Signed-off-by: Zebediah Figura zfigura@codeweavers.com --- libs/vkd3d-shader/hlsl_codegen.c | 5 +++++ 1 file changed, 5 insertions(+)
diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 7e3e4d77..398be349 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -371,6 +371,11 @@ static bool fold_constants(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, voi
switch (arg1->node.data_type->base_type) { + case HLSL_TYPE_INT: + for (i = 0; i < dimx; ++i) + res->value.f[i] = arg1->value.i[i]; + break; + case HLSL_TYPE_UINT: for (i = 0; i < dimx; ++i) res->value.f[i] = arg1->value.u[i];
Signed-off-by: Zebediah Figura zfigura@codeweavers.com --- libs/vkd3d-shader/hlsl_codegen.c | 5 +++++ 1 file changed, 5 insertions(+)
diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 398be349..bef9ea84 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -401,6 +401,11 @@ static bool fold_constants(struct hlsl_ctx *ctx, struct hlsl_ir_node *instr, voi { switch (expr->op) { + case HLSL_IR_UNOP_NEG: + for (i = 0; i < instr->data_type->dimx; ++i) + res->value.u[i] = -arg1->value.u[i]; + break; + case HLSL_IR_BINOP_ADD: for (i = 0; i < instr->data_type->dimx; ++i) res->value.u[i] = arg1->value.u[i] + arg2->value.u[i];
Signed-off-by: Matteo Bruni mbruni@codeweavers.com
Signed-off-by: Henri Verbeet hverbeet@codeweavers.com
Signed-off-by: Zebediah Figura zfigura@codeweavers.com --- Makefile.am | 1 + libs/vkd3d-shader/hlsl.c | 76 +++ libs/vkd3d-shader/hlsl.h | 13 + libs/vkd3d-shader/hlsl_codegen.c | 889 +------------------------------ libs/vkd3d-shader/hlsl_sm1.c | 828 ++++++++++++++++++++++++++++ 5 files changed, 922 insertions(+), 885 deletions(-) create mode 100644 libs/vkd3d-shader/hlsl_sm1.c
diff --git a/Makefile.am b/Makefile.am index 07d5af97..9624485b 100644 --- a/Makefile.am +++ b/Makefile.am @@ -172,6 +172,7 @@ libvkd3d_shader_la_SOURCES = \ libs/vkd3d-shader/hlsl.c \ libs/vkd3d-shader/hlsl.h \ libs/vkd3d-shader/hlsl_codegen.c \ + libs/vkd3d-shader/hlsl_sm1.c \ libs/vkd3d-shader/preproc.h \ libs/vkd3d-shader/sm4.h \ libs/vkd3d-shader/spirv.c \ diff --git a/libs/vkd3d-shader/hlsl.c b/libs/vkd3d-shader/hlsl.c index bc593f82..d0be0750 100644 --- a/libs/vkd3d-shader/hlsl.c +++ b/libs/vkd3d-shader/hlsl.c @@ -1,4 +1,6 @@ /* + * HLSL utility functions + * * Copyright 2012 Matteo Bruni for CodeWeavers * Copyright 2019-2020 Zebediah Figura for CodeWeavers * @@ -1449,6 +1451,80 @@ void hlsl_add_function(struct hlsl_ctx *ctx, char *name, struct hlsl_ir_function rb_put(&ctx->functions, func->name, &func->entry); }
+unsigned int hlsl_map_swizzle(unsigned int swizzle, unsigned int writemask) +{ + unsigned int i, ret = 0; + + /* Leave replicate swizzles alone; some instructions need them. */ + if (swizzle == HLSL_SWIZZLE(X, X, X, X) + || swizzle == HLSL_SWIZZLE(Y, Y, Y, Y) + || swizzle == HLSL_SWIZZLE(Z, Z, Z, Z) + || swizzle == HLSL_SWIZZLE(W, W, W, W)) + return swizzle; + + for (i = 0; i < 4; ++i) + { + if (writemask & (1 << i)) + { + ret |= (swizzle & 3) << (i * 2); + swizzle >>= 2; + } + } + return ret; +} + +unsigned int hlsl_swizzle_from_writemask(unsigned int writemask) +{ + static const unsigned int swizzles[16] = + { + 0, + HLSL_SWIZZLE(X, X, X, X), + HLSL_SWIZZLE(Y, Y, Y, Y), + HLSL_SWIZZLE(X, Y, X, X), + HLSL_SWIZZLE(Z, Z, Z, Z), + HLSL_SWIZZLE(X, Z, X, X), + HLSL_SWIZZLE(Y, Z, X, X), + HLSL_SWIZZLE(X, Y, Z, X), + HLSL_SWIZZLE(W, W, W, W), + HLSL_SWIZZLE(X, W, X, X), + HLSL_SWIZZLE(Y, W, X, X), + HLSL_SWIZZLE(X, Y, W, X), + HLSL_SWIZZLE(Z, W, X, X), + HLSL_SWIZZLE(X, Z, W, X), + HLSL_SWIZZLE(Y, Z, W, X), + HLSL_SWIZZLE(X, Y, Z, W), + }; + + return swizzles[writemask & 0xf]; +} + +unsigned int hlsl_combine_writemasks(unsigned int first, unsigned int second) +{ + unsigned int ret = 0, i, j = 0; + + for (i = 0; i < 4; ++i) + { + if (first & (1 << i)) + { + if (second & (1 << j++)) + ret |= (1 << i); + } + } + + return ret; +} + +unsigned int hlsl_combine_swizzles(unsigned int first, unsigned int second, unsigned int dim) +{ + unsigned int ret = 0, i; + for (i = 0; i < dim; ++i) + { + unsigned int s = (second >> (i * 2)) & 3; + ret |= ((first >> (s * 2)) & 3) << (i * 2); + } + return ret; +} + static const struct hlsl_profile_info *get_target_info(const char *target) { unsigned int i; diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index aede3a55..e566ac7d 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -22,6 +22,7 @@
#include "vkd3d_shader_private.h" #include "rbtree.h" +#include "vkd3d_d3dx9shader.h"
/* The general IR structure is inspired by Mesa GLSL hir, even though the code * ends up being quite different in practice. Anyway, here comes the relevant @@ -679,6 +680,18 @@ unsigned int hlsl_type_get_sm4_offset(const struct hlsl_type *type, unsigned int bool hlsl_type_is_void(const struct hlsl_type *type); bool hlsl_types_are_equal(const struct hlsl_type *t1, const struct hlsl_type *t2);
+unsigned int hlsl_combine_swizzles(unsigned int first, unsigned int second, unsigned int dim); +unsigned int hlsl_combine_writemasks(unsigned int first, unsigned int second); +unsigned int hlsl_map_swizzle(unsigned int swizzle, unsigned int writemask); +unsigned int hlsl_swizzle_from_writemask(unsigned int writemask); + +struct hlsl_reg hlsl_reg_from_deref(const struct hlsl_deref *deref, const struct hlsl_type *type); + +bool hlsl_sm1_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semantic *semantic, + bool output, D3DSHADER_PARAM_REGISTER_TYPE *type, unsigned int *reg); +bool hlsl_sm1_usage_from_semantic(const struct hlsl_semantic *semantic, D3DDECLUSAGE *usage, uint32_t *usage_idx); +int hlsl_sm1_write(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, struct vkd3d_shader_code *out); + int hlsl_lexer_compile(struct hlsl_ctx *ctx, const struct vkd3d_shader_code *hlsl);
#endif diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index bef9ea84..549636ca 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -20,7 +20,6 @@
#include "hlsl.h" #include <stdio.h> -#include "vkd3d_d3dx9shader.h"
/* Split uniforms into two variables representing the constant and temp * registers, and copy the former to the latter, so that writes to uniforms @@ -970,114 +969,6 @@ static void allocate_temp_registers(struct hlsl_ctx *ctx, struct hlsl_ir_functio allocate_temp_registers_recurse(ctx, entry_func->body, &liveness); }
-static bool sm1_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semantic *semantic, bool output, - D3DSHADER_PARAM_REGISTER_TYPE *type, unsigned int *reg) -{ - unsigned int i; - - static const struct - { - const char *semantic; - bool output; - enum vkd3d_shader_type shader_type; - unsigned int major_version; - D3DSHADER_PARAM_REGISTER_TYPE type; - DWORD offset; - } - register_table[] = - { - {"color", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_COLOROUT}, - {"depth", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_DEPTHOUT}, - {"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_DEPTHOUT}, - {"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_COLOROUT}, - {"color", false, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_INPUT}, - {"texcoord", false, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_TEXTURE}, - - {"color", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_COLOROUT}, - {"depth", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_DEPTHOUT}, - {"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_DEPTHOUT}, - {"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_COLOROUT}, - {"sv_position", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_POSITION}, - {"vface", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_FACE}, - {"vpos", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_POSITION}, - - {"color", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_ATTROUT}, - {"fog", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_FOG}, - {"position", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POSITION}, - {"psize", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POINT_SIZE}, - {"sv_position", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POSITION}, - {"texcoord", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_TEXCRDOUT}, - - {"color", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_ATTROUT}, - {"fog", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_FOG}, - {"position", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POSITION}, - {"psize", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POINT_SIZE}, - {"sv_position", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POSITION}, - {"texcoord", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_TEXCRDOUT}, - }; - - for (i = 0; i < ARRAY_SIZE(register_table); ++i) - { - if (!ascii_strcasecmp(semantic->name, register_table[i].semantic) - && output == register_table[i].output - && ctx->profile->type == register_table[i].shader_type - && ctx->profile->major_version == register_table[i].major_version) - { - *type = register_table[i].type; - if (register_table[i].type == D3DSPR_MISCTYPE || register_table[i].type == D3DSPR_RASTOUT) - *reg = register_table[i].offset; - else - *reg = semantic->index; - return true; - } - } - - return false; -} - -static bool sm1_usage_from_semantic(const struct hlsl_semantic *semantic, D3DDECLUSAGE *usage, uint32_t *usage_idx) -{ - static const struct - { - const char *name; - D3DDECLUSAGE usage; - } - semantics[] = - { - {"binormal", D3DDECLUSAGE_BINORMAL}, - {"blendindices", D3DDECLUSAGE_BLENDINDICES}, - {"blendweight", D3DDECLUSAGE_BLENDWEIGHT}, - {"color", D3DDECLUSAGE_COLOR}, - {"depth", D3DDECLUSAGE_DEPTH}, - {"fog", D3DDECLUSAGE_FOG}, - {"normal", D3DDECLUSAGE_NORMAL}, - {"position", D3DDECLUSAGE_POSITION}, - {"positiont", D3DDECLUSAGE_POSITIONT}, - {"psize", D3DDECLUSAGE_PSIZE}, - {"sample", D3DDECLUSAGE_SAMPLE}, - {"sv_depth", D3DDECLUSAGE_DEPTH}, - {"sv_position", D3DDECLUSAGE_POSITION}, - {"sv_target", D3DDECLUSAGE_COLOR}, - {"tangent", D3DDECLUSAGE_TANGENT}, - {"tessfactor", D3DDECLUSAGE_TESSFACTOR}, - {"texcoord", D3DDECLUSAGE_TEXCOORD}, - }; - - unsigned int i; - - for (i = 0; i < ARRAY_SIZE(semantics); ++i) - { - if (!ascii_strcasecmp(semantic->name, semantics[i].name)) - { - *usage = semantics[i].usage; - *usage_idx = semantic->index; - return true; - } - } - - return false; -} - static void allocate_semantic_register(struct hlsl_ctx *ctx, struct hlsl_ir_var *var, unsigned int *counter, bool output) { assert(var->semantic.name); @@ -1088,14 +979,14 @@ static void allocate_semantic_register(struct hlsl_ctx *ctx, struct hlsl_ir_var uint32_t reg, usage_idx; D3DDECLUSAGE usage;
- if (!sm1_usage_from_semantic(&var->semantic, &usage, &usage_idx)) + if (!hlsl_sm1_usage_from_semantic(&var->semantic, &usage, &usage_idx)) { hlsl_error(ctx, var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SEMANTIC, "Invalid semantic '%s'.", var->semantic.name); return; }
- if (sm1_register_from_semantic(ctx, &var->semantic, output, &type, ®)) + if (hlsl_sm1_register_from_semantic(ctx, &var->semantic, output, &type, ®)) { TRACE("%s %s semantic %s[%u] matches predefined register %#x[%u].\n", ctx->profile->type == VKD3D_SHADER_TYPE_PIXEL ? "Pixel" : "Vertex", output ? "output" : "input", @@ -1213,86 +1104,12 @@ static void allocate_buffers(struct hlsl_ctx *ctx) } }
-static unsigned int map_swizzle(unsigned int swizzle, unsigned int writemask) -{ - unsigned int i, ret = 0; - - /* Leave replicate swizzles alone; some instructions need them. */ - if (swizzle == HLSL_SWIZZLE(X, X, X, X) - || swizzle == HLSL_SWIZZLE(Y, Y, Y, Y) - || swizzle == HLSL_SWIZZLE(Z, Z, Z, Z) - || swizzle == HLSL_SWIZZLE(W, W, W, W)) - return swizzle; - - for (i = 0; i < 4; ++i) - { - if (writemask & (1 << i)) - { - ret |= (swizzle & 3) << (i * 2); - swizzle >>= 2; - } - } - return ret; -} - -static unsigned int swizzle_from_writemask(unsigned int writemask) -{ - static const unsigned int swizzles[16] = - { - 0, - HLSL_SWIZZLE(X, X, X, X), - HLSL_SWIZZLE(Y, Y, Y, Y), - HLSL_SWIZZLE(X, Y, X, X), - HLSL_SWIZZLE(Z, Z, Z, Z), - HLSL_SWIZZLE(X, Z, X, X), - HLSL_SWIZZLE(Y, Z, X, X), - HLSL_SWIZZLE(X, Y, Z, X), - HLSL_SWIZZLE(W, W, W, W), - HLSL_SWIZZLE(X, W, X, X), - HLSL_SWIZZLE(Y, W, X, X), - HLSL_SWIZZLE(X, Y, W, X), - HLSL_SWIZZLE(Z, W, X, X), - HLSL_SWIZZLE(X, Z, W, X), - HLSL_SWIZZLE(Y, Z, W, X), - HLSL_SWIZZLE(X, Y, Z, W), - }; - - return swizzles[writemask & 0xf]; -} - -static unsigned int combine_writemasks(unsigned int first, unsigned int second) -{ - unsigned int ret = 0, i, j = 0; - - for (i = 0; i < 4; ++i) - { - if (first & (1 << i)) - { - if (second & (1 << j++)) - ret |= (1 << i); - } - } - - return ret; -} - -static unsigned int combine_swizzles(unsigned int first, unsigned int second, unsigned int dim) -{ - unsigned int ret = 0, i; - for (i = 0; i < dim; ++i) - { - unsigned int s = (second >> (i * 2)) & 3; - ret |= ((first >> (s * 2)) & 3) << (i * 2); - } - return ret; -} - static bool type_is_single_reg(const struct hlsl_type *type) { return type->type == HLSL_CLASS_SCALAR || type->type == HLSL_CLASS_VECTOR; }
-static struct hlsl_reg hlsl_reg_from_deref(const struct hlsl_deref *deref, const struct hlsl_type *type) +struct hlsl_reg hlsl_reg_from_deref(const struct hlsl_deref *deref, const struct hlsl_type *type) { struct hlsl_ir_node *offset_node = deref->offset.node; const struct hlsl_ir_var *var = deref->var; @@ -1331,704 +1148,6 @@ static struct hlsl_reg hlsl_reg_from_deref(const struct hlsl_deref *deref, const return ret; }
-static uint32_t sm1_version(enum vkd3d_shader_type type, unsigned int major, unsigned int minor) -{ - if (type == VKD3D_SHADER_TYPE_VERTEX) - return D3DVS_VERSION(major, minor); - else - return D3DPS_VERSION(major, minor); -} - -static D3DXPARAMETER_CLASS sm1_class(const struct hlsl_type *type) -{ - switch (type->type) - { - case HLSL_CLASS_ARRAY: - return sm1_class(type->e.array.type); - case HLSL_CLASS_MATRIX: - assert(type->modifiers & HLSL_MODIFIERS_MAJORITY_MASK); - if (type->modifiers & HLSL_MODIFIER_COLUMN_MAJOR) - return D3DXPC_MATRIX_COLUMNS; - else - return D3DXPC_MATRIX_ROWS; - case HLSL_CLASS_OBJECT: - return D3DXPC_OBJECT; - case HLSL_CLASS_SCALAR: - return D3DXPC_SCALAR; - case HLSL_CLASS_STRUCT: - return D3DXPC_STRUCT; - case HLSL_CLASS_VECTOR: - return D3DXPC_VECTOR; - default: - ERR("Invalid class %#x.\n", type->type); - assert(0); - return 0; - } -} - -static D3DXPARAMETER_TYPE sm1_base_type(const struct hlsl_type *type) -{ - switch (type->base_type) - { - case HLSL_TYPE_BOOL: - return D3DXPT_BOOL; - case HLSL_TYPE_FLOAT: - case HLSL_TYPE_HALF: - return D3DXPT_FLOAT; - case HLSL_TYPE_INT: - case HLSL_TYPE_UINT: - return D3DXPT_INT; - case HLSL_TYPE_PIXELSHADER: - return D3DXPT_PIXELSHADER; - case HLSL_TYPE_SAMPLER: - switch (type->sampler_dim) - { - case HLSL_SAMPLER_DIM_1D: - return D3DXPT_SAMPLER1D; - case HLSL_SAMPLER_DIM_2D: - return D3DXPT_SAMPLER2D; - case HLSL_SAMPLER_DIM_3D: - return D3DXPT_SAMPLER3D; - case HLSL_SAMPLER_DIM_CUBE: - return D3DXPT_SAMPLERCUBE; - case HLSL_SAMPLER_DIM_GENERIC: - return D3DXPT_SAMPLER; - default: - ERR("Invalid dimension %#x.\n", type->sampler_dim); - } - break; - case HLSL_TYPE_STRING: - return D3DXPT_STRING; - case HLSL_TYPE_TEXTURE: - switch (type->sampler_dim) - { - case HLSL_SAMPLER_DIM_1D: - return D3DXPT_TEXTURE1D; - case HLSL_SAMPLER_DIM_2D: - return D3DXPT_TEXTURE2D; - case HLSL_SAMPLER_DIM_3D: - return D3DXPT_TEXTURE3D; - case HLSL_SAMPLER_DIM_CUBE: - return D3DXPT_TEXTURECUBE; - case HLSL_SAMPLER_DIM_GENERIC: - return D3DXPT_TEXTURE; - default: - ERR("Invalid dimension %#x.\n", type->sampler_dim); - } - break; - case HLSL_TYPE_VERTEXSHADER: - return D3DXPT_VERTEXSHADER; - case HLSL_TYPE_VOID: - return D3DXPT_VOID; - default: - assert(0); - } - assert(0); - return 0; -} - -static const struct hlsl_type *get_array_type(const struct hlsl_type *type) -{ - if (type->type == HLSL_CLASS_ARRAY) - return get_array_type(type->e.array.type); - return type; -} - -static unsigned int get_array_size(const struct hlsl_type *type) -{ - if (type->type == HLSL_CLASS_ARRAY) - return get_array_size(type->e.array.type) * type->e.array.elements_count; - return 1; -} - -static void write_sm1_type(struct vkd3d_bytecode_buffer *buffer, struct hlsl_type *type, unsigned int ctab_start) -{ - const struct hlsl_type *array_type = get_array_type(type); - unsigned int array_size = get_array_size(type); - struct hlsl_struct_field *field; - unsigned int field_count = 0; - size_t fields_offset = 0; - - if (type->bytecode_offset) - return; - - if (array_type->type == HLSL_CLASS_STRUCT) - { - LIST_FOR_EACH_ENTRY(field, array_type->e.elements, struct hlsl_struct_field, entry) - { - field->name_bytecode_offset = put_string(buffer, field->name); - write_sm1_type(buffer, field->type, ctab_start); - } - - fields_offset = bytecode_get_size(buffer) - ctab_start; - - LIST_FOR_EACH_ENTRY(field, array_type->e.elements, struct hlsl_struct_field, entry) - { - put_u32(buffer, field->name_bytecode_offset - ctab_start); - put_u32(buffer, field->type->bytecode_offset - ctab_start); - ++field_count; - } - } - - type->bytecode_offset = put_u32(buffer, sm1_class(type) | (sm1_base_type(type) << 16)); - put_u32(buffer, type->dimy | (type->dimx << 16)); - put_u32(buffer, array_size | (field_count << 16)); - put_u32(buffer, fields_offset); -} - -static void sm1_sort_extern(struct list *sorted, struct hlsl_ir_var *to_sort) -{ - struct hlsl_ir_var *var; - - list_remove(&to_sort->extern_entry); - - LIST_FOR_EACH_ENTRY(var, sorted, struct hlsl_ir_var, extern_entry) - { - if (strcmp(to_sort->name, var->name) < 0) - { - list_add_before(&var->extern_entry, &to_sort->extern_entry); - return; - } - } - - list_add_tail(sorted, &to_sort->extern_entry); -} - -static void sm1_sort_externs(struct hlsl_ctx *ctx) -{ - struct list sorted = LIST_INIT(sorted); - struct hlsl_ir_var *var, *next; - - LIST_FOR_EACH_ENTRY_SAFE(var, next, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) - sm1_sort_extern(&sorted, var); - list_move_tail(&ctx->extern_vars, &sorted); -} - -static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - struct hlsl_ir_function_decl *entry_func) -{ - size_t ctab_offset, ctab_start, ctab_end, vars_start, size_offset, creator_offset, offset; - unsigned int uniform_count = 0; - struct hlsl_ir_var *var; - - LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) - { - if (!var->semantic.name && var->reg.allocated) - { - ++uniform_count; - - if (var->is_param && var->is_uniform) - { - struct vkd3d_string_buffer *name; - - if (!