Mesa (main): nir/fold_16bit_sampler_conversions: add a mask for supported sampler dims

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 27a43b531bff9493be8c873533c73e5f999dd1e7
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=27a43b531bff9493be8c873533c73e5f999dd1e7

Author: Mike Blumenkrantz <michael.blumenkrantz at gmail.com>
Date:   Mon Apr 11 08:52:24 2022 -0400

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 at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/15852>

---

 src/amd/vulkan/radv_pipeline.c               | 5 ++++-
 src/compiler/nir/nir.h                       | 2 +-
 src/compiler/nir/nir_lower_mediump.c         | 6 +++---
 src/gallium/drivers/radeonsi/si_shader_nir.c | 6 +++++-
 4 files changed, 13 insertions(+), 6 deletions(-)

diff --git a/src/amd/vulkan/radv_pipeline.c b/src/amd/vulkan/radv_pipeline.c
index 81316aa8a84..87b254b8e6b 100644
--- a/src/amd/vulkan/radv_pipeline.c
+++ b/src/amd/vulkan/radv_pipeline.c
@@ -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);
           }
diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index d4f381f4300..50fc412d7fd 100644
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -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 */
diff --git a/src/compiler/nir/nir_lower_mediump.c b/src/compiler/nir/nir_lower_mediump.c
index edf6bd155ef..1be4637114d 100644
--- a/src/compiler/nir/nir_lower_mediump.c
+++ b/src/compiler/nir/nir_lower_mediump.c
@@ -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. */
diff --git a/src/gallium/drivers/radeonsi/si_shader_nir.c b/src/gallium/drivers/radeonsi/si_shader_nir.c
index 130d8682069..50094e1ea3a 100644
--- a/src/gallium/drivers/radeonsi/si_shader_nir.c
+++ b/src/gallium/drivers/radeonsi/si_shader_nir.c
@@ -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) {