diff --git a/libs/vkd3d-shader/hlsl.h b/libs/vkd3d-shader/hlsl.h index c8e267d295b2c95253085649cf0e3ad4fbafb9de..38a7715e99cedf7a09f11038ffa17e475cb846aa 100644 --- a/libs/vkd3d-shader/hlsl.h +++ b/libs/vkd3d-shader/hlsl.h @@ -342,12 +342,17 @@ struct hlsl_attribute #define HLSL_ARRAY_ELEMENTS_COUNT_IMPLICIT 0 -/* Reservation of a specific register to a variable, field, or buffer, written in the HLSL source - * using the register(·) syntax */ +/* Reservation of a register and/or an offset for objects inside constant buffers, to be used as a + * starting point of their allocation. They are available through the register(·) and the + * packoffset(·) syntaxes, respectivelly. + * The costant buffer offset is measured register components. */ struct hlsl_reg_reservation { - char type; - unsigned int index; + char reg_type; + unsigned int reg_index; + + char offset_type; + unsigned int offset_index; }; struct hlsl_ir_var @@ -360,8 +365,7 @@ struct hlsl_ir_var struct hlsl_buffer *buffer; /* Bitfield for storage modifiers (type modifiers are stored in data_type->modifiers). */ unsigned int storage_modifiers; - /* Optional register to be used as a starting point for the variable allocation, specified - * by the user via the register(·) syntax. */ + /* Optional reservations of registers and/or offsets for variables within constant buffers. */ struct hlsl_reg_reservation reg_reservation; /* Item entry in hlsl_scope.vars. Specifically hlsl_ctx.globals.vars if the variable is global. */ @@ -674,6 +678,9 @@ struct hlsl_buffer unsigned size, used_size; /* Register of type 'b' on which the buffer is allocated. */ struct hlsl_reg reg; + + bool manually_packed_elements; + bool automatically_packed_elements; }; struct hlsl_ctx diff --git a/libs/vkd3d-shader/hlsl.l b/libs/vkd3d-shader/hlsl.l index adff1da04d828f2e00e87282c8651096eda00737..10751bbe9ba0360bae4fe0dc08d29885eba8dd14 100644 --- a/libs/vkd3d-shader/hlsl.l +++ b/libs/vkd3d-shader/hlsl.l @@ -95,6 +95,7 @@ matrix {return KW_MATRIX; } namespace {return KW_NAMESPACE; } nointerpolation {return KW_NOINTERPOLATION; } out {return KW_OUT; } +packoffset {return KW_PACKOFFSET; } pass {return KW_PASS; } PixelShader {return KW_PIXELSHADER; } precise {return KW_PRECISE; } diff --git a/libs/vkd3d-shader/hlsl.y b/libs/vkd3d-shader/hlsl.y index afcc4575370a86e1536079619888e6acc6681561..79ec970a0addcef03a4ef230892e2ee3fd3ec396 100644 --- a/libs/vkd3d-shader/hlsl.y +++ b/libs/vkd3d-shader/hlsl.y @@ -983,6 +983,9 @@ static bool gen_struct_fields(struct hlsl_ctx *ctx, struct parse_fields *fields, hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_SYNTAX, "Illegal initializer on a struct field."); free_parse_initializer(&v->initializer); } + if (v->reg_reservation.offset_type) + hlsl_error(ctx, &v->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() is not allowed inside struct definitions."); vkd3d_free(v); } vkd3d_free(defs); @@ -1063,7 +1066,12 @@ static bool add_func_parameter(struct hlsl_ctx *ctx, struct hlsl_func_parameters hlsl_error(ctx, &loc, VKD3D_SHADER_ERROR_HLSL_INVALID_MODIFIER, "Parameter '%s' is declared as both \"out\" and \"uniform\".", param->name); - if (!(var = hlsl_new_var(ctx, param->name, param->type, loc, ¶m->semantic, param->modifiers, ¶m->reg_reservation))) + if (param->reg_reservation.offset_type) + hlsl_error(ctx, &loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() is not allowed on function parameters."); + + if (!(var = hlsl_new_var(ctx, param->name, param->type, loc, ¶m->semantic, param->modifiers, + ¶m->reg_reservation))) return false; var->is_param = 1; @@ -1084,12 +1092,61 @@ static struct hlsl_reg_reservation parse_reg_reservation(const char *reg_string) { struct hlsl_reg_reservation reservation = {0}; - if (!sscanf(reg_string + 1, "%u", &reservation.index)) + if (!sscanf(reg_string + 1, "%u", &reservation.reg_index)) { FIXME("Unsupported register reservation syntax.