(name = hlsl_get_string_buffer(ctx))) - { - buffer->status = VKD3D_ERROR_OUT_OF_MEMORY; - return; - } - vkd3d_string_buffer_printf(name, "$%s", var->name); - vkd3d_free((char *)var->name); - var->name = hlsl_strdup(ctx, name->buffer); - hlsl_release_string_buffer(ctx, name); - } - } - } - - sm1_sort_externs(ctx); - - size_offset = put_u32(buffer, 0); - ctab_offset = put_u32(buffer, MAKEFOURCC('C','T','A','B')); - - ctab_start = put_u32(buffer, sizeof(D3DXSHADER_CONSTANTTABLE)); - creator_offset = put_u32(buffer, 0); - put_u32(buffer, sm1_version(ctx->profile->type, ctx->profile->major_version, ctx->profile->minor_version)); - put_u32(buffer, uniform_count); - put_u32(buffer, sizeof(D3DXSHADER_CONSTANTTABLE)); /* offset of constants */ - put_u32(buffer, 0); /* FIXME: flags */ - put_u32(buffer, 0); /* FIXME: target string */ - - vars_start = bytecode_get_size(buffer); - - LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) - { - if (!var->semantic.name && var->reg.allocated) - { - put_u32(buffer, 0); /* name */ - put_u32(buffer, D3DXRS_FLOAT4 | (var->reg.id << 16)); - put_u32(buffer, var->data_type->reg_size / 4); - put_u32(buffer, 0); /* type */ - put_u32(buffer, 0); /* FIXME: default value */ - } - } - - uniform_count = 0; - - LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) - { - if (!var->semantic.name && var->reg.allocated) - { - size_t var_offset = vars_start + (uniform_count * 5 * sizeof(uint32_t)); - size_t name_offset; - - name_offset = put_string(buffer, var->name); - set_u32(buffer, var_offset, name_offset - ctab_start); - - write_sm1_type(buffer, var->data_type, ctab_start); - set_u32(buffer, var_offset + 3 * sizeof(uint32_t), var->data_type->bytecode_offset - ctab_start); - ++uniform_count; - } - } - - offset = put_string(buffer, vkd3d_shader_get_version(NULL, NULL)); - set_u32(buffer, creator_offset, offset - ctab_start); - - ctab_end = bytecode_get_size(buffer); - set_u32(buffer, size_offset, D3DSIO_COMMENT | (((ctab_end - ctab_offset) / sizeof(uint32_t)) << 16)); -} - -static uint32_t sm1_encode_register_type(D3DSHADER_PARAM_REGISTER_TYPE type) -{ - return ((type << D3DSP_REGTYPE_SHIFT) & D3DSP_REGTYPE_MASK) - | ((type << D3DSP_REGTYPE_SHIFT2) & D3DSP_REGTYPE_MASK2); -} - -struct sm1_instruction -{ - D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode; - - struct sm1_dst_register - { - D3DSHADER_PARAM_REGISTER_TYPE type; - D3DSHADER_PARAM_DSTMOD_TYPE mod; - unsigned int writemask; - uint32_t reg; - } dst; - - struct sm1_src_register - { - D3DSHADER_PARAM_REGISTER_TYPE type; - D3DSHADER_PARAM_SRCMOD_TYPE mod; - unsigned int swizzle; - uint32_t reg; - } srcs[2]; - unsigned int src_count; - - unsigned int has_dst; -}; - -static void write_sm1_dst_register(struct vkd3d_bytecode_buffer *buffer, const struct sm1_dst_register *reg) -{ - assert(reg->writemask); - put_u32(buffer, (1u << 31) | sm1_encode_register_type(reg->type) | reg->mod | (reg->writemask << 16) | reg->reg); -} - -static void write_sm1_src_register(struct vkd3d_bytecode_buffer *buffer, - const struct sm1_src_register *reg, unsigned int dst_writemask) -{ - unsigned int swizzle = map_swizzle(reg->swizzle, dst_writemask); - - put_u32(buffer, (1u << 31) | sm1_encode_register_type(reg->type) | reg->mod | (swizzle << 16) | reg->reg); -} - -static void write_sm1_instruction(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct sm1_instruction *instr) -{ - uint32_t token = instr->opcode; - unsigned int i; - - if (ctx->profile->major_version > 1) - token |= (instr->has_dst + instr->src_count) << D3DSI_INSTLENGTH_SHIFT; - put_u32(buffer, token); - - if (instr->has_dst) - write_sm1_dst_register(buffer, &instr->dst); - - for (i = 0; i < instr->src_count; ++i) - write_sm1_src_register(buffer, &instr->srcs[i], instr->dst.writemask); -}; - -static void write_sm1_binary_op(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode, const struct hlsl_reg *dst, - const struct hlsl_reg *src1, const struct hlsl_reg *src2) -{ - const struct sm1_instruction instr = - { - .opcode = opcode, - - .dst.type = D3DSPR_TEMP, - .dst.writemask = dst->writemask, - .dst.reg = dst->id, - .has_dst = 1, - - .srcs[0].type = D3DSPR_TEMP, - .srcs[0].swizzle = swizzle_from_writemask(src1->writemask), - .srcs[0].reg = src1->id, - .srcs[1].type = D3DSPR_TEMP, - .srcs[1].swizzle = swizzle_from_writemask(src2->writemask), - .srcs[1].reg = src2->id, - .src_count = 2, - }; - write_sm1_instruction(ctx, buffer, &instr); -} - -static void write_sm1_unary_op(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode, const struct hlsl_reg *dst, - const struct hlsl_reg *src, D3DSHADER_PARAM_SRCMOD_TYPE src_mod) -{ - const struct sm1_instruction instr = - { - .opcode = opcode, - - .dst.type = D3DSPR_TEMP, - .dst.writemask = dst->writemask, - .dst.reg = dst->id, - .has_dst = 1, - - .srcs[0].type = D3DSPR_TEMP, - .srcs[0].swizzle = swizzle_from_writemask(src->writemask), - .srcs[0].reg = src->id, - .srcs[0].mod = src_mod, - .src_count = 1, - }; - write_sm1_instruction(ctx, buffer, &instr); -} - -static void write_sm1_constant_defs(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer) -{ - unsigned int i, x; - - for (i = 0; i < ctx->constant_defs.count; ++i) - { - uint32_t token = D3DSIO_DEF; - const struct sm1_dst_register reg = - { - .type = D3DSPR_CONST, - .writemask = VKD3DSP_WRITEMASK_ALL, - .reg = i, - }; - - if (ctx->profile->major_version > 1) - token |= 5 << D3DSI_INSTLENGTH_SHIFT; - put_u32(buffer, token); - - write_sm1_dst_register(buffer, ®); - for (x = 0; x < 4; ++x) - put_f32(buffer, ctx->constant_defs.values[i].f[x]); - } -} - -static void write_sm1_semantic_dcl(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct hlsl_ir_var *var, bool output) -{ - struct sm1_dst_register reg = {0}; - uint32_t token, usage_idx; - D3DDECLUSAGE usage; - bool ret; - - if (sm1_register_from_semantic(ctx, &var->semantic, output, ®.type, ®.reg)) - { - usage = 0; - usage_idx = 0; - } - else - { - ret = sm1_usage_from_semantic(&var->semantic, &usage, &usage_idx); - assert(ret); - reg.type = output ? D3DSPR_OUTPUT : D3DSPR_INPUT; - reg.reg = var->reg.id; - } - - token = D3DSIO_DCL; - if (ctx->profile->major_version > 1) - token |= 2 << D3DSI_INSTLENGTH_SHIFT; - put_u32(buffer, token); - - token = (1u << 31); - token |= usage << D3DSP_DCL_USAGE_SHIFT; - token |= usage_idx << D3DSP_DCL_USAGEINDEX_SHIFT; - put_u32(buffer, token); - - reg.writemask = (1 << var->data_type->dimx) - 1; - write_sm1_dst_register(buffer, ®); -} - -static void write_sm1_semantic_dcls(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer) -{ - bool write_in = false, write_out = false; - struct hlsl_ir_var *var; - - if (ctx->profile->type == VKD3D_SHADER_TYPE_PIXEL) - write_in = true; - else if (ctx->profile->type == VKD3D_SHADER_TYPE_VERTEX && ctx->profile->major_version == 3) - write_in = write_out = true; - else if (ctx->profile->type == VKD3D_SHADER_TYPE_VERTEX && ctx->profile->major_version < 3) - write_in = true; - - LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) - { - if (write_in && var->is_input_semantic) - write_sm1_semantic_dcl(ctx, buffer, var, false); - if (write_out && var->is_output_semantic) - write_sm1_semantic_dcl(ctx, buffer, var, true); - } -} - -static void write_sm1_constant(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct hlsl_ir_node *instr) -{ - const struct hlsl_ir_constant *constant = hlsl_ir_constant(instr); - struct sm1_instruction sm1_instr = - { - .opcode = D3DSIO_MOV, - - .dst.type = D3DSPR_TEMP, - .dst.reg = instr->reg.id, - .dst.writemask = instr->reg.writemask, - .has_dst = 1, - - .srcs[0].type = D3DSPR_CONST, - .srcs[0].reg = constant->reg.id, - .srcs[0].swizzle = swizzle_from_writemask(constant->reg.writemask), - .src_count = 1, - }; - - assert(instr->reg.allocated); - assert(constant->reg.allocated); - write_sm1_instruction(ctx, buffer, &sm1_instr); -} - -static void write_sm1_expr(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr) -{ - struct hlsl_ir_expr *expr = hlsl_ir_expr(instr); - struct hlsl_ir_node *arg1 = expr->operands[0].node; - struct hlsl_ir_node *arg2 = expr->operands[1].node; - unsigned int i; - - assert(instr->reg.allocated); - - if (instr->data_type->base_type != HLSL_TYPE_FLOAT) - { - FIXME("Non-float operations need to be lowered.