intel/fs: Add FS_OPCODE_SCHEDULING_FENCE

[mesa.git] / src / intel / compiler / brw_fs_generator.cpp

diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index 036aa58cc7c7d7c426551fcad4cdc112d9c50f7d..e8bbce5b2903263c8eae368a1a36bed411e4f238 100644 (file)
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -2062,7 +2062,15 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width,
       case SHADER_OPCODE_SEND:
          generate_send(inst, dst, src[0], src[1], src[2],
                        inst->ex_mlen > 0 ? src[3] : brw_null_reg());
-         send_count++;
+         if ((inst->desc & 0xff) == BRW_BTI_STATELESS ||
+             (inst->desc & 0xff) == GEN8_BTI_STATELESS_NON_COHERENT) {
+            if (inst->size_written)
+               fill_count++;
+            else
+               spill_count++;
+         } else {
+            send_count++;
+         }
          break;
 
       case SHADER_OPCODE_GET_BUFFER_SIZE:
@@ -2169,6 +2177,11 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width,
          send_count++;
          break;
 
+      case FS_OPCODE_SCHEDULING_FENCE:
+         if (unlikely(debug_flag))
+            disasm_info->use_tail = true;
+         break;
+
       case SHADER_OPCODE_INTERLOCK:
          assert(devinfo->gen >= 9);
          /* The interlock is basically a memory fence issued via sendc */