\n"); return reservation; } - reservation.type = reg_string[0]; + reservation.reg_type = reg_string[0]; + return reservation; +} + +static struct hlsl_reg_reservation parse_packoffset(struct hlsl_ctx *ctx, const char *reg_string, + const char *swizzle, const struct vkd3d_shader_location *loc) +{ + struct hlsl_reg_reservation reservation = {0}; + char *endptr; + + if (ctx->profile->major_version < 4) + return reservation; + + reservation.offset_index = strtoul(reg_string + 1, &endptr, 10); + if (*endptr) + { + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Invalid packoffset() syntax."); + return reservation; + } + + reservation.offset_type = reg_string[0]; + if (reservation.offset_type != 'c') + { + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Only 'c' registers are allowed in packoffset()."); + return reservation; + } + + reservation.offset_index *= 4; + + if (swizzle) + { + if (strlen(swizzle) != 1) + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Invalid packoffset() component \"%s\".", swizzle); + + if (swizzle[0] == 'x' || swizzle[0] == 'r') + reservation.offset_index += 0; + else if (swizzle[0] == 'y' || swizzle[0] == 'g') + reservation.offset_index += 1; + else if (swizzle[0] == 'z' || swizzle[0] == 'b') + reservation.offset_index += 2; + else if (swizzle[0] == 'w' || swizzle[0] == 'a') + reservation.offset_index += 3; + else + hlsl_error(ctx, loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Invalid packoffset() component \"%s\".", swizzle); + } + return reservation; } @@ -2043,6 +2100,13 @@ static struct list *declare_vars(struct hlsl_ctx *ctx, struct hlsl_type *basic_t var->buffer = ctx->cur_buffer; + if (var->buffer == ctx->globals_buffer) + { + if (var->reg_reservation.offset_type) + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() is only allowed inside constant buffer declarations."); + } + if (ctx->cur_scope == ctx->globals) { local = false; @@ -3846,6 +3910,7 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type %token KW_NAMESPACE %token KW_NOINTERPOLATION %token KW_OUT +%token KW_PACKOFFSET %token KW_PASS %token KW_PIXELSHADER %token KW_PRECISE @@ -3999,6 +4064,7 @@ static void validate_texture_format_type(struct hlsl_ctx *ctx, struct hlsl_type %type <parameters> parameters %type <reg_reservation> register_opt +%type <reg_reservation> packoffset_opt %type <sampler_dim> texture_type texture_ms_type uav_type @@ -4349,8 +4415,11 @@ func_prototype_no_attrs: "Semantics are not allowed on void functions."); } - if ($7.reg_reservation.type) + if ($7.reg_reservation.reg_type) FIXME("Unexpected register reservation for a function.\n"); + if ($7.reg_reservation.offset_type) + hlsl_error(ctx, &@5, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() is not allowed on functions."); if (($$.decl = get_func_decl(&ctx->functions, $3, &$5))) { @@ -4477,18 +4546,25 @@ colon_attribute: %empty { $$.semantic.name = NULL; - $$.reg_reservation.type = 0; + $$.reg_reservation.reg_type = 0; + $$.reg_reservation.offset_type = 0; } | semantic { $$.semantic = $1; - $$.reg_reservation.type = 0; + $$.reg_reservation.reg_type = 0; + $$.reg_reservation.offset_type = 0; } | register_opt { $$.semantic.name = NULL; $$.reg_reservation = $1; } + | packoffset_opt + { + $$.semantic.name = NULL; + $$.reg_reservation = $1; + } semantic: ':' any_identifier @@ -4518,6 +4594,21 @@ register_opt: vkd3d_free($6); } +packoffset_opt: + ':' KW_PACKOFFSET '(' any_identifier ')' + { + $$ = parse_packoffset(ctx, $4, NULL, &@$); + + vkd3d_free($4); + } + | ':' KW_PACKOFFSET '(' any_identifier '.' any_identifier ')' + { + $$ = parse_packoffset(ctx, $4, $6, &@$); + + vkd3d_free($4); + vkd3d_free($6); + } + parameters: scope_start { diff --git a/libs/vkd3d-shader/hlsl_codegen.c b/libs/vkd3d-shader/hlsl_codegen.c index 6c529f1913efa7cfbe8e8d19b2637d6548e1af81..5d7e426f152814ba9427c384bb4f3840c9a0641f 100644 --- a/libs/vkd3d-shader/hlsl_codegen.c +++ b/libs/vkd3d-shader/hlsl_codegen.c @@ -2307,9 +2307,9 @@ static void allocate_register_reservations(struct hlsl_ctx *ctx) continue; regset = hlsl_type_get_regset(var->data_type); - if (var->reg_reservation.type) + if (var->reg_reservation.reg_type) { - if (var->reg_reservation.type != get_regset_name(regset)) + if (var->reg_reservation.reg_type != get_regset_name(regset)) { struct vkd3d_string_buffer *type_string; @@ -2322,8 +2322,9 @@ static void allocate_register_reservations(struct hlsl_ctx *ctx) else { var->regs[regset].allocated = true; - var->regs[regset].id = var->reg_reservation.index; - TRACE("Allocated reserved %s to %c%u.\n", var->name, var->reg_reservation.type, var->reg_reservation.index); + var->regs[regset].id = var->reg_reservation.reg_index; + TRACE("Allocated reserved %s to %c%u.\n", var->name, var->reg_reservation.reg_type, + var->reg_reservation.reg_index); } } } @@ -2898,23 +2899,117 @@ static const struct hlsl_buffer *get_reserved_buffer(struct hlsl_ctx *ctx, uint3 LIST_FOR_EACH_ENTRY(buffer, &ctx->buffers, const struct hlsl_buffer, entry) { - if (buffer->used_size && buffer->reservation.type == 'b' && buffer->reservation.index == index) + if (buffer->used_size && buffer->reservation.reg_type == 'b' && buffer->reservation.reg_index == index) return buffer; } return NULL; } -static void calculate_buffer_offset(struct hlsl_ir_var *var) +static void calculate_buffer_offset(struct hlsl_ctx *ctx, struct hlsl_ir_var *var) { + unsigned int var_reg_size = var->data_type->reg_size[HLSL_REGSET_NUMERIC]; + enum hlsl_type_class var_class = var->data_type->class; struct hlsl_buffer *buffer = var->buffer; - buffer->size = hlsl_type_get_sm4_offset(var->data_type, buffer->size); + if (var->reg_reservation.offset_type == 'c') + { + if (var->reg_reservation.offset_index % 4) + { + if (var_class == HLSL_CLASS_MATRIX) + { + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() reservations with matrix types must be aligned with the beginning of a register."); + } + else if (var_class == HLSL_CLASS_ARRAY) + { + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() reservations with array types must be aligned with the beginning of a register."); + } + else if (var_class == HLSL_CLASS_STRUCT) + { + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() reservations with struct types must be aligned with the beginning of a register."); + } + else if (var_class == HLSL_CLASS_VECTOR) + { + unsigned int aligned_offset = hlsl_type_get_sm4_offset(var->data_type, var->reg_reservation.offset_index); + + if (var->reg_reservation.offset_index != aligned_offset) + hlsl_error(ctx, &var->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() reservations with vector types cannot span multiple registers."); + } + } + var->buffer_offset = var->reg_reservation.offset_index; + } + else + { + var->buffer_offset = hlsl_type_get_sm4_offset(var->data_type, buffer->size); + } - var->buffer_offset = buffer->size; TRACE("Allocated buffer offset %u to %s.