\n"); - return; - } - - switch (expr->op) - { - case HLSL_IR_BINOP_ADD: - write_sm1_binary_op(ctx, buffer, D3DSIO_ADD, &instr->reg, &arg1->reg, &arg2->reg); - break; - - case HLSL_IR_BINOP_MUL: - write_sm1_binary_op(ctx, buffer, D3DSIO_MUL, &instr->reg, &arg1->reg, &arg2->reg); - break; - - case HLSL_IR_BINOP_SUB: - write_sm1_binary_op(ctx, buffer, D3DSIO_SUB, &instr->reg, &arg1->reg, &arg2->reg); - break; - - case HLSL_IR_UNOP_NEG: - write_sm1_unary_op(ctx, buffer, D3DSIO_MOV, &instr->reg, &arg1->reg, D3DSPSM_NEG); - break; - - case HLSL_IR_UNOP_RCP: - for (i = 0; i < instr->data_type->dimx; ++i) - { - struct hlsl_reg src = arg1->reg, dst = instr->reg; - - src.writemask = combine_writemasks(src.writemask, 1u << i); - dst.writemask = combine_writemasks(dst.writemask, 1u << i); - write_sm1_unary_op(ctx, buffer, D3DSIO_RCP, &dst, &src, 0); - } - break; - - default: - FIXME("Unhandled op %u.\n", expr->op); - break; - } -} - -static void write_sm1_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr) -{ - const struct hlsl_ir_load *load = hlsl_ir_load(instr); - const struct hlsl_reg reg = hlsl_reg_from_deref(&load->src, instr->data_type); - struct sm1_instruction sm1_instr = - { - .opcode = D3DSIO_MOV, - - .dst.type = D3DSPR_TEMP, - .dst.reg = instr->reg.id, - .dst.writemask = instr->reg.writemask, - .has_dst = 1, - - .srcs[0].type = D3DSPR_TEMP, - .srcs[0].reg = reg.id, - .srcs[0].swizzle = swizzle_from_writemask(reg.writemask), - .src_count = 1, - }; - - assert(instr->reg.allocated); - - if (load->src.var->is_uniform) - { - assert(reg.allocated); - sm1_instr.srcs[0].type = D3DSPR_CONST; - } - else if (load->src.var->is_input_semantic) - { - if (!sm1_register_from_semantic(ctx, &load->src.var->semantic, - false, &sm1_instr.srcs[0].type, &sm1_instr.srcs[0].reg)) - { - assert(reg.allocated); - sm1_instr.srcs[0].type = D3DSPR_INPUT; - sm1_instr.srcs[0].reg = reg.id; - } - else - sm1_instr.srcs[0].swizzle = swizzle_from_writemask((1 << load->src.var->data_type->dimx) - 1); - } - - write_sm1_instruction(ctx, buffer, &sm1_instr); -} - -static void write_sm1_store(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct hlsl_ir_node *instr) -{ - const struct hlsl_ir_store *store = hlsl_ir_store(instr); - const struct hlsl_ir_node *rhs = store->rhs.node; - const struct hlsl_reg reg = hlsl_reg_from_deref(&store->lhs, rhs->data_type); - struct sm1_instruction sm1_instr = - { - .opcode = D3DSIO_MOV, - - .dst.type = D3DSPR_TEMP, - .dst.reg = reg.id, - .dst.writemask = combine_writemasks(reg.writemask, store->writemask), - .has_dst = 1, - - .srcs[0].type = D3DSPR_TEMP, - .srcs[0].reg = rhs->reg.id, - .srcs[0].swizzle = swizzle_from_writemask(rhs->reg.writemask), - .src_count = 1, - }; - - if (store->lhs.var->data_type->type == HLSL_CLASS_MATRIX) - { - FIXME("Matrix writemasks need to be lowered.\n"); - return; - } - - if (store->lhs.var->is_output_semantic) - { - if (!sm1_register_from_semantic(ctx, &store->lhs.var->semantic, true, &sm1_instr.dst.type, &sm1_instr.dst.reg)) - { - assert(reg.allocated); - sm1_instr.dst.type = D3DSPR_OUTPUT; - sm1_instr.dst.reg = reg.id; - } - else - sm1_instr.dst.writemask = (1u << store->lhs.var->data_type->dimx) - 1; - } - else - assert(reg.allocated); - - write_sm1_instruction(ctx, buffer, &sm1_instr); -} - -static void write_sm1_swizzle(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct hlsl_ir_node *instr) -{ - const struct hlsl_ir_swizzle *swizzle = hlsl_ir_swizzle(instr); - const struct hlsl_ir_node *val = swizzle->val.node; - struct sm1_instruction sm1_instr = - { - .opcode = D3DSIO_MOV, - - .dst.type = D3DSPR_TEMP, - .dst.reg = instr->reg.id, - .dst.writemask = instr->reg.writemask, - .has_dst = 1, - - .srcs[0].type = D3DSPR_TEMP, - .srcs[0].reg = val->reg.id, - .srcs[0].swizzle = combine_swizzles(swizzle_from_writemask(val->reg.writemask), - swizzle->swizzle, instr->data_type->dimx), - .src_count = 1, - }; - - assert(instr->reg.allocated); - assert(val->reg.allocated); - write_sm1_instruction(ctx, buffer, &sm1_instr); -} - -static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, - const struct hlsl_ir_function_decl *entry_func) -{ - const struct hlsl_ir_node *instr; - - LIST_FOR_EACH_ENTRY(instr, entry_func->body, struct hlsl_ir_node, entry) - { - if (instr->data_type) - { - if (instr->data_type->type == HLSL_CLASS_MATRIX) - { - FIXME("Matrix operations need to be lowered.\n"); - break; - } - - assert(instr->data_type->type == HLSL_CLASS_SCALAR || instr->data_type->type == HLSL_CLASS_VECTOR); - } - - switch (instr->type) - { - case HLSL_IR_CONSTANT: - write_sm1_constant(ctx, buffer, instr); - break; - - case HLSL_IR_EXPR: - write_sm1_expr(ctx, buffer, instr); - break; - - case HLSL_IR_LOAD: - write_sm1_load(ctx, buffer, instr); - break; - - case HLSL_IR_STORE: - write_sm1_store(ctx, buffer, instr); - break; - - case HLSL_IR_SWIZZLE: - write_sm1_swizzle(ctx, buffer, instr); - break; - - default: - FIXME("Unhandled instruction type %s.\n", hlsl_node_type_to_string(instr->type)); - } - } -} - -static int write_sm1_shader(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, - struct vkd3d_shader_code *out) -{ - struct vkd3d_bytecode_buffer buffer = {0}; - int ret; - - put_u32(&buffer, sm1_version(ctx->profile->type, ctx->profile->major_version, ctx->profile->minor_version)); - - write_sm1_uniforms(ctx, &buffer, entry_func); - - write_sm1_constant_defs(ctx, &buffer); - write_sm1_semantic_dcls(ctx, &buffer); - write_sm1_instructions(ctx, &buffer, entry_func); - - put_u32(&buffer, D3DSIO_END); - - if (!(ret = buffer.status)) - { - out->code = buffer.data; - out->size = buffer.size; - } - return ret; -} - int hlsl_emit_dxbc(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, struct vkd3d_shader_code *out) { struct hlsl_ir_var *var; @@ -2104,7 +1223,7 @@ int hlsl_emit_dxbc(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_fun return ctx->result;
if (ctx->profile->major_version < 4) - return write_sm1_shader(ctx, entry_func, out); + return hlsl_sm1_write(ctx, entry_func, out); else return VKD3D_ERROR_NOT_IMPLEMENTED; } diff --git a/libs/vkd3d-shader/hlsl_sm1.c b/libs/vkd3d-shader/hlsl_sm1.c new file mode 100644 index 00000000..6f9df654 --- /dev/null +++ b/libs/vkd3d-shader/hlsl_sm1.c @@ -0,0 +1,828 @@ +/* + * HLSL code generation for DXBC shader models 1-3 + * + * Copyright 2019-2020 Zebediah Figura for CodeWeavers + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA + */ + +#include "hlsl.h" +#include <stdio.h> + +bool hlsl_sm1_register_from_semantic(struct hlsl_ctx *ctx, const struct hlsl_semantic *semantic, + bool output, D3DSHADER_PARAM_REGISTER_TYPE *type, unsigned int *reg) +{ + unsigned int i; + + static const struct + { + const char *semantic; + bool output; + enum vkd3d_shader_type shader_type; + unsigned int major_version; + D3DSHADER_PARAM_REGISTER_TYPE type; + DWORD offset; + } + register_table[] = + { + {"color", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_COLOROUT}, + {"depth", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_DEPTHOUT}, + {"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_DEPTHOUT}, + {"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_COLOROUT}, + {"color", false, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_INPUT}, + {"texcoord", false, VKD3D_SHADER_TYPE_PIXEL, 2, D3DSPR_TEXTURE}, + + {"color", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_COLOROUT}, + {"depth", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_DEPTHOUT}, + {"sv_depth", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_DEPTHOUT}, + {"sv_target", true, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_COLOROUT}, + {"sv_position", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_POSITION}, + {"vface", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_FACE}, + {"vpos", false, VKD3D_SHADER_TYPE_PIXEL, 