ enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser, if (result >= 0) remove_dead_code(parser); + if (result >= 0) + { + struct cf_flattener flattener = {0}; + result = flatten_control_flow_constructs(parser, &flattener); + vkd3d_free(flattener.control_flow_info); + if (result >= 0) + { + vkd3d_free(parser->instructions.elements); + parser->instructions.elements = flattener.instructions; + parser->instructions.capacity = flattener.instruction_capacity; + parser->instructions.count = flattener.instruction_count; + parser->shader_desc.block_count = flattener.block_id; + } + else + { + vkd3d_free(flattener.instructions); + } + } + if (result >= 0) result = normalise_combined_samplers(parser); @@ -1699,6 +2256,9 @@ static void vsir_validate_instruction(struct validation_context *ctx) case VKD3DSIH_SWITCH: vsir_validate_dst_count(ctx, instruction, 0); vsir_validate_src_count(ctx, instruction, 1); + /* New monolithic switch variant. */ + if (instruction->declaration.switch_.case_count) + break; if (!vkd3d_array_reserve((void **)&ctx->blocks, &ctx->blocks_capacity, ctx->depth + 1, sizeof(*ctx->blocks))) return; ctx->blocks[ctx->depth++] = instruction->handler_idx; diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 1745d521..54436bd3 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -348,6 +348,12 @@ static bool vkd3d_spirv_stream_append(struct vkd3d_spirv_stream *dst_stream, return true; } +static void vkd3d_spirv_stream_add_word_count(struct vkd3d_spirv_stream *stream, size_t location, + unsigned int word_count) +{ + stream->words[location] += word_count << SpvWordCountShift; +} + struct vkd3d_spirv_builder { uint64_t capability_mask; @@ -851,20 +857,6 @@ static void vkd3d_spirv_end_function_stream_insertion(struct vkd3d_spirv_builder builder->insertion_location = ~(size_t)0; } -struct vkd3d_spirv_op_branch_conditional -{ - uint32_t opcode; - uint32_t condition_id; - uint32_t true_label; - uint32_t false_label; -}; - -static struct vkd3d_spirv_op_branch_conditional *vkd3d_spirv_as_op_branch_conditional( - struct vkd3d_spirv_stream *stream, size_t location) -{ - return (struct vkd3d_spirv_op_branch_conditional *)&stream->words[location]; -} - static void vkd3d_spirv_build_op_capability(struct vkd3d_spirv_stream *stream, SpvCapability cap) { @@ -2206,51 +2198,6 @@ static const char *debug_vkd3d_symbol(const struct vkd3d_symbol *symbol) } } -struct vkd3d_if_cf_info -{ - size_t stream_location; - unsigned int id; - uint32_t merge_block_id; - uint32_t else_block_id; -}; - -struct vkd3d_loop_cf_info -{ - uint32_t header_block_id; - uint32_t continue_block_id; - uint32_t merge_block_id; -}; - -struct vkd3d_switch_cf_info -{ - size_t stream_location; - unsigned int id; - uint32_t selector_id; - uint32_t merge_block_id; - uint32_t default_block_id; - uint32_t *case_blocks; - size_t case_blocks_size; - unsigned int case_block_count; -}; - -struct vkd3d_control_flow_info -{ - union - { - struct vkd3d_if_cf_info if_; - struct vkd3d_loop_cf_info loop; - struct vkd3d_switch_cf_info switch_; - } u; - - enum - { - VKD3D_BLOCK_IF, - VKD3D_BLOCK_LOOP, - VKD3D_BLOCK_SWITCH, - } current_block; - bool inside_block; -}; - struct vkd3d_push_constant_buffer_binding { struct vkd3d_shader_register reg; @@ -2297,13 +2244,6 @@ struct spirv_compiler enum vkd3d_shader_type shader_type; - unsigned int branch_id; - unsigned int loop_id; - unsigned int switch_id; - unsigned int control_flow_depth; - struct vkd3d_control_flow_info *control_flow_info; - size_t control_flow_info_size; - struct vkd3d_shader_interface_info shader_interface; struct vkd3d_shader_descriptor_offset_info offset_info; uint32_t descriptor_offsets_member_id; @@ -2312,7 +2252,6 @@ struct spirv_compiler struct vkd3d_push_constant_buffer_binding *push_constants; const struct vkd3d_shader_spirv_target_info *spirv_target_info; - bool main_block_open; bool after_declarations_section; struct shader_signature input_signature; struct shader_signature output_signature; @@ -2352,6 +2291,9 @@ struct spirv_compiler uint32_t *ssa_register_ids; unsigned int ssa_register_count; + + uint32_t *block_label_ids; + unsigned int block_count; }; static bool is_in_default_phase(const struct spirv_compiler *compiler) @@ -2381,8 +2323,6 @@ static const char *spirv_compiler_get_entry_point_name(const struct spirv_compil static void spirv_compiler_destroy(struct spirv_compiler *compiler) { - vkd3d_free(compiler->control_flow_info); - vkd3d_free(compiler->output_info); vkd3d_free(compiler->push_constants); @@ -2401,6 +2341,7 @@ static void spirv_compiler_destroy(struct spirv_compiler *compiler) shader_signature_cleanup(&compiler->patch_constant_signature); vkd3d_free(compiler->ssa_register_ids); + vkd3d_free(compiler->block_label_ids); vkd3d_free(compiler); } @@ -2720,6 +2661,15 @@ static struct vkd3d_string_buffer *vkd3d_shader_register_range_string(struct spi return buffer; } +static uint32_t spirv_compiler_get_label_id(struct spirv_compiler *compiler, unsigned int block_id) +{ + assert(block_id); + --block_id; + if (!compiler->block_label_ids[block_id]) + compiler->block_label_ids[block_id] = vkd3d_spirv_alloc_id(&compiler->spirv_builder); + return compiler->block_label_ids[block_id]; +} + static struct vkd3d_shader_descriptor_binding spirv_compiler_get_descriptor_binding( struct spirv_compiler *compiler, const struct vkd3d_shader_register *reg, const struct vkd3d_shader_register_range *range, enum vkd3d_shader_resource_type resource_type, @@ -5379,7 +5329,6 @@ static void spirv_compiler_emit_initial_declarations(struct spirv_compiler *comp if (compiler->shader_type != VKD3D_SHADER_TYPE_HULL) { vkd3d_spirv_builder_begin_main_function(builder); - compiler->main_block_open = true; } } @@ -7354,379 +7303,114 @@ static void spirv_compiler_emit_kill(struct spirv_compiler *compiler, vkd3d_spirv_build_op_label(builder, merge_block_id); } -static struct vkd3d_control_flow_info *spirv_compiler_push_control_flow_level( - struct spirv_compiler *compiler) +static bool spirv_compiler_init_blocks(struct spirv_compiler *compiler, unsigned int block_count) { - if (!vkd3d_array_reserve((void **)&compiler->control_flow_info, &compiler->control_flow_info_size, - compiler->control_flow_depth + 1, sizeof(*compiler->control_flow_info))) - { - ERR("Failed to allocate control flow info structure.\n"); - return NULL; - } + compiler->block_count = block_count; + + if (!(compiler->block_label_ids = vkd3d_calloc(block_count, sizeof(*compiler->block_label_ids)))) + return false; - return &compiler->control_flow_info[compiler->control_flow_depth++]; + return true; } -static void spirv_compiler_pop_control_flow_level(struct spirv_compiler *compiler) +static void spirv_compiler_emit_label(struct spirv_compiler *compiler, + const struct vkd3d_shader_instruction *instruction) { - struct vkd3d_control_flow_info *cf_info; - - assert(compiler->control_flow_depth); + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + const struct vkd3d_shader_src_param *src = instruction->src; + unsigned int block_id = src->reg.idx[0].offset; + uint32_t label_id; - cf_info = &compiler->control_flow_info[--compiler->control_flow_depth]; - memset(cf_info, 0, sizeof(*cf_info)); + assert(src->reg.type == VKD3DSPR_LABEL); + label_id = spirv_compiler_get_label_id(compiler, block_id); + vkd3d_spirv_build_op_label(builder, label_id); } -static struct vkd3d_control_flow_info *spirv_compiler_find_innermost_loop( - struct spirv_compiler *compiler) +static void spirv_compiler_emit_merge(struct spirv_compiler *compiler, + uint32_t merge_block_id, uint32_t continue_block_id) { - int depth; + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + + if (!merge_block_id) + return; - for (depth = compiler->control_flow_depth - 1; depth >= 0; --depth) + merge_block_id = spirv_compiler_get_label_id(compiler, merge_block_id); + if (!continue_block_id) { - if (compiler->control_flow_info[depth].current_block == VKD3D_BLOCK_LOOP) - return &compiler->control_flow_info[depth]; + vkd3d_spirv_build_op_selection_merge(builder, merge_block_id, SpvSelectionControlMaskNone); + } + else + { + continue_block_id = spirv_compiler_get_label_id(compiler, continue_block_id); + vkd3d_spirv_build_op_loop_merge(builder, merge_block_id, continue_block_id, SpvLoopControlMaskNone); } - - return NULL; } -static struct vkd3d_control_flow_info *spirv_compiler_find_innermost_breakable_cf_construct( - struct spirv_compiler *compiler) +static void spirv_compiler_emit_branch(struct spirv_compiler *compiler, + const struct vkd3d_shader_instruction *instruction) { - int depth; + const struct vkd3d_shader_branch *branch = &instruction->declaration.branch; + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + const struct vkd3d_shader_src_param *src = instruction->src; + uint32_t condition_id; - for (depth = compiler->control_flow_depth - 1; depth >= 0; --depth) + if (!instruction->src_count) { - if (compiler->control_flow_info[depth].current_block == VKD3D_BLOCK_LOOP - || compiler->control_flow_info[depth].current_block == VKD3D_BLOCK_SWITCH) - return &compiler->control_flow_info[depth]; + spirv_compiler_emit_merge(compiler, branch->merge_block_id, branch->continue_block_id); + vkd3d_spirv_build_op_branch(builder, spirv_compiler_get_label_id(compiler, branch->true_id)); + return; } - return NULL; + assert(instruction->src_count == 1); + if (!vkd3d_swizzle_is_scalar(src->swizzle)) + { + WARN("Unexpected src swizzle %#x.