\n", var->buffer_offset, var->name); - buffer->size += var->data_type->reg_size[HLSL_REGSET_NUMERIC]; + buffer->size = max(buffer->size, var->buffer_offset + var_reg_size); if (var->last_read) - buffer->used_size = buffer->size; + buffer->used_size = max(buffer->used_size, var->buffer_offset + var_reg_size); +} + +static void validate_buffer_offsets(struct hlsl_ctx *ctx) +{ + struct hlsl_ir_var *var1, *var2; + struct hlsl_buffer *buffer; + + LIST_FOR_EACH_ENTRY(var1, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + if (!var1->is_uniform || var1->data_type->class == HLSL_CLASS_OBJECT) + continue; + + buffer = var1->buffer; + if (!buffer->used_size) + continue; + + LIST_FOR_EACH_ENTRY(var2, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + unsigned int var1_reg_size, var2_reg_size; + + if (!var2->is_uniform || var2->data_type->class == HLSL_CLASS_OBJECT) + continue; + + if (var1 == var2 || var1->buffer != var2->buffer) + continue; + + /* This is to avoid reporting the error twice for the same pair of overlapping variables. */ + if (strcmp(var1->name, var2->name) >= 0) + continue; + + var1_reg_size = var1->data_type->reg_size[HLSL_REGSET_NUMERIC]; + var2_reg_size = var2->data_type->reg_size[HLSL_REGSET_NUMERIC]; + + if (var1->buffer_offset < var2->buffer_offset + var2_reg_size + && var2->buffer_offset < var1->buffer_offset + var1_reg_size) + hlsl_error(ctx, &buffer->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "Invalid packoffset() reservation: Variables %s and %s overlap.", + var1->name, var2->name); + } + } + + LIST_FOR_EACH_ENTRY(var1, &ctx->extern_vars, struct hlsl_ir_var, extern_entry) + { + buffer = var1->buffer; + if (!buffer || buffer == ctx->globals_buffer) + continue; + + if (var1->reg_reservation.offset_type + || (var1->data_type->class == HLSL_CLASS_OBJECT && var1->reg_reservation.reg_type)) + buffer->manually_packed_elements = true; + else + buffer->automatically_packed_elements = true; + + if (buffer->manually_packed_elements && buffer->automatically_packed_elements) + { + hlsl_error(ctx, &buffer->loc, VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION, + "packoffset() must be specified for all the buffer elements, or none of them."); + break; + } + } } static void allocate_buffers(struct hlsl_ctx *ctx) @@ -2930,10 +3025,12 @@ static void allocate_buffers(struct hlsl_ctx *ctx) if (var->is_param) var->buffer = ctx->params_buffer; - calculate_buffer_offset(var); + calculate_buffer_offset(ctx, var); } } + validate_buffer_offsets(ctx); + LIST_FOR_EACH_ENTRY(buffer, &ctx->buffers, struct hlsl_buffer, entry) { if (!buffer->used_size) @@ -2941,23 +3038,23 @@ static void allocate_buffers(struct hlsl_ctx *ctx) if (buffer->type == HLSL_BUFFER_CONSTANT) { - if (buffer->reservation.type == 'b') + if (buffer->reservation.reg_type == 'b') { - const struct hlsl_buffer *reserved_buffer = get_reserved_buffer(ctx, buffer->reservation.index); + const struct hlsl_buffer *reserved_buffer = get_reserved_buffer(ctx, buffer->reservation.reg_index); if (reserved_buffer && reserved_buffer != buffer) { hlsl_error(ctx, &buffer->loc, VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS, - "Multiple buffers bound to cb%u.", buffer->reservation.index); + "Multiple buffers bound to cb%u.", buffer->reservation.reg_index); hlsl_note(ctx, &reserved_buffer->loc, VKD3D_SHADER_LOG_ERROR, - "Buffer %s is already bound to cb%u.", reserved_buffer->name, buffer->reservation.index); + "Buffer %s is already bound to cb%u.", reserved_buffer->name, buffer->reservation.reg_index); } - buffer->reg.id = buffer->reservation.index; + buffer->reg.id = buffer->reservation.reg_index; buffer->reg.allocated = true; TRACE("Allocated reserved %s to cb%u.