nir/fold_16bit_sampler_conversions: add a mask for supported sampler dims
AMD might not support cubes, but that doesn't mean cubes can't be used on other drivers Reviewed-by: Samuel Pitoiset <samuel.pitoiset@gmail.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15852>
This commit is contained in:

committed by
Marge Bot

parent
324b2ae5f2
commit
27a43b531b
@@ -4485,10 +4485,13 @@ radv_create_shaders(struct radv_pipeline *pipeline, struct radv_pipeline_layout
|
||||
}
|
||||
if (((stages[i].nir->info.bit_sizes_int | stages[i].nir->info.bit_sizes_float) & 16) &&
|
||||
device->physical_device->rad_info.chip_class >= GFX9) {
|
||||
uint32_t sampler_dims = UINT32_MAX;
|
||||
/* Skip because AMD doesn't support 16-bit types with these. */
|
||||
sampler_dims &= ~BITFIELD_BIT(GLSL_SAMPLER_DIM_CUBE);
|
||||
// TODO: also optimize the tex srcs. see radeonSI for reference */
|
||||
/* Skip if there are potentially conflicting rounding modes */
|
||||
if (!nir_has_any_rounding_mode_enabled(stages[i].nir->info.float_controls_execution_mode))
|
||||
NIR_PASS_V(stages[i].nir, nir_fold_16bit_sampler_conversions, 0);
|
||||
NIR_PASS_V(stages[i].nir, nir_fold_16bit_sampler_conversions, 0, sampler_dims);
|
||||
|
||||
NIR_PASS_V(stages[i].nir, nir_opt_vectorize, opt_vectorize_callback, NULL);
|
||||
}
|
||||
|
@@ -5219,7 +5219,7 @@ bool nir_force_mediump_io(nir_shader *nir, nir_variable_mode modes,
|
||||
nir_alu_type types);
|
||||
bool nir_unpack_16bit_varying_slots(nir_shader *nir, nir_variable_mode modes);
|
||||
bool nir_fold_16bit_sampler_conversions(nir_shader *nir,
|
||||
unsigned tex_src_types);
|
||||
unsigned tex_src_types, uint32_t sampler_dims);
|
||||
|
||||
typedef struct {
|
||||
bool legalize_type; /* whether this src should be legalized */
|
||||
|
@@ -422,7 +422,8 @@ replace_with_mov(nir_builder *b, nir_instr *instr, nir_src *src,
|
||||
*/
|
||||
bool
|
||||
nir_fold_16bit_sampler_conversions(nir_shader *nir,
|
||||
unsigned tex_src_types)
|
||||
unsigned tex_src_types,
|
||||
uint32_t sampler_dims)
|
||||
{
|
||||
bool changed = false;
|
||||
nir_function_impl *impl = nir_shader_get_entrypoint(nir);
|
||||
@@ -444,10 +445,9 @@ nir_fold_16bit_sampler_conversions(nir_shader *nir,
|
||||
if (tex->is_sparse)
|
||||
continue;
|
||||
|
||||
/* Skip because AMD doesn't support 16-bit types with these. */
|
||||
if ((tex->op == nir_texop_txs ||
|
||||
tex->op == nir_texop_query_levels) ||
|
||||
tex->sampler_dim == GLSL_SAMPLER_DIM_CUBE)
|
||||
!(sampler_dims & BITFIELD_BIT(tex->sampler_dim)))
|
||||
continue;
|
||||
|
||||
/* Optimize source operands. */
|
||||
|
@@ -166,9 +166,13 @@ static void si_late_optimize_16bit_samplers(struct si_screen *sscreen, nir_shade
|
||||
};
|
||||
bool changed = false;
|
||||
|
||||
uint32_t sampler_dims = UINT32_MAX;
|
||||
/* Skip because AMD doesn't support 16-bit types with these. */
|
||||
sampler_dims &= ~BITFIELD_BIT(GLSL_SAMPLER_DIM_CUBE);
|
||||
NIR_PASS(changed, nir, nir_fold_16bit_sampler_conversions,
|
||||
(1 << nir_tex_src_coord) |
|
||||
(has_g16 ? 1 << nir_tex_src_ddx : 0));
|
||||
(has_g16 ? 1 << nir_tex_src_ddx : 0),
|
||||
sampler_dims);
|
||||
NIR_PASS(changed, nir, nir_legalize_16bit_sampler_srcs, tex_constraints);
|
||||
|
||||
if (changed) {
|
||||
|
Reference in New Issue
Block a user