\n", src->swizzle); + spirv_compiler_warning(compiler, VKD3D_SHADER_WARNING_SPV_INVALID_SWIZZLE, + "The swizzle for a branch condition value is not scalar."); + } + + condition_id = spirv_compiler_emit_load_src(compiler, src, VKD3DSP_WRITEMASK_0); + condition_id = spirv_compiler_emit_int_to_bool(compiler, instruction->flags, 1, condition_id); + /* Emit the merge immediately before the branch instruction. */ + spirv_compiler_emit_merge(compiler, branch->merge_block_id, branch->continue_block_id); + vkd3d_spirv_build_op_branch_conditional(builder, condition_id, + spirv_compiler_get_label_id(compiler, branch->true_id), + spirv_compiler_get_label_id(compiler, branch->false_id)); } -static int spirv_compiler_emit_control_flow_instruction(struct spirv_compiler *compiler, +static void spirv_compiler_emit_switch(struct spirv_compiler *compiler, const struct vkd3d_shader_instruction *instruction) { - uint32_t loop_header_block_id, loop_body_block_id, continue_block_id; + const struct vkd3d_shader_switch *switch_ = &instruction->declaration.switch_; struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + struct vkd3d_spirv_stream *stream = &builder->function_stream; const struct vkd3d_shader_src_param *src = instruction->src; - uint32_t merge_block_id, val_id, condition_id, true_label; - struct vkd3d_control_flow_info *cf_info; - - cf_info = compiler->control_flow_depth - ? &compiler->control_flow_info[compiler->control_flow_depth - 1] : NULL; + const struct vkd3d_shader_switch_case *switch_case; + uint32_t val_id; + size_t location; + unsigned int i; - switch (instruction->handler_idx) + assert(instruction->src_count == 1); + if (!vkd3d_swizzle_is_scalar(src->swizzle)) { - case VKD3DSIH_IF: - if (!(cf_info = spirv_compiler_push_control_flow_level(compiler))) - return VKD3D_ERROR_OUT_OF_MEMORY; - - val_id = spirv_compiler_emit_load_src(compiler, src, VKD3DSP_WRITEMASK_0); - condition_id = spirv_compiler_emit_int_to_bool(compiler, instruction->flags, 1, val_id); - - true_label = vkd3d_spirv_alloc_id(builder); - merge_block_id = vkd3d_spirv_alloc_id(builder); - vkd3d_spirv_build_op_selection_merge(builder, merge_block_id, SpvSelectionControlMaskNone); - cf_info->u.if_.stream_location = vkd3d_spirv_stream_current_location(&builder->function_stream); - vkd3d_spirv_build_op_branch_conditional(builder, condition_id, true_label, merge_block_id); - - vkd3d_spirv_build_op_label(builder, true_label); - - cf_info->u.if_.id = compiler->branch_id; - cf_info->u.if_.merge_block_id = merge_block_id; - cf_info->u.if_.else_block_id = 0; - cf_info->inside_block = true; - cf_info->current_block = VKD3D_BLOCK_IF; - - vkd3d_spirv_build_op_name(builder, merge_block_id, "branch%u_merge", compiler->branch_id); - vkd3d_spirv_build_op_name(builder, true_label, "branch%u_true", compiler->branch_id); - ++compiler->branch_id; - break; - - case VKD3DSIH_ELSE: - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_IF); - - if (cf_info->inside_block) - vkd3d_spirv_build_op_branch(builder, cf_info->u.if_.merge_block_id); - - cf_info->u.if_.else_block_id = vkd3d_spirv_alloc_id(builder); - vkd3d_spirv_as_op_branch_conditional(&builder->function_stream, - cf_info->u.if_.stream_location)->false_label = cf_info->u.if_.else_block_id; - vkd3d_spirv_build_op_name(builder, - cf_info->u.if_.else_block_id, "branch%u_false", cf_info->u.if_.id); - vkd3d_spirv_build_op_label(builder, cf_info->u.if_.else_block_id); - cf_info->inside_block = true; - break; - - case VKD3DSIH_ENDIF: - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_IF); - - if (cf_info->inside_block) - vkd3d_spirv_build_op_branch(builder, cf_info->u.if_.merge_block_id); - - vkd3d_spirv_build_op_label(builder, cf_info->u.if_.merge_block_id); - - spirv_compiler_pop_control_flow_level(compiler); - break; - - case VKD3DSIH_LOOP: - if (!