summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorKonstantin Zhuravlyov <kzhuravl_dev@outlook.com>2019-05-28 21:18:34 +0000
committerKonstantin Zhuravlyov <kzhuravl_dev@outlook.com>2019-05-28 21:18:34 +0000
commitfe23ed2c681413e7baf517c79aee9be130579873 (patch)
tree93e7fd028e75212807a54211c1e74619bf6dd01f
parent14689910737b8e63a0ef7caf407d13aa68bbd6f8 (diff)
downloadbcm5719-llvm-fe23ed2c681413e7baf517c79aee9be130579873.tar.gz
bcm5719-llvm-fe23ed2c681413e7baf517c79aee9be130579873.zip
AMDGPU: Temporary drop s_mul_hi_i/u32 patterns
It introduces performance regressions in several applications. This has already been submitted downstream. llvm-svn: 361879
-rw-r--r--llvm/lib/Target/AMDGPU/SOPInstructions.td8
-rw-r--r--llvm/test/CodeGen/AMDGPU/mul.ll5
2 files changed, 2 insertions, 11 deletions
diff --git a/llvm/lib/Target/AMDGPU/SOPInstructions.td b/llvm/lib/Target/AMDGPU/SOPInstructions.td
index 401a560a683..342293851c3 100644
--- a/llvm/lib/Target/AMDGPU/SOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SOPInstructions.td
@@ -558,12 +558,8 @@ let SubtargetPredicate = isGFX9Plus in {
def S_LSHL4_ADD_U32 : SOP2_32<"s_lshl4_add_u32">;
} // End Defs = [SCC]
- let isCommutable = 1 in {
- def S_MUL_HI_U32 : SOP2_32<"s_mul_hi_u32",
- [(set i32:$sdst, (UniformBinFrag<mulhu> SSrc_b32:$src0, SSrc_b32:$src1))]>;
- def S_MUL_HI_I32 : SOP2_32<"s_mul_hi_i32",
- [(set i32:$sdst, (UniformBinFrag<mulhs> SSrc_b32:$src0, SSrc_b32:$src1))]>;
- }
+ def S_MUL_HI_U32 : SOP2_32<"s_mul_hi_u32">;
+ def S_MUL_HI_I32 : SOP2_32<"s_mul_hi_i32">;
} // End SubtargetPredicate = isGFX9Plus
//===----------------------------------------------------------------------===//
diff --git a/llvm/test/CodeGen/AMDGPU/mul.ll b/llvm/test/CodeGen/AMDGPU/mul.ll
index d9d51597891..f8b4ac906c0 100644
--- a/llvm/test/CodeGen/AMDGPU/mul.ll
+++ b/llvm/test/CodeGen/AMDGPU/mul.ll
@@ -141,11 +141,6 @@ define amdgpu_kernel void @v_mul_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %
; crash with a 'failed to select' error.
; FUNC-LABEL: {{^}}s_mul_i64:
-; GFX9_10-DAG: s_mul_i32
-; GFX9_10-DAG: s_mul_hi_u32
-; GFX9_10-DAG: s_mul_i32
-; GFX9_10-DAG: s_mul_i32
-; GFX9_10: s_endpgm
define amdgpu_kernel void @s_mul_i64(i64 addrspace(1)* %out, i64 %a, i64 %b) nounwind {
%mul = mul i64 %a, %b
store i64 %mul, i64 addrspace(1)* %out, align 8
OpenPOWER on IntegriCloud