\n", buffer->name, index); } - else if (!buffer->reservation.type) + else if (!buffer->reservation.reg_type) { while (get_reserved_buffer(ctx, index)) ++index; @@ -3043,7 +3140,7 @@ static void allocate_objects(struct hlsl_ctx *ctx, enum hlsl_regset regset) regset_name, index); } - var->regs[regset].id = var->reg_reservation.index; + var->regs[regset].id = var->reg_reservation.reg_index; var->regs[regset].allocated = true; TRACE("Allocated reserved %s to %c%u.\n", var->name, regset_name, var->regs[regset].id); } diff --git a/libs/vkd3d-shader/hlsl_sm4.c b/libs/vkd3d-shader/hlsl_sm4.c index 3fc44d9c566ae84d7bfc945541b10b42b28f0902..6dd76d67faff892d5c394940690b0bc92b3548ed 100644 --- a/libs/vkd3d-shader/hlsl_sm4.c +++ b/libs/vkd3d-shader/hlsl_sm4.c @@ -594,7 +594,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc) var = extern_resources[i]; regset = hlsl_type_get_regset(var->data_type); - if (var->reg_reservation.type) + if (var->reg_reservation.reg_type) flags |= D3D_SIF_USERPACKED; put_u32(&buffer, 0); /* name */ @@ -624,7 +624,7 @@ static void write_sm4_rdef(struct hlsl_ctx *ctx, struct dxbc_writer *dxbc) if (!cbuffer->reg.allocated) continue; - if (cbuffer->reservation.type) + if (cbuffer->reservation.reg_type) flags |= D3D_SIF_USERPACKED; put_u32(&buffer, 0); /* name */ diff --git a/tests/cbuffer.shader_test b/tests/cbuffer.shader_test index 07ef18dba0b85ab498259c8da667dcba2127cfa4..9cf42d9d3de6fdb845bc0fe4ad07eae244357957 100644 --- a/tests/cbuffer.shader_test +++ b/tests/cbuffer.shader_test @@ -15,3 +15,691 @@ float4 main() : sv_target uniform 0 float4 1.0 2.0 3.0 4.0 draw quad probe all rgba (1.0, 2.0, 3.0, 4.0) + + +% SM1 buffer offset allocation follows different rules than SM4. +% Those would have to be tested separately. +[require] +shader model >= 4.0 + + +[pixel shader fail] +cbuffer buffer +{ + float4 a : packoffset(c1invalid_extra_chars); +} + +float4 main() : sv_target +{ + return 0; +} + + +% Respect register boundaries +[pixel shader] +cbuffer buffer +{ + float2 a; + float2 b; + float2 c; + float3 d; +} + +float4 main() : sv_target +{ + return float4(a.x, b.x, c.x, d.x); +} + +[test] +uniform 0 float4 0.0 1.0 2.0 3.0 +uniform 4 float4 4.0 5.0 6.0 7.0 +uniform 8 float4 8.0 9.0 10.0 11.0 +uniform 12 float4 12.0 13.0 14.0 15.0 +draw quad +probe all rgba (0.0, 2.0, 4.0, 8.0) + + +[pixel shader] +cbuffer buffer +{ + float a; + float b[2]; + float c; +} + +float4 main() : sv_target +{ + return float4(a, b[0], b[1], c); +} + +[test] +uniform 0 float4 0.0 1.0 2.0 3.0 +uniform 4 float4 4.0 5.0 6.0 7.0 +uniform 8 float4 8.0 9.0 10.0 11.0 +draw quad +probe all rgba (0.0, 4.0, 8.0, 9.0) + + +[pixel shader] +cbuffer buffer +{ + float a; + struct + { + float b; + float c; + } p; + float d; +} + +float4 main() : sv_target +{ + return float4(a, p.b, p.c, d); +} + +[test] +uniform 0 float4 0.0 1.0 2.0 3.0 +uniform 4 float4 4.0 5.0 6.0 7.0 +uniform 8 float4 8.0 9.0 10.0 11.0 +uniform 12 float4 12.0 13.0 14.0 15.0 +draw quad +probe all rgba (0.0, 4.0, 5.0, 6.0) + + +[pixel shader fail] +// Elements cannot overlap if buffer is used. +cbuffer buffer +{ + float c : packoffset(c0); + float4 a[2] : packoffset(c1); + float4 b[2] : packoffset(c2); +} + +float4 main() : sv_target +{ + return c; +} + + +[pixel shader] +// Elements can overlap if buffer is not used. +cbuffer buffer +{ + float4 a[2] : packoffset(c1); + float4 b[2] : packoffset(c2); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader] +cbuffer buffer +{ + float4 a : packoffset(c1); + float4 b : packoffset(c2); +} + +float4 main() : sv_target +{ + return 100 * a + b; +} + +[test] +uniform 0 float4 1.0 2.0 3.0 4.0 +uniform 4 float4 5.0 6.0 7.0 8.0 +uniform 8 float4 9.0 10.0 11.0 12.0 +draw quad +probe all rgba (509, 610, 711, 812) + + +[pixel shader] +struct apple +{ + float2 a; + float b; + float4 c; +}; + +cbuffer buffer +{ + float4 foo : packoffset(c3); + struct apple bar : packoffset(c1); +} + +float4 main() : sv_target +{ + return 1000 * foo + 100 * float4(bar.a, 0, 0) + 10 * float4(bar.b, 0, 0, 0) + bar.c; +} + +[test] +uniform 0 float4 0.0 1.0 2.0 3.0 +uniform 4 float4 4.0 5.0 6.0 7.0 +uniform 8 float4 8.0 9.0 10.0 11.0 +uniform 12 float4 12.0 13.0 14.0 15.0 +draw quad +probe all rgba (12468.0, 13509.0, 14010.0, 15011.0) + + +[pixel shader] +cbuffer buffer +{ + float2 c : packoffset(c0.y); +} + +float4 main() : sv_target +{ + return float4(c, c); +} + +[test] +uniform 0 float4 1.0 2.0 3.0 4.0 +draw quad +probe all rgba (2.0, 3.0, 2.0, 3.0) + + +[pixel shader fail] +// Elements must respect register boundaries. +cbuffer buffer +{ + float4 c : packoffset(c0.z); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +// Matrices must be aligned. +cbuffer buffer +{ + float1x1 m : packoffset(c0.y); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +// Arrays must be aligned. +cbuffer buffer +{ + float a[1] : packoffset(c0.y); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +// Structs must be aligned. +struct apple +{ + float p; +}; + +cbuffer buffer +{ + struct apple a : packoffset(c0.y); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +// Invalid offset on unused buffer. +cbuffer buffer +{ + float3 a : packoffset(c0.z); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +// Invalid offset on unused variable. +cbuffer buffer +{ + float a : packoffset(c0); + float3 b : packoffset(c0.z); +} + +float4 main() : sv_target +{ + return a; +} + + +[pixel shader] +cbuffer buffer +{ + float4 a : packoffset(c1); + float1 b : packoffset(c0); + float1 c : packoffset(c0.y); +} + +float4 main() : sv_target +{ + return 100 * a + 10 * b + c; +} + +[test] +uniform 0 float 1.0 +uniform 1 float 2.0 +uniform 4 float4 5.0 6.0 7.0 8.0 +draw quad +probe all rgba (512.0, 612.0, 712.0, 812.0) + + +[pixel shader fail] +// packoffset cannot be used unless all elements use it. +cbuffer buffer +{ + float4 a : packoffset(c0); + float4 b; +} + +float4 main() : sv_target +{ + return a + b; +} + + +[pixel shader] +cbuffer buffer +{ + float2 c : packoffset(c0.b); +} + +float4 main() : sv_target +{ + return float4(c, c); +} + +[test] +uniform 0 float4 1.0 2.0 3.0 4.0 +draw quad +probe all rgba (3.0, 4.0, 3.0, 4.0) + + +[pixel shader fail] +cbuffer buffer +{ + float2 c : packoffset(c0.xy); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +cbuffer buffer +{ + float4x4 mat : packoffset(c0._m00); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +cbuffer buffer +{ + float4 a : packoffset(c0._m00); +} + +float4 main() : sv_target +{ + return 0; +} + + +[pixel shader fail] +cbuffer buffer +{ + float2 c : packoffset(c0.xz); +} + +float4 main() : sv_target +{ + return 0; +} + + +% packoffset cannot be used outside a constant buffer. +[pixel shader fail] +float4 a : packoffset(c0); + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +float4 foo(float4 a : packoffset(c0)) +{ + return a; +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +float4 foo(float4 a) : packoffset(c0) +{ + return a; +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +struct apple +{ + float4 a : packoffset(c0); +}; + +cbuffer buffer +{ + struct apple a; +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +cbuffer buffer +{ + struct + { + float4 a : packoffset(c0); + } s; +} + +float4 main() : sv_target +{ + return 0; +} + + +[texture 0] +size (1, 1) +0.0 0.0 0.0 4.0 + +[sampler 0] +filter linear linear linear +address clamp clamp clamp + + +[pixel shader] +// Resources are allowed inside constant buffers but they behave as regular resources. +cbuffer buffer +{ + float4 a; + Texture2D tex; + sampler sam; + float4 b; +} + +float4 