(cf_info = spirv_compiler_push_control_flow_level(compiler))) - return VKD3D_ERROR_OUT_OF_MEMORY; - - loop_header_block_id = vkd3d_spirv_alloc_id(builder); - loop_body_block_id = vkd3d_spirv_alloc_id(builder); - continue_block_id = vkd3d_spirv_alloc_id(builder); - merge_block_id = vkd3d_spirv_alloc_id(builder); - - vkd3d_spirv_build_op_branch(builder, loop_header_block_id); - vkd3d_spirv_build_op_label(builder, loop_header_block_id); - vkd3d_spirv_build_op_loop_merge(builder, merge_block_id, continue_block_id, SpvLoopControlMaskNone); - vkd3d_spirv_build_op_branch(builder, loop_body_block_id); - - vkd3d_spirv_build_op_label(builder, loop_body_block_id); - - cf_info->u.loop.header_block_id = loop_header_block_id; - cf_info->u.loop.continue_block_id = continue_block_id; - cf_info->u.loop.merge_block_id = merge_block_id; - cf_info->current_block = VKD3D_BLOCK_LOOP; - cf_info->inside_block = true; - - vkd3d_spirv_build_op_name(builder, loop_header_block_id, "loop%u_header", compiler->loop_id); - vkd3d_spirv_build_op_name(builder, loop_body_block_id, "loop%u_body", compiler->loop_id); - vkd3d_spirv_build_op_name(builder, continue_block_id, "loop%u_continue", compiler->loop_id); - vkd3d_spirv_build_op_name(builder, merge_block_id, "loop%u_merge", compiler->loop_id); - ++compiler->loop_id; - break; - - case VKD3DSIH_ENDLOOP: - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_LOOP); - - /* The loop block may have already been ended by an unconditional - * break instruction right before the end of the loop. */ - if (cf_info->inside_block) - vkd3d_spirv_build_op_branch(builder, cf_info->u.loop.continue_block_id); - - vkd3d_spirv_build_op_label(builder, cf_info->u.loop.continue_block_id); - vkd3d_spirv_build_op_branch(builder, cf_info->u.loop.header_block_id); - vkd3d_spirv_build_op_label(builder, cf_info->u.loop.merge_block_id); - - spirv_compiler_pop_control_flow_level(compiler); - break; - - case VKD3DSIH_SWITCH: - if (!(cf_info = spirv_compiler_push_control_flow_level(compiler))) - return VKD3D_ERROR_OUT_OF_MEMORY; - - merge_block_id = vkd3d_spirv_alloc_id(builder); - - assert(src->reg.data_type == VKD3D_DATA_INT); - val_id = spirv_compiler_emit_load_src(compiler, src, VKD3DSP_WRITEMASK_0); - - vkd3d_spirv_build_op_selection_merge(builder, merge_block_id, SpvSelectionControlMaskNone); - - cf_info->u.switch_.id = compiler->switch_id; - cf_info->u.switch_.merge_block_id = merge_block_id; - cf_info->u.switch_.stream_location = vkd3d_spirv_stream_current_location(&builder->function_stream); - cf_info->u.switch_.selector_id = val_id; - cf_info->u.switch_.case_blocks = NULL; - cf_info->u.switch_.case_blocks_size = 0; - cf_info->u.switch_.case_block_count = 0; - cf_info->u.switch_.default_block_id = 0; - cf_info->inside_block = false; - cf_info->current_block = VKD3D_BLOCK_SWITCH; - - vkd3d_spirv_build_op_name(builder, merge_block_id, "switch%u_merge", compiler->switch_id); - - ++compiler->switch_id; - - if (!vkd3d_array_reserve((void **)&cf_info->u.switch_.case_blocks, &cf_info->u.switch_.case_blocks_size, - 10, sizeof(*cf_info->u.switch_.case_blocks))) - return VKD3D_ERROR_OUT_OF_MEMORY; - - break; - - case VKD3DSIH_ENDSWITCH: - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); - assert(!cf_info->inside_block); - - if (!cf_info->u.switch_.default_block_id) - cf_info->u.switch_.default_block_id = cf_info->u.switch_.merge_block_id; - - vkd3d_spirv_build_op_label(builder, cf_info->u.switch_.merge_block_id); - - /* The OpSwitch instruction is inserted when the endswitch - * instruction is processed because we do not know the number - * of case statements in advance.*/ - vkd3d_spirv_begin_function_stream_insertion(builder, cf_info->u.switch_.stream_location); - vkd3d_spirv_build_op_switch(builder, cf_info->u.switch_.selector_id, - cf_info->u.switch_.default_block_id, cf_info->u.switch_.case_blocks, - cf_info->u.switch_.case_block_count); - vkd3d_spirv_end_function_stream_insertion(builder); - - vkd3d_free(cf_info->u.switch_.case_blocks); - spirv_compiler_pop_control_flow_level(compiler); - break; - - case VKD3DSIH_CASE: - { - uint32_t label_id, value; - - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); - - if (src->swizzle != VKD3D_SHADER_SWIZZLE(X, X, X, X)) - { - WARN("Unexpected src swizzle %#x.