[Mesa-stable] [PATCH 2/3] i965/vec4: Fix swizzles on atomic sources.
Kenneth Graunke
kenneth at whitecape.org
Sun Sep 24 22:09:49 UTC 2017
Atomic operation sources are scalar values, but we were failing to
select the .x component of the second operand. For example,
atomicCounterCompSwapARB(counter, 5u, 10u)
would generate
mov(8) vgrf4.x:D, 5D
mov(8) vgrf5.x:D, 10D
mov(8) vgrf9.x:UD, vgrf4.xyzw:D
mov(8) vgrf9.y:UD, vgrf5.xyzw:D
which wrongly selects the .y component of vgrf5, so the actual 10u value
would get dead code eliminated. The swizzle works for the other source,
but both of them ought to be .xxxx.
Fixes the compare and swap CTS tests in:
KHR-GL45.shader_atomic_counter_ops_tests.ShaderAtomicCounterOpsExchangeTestCase
Cc: "17.2 17.1 17.0 13.0" <mesa-stable at lists.freedesktop.org>
---
src/intel/compiler/brw_vec4_surface_builder.cpp | 13 +++++++++----
1 file changed, 9 insertions(+), 4 deletions(-)
diff --git a/src/intel/compiler/brw_vec4_surface_builder.cpp b/src/intel/compiler/brw_vec4_surface_builder.cpp
index 5029cdce558..0e02aaf933a 100644
--- a/src/intel/compiler/brw_vec4_surface_builder.cpp
+++ b/src/intel/compiler/brw_vec4_surface_builder.cpp
@@ -212,10 +212,15 @@ namespace brw {
const unsigned size = (src0.file != BAD_FILE) + (src1.file != BAD_FILE);
const dst_reg srcs = bld.vgrf(BRW_REGISTER_TYPE_UD);
- if (size >= 1)
- bld.MOV(writemask(srcs, WRITEMASK_X), src0);
- if (size >= 2)
- bld.MOV(writemask(srcs, WRITEMASK_Y), src1);
+ if (size >= 1) {
+ bld.MOV(writemask(srcs, WRITEMASK_X),
+ swizzle(src0, BRW_SWIZZLE_XXXX));
+ }
+
+ if (size >= 2) {
+ bld.MOV(writemask(srcs, WRITEMASK_Y),
+ swizzle(src1, BRW_SWIZZLE_XXXX));
+ }
return emit_send(bld, SHADER_OPCODE_UNTYPED_ATOMIC, src_reg(),
emit_insert(bld, addr, dims, has_simd4x2),
--
2.11.0
More information about the mesa-stable
mailing list