main() : sv_target +{ + return a + b + tex.Sample(sam, float2(0, 0)); +} + +[test] +uniform 0 float4 1.0 0.0 0.0 0.0 +uniform 4 float4 0.0 2.0 0.0 0.0 +uniform 8 float4 0.0 0.0 3.0 0.0 +draw quad +probe all rgba (1.0, 2.0, 0.0, 4.0) + + +% packoffset() cannot be used to specify other types of registers +[pixel shader fail] +cbuffer buffer +{ + sampler sam : packoffset(s0); +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +cbuffer buffer +{ + Texture2D tex : packoffset(t0); +} + +float4 main() : sv_target +{ + return 0; +} + + +[texture 0] +size (1, 1) +1.0 1.0 1.0 1.0 + +[texture 1] +size (1, 1) +2.0 2.0 2.0 2.0 + + +[pixel shader] +// packoffset() can be used in Textures, doesn't change the allocated t register. +cbuffer buffer +{ + Texture2D tex : packoffset(c1); +} + +float4 main() : sv_target +{ + return tex.Load(int3(0, 0, 0)); +} + +[test] +draw quad +probe all rgba (1.0, 1.0, 1.0, 1.0) + + +% Samplers cannot have packoffset(), unless register() is also specified, or they are not used. +% Note: In SM1 the rules are different: packoffset() is allowed for samplers, but they cannot be +% used together with other numeric fields, which seems like a bug. +[pixel shader fail todo] +Texture2D tex; + +cbuffer buffer +{ + sampler sam : packoffset(c1); +} + +float4 main() : sv_target +{ + return tex.Sample(sam, float2(0, 0)); +} + +[pixel shader todo] +Texture2D tex; + +cbuffer buffer +{ + sampler sam : packoffset(c1) : register(s0); +} + +float4 main() : sv_target +{ + return tex.Sample(sam, float2(0, 0)); +} + +[pixel shader] +cbuffer buffer +{ + sampler sam : packoffset(c1); +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader] +cbuffer buffer +{ + sampler sam : packoffset(c1); + float4 a : packoffset(c0); +} + +float4 main() : sv_target +{ + return a; +} + + +% When packoffset is used in one field, resources are also expected to have a reservation. +[pixel shader fail] +cbuffer buffer +{ + float4 foo : packoffset(c0); + Texture2D tex; +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +cbuffer buffer +{ + float4 foo : packoffset(c0); + sampler sam; +} + +float4 main() : sv_target +{ + return 0; +} + +% register() can be used instead of packoffset(). +[pixel shader] +cbuffer buffer +{ + float4 foo : packoffset(c0); + Texture2D tex : register(t1); +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader] +cbuffer buffer +{ + float4 foo : packoffset(c0); + sampler sam : register(s1); +} + +float4 main() : sv_target +{ + return 0; +} + +% Using register() alone is considered manual packing for resources, so the other fields expect packoffset(). +[pixel shader fail] +cbuffer buffer +{ + float4 foo; + Texture2D tex : register(t1); +} + +float4 main() : sv_target +{ + return 0; +} + +[pixel shader fail] +cbuffer buffer +{ + float4 foo; + sampler sam : register(s1); +} + +float4 main() : sv_target +{ + return 0; +} + +% Using register() alone is not considered manual packing for non-resources. +[pixel shader] +cbuffer buffer +{ + float4 foo : register(c0); + Texture2D tex; +} + +float4 main() : sv_target +{ + return 0; +} + + +[require] +shader model >= 5.0 + +[texture 0] +size (1, 1) +0.0 0.0 0.0 0.5 + +[pixel shader todo] +struct apple +{ + float2 a; + Texture2D tex; + float b; +}; + +cbuffer buffer +{ + float4 foo : packoffset(c3); + struct apple bar : packoffset(c1); +} + +float4 main() : sv_target +{ + return 10 * foo + float4(bar.a, 0, 0) + float4(0, 0, bar.b, 0) + bar.tex.Load(int3(0, 0, 0)); +} + +[test] +uniform 0 float4 0.0 1.0 2.0 3.0 +uniform 4 float4 4.0 5.0 6.0 7.0 +uniform 8 float4 8.0 9.0 10.0 11.0 +uniform 12 float4 12.0 13.0 14.0 15.0 +todo draw quad +todo probe all rgba (124.0, 135.0, 146.0, 150.5)