\n", src->swizzle); - spirv_compiler_warning(compiler, VKD3D_SHADER_WARNING_SPV_INVALID_SWIZZLE, - "The swizzle for a switch case value is not scalar."); - } - assert(src->reg.type == VKD3DSPR_IMMCONST); - value = *src->reg.u.immconst_uint; - - if (!vkd3d_array_reserve((void **)&cf_info->u.switch_.case_blocks, &cf_info->u.switch_.case_blocks_size, - 2 * (cf_info->u.switch_.case_block_count + 1), sizeof(*cf_info->u.switch_.case_blocks))) - return VKD3D_ERROR_OUT_OF_MEMORY; - - label_id = vkd3d_spirv_alloc_id(builder); - if (cf_info->inside_block) /* fall-through */ - vkd3d_spirv_build_op_branch(builder, label_id); - - cf_info->u.switch_.case_blocks[2 * cf_info->u.switch_.case_block_count + 0] = value; - cf_info->u.switch_.case_blocks[2 * cf_info->u.switch_.case_block_count + 1] = label_id; - ++cf_info->u.switch_.case_block_count; - - vkd3d_spirv_build_op_label(builder, label_id); - cf_info->inside_block = true; - vkd3d_spirv_build_op_name(builder, label_id, "switch%u_case%u", cf_info->u.switch_.id, value); - break; - } - - case VKD3DSIH_DEFAULT: - assert(compiler->control_flow_depth); - assert(cf_info->current_block == VKD3D_BLOCK_SWITCH); - assert(!cf_info->u.switch_.default_block_id); - - cf_info->u.switch_.default_block_id = vkd3d_spirv_alloc_id(builder); - if (cf_info->inside_block) /* fall-through */ - vkd3d_spirv_build_op_branch(builder, cf_info->u.switch_.default_block_id); - - vkd3d_spirv_build_op_label(builder, cf_info->u.switch_.default_block_id); - vkd3d_spirv_build_op_name(builder, cf_info->u.switch_.default_block_id, - "switch%u_default", cf_info->u.switch_.id); - cf_info->inside_block = true; - break; - - case VKD3DSIH_BREAK: - { - struct vkd3d_control_flow_info *breakable_cf_info; - - assert(compiler->control_flow_depth); - - if (!(breakable_cf_info = spirv_compiler_find_innermost_breakable_cf_construct(compiler))) - { - FIXME("Unhandled break instruction.\n"); - return VKD3D_ERROR_INVALID_SHADER; - } - - if (breakable_cf_info->current_block == VKD3D_BLOCK_LOOP) - { - vkd3d_spirv_build_op_branch(builder, breakable_cf_info->u.loop.merge_block_id); - } - else if (breakable_cf_info->current_block == VKD3D_BLOCK_SWITCH) - { - /* The current case block may have already been ended by an - * unconditional continue instruction. */ - if (breakable_cf_info->inside_block) - vkd3d_spirv_build_op_branch(builder, breakable_cf_info->u.switch_.merge_block_id); - } - - cf_info->inside_block = false; - break; - } - - case VKD3DSIH_BREAKP: - { - struct vkd3d_control_flow_info *loop_cf_info; - - assert(compiler->control_flow_depth); - - if (!(loop_cf_info = spirv_compiler_find_innermost_loop(compiler))) - { - ERR("Invalid 'breakc' instruction outside loop.\n"); - return VKD3D_ERROR_INVALID_SHADER; - } - - merge_block_id = spirv_compiler_emit_conditional_branch(compiler, - instruction, loop_cf_info->u.loop.merge_block_id); - vkd3d_spirv_build_op_label(builder, merge_block_id); - break; - } - - case VKD3DSIH_CONTINUE: - { - struct vkd3d_control_flow_info *loop_cf_info; - - assert(compiler->control_flow_depth); - - if (!(loop_cf_info = spirv_compiler_find_innermost_loop(compiler))) - { - ERR("Invalid 'continue' instruction outside loop.\n"); - return VKD3D_ERROR_INVALID_SHADER; - } - - vkd3d_spirv_build_op_branch(builder, loop_cf_info->u.loop.continue_block_id); - - cf_info->inside_block = false; - break; - } - - case VKD3DSIH_CONTINUEP: - { - struct vkd3d_control_flow_info *loop_cf_info; - - if (!(loop_cf_info = spirv_compiler_find_innermost_loop(compiler))) - { - ERR("Invalid 'continuec' instruction outside loop.\n"); - return VKD3D_ERROR_INVALID_SHADER; - } - - merge_block_id = spirv_compiler_emit_conditional_branch(compiler, - instruction, loop_cf_info->u.loop.continue_block_id); - vkd3d_spirv_build_op_label(builder, merge_block_id); - break; - } - - case VKD3DSIH_RET: - spirv_compiler_emit_return(compiler, instruction); - - if (cf_info) - cf_info->inside_block = false; - else - compiler->main_block_open = false; - break; - - case VKD3DSIH_RETP: - spirv_compiler_emit_retc(compiler, instruction); - break; - - case VKD3DSIH_DISCARD: - case VKD3DSIH_TEXKILL: - spirv_compiler_emit_kill(compiler, instruction); - break; - - default: - ERR("Unexpected instruction %#x.\n", instruction->handler_idx); - break; + WARN("Unexpected src swizzle %#x.