intel/fs: Rework INTERPOLATE_AT_PER_SLOT_OFFSET

This reworks INTERPOLATE_AT_PER_SLOT_OFFSET to work more like an ALU
operation and less like a send.  This is less code over-all and, as a
side-effect, it now properly handles execution groups and lowering so
SIMD32 support just falls out.

Reviewed-by: Jason Ekstrand <jason@jlekstrand.net>
Reviewed-by: Matt Turner <mattst88@gmail.com>

diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index 07be2e3..84becb3 100644 (file)
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -805,6 +805,8 @@ fs_inst::components_read(unsigned i) const
       else
          return 1;
    }
+   case FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET:
+      return (i == 0 ? 2 : 1);
 
    default:
       return 1;
@@ -840,7 +842,6 @@ fs_inst::size_read(int arg) const
    case SHADER_OPCODE_TYPED_SURFACE_WRITE:
    case FS_OPCODE_INTERPOLATE_AT_SAMPLE:
    case FS_OPCODE_INTERPOLATE_AT_SHARED_OFFSET:
-   case FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET:
    case SHADER_OPCODE_BYTE_SCATTERED_WRITE:
    case SHADER_OPCODE_BYTE_SCATTERED_READ:
       if (arg == 0)

diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index a01d320..c352f87 100644 (file)
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -1579,16 +1579,18 @@ fs_generator::generate_pixel_interpolator_query(fs_inst *inst,
                                                 struct brw_reg msg_data,
                                                 unsigned msg_type)
 {
-   assert(inst->size_written % REG_SIZE == 0);
+   const bool has_payload = inst->src[0].file != BAD_FILE;
    assert(msg_data.type == BRW_REGISTER_TYPE_UD);
+   assert(inst->size_written % REG_SIZE == 0);
 
    brw_pixel_interpolator_query(p,
          retype(dst, BRW_REGISTER_TYPE_UW),
-         src,
+         /* If we don't have a payload, what we send doesn't matter */
+         has_payload ? src : brw_vec8_grf(0, 0),
          inst->pi_noperspective,
          msg_type,
          msg_data,
-         inst->mlen,
+         has_payload ? 2 * inst->exec_size / 8 : 1,
          inst->size_written / REG_SIZE);
 }
 

diff --git a/src/intel/compiler/brw_fs_nir.cpp b/src/intel/compiler/brw_fs_nir.cpp
index 7ee68de..b881975 100644 (file)
--- a/src/intel/compiler/brw_fs_nir.cpp
+++ b/src/intel/compiler/brw_fs_nir.cpp
@@ -1795,21 +1795,8 @@ emit_pixel_interpolater_send(const fs_builder &bld,
 {
    struct brw_wm_prog_data *wm_prog_data =
       brw_wm_prog_data(bld.shader->stage_prog_data);
-   fs_inst *inst;
-   fs_reg payload;
-   int mlen;
 
-   if (src.file == BAD_FILE) {
-      /* Dummy payload */
-      payload = bld.vgrf(BRW_REGISTER_TYPE_F, 1);
-      mlen = 1;
-   } else {
-      payload = src;
-      mlen = 2 * bld.dispatch_width() / 8;
-   }
-
-   inst = bld.emit(opcode, dst, payload, desc);
-   inst->mlen = mlen;
+   fs_inst *inst = bld.emit(opcode, dst, src, desc);
    /* 2 floats per slot returned */
    inst->size_written = 2 * dst.component_size(inst->exec_size);
    inst->pi_noperspective = interpolation == INTERP_MODE_NOPERSPECTIVE;
@@ -3464,7 +3451,7 @@ fs_visitor::nir_emit_fs_intrinsic(const fs_builder &bld,
                                             FS_OPCODE_INTERPOLATE_AT_SAMPLE,
                                             dest,
                                             fs_reg(), /* src */
-                                            msg_data,
+                                            component(msg_data, 0),
                                             interpolation);
             set_predicate(BRW_PREDICATE_NORMAL, inst);