pan: Add a pass to resize I/O load/stores as needed by the varying layout

Reviewed-by: Lorenzo Rossi <lorenzo.rossi@collabora.com>
Acked-by: Eric R. Smith <eric.smith@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38681>

src/panfrost/compiler/meson.build

@@ -18,6 +18,7 @@ libpanfrost_compiler_files = files(
   'pan_nir_lower_texel_buffer_index.c',
   'pan_nir_lower_vertex_id.c',
   'pan_nir_lower_xfb.c',
+  'pan_nir_resize_varying_io.c',
 )
 
 subdir('bifrost')

src/panfrost/compiler/pan_nir.h

@@ -84,4 +84,7 @@ uint32_t pan_nir_collect_noperspective_varyings_fs(nir_shader *s);
 
 void pan_nir_collect_varyings(nir_shader *s, struct pan_shader_info *info);
 
+bool pan_nir_resize_varying_io(nir_shader *nir,
+                               const struct pan_varying_layout *varying_layout);
+
 #endif /* __PAN_NIR_H__ */

src/panfrost/compiler/pan_nir_resize_varying_io.c

@@ -0,0 +1,133 @@
+/*
+ * Copyright (c) 2025 Collabora, Ltd.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include "pan_nir.h"
+#include "nir_builder.h"
+
+static bool
+resize_io_intr(nir_builder *b, nir_intrinsic_instr *intr, void *data)
+{
+   const struct pan_varying_layout *varying_layout = data;
+
+   bool is_load;
+   switch (intr->intrinsic) {
+   case nir_intrinsic_store_output:
+   case nir_intrinsic_store_per_view_output:
+      if (b->shader->info.stage != MESA_SHADER_VERTEX)
+         return false;
+      is_load = false;
+      break;
+
+   case nir_intrinsic_load_input:
+   case nir_intrinsic_load_interpolated_input:
+      if (b->shader->info.stage != MESA_SHADER_FRAGMENT)
+         return false;
+      is_load = true;
+      break;
+
+   default:
+      return false;
+   }
+
+   nir_io_semantics sem = nir_intrinsic_io_semantics(intr);
+   const struct pan_varying_slot *slot =
+      pan_varying_layout_find_slot(varying_layout, sem.location);
+   if (slot == NULL) {
+      if (is_load) {
+         b->cursor = nir_after_instr(&intr->instr);
+
+         nir_def *zero = nir_imm_zero(b, intr->def.num_components,
+                                      intr->def.bit_size);
+         nir_def_replace(&intr->def, zero);
+      } else {
+         assert(!"We should never have a mismatch on outputs");
+         nir_instr_remove(&intr->instr);
+      }
+      return true;
+   }
+
+   nir_alu_type data_type;
+   if (is_load) {
+      data_type = nir_intrinsic_dest_type(intr);
+      assert(intr->def.bit_size == nir_alu_type_get_type_size(data_type));
+   } else {
+      data_type = nir_intrinsic_src_type(intr);
+      assert(nir_src_bit_size(intr->src[0]) ==
+             nir_alu_type_get_type_size(data_type));
+   }
+
+   const unsigned slot_bit_size = nir_alu_type_get_type_size(slot->alu_type);
+
+   /* We trust the base type in the shader and only adjust the bit size */
+   const nir_alu_type slot_type =
+      nir_alu_type_get_base_type(data_type) | slot_bit_size;
+
+   if (slot_bit_size == nir_alu_type_get_type_size(data_type)) {
+      if (!sem.medium_precision)
+         return false;
+
+      /* There's nothing to actually lower but we still want to smash off
+       * mediump so the back-end doesn't screw anything up on us.
+       *
+       * TODO: This is a hack to work around the back-end.  It really
+       * shouldn't care and should just do whatever load it's told.
+       */
+      sem.medium_precision = false;
+      nir_intrinsic_set_io_semantics(intr, sem);
+      return true;
+   }
+
+   sem.medium_precision = false;
+   nir_intrinsic_set_io_semantics(intr, sem);
+
+   if (is_load) {
+      b->cursor = nir_after_instr(&intr->instr);
+
+      intr->def.bit_size = slot_bit_size;
+      nir_intrinsic_set_dest_type(intr, slot_type);
+
+      data = nir_type_convert(b, &intr->def, slot_type, data_type,
+                              nir_rounding_mode_undef);
+
+      nir_def_rewrite_uses_after(&intr->def, data);
+   } else {
+      b->cursor = nir_before_instr(&intr->instr);
+
+      nir_def *data = nir_type_convert(b, intr->src[0].ssa, data_type,
+                                       slot_type, nir_rounding_mode_undef);
+
+      nir_src_rewrite(&intr->src[0], data);
+      nir_intrinsic_set_src_type(intr, slot_type);
+   }
+
+   return true;
+}
+
+bool
+pan_nir_resize_varying_io(nir_shader *nir,
+                          const struct pan_varying_layout *varying_layout)
+{
+   return nir_shader_intrinsics_pass(nir, resize_io_intr,
+                                     nir_metadata_control_flow,
+                                     (void *)varying_layout);
+}