ac/nir: lower sample_pos in ac_nir_lower_ps_early

i.e. before future linking optimizations and shader_info gathering

The ac_nir_lower_ps_early call is also added for non-monolithic shaders
because it's required now.

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33024>

src/amd/common/nir/ac_nir_lower_ps_early.c

@@ -331,6 +331,14 @@ lower_ps_intrinsic(nir_builder *b, nir_instr *instr, void *state)
          return true;
       }
       break;
+   case nir_intrinsic_load_sample_pos:
+      /* Even though we could lower this by unpacking the sample position from user SGPRs, doing
+       * that hasn't shown any performance improvement in piglit/pixel-rate. Note that this is
+       * only used by sample shading.
+       */
+      /* sample_pos = ffract(frag_coord.xy); */
+      nir_def_replace(&intrin->def, nir_ffract(b, nir_channels(b, nir_load_frag_coord(b), 0x3)));
+      return true;
    default:
       break;
    }

src/gallium/drivers/radeonsi/si_shader.c

@@ -2564,6 +2564,13 @@ static struct nir_shader *si_get_nir_shader(struct si_shader *shader, struct si_
          NIR_PASS_V(nir, si_nir_emit_polygon_stipple, args);
 
       progress = true;
+   } else if (nir->info.stage == MESA_SHADER_FRAGMENT) {
+      ac_nir_lower_ps_early_options early_options = {
+         .alpha_func = COMPARE_FUNC_ALWAYS,
+         .spi_shader_col_format_hint = ~0,
+      };
+      NIR_PASS_V(nir, ac_nir_lower_ps_early, &early_options);
+      progress = true;
    }
 
    assert(shader->wave_size == 32 || shader->wave_size == 64);