wine/vkd3d
0
0
Fork 0
mirror of https://gitlab.winehq.org/wine/vkd3d.git synced 2024-11-21 16:46:41 -08:00
Code Issues Packages Projects Releases Wiki Activity

vkd3d-shader/hlsl: Turn register(cX) reservations into buffer offset for SM4.

Browse Source
This commit is contained in:
Francisco Casas 2023-11-08 18:45:26 -03:00 committed by Alexandre Julliard
parent c7f7636534
commit 13f713f74b
Notes: Alexandre Julliard 2024-01-29 22:52:16 +01:00 
Approved-by: Giovanni Mascellani (@giomasce)
Approved-by: Zebediah Figura (@zfigura)
Approved-by: Henri Verbeet (@hverbeet)
Approved-by: Alexandre Julliard (@julliard)
Merge-Request: https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/458
2 changed files with 69 additions and 41 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

100
libs/vkd3d-shader/hlsl_codegen.c
View File

@ -4302,45 +4302,52 @@ static const struct hlsl_buffer *get_reserved_buffer(struct hlsl_ctx *ctx, uint3
return NULL;
}
static void calculate_buffer_offset(struct hlsl_ctx *ctx, struct hlsl_ir_var *var)
static void calculate_buffer_offset(struct hlsl_ctx *ctx, struct hlsl_ir_var *var, bool register_reservation)
{
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;
if (var->reg_reservation.offset_type == 'c')
if (register_reservation)
{
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;
var->buffer_offset = 4 * var->reg_reservation.reg_index;
}
else
{
var->buffer_offset = 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);
}
}
TRACE("Allocated buffer offset %u to %s.\n", var->buffer_offset, var->name);
@ -4409,6 +4416,11 @@ static void validate_buffer_offsets(struct hlsl_ctx *ctx)
}
}
static bool var_has_buffer_offset_register_reservation(struct hlsl_ctx *ctx, const struct hlsl_ir_var *var)
{
return var->reg_reservation.reg_type == 'c' && var->buffer == ctx->globals_buffer;
}
static void allocate_buffers(struct hlsl_ctx *ctx)
{
struct hlsl_buffer *buffer;
@ -4417,13 +4429,29 @@ static void allocate_buffers(struct hlsl_ctx *ctx)
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
if (var->is_uniform && !hlsl_type_is_resource(var->data_type))
{
if (var->is_param)
var->buffer = ctx->params_buffer;
if (!var->is_uniform || hlsl_type_is_resource(var->data_type))
continue;
calculate_buffer_offset(ctx, var);
}
if (var->is_param)
var->buffer = ctx->params_buffer;
}
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
if (!var->is_uniform || hlsl_type_is_resource(var->data_type))
continue;
if (var_has_buffer_offset_register_reservation(ctx, var))
calculate_buffer_offset(ctx, var, true);
}
LIST_FOR_EACH_ENTRY(var, &ctx->extern_vars, struct hlsl_ir_var, extern_entry)
{
if (!var->is_uniform || hlsl_type_is_resource(var->data_type))
continue;
if (!var_has_buffer_offset_register_reservation(ctx, var))
calculate_buffer_offset(ctx, var, false);
}
validate_buffer_offsets(ctx);

10
tests/hlsl/register-reservations-numeric.shader_test
View File

@ -1,4 +1,4 @@
[pixel shader fail(sm<6) todo(sm>=4)]
[pixel shader fail(sm<6)]
// Overlapping register(cX) reservations are not allowed except on SM6, where they are aliased.
// On SM1 this gives hr 0x88760b59.
float a : register(c0);
@ -24,7 +24,7 @@ float4 main() : sv_target
uniform 0 float4 0.1 0.2 0.3 0.4
uniform 4 float4 1.1 1.2 1.3 1.4
draw quad
todo(sm>=4 & sm<6) probe all rgba (1.1, 1.4, 0.2, 0.3)
probe all rgba (1.1, 1.4, 0.2, 0.3)
[pixel shader]
@ -43,7 +43,7 @@ uniform 8 float4 2.1 2.2 2.3 2.4
uniform 12 float4 3.1 3.2 3.3 3.4
uniform 16 float4 4.1 4.2 4.3 4.4
draw quad
todo(sm>=4 & sm<6) probe all rgba (4.1, 4.2, 1.3, 1.4)
probe all rgba (4.1, 4.2, 1.3, 1.4)
[require]
@ -85,7 +85,7 @@ uniform 4 float4 1.1 1.2 1.3 1.4
uniform 8 float4 2.1 2.2 2.3 2.4
uniform 12 float4 3.1 3.2 3.3 3.4
draw quad
todo(sm<6) probe all rgba (2.1, 2.2, 0.0, 0.0)
probe all rgba (2.1, 2.2, 0.0, 0.0)
[require]
@ -156,7 +156,7 @@ uniform 4 float4 1.1 1.2 1.3 1.4
uniform 8 float4 2.1 2.2 2.3 2.4
uniform 12 float4 3.1 3.2 3.3 3.4
draw quad
todo(sm<6) probe all rgba (2.1, 2.2, 2.3, 0.0)
probe all rgba (2.1, 2.2, 2.3, 0.0)
[pixel shader]