intel/fs: Remove FS_OPCODE_UNPACK_HALF_2x16_SPLIT opcodes.

These are broken on a future platform, but it turns out we don't need
to fix them, since they're just type-converting moves with strided
source.  Kill them.

Reviewed-by: Iago Toral Quiroga <[email protected]>

6 files changed, 4 insertions, 47 deletions

diff --git a/src/intel/compiler/brw_eu_defines.h b/src/intel/compiler/brw_eu_defines.h
index affe977835b..b7bd104be59 100644
--- a/src/intel/compiler/brw_eu_defines.h
+++ b/src/intel/compiler/brw_eu_defines.h
@@ -523,8 +523,6 @@ enum opcode {
    FS_OPCODE_DISCARD_JUMP,
    FS_OPCODE_SET_SAMPLE_ID,
    FS_OPCODE_PACK_HALF_2x16_SPLIT,
-   FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X,
-   FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y,
    FS_OPCODE_PLACEHOLDER_HALT,
    FS_OPCODE_INTERPOLATE_AT_SAMPLE,
    FS_OPCODE_INTERPOLATE_AT_SHARED_OFFSET,
diff --git a/src/intel/compiler/brw_fs.cpp b/src/intel/compiler/brw_fs.cpp
index cfc1a470886..c4052be978f 100644
--- a/src/intel/compiler/brw_fs.cpp
+++ b/src/intel/compiler/brw_fs.cpp
@@ -5475,8 +5475,6 @@ get_lowered_simd_width(const struct gen_device_info *devinfo,
    case FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD:
    case FS_OPCODE_VARYING_PULL_CONSTANT_LOAD_GEN7:
    case FS_OPCODE_PACK_HALF_2x16_SPLIT:
-   case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X:
-   case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y:
    case FS_OPCODE_INTERPOLATE_AT_SAMPLE:
    case FS_OPCODE_INTERPOLATE_AT_SHARED_OFFSET:
    case FS_OPCODE_INTERPOLATE_AT_PER_SLOT_OFFSET:
diff --git a/src/intel/compiler/brw_fs.h b/src/intel/compiler/brw_fs.h
index 7edaa3af43c..d141a9237df 100644
--- a/src/intel/compiler/brw_fs.h
+++ b/src/intel/compiler/brw_fs.h
@@ -461,9 +461,6 @@ private:
                                       struct brw_reg dst,
                                       struct brw_reg x,
                                       struct brw_reg y);
-   void generate_unpack_half_2x16_split(fs_inst *inst,
-                                        struct brw_reg dst,
-                                        struct brw_reg src);
 
    void generate_shader_time_add(fs_inst *inst,
                                  struct brw_reg payload,
diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index 84627e83132..9088c97d92b 100644
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -1756,35 +1756,6 @@ fs_generator::generate_pack_half_2x16_split(fs_inst *,
 }
 
 void
-fs_generator::generate_unpack_half_2x16_split(fs_inst *inst,
-                                              struct brw_reg dst,
-                                              struct brw_reg src)
-{
-   assert(devinfo->gen >= 7);
-   assert(dst.type == BRW_REGISTER_TYPE_F);
-   assert(src.type == BRW_REGISTER_TYPE_UD);
-
-   /* From the Ivybridge PRM, Vol4, Part3, Section 6.26 f16to32:
-    *
-    *   Because this instruction does not have a 16-bit floating-point type,
-    *   the source data type must be Word (W). The destination type must be
-    *   F (Float).
-    */
-   struct brw_reg src_w = spread(retype(src, BRW_REGISTER_TYPE_W), 2);
-
-   /* Each channel of src has the form of unpackHalf2x16's input: 0xhhhhllll.
-    * For the Y case, we wish to access only the upper word; therefore
-    * a 16-bit subregister offset is needed.
-    */
-   assert(inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X ||
-          inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y);
-   if (inst->opcode == FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y)
-      src_w.subnr += 2;
-
-   brw_F16TO32(p, dst, src_w);
-}
-
-void
 fs_generator::generate_shader_time_add(fs_inst *,
                                        struct brw_reg payload,
                                        struct brw_reg offset,
@@ -2421,11 +2392,6 @@ fs_generator::generate_code(const cfg_t *cfg, int dispatch_width)
           generate_pack_half_2x16_split(inst, dst, src[0], src[1]);
           break;
 
-      case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X:
-      case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y:
-         generate_unpack_half_2x16_split(inst, dst, src[0]);
-         break;
-
       case FS_OPCODE_PLACEHOLDER_HALT:
          /* This is the place where the final HALT needs to be inserted if
           * we've emitted any discards.  If not, this will emit no code.
diff --git a/src/intel/compiler/brw_fs_nir.cpp b/src/intel/compiler/brw_fs_nir.cpp
index 312cd22de79..fe3dff016ad 100644
--- a/src/intel/compiler/brw_fs_nir.cpp
+++ b/src/intel/compiler/brw_fs_nir.cpp
@@ -1319,11 +1319,13 @@ fs_visitor::nir_emit_alu(const fs_builder &bld, nir_alu_instr *instr)
       unreachable("not reached: should be handled by lower_packing_builtins");
 
    case nir_op_unpack_half_2x16_split_x:
-      inst = bld.emit(FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X, result, op[0]);
+      inst = bld.emit(BRW_OPCODE_F16TO32, result,
+                      subscript(op[0], BRW_REGISTER_TYPE_UW, 0));
       inst->saturate = instr->dest.saturate;
       break;
    case nir_op_unpack_half_2x16_split_y:
-      inst = bld.emit(FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y, result, op[0]);
+      inst = bld.emit(BRW_OPCODE_F16TO32, result,
+                      subscript(op[0], BRW_REGISTER_TYPE_UW, 1));
       inst->saturate = instr->dest.saturate;
       break;
 
diff --git a/src/intel/compiler/brw_shader.cpp b/src/intel/compiler/brw_shader.cpp
index b77bd798d17..97966c951a1 100644
--- a/src/intel/compiler/brw_shader.cpp
+++ b/src/intel/compiler/brw_shader.cpp
@@ -415,10 +415,6 @@ brw_instruction_name(const struct gen_device_info *devinfo, enum opcode op)
 
    case FS_OPCODE_PACK_HALF_2x16_SPLIT:
       return "pack_half_2x16_split";
-   case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_X:
-      return "unpack_half_2x16_split_x";
-   case FS_OPCODE_UNPACK_HALF_2x16_SPLIT_Y:
-      return "unpack_half_2x16_split_y";
 
    case FS_OPCODE_PLACEHOLDER_HALT:
       return "placeholder_halt";