3, D3DSPR_MISCTYPE, D3DSMO_POSITION}, + + {"color", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_ATTROUT}, + {"fog", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_FOG}, + {"position", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POSITION}, + {"psize", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POINT_SIZE}, + {"sv_position", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_RASTOUT, D3DSRO_POSITION}, + {"texcoord", true, VKD3D_SHADER_TYPE_VERTEX, 1, D3DSPR_TEXCRDOUT}, + + {"color", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_ATTROUT}, + {"fog", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_FOG}, + {"position", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POSITION}, + {"psize", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POINT_SIZE}, + {"sv_position", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_RASTOUT, D3DSRO_POSITION}, + {"texcoord", true, VKD3D_SHADER_TYPE_VERTEX, 2, D3DSPR_TEXCRDOUT}, + }; + + for (i = 0; i < ARRAY_SIZE(register_table); ++i) + { + if (!ascii_strcasecmp(semantic->name, register_table[i].semantic) + && output == register_table[i].output + && ctx->profile->type == register_table[i].shader_type + && ctx->profile->major_version == register_table[i].major_version) + { + *type = register_table[i].type; + if (register_table[i].type == D3DSPR_MISCTYPE || register_table[i].type == D3DSPR_RASTOUT) + *reg = register_table[i].offset; + else + *reg = semantic->index; + return true; + } + } + + return false; +} + +bool hlsl_sm1_usage_from_semantic(const struct hlsl_semantic *semantic, D3DDECLUSAGE *usage, uint32_t *usage_idx) +{ + static const struct + { + const char *name; + D3DDECLUSAGE usage; + } + semantics[] = + { + {"binormal", D3DDECLUSAGE_BINORMAL}, + {"blendindices", D3DDECLUSAGE_BLENDINDICES}, + {"blendweight", D3DDECLUSAGE_BLENDWEIGHT}, + {"color", D3DDECLUSAGE_COLOR}, + {"depth", D3DDECLUSAGE_DEPTH}, + {"fog", D3DDECLUSAGE_FOG}, + {"normal", D3DDECLUSAGE_NORMAL}, + {"position", D3DDECLUSAGE_POSITION}, + {"positiont", D3DDECLUSAGE_POSITIONT}, + {"psize", D3DDECLUSAGE_PSIZE}, + {"sample", D3DDECLUSAGE_SAMPLE}, + {"sv_depth", D3DDECLUSAGE_DEPTH}, + {"sv_position", D3DDECLUSAGE_POSITION}, + {"sv_target", D3DDECLUSAGE_COLOR}, + {"tangent", D3DDECLUSAGE_TANGENT}, + {"tessfactor", D3DDECLUSAGE_TESSFACTOR}, + {"texcoord", D3DDECLUSAGE_TEXCOORD}, + }; + + unsigned int i; + + for (i = 0; i < ARRAY_SIZE(semantics); ++i) + { + if (!ascii_strcasecmp(semantic->name, semantics[i].name)) + { + *usage = semantics[i].usage; + *usage_idx = semantic->index; + return true; + } + } + + return false; +} + +static uint32_t sm1_version(enum vkd3d_shader_type type, unsigned int major, unsigned int minor) +{ + if (type == VKD3D_SHADER_TYPE_VERTEX) + return D3DVS_VERSION(major, minor); + else + return D3DPS_VERSION(major, minor); +} + +static D3DXPARAMETER_CLASS sm1_class(const struct hlsl_type *type) +{ + switch (type->type) + { + case HLSL_CLASS_ARRAY: + return sm1_class(type->e.array.type); + case HLSL_CLASS_MATRIX: + assert(type->modifiers & HLSL_MODIFIERS_MAJORITY_MASK); + if (type->modifiers & HLSL_MODIFIER_COLUMN_MAJOR) + return D3DXPC_MATRIX_COLUMNS; + else + return D3DXPC_MATRIX_ROWS; + case HLSL_CLASS_OBJECT: + return D3DXPC_OBJECT; + case HLSL_CLASS_SCALAR: + return D3DXPC_SCALAR; + case HLSL_CLASS_STRUCT: + return D3DXPC_STRUCT; + case HLSL_CLASS_VECTOR: + return D3DXPC_VECTOR; + default: + ERR("Invalid class %#x.\n", type->type); + assert(0); + return 0; + } +} + +static D3DXPARAMETER_TYPE sm1_base_type(const struct hlsl_type *type) +{ + switch (type->base_type) + { + case HLSL_TYPE_BOOL: + return D3DXPT_BOOL; + case HLSL_TYPE_FLOAT: + case HLSL_TYPE_HALF: + return D3DXPT_FLOAT; + case HLSL_TYPE_INT: + case HLSL_TYPE_UINT: + return D3DXPT_INT; + case HLSL_TYPE_PIXELSHADER: + return D3DXPT_PIXELSHADER; + case HLSL_TYPE_SAMPLER: + switch (type->sampler_dim) + { + case HLSL_SAMPLER_DIM_1D: + return D3DXPT_SAMPLER1D; + case HLSL_SAMPLER_DIM_2D: + return D3DXPT_SAMPLER2D; + case HLSL_SAMPLER_DIM_3D: + return D3DXPT_SAMPLER3D; + case HLSL_SAMPLER_DIM_CUBE: + return D3DXPT_SAMPLERCUBE; + case HLSL_SAMPLER_DIM_GENERIC: + return D3DXPT_SAMPLER; + default: + ERR("Invalid dimension %#x.\n", type->sampler_dim); + } + break; + case HLSL_TYPE_STRING: + return D3DXPT_STRING; + case HLSL_TYPE_TEXTURE: + switch (type->sampler_dim) + { + case HLSL_SAMPLER_DIM_1D: + return D3DXPT_TEXTURE1D; + case HLSL_SAMPLER_DIM_2D: + return D3DXPT_TEXTURE2D; + case HLSL_SAMPLER_DIM_3D: + return D3DXPT_TEXTURE3D; + case HLSL_SAMPLER_DIM_CUBE: + return D3DXPT_TEXTURECUBE; + case HLSL_SAMPLER_DIM_GENERIC: + return D3DXPT_TEXTURE; + default: + ERR("Invalid dimension %#x.\n", type->sampler_dim); + } + break; + case HLSL_TYPE_VERTEXSHADER: + return D3DXPT_VERTEXSHADER; + case HLSL_TYPE_VOID: + return D3DXPT_VOID; + default: + assert(0); + } + assert(0); + return 0; +} + +static const struct hlsl_type *get_array_type(const struct hlsl_type *type) +{ + if (type->type == HLSL_CLASS_ARRAY) + return get_array_type(type->e.array.type); + return type; +} + +static unsigned int get_array_size(const struct hlsl_type *type) +{ + if (type->type == HLSL_CLASS_ARRAY) + return get_array_size(type->e.array.type) * type->e.array.elements_count; + return 1; +} + +static void write_sm1_type(struct vkd3d_bytecode_buffer *buffer, struct hlsl_type *type, unsigned int ctab_start) +{ + const struct hlsl_type *array_type = get_array_type(type); + unsigned int array_size = get_array_size(type); + struct hlsl_struct_field *field; + unsigned int field_count = 0; + size_t fields_offset = 0; + + if (type->bytecode_offset) + return; + + if (array_type->type == HLSL_CLASS_STRUCT) + { + LIST_FOR_EACH_ENTRY(field, array_type->e.elements, struct hlsl_struct_field, entry) + { + field->name_bytecode_offset = put_string(buffer, field->name); + write_sm1_type(buffer, field->type, ctab_start); + } + + fields_offset = bytecode_get_size(buffer) - ctab_start; + + LIST_FOR_EACH_ENTRY(field, array_type->e.elements, struct hlsl_struct_field, entry) + { + put_u32(buffer, field->name_bytecode_offset - ctab_start); + put_u32(buffer, field->type->bytecode_offset - ctab_start); + ++field_count; + } + } + + type->bytecode_offset = put_u32(buffer, sm1_class(type) | (sm1_base_type(type) << 16)); + put_u32(buffer, type->dimy | (type->dimx << 16)); + put_u32(buffer, array_size | (field_count << 16)); + put_u32(buffer, fields_offset); +} + +static void sm1_sort_extern(struct list *sorted, struct hlsl_ir_var *to_sort) +{ + struct hlsl_ir_var *var; + + list_remove(&to_sort->extern_entry); + + LIST_FOR_EACH_ENTRY(var, sorted, struct hlsl_ir_var, extern_entry) + { + if (strcmp(to_sort->name, var->name) < 0) + { + list_add_before(&var->extern_entry, &to_sort->extern_entry); + return; + } + } + + list_add_tail(sorted, &to_sort->extern_entry); +} + +static void sm1_sort_externs(struct hlsl_ctx *ctx) +{ + struct list sorted = LIST_INIT(sorted); + struct hlsl_ir_var *var, *next; + + LIST_FOR_EACH_ENTRY_SAFE(var, next, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + sm1_sort_extern(&sorted, var); + list_move_tail(&ctx->extern_vars, &sorted); +} + +static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + struct hlsl_ir_function_decl *entry_func) +{ + size_t ctab_offset, ctab_start, ctab_end, vars_start, size_offset, creator_offset, offset; + unsigned int uniform_count = 0; + struct hlsl_ir_var *var; + + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (!var->semantic.name && var->reg.allocated) + { + ++uniform_count; + + if (var->is_param && var->is_uniform) + { + struct vkd3d_string_buffer *name; + + if (!