\n", src->swizzle); + spirv_compiler_warning(compiler, VKD3D_SHADER_WARNING_SPV_INVALID_SWIZZLE, + "The swizzle for a switch value is not scalar."); } - return VKD3D_OK; + val_id = spirv_compiler_emit_load_src(compiler, src, VKD3DSP_WRITEMASK_0); + /* Emit the merge immediately before the switch instruction. */ + spirv_compiler_emit_merge(compiler, switch_->merge_block_id, 0); + location = vkd3d_spirv_stream_current_location(stream); + vkd3d_spirv_build_op_switch(builder, val_id, + spirv_compiler_get_label_id(compiler, switch_->default_id), NULL, 0); + for (i = 0; i < switch_->case_count; ++i) + { + switch_case = &switch_->cases[i]; + vkd3d_spirv_build_word(stream, switch_case->value); + vkd3d_spirv_build_word(stream, spirv_compiler_get_label_id(compiler, switch_case->block_id)); + } + vkd3d_spirv_stream_add_word_count(stream, location, switch_->case_count * 2u); } static void spirv_compiler_emit_deriv_instruction(struct spirv_compiler *compiler, @@ -9476,24 +9160,24 @@ static int spirv_compiler_handle_instruction(struct spirv_compiler *compiler, case VKD3DSIH_F32TOF16: spirv_compiler_emit_f32tof16(compiler, instruction); break; - case VKD3DSIH_BREAK: - case VKD3DSIH_BREAKP: - case VKD3DSIH_CASE: - case VKD3DSIH_CONTINUE: - case VKD3DSIH_CONTINUEP: - case VKD3DSIH_DEFAULT: - case VKD3DSIH_DISCARD: - case VKD3DSIH_ELSE: - case VKD3DSIH_ENDIF: - case VKD3DSIH_ENDLOOP: - case VKD3DSIH_ENDSWITCH: - case VKD3DSIH_IF: - case VKD3DSIH_LOOP: case VKD3DSIH_RET: + spirv_compiler_emit_return(compiler, instruction); + break; case VKD3DSIH_RETP: - case VKD3DSIH_SWITCH: + spirv_compiler_emit_retc(compiler, instruction); + break; + case VKD3DSIH_DISCARD: case VKD3DSIH_TEXKILL: - ret = spirv_compiler_emit_control_flow_instruction(compiler, instruction); + spirv_compiler_emit_kill(compiler, instruction); + break; + case VKD3DSIH_LABEL: + spirv_compiler_emit_label(compiler, instruction); + break; + case VKD3DSIH_BRANCH: + spirv_compiler_emit_branch(compiler, instruction); + break; + case VKD3DSIH_SWITCH: + spirv_compiler_emit_switch(compiler, instruction); break; case VKD3DSIH_DSX: case VKD3DSIH_DSX_COARSE: @@ -9683,6 +9367,9 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler, if ((result = vkd3d_shader_normalise(parser, compile_info)) < 0) return result; + if (parser->shader_desc.block_count && !spirv_compiler_init_blocks(compiler, parser->shader_desc.block_count)) + return VKD3D_ERROR_OUT_OF_MEMORY; + instructions = parser->instructions; memset(&parser->instructions, 0, sizeof(parser->instructions)); @@ -9711,9 +9398,6 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler, if (result < 0) return result; - if (compiler->main_block_open) - vkd3d_spirv_build_op_return(builder); - if (!is_in_default_phase(compiler)) spirv_compiler_leave_shader_phase(compiler); else diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index fd812a6c..40934171 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -230,6 +230,7 @@ enum vkd3d_shader_opcode VKD3DSIH_BEM, VKD3DSIH_BFI, VKD3DSIH_BFREV, + VKD3DSIH_BRANCH, VKD3DSIH_BREAK, VKD3DSIH_BREAKC, VKD3DSIH_BREAKP, @@ -965,6 +966,7 @@ struct vkd3d_shader_desc uint32_t temp_count; unsigned int ssa_count; + unsigned int block_count; struct { @@ -1069,6 +1071,28 @@ struct vkd3d_shader_location unsigned int line, column; }; +struct vkd3d_shader_branch +{ + unsigned int merge_block_id; + unsigned int continue_block_id; + unsigned int true_id; + unsigned int false_id; +}; + +struct vkd3d_shader_switch_case +{ + unsigned int value; + unsigned int block_id; +}; + +struct vkd3d_shader_switch +{ + unsigned int merge_block_id; + const struct vkd3d_shader_switch_case *cases; + unsigned int case_count; + unsigned int default_id; +}; + struct vkd3d_shader_instruction { struct vkd3d_shader_location location; @@ -1108,6 +1132,8 @@ struct vkd3d_shader_instruction struct vkd3d_shader_index_range index_range; struct vkd3d_shader_indexable_temp indexable_temp; struct vkd3d_shader_function_table_pointer fp; + struct vkd3d_shader_branch branch; + struct vkd3d_shader_switch switch_; } declaration; }; -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450
From: Conor McCarthy <cmccarthy(a)codeweavers.com> --- libs/vkd3d-shader/dxbc.c | 5 +++ libs/vkd3d-shader/ir.c | 45 ++++++++++++++++++++++++ libs/vkd3d-shader/spirv.c | 8 +++++ libs/vkd3d-shader/vkd3d_shader_private.h | 3 ++ 4 files changed, 61 insertions(+) diff --git a/libs/vkd3d-shader/dxbc.c b/libs/vkd3d-shader/dxbc.c index a9a7aefe..5edf955c 100644 --- a/libs/vkd3d-shader/dxbc.c +++ b/libs/vkd3d-shader/dxbc.c @@ -547,9 +547,14 @@ static int shdr_handler(const struct vkd3d_shader_dxbc_section_desc *section, void free_shader_desc(struct vkd3d_shader_desc *desc) { + size_t i; + shader_signature_cleanup(&desc->input_signature); shader_signature_cleanup(&desc->output_signature); shader_signature_cleanup(&desc->patch_constant_signature); + for (i = 0; i < desc->block_name_count; ++i) + vkd3d_free((void *)desc->block_names[i]); + vkd3d_free(desc->block_names); } int shader_extract_from_dxbc(const struct vkd3d_shader_code *dxbc, diff --git a/libs/vkd3d-shader/ir.c b/libs/vkd3d-shader/ir.c index 3c535d56..c3d9cd9b 100644 --- a/libs/vkd3d-shader/ir.c +++ b/libs/vkd3d-shader/ir.c @@ -1480,6 +1480,9 @@ struct cf_flattener size_t instruction_count; unsigned int block_id; + const char **block_names; + size_t block_name_capacity; + size_t block_name_count; unsigned int branch_id; unsigned int loop_id; @@ -1634,6 +1637,32 @@ static struct vkd3d_control_flow_info *cf_flattener_find_innermost_breakable_cf_ return NULL; } +static void cf_flattener_create_block_name(struct cf_flattener *flattener, unsigned int block_id, + const char *fmt, ...) +{ + struct vkd3d_string_buffer buffer; + size_t block_name_count; + va_list args; + + assert(block_id); + --block_id; + + block_name_count = max(flattener->block_name_count, block_id + 1); + if (!vkd3d_array_reserve((void **)&flattener->block_names, &flattener->block_name_capacity, + block_name_count, sizeof(*flattener->block_names))) + return; + memset(&flattener->block_names[flattener->block_name_count], 0, + (block_name_count - flattener->block_name_count) * sizeof(*flattener->block_names)); + flattener->block_name_count = block_name_count; + + vkd3d_string_buffer_init(&buffer); + va_start(args, fmt); + vkd3d_string_buffer_vprintf(&buffer, fmt, args); + va_end(args); + + flattener->block_names[block_id] = buffer.buffer; +} + static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_parser *parser, struct cf_flattener *flattener) { @@ -1680,6 +1709,8 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par cf_info->inside_block = true; cf_info->current_block = VKD3D_BLOCK_IF; + cf_flattener_create_block_name(flattener, merge_block_id, "branch%u_merge", flattener->branch_id); + cf_flattener_create_block_name(flattener, true_block_id, "branch%u_true", flattener->branch_id); ++flattener->branch_id; break; @@ -1694,6 +1725,8 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par dst_ins = &flattener->instructions[cf_info->u.if_.ins_location]; dst_ins->declaration.branch.false_id = cf_info->u.if_.else_block_id; + cf_flattener_create_block_name(flattener, + cf_info->u.if_.else_block_id, "branch%u_false", cf_info->u.if_.id); cf_flattener_emit_label(flattener, cf_info->u.if_.else_block_id); cf_info->inside_block = true; @@ -1733,6 +1766,10 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par cf_info->current_block = VKD3D_BLOCK_LOOP; cf_info->inside_block = true; + cf_flattener_create_block_name(flattener, loop_header_block_id, "loop%u_header", flattener->loop_id); + cf_flattener_create_block_name(flattener, loop_body_block_id, "loop%u_body", flattener->loop_id); + cf_flattener_create_block_name(flattener, continue_block_id, "loop%u_continue", flattener->loop_id); + cf_flattener_create_block_name(flattener, merge_block_id, "loop%u_merge", flattener->loop_id); ++flattener->loop_id; break; @@ -1780,6 +1817,8 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par cf_info->inside_block = false; cf_info->current_block = VKD3D_BLOCK_SWITCH; + cf_flattener_create_block_name(flattener, merge_block_id, "switch%u_merge", flattener->switch_id); + ++flattener->switch_id; if (!vkd3d_array_reserve((void **)&cf_info->u.switch_.cases, &cf_info->u.switch_.cases_size, @@ -1839,6 +1878,7 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par ++cf_info->u.switch_.cases_count; cf_flattener_emit_label(flattener, label_id); + cf_flattener_create_block_name(flattener, label_id, "switch%u_case%u", cf_info->u.switch_.id, value); cf_info->inside_block = true; break; } @@ -1853,6 +1893,9 @@ static enum vkd3d_result flatten_control_flow_constructs(struct vkd3d_shader_par cf_flattener_emit_unconditional_branch(flattener, cf_info->u.switch_.default_block_id); cf_flattener_emit_label(flattener, cf_info->u.switch_.default_block_id); + + cf_flattener_create_block_name(flattener, cf_info->u.switch_.default_block_id, + "switch%u_default", cf_info->u.switch_.id); cf_info->inside_block = true; break; @@ -1995,6 +2038,8 @@ enum vkd3d_result vkd3d_shader_normalise(struct vkd3d_shader_parser *parser, struct cf_flattener flattener = {0}; result = flatten_control_flow_constructs(parser, &flattener); vkd3d_free(flattener.control_flow_info); + parser->shader_desc.block_names = flattener.block_names; + parser->shader_desc.block_name_count = flattener.block_name_count; if (result >= 0) { vkd3d_free(parser->instructions.elements); diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 54436bd3..0cef8284 