i965/nir: Do not scalarize phis in non-scalar setups

[mesa.git] / src / mesa / drivers / dri / i965 / brw_fs_register_coalesce.cpp

diff --git a/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp b/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp
index 09f0faddbdbffe022279cccbd6f174ded955c543..72e873857ced950272a885b15db2914b50ba3e2c 100644 (file)
--- a/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_register_coalesce.cpp
@@ -63,21 +63,6 @@ is_nop_mov(const fs_inst *inst)
    return false;
 }
 
-static bool
-is_copy_payload(const fs_visitor *v, const fs_inst *inst)
-{
-   if (v->alloc.sizes[inst->src[0].reg] != inst->regs_written)
-      return false;
-
-   fs_reg reg = inst->src[0];
-
-   for (int i = 0; i < inst->sources; i++)
-      if (!inst->src[i].equals(offset(reg, i)))
-         return false;
-
-   return true;
-}
-
 static bool
 is_coalesce_candidate(const fs_visitor *v, const fs_inst *inst)
 {
@@ -99,7 +84,7 @@ is_coalesce_candidate(const fs_visitor *v, const fs_inst *inst)
       return false;
 
    if (inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD) {
-      if (!is_copy_payload(v, inst)) {
+      if (!inst->is_copy_payload(v->alloc)) {
          return false;
       }
    }
@@ -182,7 +167,6 @@ fs_visitor::register_coalesce()
          src_size = alloc.sizes[inst->src[0].reg];
          assert(src_size <= MAX_VGRF_SIZE);
 
-         assert(inst->src[0].width % 8 == 0);
          channels_remaining = src_size;
          memset(mov, 0, sizeof(mov));
 
@@ -211,7 +195,7 @@ fs_visitor::register_coalesce()
             continue;
          }
          reg_to_offset[offset] = inst->dst.reg_offset;
-         if (inst->src[0].width == 16)
+         if (inst->regs_written > 1)
             reg_to_offset[offset + 1] = inst->dst.reg_offset + 1;
          mov[offset] = inst;
          channels_remaining -= inst->regs_written;
@@ -244,7 +228,6 @@ fs_visitor::register_coalesce()
          continue;
 
       progress = true;
-      bool was_load_payload = inst->opcode == SHADER_OPCODE_LOAD_PAYLOAD;
 
       for (int i = 0; i < src_size; i++) {
          if (mov[i]) {
@@ -258,22 +241,19 @@ fs_visitor::register_coalesce()
       }
 
       foreach_block_and_inst(block, fs_inst, scan_inst, cfg) {
-         for (int i = 0; i < src_size; i++) {
-            if (mov[i] || was_load_payload) {
-               if (scan_inst->dst.file == GRF &&
-                   scan_inst->dst.reg == reg_from &&
-                   scan_inst->dst.reg_offset == i) {
-                  scan_inst->dst.reg = reg_to;
-                  scan_inst->dst.reg_offset = reg_to_offset[i];
-               }
-               for (int j = 0; j < scan_inst->sources; j++) {
-                  if (scan_inst->src[j].file == GRF &&
-                      scan_inst->src[j].reg == reg_from &&
-                      scan_inst->src[j].reg_offset == i) {
-                     scan_inst->src[j].reg = reg_to;
-                     scan_inst->src[j].reg_offset = reg_to_offset[i];
-                  }
-               }
+         if (scan_inst->dst.file == GRF &&
+             scan_inst->dst.reg == reg_from) {
+            scan_inst->dst.reg = reg_to;
+            scan_inst->dst.reg_offset =
+               reg_to_offset[scan_inst->dst.reg_offset];
+         }
+
+         for (int j = 0; j < scan_inst->sources; j++) {
+            if (scan_inst->src[j].file == GRF &&
+                scan_inst->src[j].reg == reg_from) {
+               scan_inst->src[j].reg = reg_to;
+               scan_inst->src[j].reg_offset =
+                  reg_to_offset[scan_inst->src[j].reg_offset];
             }
          }
       }