(name = hlsl_get_string_buffer(ctx))) + { + buffer->status = VKD3D_ERROR_OUT_OF_MEMORY; + return; + } + vkd3d_string_buffer_printf(name, "$%s", var->name); + vkd3d_free((char *)var->name); + var->name = hlsl_strdup(ctx, name->buffer); + hlsl_release_string_buffer(ctx, name); + } + } + } + + sm1_sort_externs(ctx); + + size_offset = put_u32(buffer, 0); + ctab_offset = put_u32(buffer, MAKEFOURCC('C','T','A','B')); + + ctab_start = put_u32(buffer, sizeof(D3DXSHADER_CONSTANTTABLE)); + creator_offset = put_u32(buffer, 0); + put_u32(buffer, sm1_version(ctx->profile->type, ctx->profile->major_version, ctx->profile->minor_version)); + put_u32(buffer, uniform_count); + put_u32(buffer, sizeof(D3DXSHADER_CONSTANTTABLE)); /* offset of constants */ + put_u32(buffer, 0); /* FIXME: flags */ + put_u32(buffer, 0); /* FIXME: target string */ + + vars_start = bytecode_get_size(buffer); + + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (!var->semantic.name && var->reg.allocated) + { + put_u32(buffer, 0); /* name */ + put_u32(buffer, D3DXRS_FLOAT4 | (var->reg.id << 16)); + put_u32(buffer, var->data_type->reg_size / 4); + put_u32(buffer, 0); /* type */ + put_u32(buffer, 0); /* FIXME: default value */ + } + } + + uniform_count = 0; + + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (!var->semantic.name && var->reg.allocated) + { + size_t var_offset = vars_start + (uniform_count * 5 * sizeof(uint32_t)); + size_t name_offset; + + name_offset = put_string(buffer, var->name); + set_u32(buffer, var_offset, name_offset - ctab_start); + + write_sm1_type(buffer, var->data_type, ctab_start); + set_u32(buffer, var_offset + 3 * sizeof(uint32_t), var->data_type->bytecode_offset - ctab_start); + ++uniform_count; + } + } + + offset = put_string(buffer, vkd3d_shader_get_version(NULL, NULL)); + set_u32(buffer, creator_offset, offset - ctab_start); + + ctab_end = bytecode_get_size(buffer); + set_u32(buffer, size_offset, D3DSIO_COMMENT | (((ctab_end - ctab_offset) / sizeof(uint32_t)) << 16)); +} + +static uint32_t sm1_encode_register_type(D3DSHADER_PARAM_REGISTER_TYPE type) +{ + return ((type << D3DSP_REGTYPE_SHIFT) & D3DSP_REGTYPE_MASK) + | ((type << D3DSP_REGTYPE_SHIFT2) & D3DSP_REGTYPE_MASK2); +} + +struct sm1_instruction +{ + D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode; + + struct sm1_dst_register + { + D3DSHADER_PARAM_REGISTER_TYPE type; + D3DSHADER_PARAM_DSTMOD_TYPE mod; + unsigned int writemask; + uint32_t reg; + } dst; + + struct sm1_src_register + { + D3DSHADER_PARAM_REGISTER_TYPE type; + D3DSHADER_PARAM_SRCMOD_TYPE mod; + unsigned int swizzle; + uint32_t reg; + } srcs[2]; + unsigned int src_count; + + unsigned int has_dst; +}; + +static void write_sm1_dst_register(struct vkd3d_bytecode_buffer *buffer, const struct sm1_dst_register *reg) +{ + assert(reg->writemask); + put_u32(buffer, (1u << 31) | sm1_encode_register_type(reg->type) | reg->mod | (reg->writemask << 16) | reg->reg); +} + +static void write_sm1_src_register(struct vkd3d_bytecode_buffer *buffer, + const struct sm1_src_register *reg, unsigned int dst_writemask) +{ + unsigned int swizzle = hlsl_map_swizzle(reg->swizzle, dst_writemask); + + put_u32(buffer, (1u << 31) | sm1_encode_register_type(reg->type) | reg->mod | (swizzle << 16) | reg->reg); +} + +static void write_sm1_instruction(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct sm1_instruction *instr) +{ + uint32_t token = instr->opcode; + unsigned int i; + + if (ctx->profile->major_version > 1) + token |= (instr->has_dst + instr->src_count) << D3DSI_INSTLENGTH_SHIFT; + put_u32(buffer, token); + + if (instr->has_dst) + write_sm1_dst_register(buffer, &instr->dst); + + for (i = 0; i < instr->src_count; ++i) + write_sm1_src_register(buffer, &instr->srcs[i], instr->dst.writemask); +}; + +static void write_sm1_binary_op(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode, const struct hlsl_reg *dst, + const struct hlsl_reg *src1, const struct hlsl_reg *src2) +{ + const struct sm1_instruction instr = + { + .opcode = opcode, + + .dst.type = D3DSPR_TEMP, + .dst.writemask = dst->writemask, + .dst.reg = dst->id, + .has_dst = 1, + + .srcs[0].type = D3DSPR_TEMP, + .srcs[0].swizzle = hlsl_swizzle_from_writemask(src1->writemask), + .srcs[0].reg = src1->id, + .srcs[1].type = D3DSPR_TEMP, + .srcs[1].swizzle = hlsl_swizzle_from_writemask(src2->writemask), + .srcs[1].reg = src2->id, + .src_count = 2, + }; + write_sm1_instruction(ctx, buffer, &instr); +} + +static void write_sm1_unary_op(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + D3DSHADER_INSTRUCTION_OPCODE_TYPE opcode, const struct hlsl_reg *dst, + const struct hlsl_reg *src, D3DSHADER_PARAM_SRCMOD_TYPE src_mod) +{ + const struct sm1_instruction instr = + { + .opcode = opcode, + + .dst.type = D3DSPR_TEMP, + .dst.writemask = dst->writemask, + .dst.reg = dst->id, + .has_dst = 1, + + .srcs[0].type = D3DSPR_TEMP, + .srcs[0].swizzle = hlsl_swizzle_from_writemask(src->writemask), + .srcs[0].reg = src->id, + .srcs[0].mod = src_mod, + .src_count = 1, + }; + write_sm1_instruction(ctx, buffer, &instr); +} + +static void write_sm1_constant_defs(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer) +{ + unsigned int i, x; + + for (i = 0; i < ctx->constant_defs.count; ++i) + { + uint32_t token = D3DSIO_DEF; + const struct sm1_dst_register reg = + { + .type = D3DSPR_CONST, + .writemask = VKD3DSP_WRITEMASK_ALL, + .reg = i, + }; + + if (ctx->profile->major_version > 1) + token |= 5 << D3DSI_INSTLENGTH_SHIFT; + put_u32(buffer, token); + + write_sm1_dst_register(buffer, ®); + for (x = 0; x < 4; ++x) + put_f32(buffer, ctx->constant_defs.values[i].f[x]); + } +} + +static void write_sm1_semantic_dcl(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct hlsl_ir_var *var, bool output) +{ + struct sm1_dst_register reg = {0}; + uint32_t token, usage_idx; + D3DDECLUSAGE usage; + bool ret; + + if (hlsl_sm1_register_from_semantic(ctx, &var->semantic, output, ®.type, ®.reg)) + { + usage = 0; + usage_idx = 0; + } + else + { + ret = hlsl_sm1_usage_from_semantic(&var->semantic, &usage, &usage_idx); + assert(ret); + reg.type = output ? D3DSPR_OUTPUT : D3DSPR_INPUT; + reg.reg = var->reg.id; + } + + token = D3DSIO_DCL; + if (ctx->profile->major_version > 1) + token |= 2 << D3DSI_INSTLENGTH_SHIFT; + put_u32(buffer, token); + + token = (1u << 31); + token |= usage << D3DSP_DCL_USAGE_SHIFT; + token |= usage_idx << D3DSP_DCL_USAGEINDEX_SHIFT; + put_u32(buffer, token); + + reg.writemask = (1 << var->data_type->dimx) - 1; + write_sm1_dst_register(buffer, ®); +} + +static void write_sm1_semantic_dcls(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer) +{ + bool write_in = false, write_out = false; + struct hlsl_ir_var *var; + + if (ctx->profile->type == VKD3D_SHADER_TYPE_PIXEL) + write_in = true; + else if (ctx->profile->type == VKD3D_SHADER_TYPE_VERTEX && ctx->profile->major_version == 3) + write_in = write_out = true; + else if (ctx->profile->type == VKD3D_SHADER_TYPE_VERTEX && ctx->profile->major_version < 3) + write_in = true; + + LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (write_in && var->is_input_semantic) + write_sm1_semantic_dcl(ctx, buffer, var, false); + if (write_out && var->is_output_semantic) + write_sm1_semantic_dcl(ctx, buffer, var, true); + } +} + +static void write_sm1_constant(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct hlsl_ir_node *instr) +{ + const struct hlsl_ir_constant *constant = hlsl_ir_constant(instr); + struct sm1_instruction sm1_instr = + { + .opcode = D3DSIO_MOV, + + .dst.type = D3DSPR_TEMP, + .dst.reg = instr->reg.id, + .dst.writemask = instr->reg.writemask, + .has_dst = 1, + + .srcs[0].type = D3DSPR_CONST, + .srcs[0].reg = constant->reg.id, + .srcs[0].swizzle = hlsl_swizzle_from_writemask(constant->reg.writemask), + .src_count = 1, + }; + + assert(instr->reg.allocated); + assert(constant->reg.allocated); + write_sm1_instruction(ctx, buffer, &sm1_instr); +} + +static void write_sm1_expr(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr) +{ + struct hlsl_ir_expr *expr = hlsl_ir_expr(instr); + struct hlsl_ir_node *arg1 = expr->operands[0].node; + struct hlsl_ir_node *arg2 = expr->operands[1].node; + unsigned int i; + + assert(instr->reg.allocated); + + if (instr->data_type->base_type != HLSL_TYPE_FLOAT) + { + FIXME("Non-float operations need to be lowered.\n"); + return; + } + + switch (expr->op) + { + case HLSL_IR_BINOP_ADD: + write_sm1_binary_op(ctx, buffer, D3DSIO_ADD, &instr->reg, &arg1->reg, &arg2->reg); + break; + + case HLSL_IR_BINOP_MUL: + write_sm1_binary_op(ctx, buffer, D3DSIO_MUL, &instr->reg, &arg1->reg, &arg2->reg); + break; + + case HLSL_IR_BINOP_SUB: + write_sm1_binary_op(ctx, buffer, D3DSIO_SUB, &instr->reg, &arg1->reg, &arg2->reg); + break; + + case HLSL_IR_UNOP_NEG: + write_sm1_unary_op(ctx, buffer, D3DSIO_MOV, &instr->reg, &arg1->reg, D3DSPSM_NEG); + break; + + case HLSL_IR_UNOP_RCP: + for (i = 0; i < instr->data_type->dimx; ++i) + { + struct hlsl_reg src = arg1->reg, dst = instr->reg; + + src.writemask = hlsl_combine_writemasks(src.writemask, 1u << i); + dst.writemask = hlsl_combine_writemasks(dst.writemask, 1u << i); + write_sm1_unary_op(ctx, buffer, D3DSIO_RCP, &dst, &src, 0); + } + break; + + default: + FIXME("Unhandled op %u.