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -2294,6 +2294,8 @@ struct spirv_compiler uint32_t *block_label_ids; unsigned int block_count; + const char **block_names; + size_t block_name_count; }; static bool is_in_default_phase(const struct spirv_compiler *compiler) @@ -7324,6 +7326,10 @@ static void spirv_compiler_emit_label(struct spirv_compiler *compiler, assert(src->reg.type == VKD3DSPR_LABEL); label_id = spirv_compiler_get_label_id(compiler, block_id); vkd3d_spirv_build_op_label(builder, label_id); + + --block_id; + if (block_id < compiler->block_name_count && compiler->block_names[block_id]) + vkd3d_spirv_build_op_name(builder, label_id, compiler->block_names[block_id]); } static void spirv_compiler_emit_merge(struct spirv_compiler *compiler, @@ -9380,6 +9386,8 @@ static int spirv_compiler_generate_spirv(struct spirv_compiler *compiler, memset(&shader_desc->output_signature, 0, sizeof(shader_desc->output_signature)); memset(&shader_desc->patch_constant_signature, 0, sizeof(shader_desc->patch_constant_signature)); compiler->use_vocp = parser->shader_desc.use_vocp; + compiler->block_names = parser->shader_desc.block_names; + compiler->block_name_count = parser->shader_desc.block_name_count; compiler->input_control_point_count = shader_desc->input_control_point_count; compiler->output_control_point_count = shader_desc->output_control_point_count; diff --git a/libs/vkd3d-shader/vkd3d_shader_private.h b/libs/vkd3d-shader/vkd3d_shader_private.h index 40934171..4d20f8d5 100644 --- a/libs/vkd3d-shader/vkd3d_shader_private.h +++ b/libs/vkd3d-shader/vkd3d_shader_private.h @@ -974,6 +974,9 @@ struct vkd3d_shader_desc } flat_constant_count[3]; bool use_vocp; + + const char **block_names; + size_t block_name_count; }; struct vkd3d_shader_register_semantic -- GitLab https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450
Giovanni Mascellani (@giomasce) commented about libs/vkd3d-shader/ir.c:
+ struct cf_flattener flattener = {0}; + result = flatten_control_flow_constructs(parser, &flattener); + vkd3d_free(flattener.control_flow_info); + if (result >= 0) + { + vkd3d_free(parser->instructions.elements); + parser->instructions.elements = flattener.instructions; + parser->instructions.capacity = flattener.instruction_capacity; + parser->instructions.count = flattener.instruction_count; + parser->shader_desc.block_count = flattener.block_id; + } + else + { + vkd3d_free(flattener.instructions); + } + } For homogeneity with the other passes, could this be moved to a dedicated function?
-- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450#note_51488
Giovanni Mascellani (@giomasce) commented about libs/vkd3d-shader/ir.c:
return VKD3D_OK; }
+struct vkd3d_if_cf_info
Maybe it makes sense to rename the structures with prefix `cf_flattener` now? -- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450#note_51489
Giovanni Mascellani (@giomasce) commented about libs/vkd3d-shader/ir.c:
case VKD3DSIH_SWITCH: vsir_validate_dst_count(ctx, instruction, 0); vsir_validate_src_count(ctx, instruction, 1); + /* New monolithic switch variant. */ + if (instruction->declaration.switch_.case_count) + break;
It seems the new `switch` should be considered an entirely different instruction, given how different it is. It could be renamed to `branch_switch`. Also, comments with the word "new" in them are likely to eventually become false, or at least misleading. -- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450#note_51490
So, it seems that the idea is that at any given time a VSIR program either has control flow based on `if`, `loop`, etc or based on `label` and `branch` (which are endowed with CFG structure information anyway). Is that true, or are cases in which the same program can legally have both types of instructions? -- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450#note_51491
On Wed Nov 8 13:44:26 2023 +0000, Giovanni Mascellani wrote:
So, it seems that the idea is that at any given time a VSIR program either has control flow based on `if`, `loop`, etc or based on `label` and `branch` (which are endowed with CFG structure information anyway). Is that true, or are cases in which the same program can legally have both types of instructions? I expect it would work to have both and then pass it through the flattener, but it should not occur in practice.
-- https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/450#note_51492
participants (3)
-
Conor McCarthy -
Conor McCarthy (＠cmccarthy)
-
Giovanni Mascellani (＠giomasce)