[Mesa-dev] [PATCH 10/18] i965: add indirect packing support for tcs and tes

[Timothy Arceri]

Reviewed-by: Kenneth Graunke <kenneth at whitecape.org>
---
 src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 33 ++++++++++++++++++++++++++++----
 1 file changed, 29 insertions(+), 4 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
index 587549f..4eaf5ea 100644
--- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
+++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp
@@ -2428,8 +2428,19 @@ fs_visitor::nir_emit_tcs_intrinsic(const fs_builder &bld,
             const fs_reg srcs[] = { icp_handle, indirect_offset };
             fs_reg payload = bld.vgrf(BRW_REGISTER_TYPE_UD, 2);
             bld.LOAD_PAYLOAD(payload, srcs, ARRAY_SIZE(srcs), 0);
-
-            inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, dst, payload);
+            if (first_component != 0) {
+               unsigned read_components = num_components + first_component;
+               fs_reg tmp = bld.vgrf(dst.type, read_components);
+               inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, tmp,
+                               payload);
+               for (unsigned i = 0; i < num_components; i++) {
+                  bld.MOV(offset(dst, bld, i),
+                          offset(tmp, bld, i + first_component));
+               }
+            } else {
+               inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, dst,
+                               payload);
+            }
             inst->offset = imm_offset;
             inst->base_mrf = -1;
             inst->mlen = 2;
@@ -2899,11 +2910,25 @@ fs_visitor::nir_emit_tes_intrinsic(const fs_builder &bld,
          fs_reg payload = bld.vgrf(BRW_REGISTER_TYPE_UD, 2);
          bld.LOAD_PAYLOAD(payload, srcs, ARRAY_SIZE(srcs), 0);
 
-         inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, dest, payload);
+         if (first_component != 0) {
+            unsigned read_components =
+                instr->num_components + first_component;
+            fs_reg tmp = bld.vgrf(dest.type, read_components);
+            inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, tmp,
+                            payload);
+            inst->regs_written = read_components;
+            for (unsigned i = 0; i < instr->num_components; i++) {
+               bld.MOV(offset(dest, bld, i),
+                       offset(tmp, bld, i + first_component));
+            }
+         } else {
+            inst = bld.emit(SHADER_OPCODE_URB_READ_SIMD8_PER_SLOT, dest,
+                            payload);
+            inst->regs_written = instr->num_components;
+         }
          inst->mlen = 2;
          inst->offset = imm_offset;
          inst->base_mrf = -1;
-         inst->regs_written = instr->num_components;
       }
       break;
    }
-- 
2.5.5