\n", expr->op); + break; + } +} + +static void write_sm1_load(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, const struct hlsl_ir_node *instr) +{ + const struct hlsl_ir_load *load = hlsl_ir_load(instr); + const struct hlsl_reg reg = hlsl_reg_from_deref(&load->src, instr->data_type); + struct sm1_instruction sm1_instr = + { + .opcode = D3DSIO_MOV, + + .dst.type = D3DSPR_TEMP, + .dst.reg = instr->reg.id, + .dst.writemask = instr->reg.writemask, + .has_dst = 1, + + .srcs[0].type = D3DSPR_TEMP, + .srcs[0].reg = reg.id, + .srcs[0].swizzle = hlsl_swizzle_from_writemask(reg.writemask), + .src_count = 1, + }; + + assert(instr->reg.allocated); + + if (load->src.var->is_uniform) + { + assert(reg.allocated); + sm1_instr.srcs[0].type = D3DSPR_CONST; + } + else if (load->src.var->is_input_semantic) + { + if (!hlsl_sm1_register_from_semantic(ctx, &load->src.var->semantic, + false, &sm1_instr.srcs[0].type, &sm1_instr.srcs[0].reg)) + { + assert(reg.allocated); + sm1_instr.srcs[0].type = D3DSPR_INPUT; + sm1_instr.srcs[0].reg = reg.id; + } + else + sm1_instr.srcs[0].swizzle = hlsl_swizzle_from_writemask((1 << load->src.var->data_type->dimx) - 1); + } + + write_sm1_instruction(ctx, buffer, &sm1_instr); +} + +static void write_sm1_store(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct hlsl_ir_node *instr) +{ + const struct hlsl_ir_store *store = hlsl_ir_store(instr); + const struct hlsl_ir_node *rhs = store->rhs.node; + const struct hlsl_reg reg = hlsl_reg_from_deref(&store->lhs, rhs->data_type); + struct sm1_instruction sm1_instr = + { + .opcode = D3DSIO_MOV, + + .dst.type = D3DSPR_TEMP, + .dst.reg = reg.id, + .dst.writemask = hlsl_combine_writemasks(reg.writemask, store->writemask), + .has_dst = 1, + + .srcs[0].type = D3DSPR_TEMP, + .srcs[0].reg = rhs->reg.id, + .srcs[0].swizzle = hlsl_swizzle_from_writemask(rhs->reg.writemask), + .src_count = 1, + }; + + if (store->lhs.var->data_type->type == HLSL_CLASS_MATRIX) + { + FIXME("Matrix writemasks need to be lowered.\n"); + return; + } + + if (store->lhs.var->is_output_semantic) + { + if (!hlsl_sm1_register_from_semantic(ctx, &store->lhs.var->semantic, + true, &sm1_instr.dst.type, &sm1_instr.dst.reg)) + { + assert(reg.allocated); + sm1_instr.dst.type = D3DSPR_OUTPUT; + sm1_instr.dst.reg = reg.id; + } + else + sm1_instr.dst.writemask = (1u << store->lhs.var->data_type->dimx) - 1; + } + else + assert(reg.allocated); + + write_sm1_instruction(ctx, buffer, &sm1_instr); +} + +static void write_sm1_swizzle(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct hlsl_ir_node *instr) +{ + const struct hlsl_ir_swizzle *swizzle = hlsl_ir_swizzle(instr); + const struct hlsl_ir_node *val = swizzle->val.node; + struct sm1_instruction sm1_instr = + { + .opcode = D3DSIO_MOV, + + .dst.type = D3DSPR_TEMP, + .dst.reg = instr->reg.id, + .dst.writemask = instr->reg.writemask, + .has_dst = 1, + + .srcs[0].type = D3DSPR_TEMP, + .srcs[0].reg = val->reg.id, + .srcs[0].swizzle = hlsl_combine_swizzles(hlsl_swizzle_from_writemask(val->reg.writemask), + swizzle->swizzle, instr->data_type->dimx), + .src_count = 1, + }; + + assert(instr->reg.allocated); + assert(val->reg.allocated); + write_sm1_instruction(ctx, buffer, &sm1_instr); +} + +static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *buffer, + const struct hlsl_ir_function_decl *entry_func) +{ + const struct hlsl_ir_node *instr; + + LIST_FOR_EACH_ENTRY(instr, entry_func->body, struct hlsl_ir_node, entry) + { + if (instr->data_type) + { + if (instr->data_type->type == HLSL_CLASS_MATRIX) + { + FIXME("Matrix operations need to be lowered.\n"); + break; + } + + assert(instr->data_type->type == HLSL_CLASS_SCALAR || instr->data_type->type == HLSL_CLASS_VECTOR); + } + + switch (instr->type) + { + case HLSL_IR_CONSTANT: + write_sm1_constant(ctx, buffer, instr); + break; + + case HLSL_IR_EXPR: + write_sm1_expr(ctx, buffer, instr); + break; + + case HLSL_IR_LOAD: + write_sm1_load(ctx, buffer, instr); + break; + + case HLSL_IR_STORE: + write_sm1_store(ctx, buffer, instr); + break; + + case HLSL_IR_SWIZZLE: + write_sm1_swizzle(ctx, buffer, instr); + break; + + default: + FIXME("Unhandled instruction type %s.\n", hlsl_node_type_to_string(instr->type)); + } + } +} + +int hlsl_sm1_write(struct hlsl_ctx *ctx, struct hlsl_ir_function_decl *entry_func, struct vkd3d_shader_code *out) +{ + struct vkd3d_bytecode_buffer buffer = {0}; + int ret; + + put_u32(&buffer, sm1_version(ctx->profile->type, ctx->profile->major_version, ctx->profile->minor_version)); + + write_sm1_uniforms(ctx, &buffer, entry_func); + + write_sm1_constant_defs(ctx, &buffer); + write_sm1_semantic_dcls(ctx, &buffer); + write_sm1_instructions(ctx, &buffer, entry_func); + + put_u32(&buffer, D3DSIO_END); + + if (!(ret = buffer.status)) + { + out->code = buffer.data; + out->size = buffer.size; + } + return ret; +}
Signed-off-by: Matteo Bruni mbruni@codeweavers.com
Signed-off-by: Henri Verbeet hverbeet@codeweavers.com
Signed-off-by: Zebediah Figura zfigura@codeweavers.com --- include/private/vkd3d_common.h | 11 +++++++++++ libs/vkd3d-shader/checksum.c | 3 +-- libs/vkd3d-shader/hlsl_sm1.c | 10 +++++----- libs/vkd3d-shader/spirv.c | 2 +- 4 files changed, 18 insertions(+), 8 deletions(-)
diff --git a/include/private/vkd3d_common.h b/include/private/vkd3d_common.h index b639c0fe..46a39b4d 100644 --- a/include/private/vkd3d_common.h +++ b/include/private/vkd3d_common.h @@ -26,6 +26,7 @@ #include <ctype.h> #include <limits.h> #include <stdbool.h> +#include <stdint.h>
#ifdef _MSC_VER #include <intrin.h> @@ -149,6 +150,16 @@ static inline bool vkd3d_bound_range(size_t start, size_t count, size_t limit) #endif }
+static inline uint16_t vkd3d_make_u16(uint8_t low, uint8_t high) +{ + return low | ((uint16_t)high << 8); +} + +static inline uint32_t vkd3d_make_u32(uint16_t low, uint16_t high) +{ + return low | ((uint32_t)high << 16); +} + static inline int ascii_isupper(int c) { return 'A' <= c && c <= 'Z'; diff --git a/libs/vkd3d-shader/checksum.c b/libs/vkd3d-shader/checksum.c index b9a71402..0910729a 100644 --- a/libs/vkd3d-shader/checksum.c +++ b/libs/vkd3d-shader/checksum.c @@ -156,8 +156,7 @@ static void byte_reverse(unsigned char *buf, unsigned longs)
do { - t = ((unsigned)buf[3] << 8 | buf[2]) << 16 | - ((unsigned)buf[1] << 8 | buf[0]); + t = vkd3d_make_u32(vkd3d_make_u16(buf[0], buf[1]), vkd3d_make_u16(buf[2], buf[3])); *(unsigned int *)buf = t; buf += 4; } while (--longs); diff --git a/libs/vkd3d-shader/hlsl_sm1.c b/libs/vkd3d-shader/hlsl_sm1.c index 6f9df654..ca02f01c 100644 --- a/libs/vkd3d-shader/hlsl_sm1.c +++ b/libs/vkd3d-shader/hlsl_sm1.c @@ -268,9 +268,9 @@ static void write_sm1_type(struct vkd3d_bytecode_buffer *buffer, struct hlsl_typ } }
- type->bytecode_offset = put_u32(buffer, sm1_class(type) | (sm1_base_type(type) << 16)); - put_u32(buffer, type->dimy | (type->dimx << 16)); - put_u32(buffer, array_size | (field_count << 16)); + type->bytecode_offset = put_u32(buffer, vkd3d_make_u32(sm1_class(type), sm1_base_type(type))); + put_u32(buffer, vkd3d_make_u32(type->dimy, type->dimx)); + put_u32(buffer, vkd3d_make_u32(array_size, field_count)); put_u32(buffer, fields_offset); }
@@ -352,7 +352,7 @@ static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffe if (!var->semantic.name && var->reg.allocated) { put_u32(buffer, 0); /* name */ - put_u32(buffer, D3DXRS_FLOAT4 | (var->reg.id << 16)); + put_u32(buffer, vkd3d_make_u32(D3DXRS_FLOAT4, var->reg.id)); put_u32(buffer, var->data_type->reg_size / 4); put_u32(buffer, 0); /* type */ put_u32(buffer, 0); /* FIXME: default value */ @@ -381,7 +381,7 @@ static void write_sm1_uniforms(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffe set_u32(buffer, creator_offset, offset - ctab_start);
ctab_end = bytecode_get_size(buffer); - set_u32(buffer, size_offset, D3DSIO_COMMENT | (((ctab_end - ctab_offset) / sizeof(uint32_t)) << 16)); + set_u32(buffer, size_offset, vkd3d_make_u32(D3DSIO_COMMENT, (ctab_end - ctab_offset) / sizeof(uint32_t))); }
static uint32_t sm1_encode_register_type(D3DSHADER_PARAM_REGISTER_TYPE type) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 08709e44..022c1c47 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -215,7 +215,7 @@ static enum vkd3d_shader_input_sysval_semantic vkd3d_siv_from_sysval(enum vkd3d_ #define VKD3D_SPIRV_VERSION 0x00010000 #define VKD3D_SPIRV_GENERATOR_ID 18 #define VKD3D_SPIRV_GENERATOR_VERSION 2 -#define VKD3D_SPIRV_GENERATOR_MAGIC ((VKD3D_SPIRV_GENERATOR_ID << 16) | VKD3D_SPIRV_GENERATOR_VERSION) +#define VKD3D_SPIRV_GENERATOR_MAGIC vkd3d_make_u32(VKD3D_SPIRV_GENERATOR_VERSION, VKD3D_SPIRV_GENERATOR_ID)
struct vkd3d_spirv_stream {
Signed-off-by: Matteo Bruni mbruni@codeweavers.com --- Perhaps one could argue for switching around the arguments to have the (high, low) order but this also works.
