nir: add nir_lower_single_sampled::lower_sample_mask_in option

GLSL defines gl_SampleMaskIn as :
   "a fragment language that indicates the set of samples covered
    by the primitive generating the fragment during multisample
    rasterization"

when variable rate shading is enabled, a single invocation might cover
multiple samples. The lowering done in nir_lower_single_sampled() does
not account for that case, so add an option to selectively disable it.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Alyssa Rosenzweig <alyssa.rosenzweig@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38641>

src/asahi/lib/agx_nir_lower_msaa.c

@@ -126,9 +126,12 @@ agx_nir_lower_monolithic_msaa(nir_shader *shader, uint8_t nr_samples)
    /* In single sampled programs, interpolateAtSample needs to return the
     * center pixel.
     */
-   if (nr_samples == 1)
-      nir_lower_single_sampled(shader);
-   else if (shader->info.fs.uses_sample_shading) {
+   if (nr_samples == 1) {
+      nir_lower_single_sampled_options lss_opts = {
+         .lower_sample_mask_in = true,
+      };
+      nir_lower_single_sampled(shader, &lss_opts);
+   } else if (shader->info.fs.uses_sample_shading) {
       agx_nir_lower_to_per_sample(shader);
       agx_nir_wrap_per_sample_loop(shader, nr_samples);
    }

src/compiler/nir/nir.h

@@ -5631,7 +5631,12 @@ bool nir_lower_uniforms_to_ubo(nir_shader *shader, bool dword_packed, bool load_
 
 bool nir_lower_is_helper_invocation(nir_shader *shader);
 
-bool nir_lower_single_sampled(nir_shader *shader);
+typedef struct nir_lower_single_sampled_options {
+   bool lower_sample_mask_in;
+} nir_lower_single_sampled_options;
+
+bool nir_lower_single_sampled(nir_shader *shader,
+                              const nir_lower_single_sampled_options *options);
 bool nir_lower_sample_shading(nir_shader *shader);
 
 bool nir_lower_atomics(nir_shader *shader, nir_instr_filter_cb filter);

src/compiler/nir/nir_lower_single_sampled.c

@@ -27,8 +27,10 @@
 static bool
 lower_single_sampled_instr(nir_builder *b,
                            nir_intrinsic_instr *intrin,
-                           UNUSED void *cb_data)
+                           void *cb_data)
 {
+   const nir_lower_single_sampled_options *options = cb_data;
+
    nir_def *lowered;
    switch (intrin->intrinsic) {
    case nir_intrinsic_load_sample_id:
@@ -45,7 +47,7 @@ lower_single_sampled_instr(nir_builder *b,
       /* Don't lower to helper invocations if helper invocations are going
        * to be lowered right back to sample mask.
        */
-      if (b->shader->options->lower_helper_invocation)
+      if (!options->lower_sample_mask_in || b->shader->options->lower_helper_invocation)
          return false;
 
       b->cursor = nir_before_instr(&intrin->instr);
@@ -90,7 +92,7 @@ lower_single_sampled_instr(nir_builder *b,
  * barycentrics to pixel, and constant-folds various built-ins.
  */
 bool
-nir_lower_single_sampled(nir_shader *shader)
+nir_lower_single_sampled(nir_shader *shader, const nir_lower_single_sampled_options *options)
 {
    assert(shader->info.stage == MESA_SHADER_FRAGMENT);
 
@@ -118,6 +120,6 @@ nir_lower_single_sampled(nir_shader *shader)
 
    return nir_shader_intrinsics_pass(shader, lower_single_sampled_instr,
                                      nir_metadata_control_flow,
-                                     NULL) ||
+                                     (void *)options) ||
           progress;
 }

src/intel/compiler/brw/brw_nir.c

@@ -1064,7 +1064,10 @@ brw_nir_lower_fs_inputs(nir_shader *nir,
       brw_nir_lower_fs_barycentrics(nir);
 
    if (key->multisample_fbo == INTEL_NEVER) {
-      NIR_PASS(_, nir, nir_lower_single_sampled);
+      nir_lower_single_sampled_options lss_opts = {
+         .lower_sample_mask_in = key->coarse_pixel == INTEL_NEVER,
+      };
+      NIR_PASS(_, nir, nir_lower_single_sampled, &lss_opts);
    } else if (key->persample_interp == INTEL_ALWAYS) {
       NIR_PASS(_, nir, nir_shader_intrinsics_pass,
                lower_barycentric_per_sample,

src/intel/compiler/elk/elk_nir.c

@@ -594,7 +594,10 @@ elk_nir_lower_fs_inputs(nir_shader *nir,
                 nir_lower_io_use_interpolated_input_intrinsics);
 
    if (key->multisample_fbo == ELK_NEVER) {
-      nir_lower_single_sampled(nir);
+      nir_lower_single_sampled_options lss_opts = {
+         .lower_sample_mask_in = true,
+      };
+      nir_lower_single_sampled(nir, &lss_opts);
    } else if (key->persample_interp == ELK_ALWAYS) {
       nir_shader_intrinsics_pass(nir, lower_barycentric_per_sample,
                                    nir_metadata_control_flow,