Signed-off-by: Henri Verbeet hverbeet@codeweavers.com
Signed-off-by: Zebediah Figura zfigura@codeweavers.com --- libs/vkd3d-shader/hlsl.c | 1 - libs/vkd3d-shader/hlsl.h | 1 - libs/vkd3d-shader/hlsl.y | 20 ++++++++++++++++++-- libs/vkd3d-shader/hlsl_sm1.c | 4 ---- 4 files changed, 18 insertions(+), 8 deletions(-)
diff --git a/libs/vkd3d-shader/hlsl.c b/libs/vkd3d-shader/hlsl.c index d0be0750..e4070572 100644 --- a/libs/vkd3d-shader/hlsl.c +++ b/libs/vkd3d-shader/hlsl.c @@ -1062,7 +1062,6 @@ static const char *debug_expr_op(const struct hlsl_ir_expr *expr) "sat",
"+", - "-", "*", "/",
diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index e566ac7d..3b0e5207 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -302,7 +302,6 @@ enum hlsl_ir_expr_op HLSL_IR_UNOP_SAT,
HLSL_IR_BINOP_ADD, - HLSL_IR_BINOP_SUB, HLSL_IR_BINOP_MUL, HLSL_IR_BINOP_DIV,
diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index 34cdb543..2754514d 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -1118,7 +1118,7 @@ static enum hlsl_ir_expr_op op_from_assignment(enum parse_assign_op op) { 0, HLSL_IR_BINOP_ADD, - HLSL_IR_BINOP_SUB, + 0, HLSL_IR_BINOP_MUL, HLSL_IR_BINOP_DIV, HLSL_IR_BINOP_MOD, @@ -1176,12 +1176,23 @@ static struct hlsl_ir_node *add_assignment(struct hlsl_ctx *ctx, struct list *in struct hlsl_ir_expr *copy; DWORD writemask = 0;
+ if (assign_op == ASSIGN_OP_SUB) + { + struct hlsl_ir_node *args[3] = {rhs}; + struct hlsl_ir_expr *expr; + + if (!(expr = add_expr(ctx, instrs, HLSL_IR_UNOP_NEG, args, &rhs->loc))) + return NULL; + rhs = &expr->node; + assign_op = ASSIGN_OP_ADD; + } if (assign_op != ASSIGN_OP_ASSIGN) { enum hlsl_ir_expr_op op = op_from_assignment(assign_op); struct hlsl_ir_node *args[3] = {lhs, rhs}; struct hlsl_ir_expr *expr;
+ assert(op); if (!(expr = add_expr(ctx, instrs, op, args, &rhs->loc))) return NULL; rhs = &expr->node; @@ -2909,7 +2920,12 @@ add_expr: } | add_expr '-' mul_expr { - $$ = add_binary_expr(ctx, $1, $3, HLSL_IR_BINOP_SUB, @2); + struct hlsl_ir_node *neg; + + if (!(neg = hlsl_new_unary_expr(ctx, HLSL_IR_UNOP_NEG, node_from_list($3), @2))) + YYABORT; + list_add_tail($3, &neg->entry); + $$ = add_binary_expr(ctx, $1, $3, HLSL_IR_BINOP_ADD, @2); }
shift_expr: diff --git a/libs/vkd3d-shader/hlsl_sm1.c b/libs/vkd3d-shader/hlsl_sm1.c index ca02f01c..8ea663dc 100644 --- a/libs/vkd3d-shader/hlsl_sm1.c +++ b/libs/vkd3d-shader/hlsl_sm1.c @@ -620,10 +620,6 @@ static void write_sm1_expr(struct hlsl_ctx *ctx, struct vkd3d_bytecode_buffer *b write_sm1_binary_op(ctx, buffer, D3DSIO_MUL, &instr->reg, &arg1->reg, &arg2->reg); break;
- case HLSL_IR_BINOP_SUB: - write_sm1_binary_op(ctx, buffer, D3DSIO_SUB, &instr->reg, &arg1->reg, &arg2->reg); - break; - case HLSL_IR_UNOP_NEG: write_sm1_unary_op(ctx, buffer, D3DSIO_MOV, &instr->reg, &arg1->reg, D3DSPSM_NEG); break;
Signed-off-by: Matteo Bruni mbruni@codeweavers.com
Eventually they will be split or otherwise transformed before getting here.
Signed-off-by: Matteo Bruni mbruni@codeweavers.com --- libs/vkd3d-shader/hlsl_sm1.c | 5 +++++ 1 file changed, 5 insertions(+)
diff --git a/libs/vkd3d-shader/hlsl_sm1.c b/libs/vkd3d-shader/hlsl_sm1.c index 8ea663dc..f351db9a 100644 --- a/libs/vkd3d-shader/hlsl_sm1.c +++ b/libs/vkd3d-shader/hlsl_sm1.c @@ -768,6 +768,11 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b FIXME("Matrix operations need to be lowered.\n"); break; } + if (instr->data_type->type == HLSL_CLASS_ARRAY) + { + FIXME("Array operations need to be split.\n"); + break; + }
assert(instr->data_type->type == HLSL_CLASS_SCALAR || instr->data_type->type == HLSL_CLASS_VECTOR); }
On Wed, 11 Aug 2021 at 13:50, Matteo Bruni mbruni@codeweavers.com wrote:
@@ -768,6 +768,11 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b FIXME("Matrix operations need to be lowered.\n"); break; }
if (instr->data_type->type == HLSL_CLASS_ARRAY){FIXME("Array operations need to be split.\n");break;}
Should that be a compiler error/warning instead? Arguably that would apply to FIXMEs/ERRs in compiler code in general; I think there are very few cases where we don't want those to be compiler messages instead.
On Wed, Aug 11, 2021 at 2:31 PM Henri Verbeet hverbeet@gmail.com wrote:
On Wed, 11 Aug 2021 at 13:50, Matteo Bruni mbruni@codeweavers.com wrote:
@@ -768,6 +768,11 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b FIXME("Matrix operations need to be lowered.\n"); break; }
if (instr->data_type->type == HLSL_CLASS_ARRAY){FIXME("Array operations need to be split.\n");break;}Should that be a compiler error/warning instead? Arguably that would apply to FIXMEs/ERRs in compiler code in general; I think there are very few cases where we don't want those to be compiler messages instead.
I guess that's a possibility. In general we're currently ignoring a lot of error conditions caused by missing compiler features. We could certainly start switching it over to spewing errors instead (or better, implementing the missing features). It feels to me that we will eventually want to do that for sure but maybe not right away. At this point I'd push the compiler a bit further, then e.g. as soon as trivial shaders start to be compiled correctly for both SM1 and SM4 we can start tightening it up so that the compiler either generates correct bytecode or errors out with some compilation message.
I can certainly be convinced otherwise though.
On Wed, 11 Aug 2021 at 14:49, Matteo Bruni matteo.mystral@gmail.com wrote:
On Wed, Aug 11, 2021 at 2:31 PM Henri Verbeet hverbeet@gmail.com wrote:
On Wed, 11 Aug 2021 at 13:50, Matteo Bruni mbruni@codeweavers.com wrote:
@@ -768,6 +768,11 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b FIXME("Matrix operations need to be lowered.\n"); break; }
if (instr->data_type->type == HLSL_CLASS_ARRAY){FIXME("Array operations need to be split.\n");break;}Should that be a compiler error/warning instead? Arguably that would apply to FIXMEs/ERRs in compiler code in general; I think there are very few cases where we don't want those to be compiler messages instead.
I guess that's a possibility. In general we're currently ignoring a lot of error conditions caused by missing compiler features. We could certainly start switching it over to spewing errors instead (or better, implementing the missing features). It feels to me that we will eventually want to do that for sure but maybe not right away. At this point I'd push the compiler a bit further, then e.g. as soon as trivial shaders start to be compiled correctly for both SM1 and SM4 we can start tightening it up so that the compiler either generates correct bytecode or errors out with some compilation message.
I can certainly be convinced otherwise though.
It may not be worth the effort of changing the existing FIXMEs at this point, but it seems more questionable to add new ones if we already know we'll have to replace them in the future. If the concern is that generating an error means that the generated code won't be returned to the caller we could generate a warning instead and upgrade it to an error later, although I'm not sure how useful invalid byte code is going to be to the caller.
On Wed, Aug 11, 2021 at 3:15 PM Henri Verbeet hverbeet@gmail.com wrote:
On Wed, 11 Aug 2021 at 14:49, Matteo Bruni matteo.mystral@gmail.com wrote:
On Wed, Aug 11, 2021 at 2:31 PM Henri Verbeet hverbeet@gmail.com wrote:
On Wed, 11 Aug 2021 at 13:50, Matteo Bruni mbruni@codeweavers.com wrote:
@@ -768,6 +768,11 @@ static void write_sm1_instructions(struct hlsl_ctx *ctx, struct vkd3d_bytecode_b FIXME("Matrix operations need to be lowered.\n"); break; }
if (instr->data_type->type == HLSL_CLASS_ARRAY){FIXME("Array operations need to be split.\n");break;}Should that be a compiler error/warning instead? Arguably that would apply to FIXMEs/ERRs in compiler code in general; I think there are very few cases where we don't want those to be compiler messages instead.
I guess that's a possibility. In general we're currently ignoring a lot of error conditions caused by missing compiler features. We could certainly start switching it over to spewing errors instead (or better, implementing the missing features). It feels to me that we will eventually want to do that for sure but maybe not right away. At this point I'd push the compiler a bit further, then e.g. as soon as trivial shaders start to be compiled correctly for both SM1 and SM4 we can start tightening it up so that the compiler either generates correct bytecode or errors out with some compilation message.
I can certainly be convinced otherwise though.
It may not be worth the effort of changing the existing FIXMEs at this point, but it seems more questionable to add new ones if we already know we'll have to replace them in the future. If the concern is that generating an error means that the generated code won't be returned to the caller we could generate a warning instead and upgrade it to an error later, although I'm not sure how useful invalid byte code is going to be to the caller.
It can be valid bytecode, although not necessarily for the HLSL that was given as source :D
I'm sending a new patch flagging a compilation error.
Signed-off-by: Henri Verbeet hverbeet@codeweavers.com
-
Henri Verbeet -
Henri Verbeet -
Matteo Bruni -
Matteo Bruni -
Zebediah Figura