diff options
| author | Craig Topper <craig.topper@intel.com> | 2018-09-27 21:28:59 +0000 |
|---|---|---|
| committer | Craig Topper <craig.topper@intel.com> | 2018-09-27 21:28:59 +0000 |
| commit | 6911bfe2634e1cf180699ab0f977f04e8bb066db (patch) | |
| tree | b13c9a7b8b252f965589c74449a13ae2343b1a2b /llvm/test/Transforms/ScalarizeMaskedMemIntrin | |
| parent | 45ad631b4c22bdd37222ef8033286e7d22a7a6d5 (diff) | |
| download | bcm5719-llvm-6911bfe2634e1cf180699ab0f977f04e8bb066db.tar.gz bcm5719-llvm-6911bfe2634e1cf180699ab0f977f04e8bb066db.zip | |
[ScalarizeMaskedMemIntrin] When expanding masked gathers, start with the passthru vector and insert the new load results into it.
Previously we started with undef and did a final merge with the passthru at the end.
llvm-svn: 343273
Diffstat (limited to 'llvm/test/Transforms/ScalarizeMaskedMemIntrin')
| -rw-r--r-- | llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll | 22 |
1 files changed, 9 insertions, 13 deletions
diff --git a/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll b/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll index 53d5243615f..40c0fc4ef6d 100644 --- a/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll +++ b/llvm/test/Transforms/ScalarizeMaskedMemIntrin/X86/expand-masked-gather.ll @@ -8,10 +8,10 @@ define <2 x i64> @scalarize_v2i64(<2 x i64*> %p, <2 x i1> %mask, <2 x i64> %pass ; CHECK: cond.load: ; CHECK-NEXT: [[PTR0:%.*]] = extractelement <2 x i64*> [[P:%.*]], i32 0 ; CHECK-NEXT: [[LOAD0:%.*]] = load i64, i64* [[PTR0]], align 8 -; CHECK-NEXT: [[RES0:%.*]] = insertelement <2 x i64> undef, i64 [[LOAD0]], i32 0 +; CHECK-NEXT: [[RES0:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[LOAD0]], i32 0 ; CHECK-NEXT: br label [[ELSE]] ; CHECK: else: -; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[RES0]], [[COND_LOAD]] ], [ undef, [[TMP0:%.*]] ] +; CHECK-NEXT: [[RES_PHI_ELSE:%.*]] = phi <2 x i64> [ [[RES0]], [[COND_LOAD]] ], [ [[PASSTHRU]], [[TMP0:%.*]] ] ; CHECK-NEXT: [[MASK1:%.*]] = extractelement <2 x i1> [[MASK]], i32 1 ; CHECK-NEXT: br i1 [[MASK1]], label [[COND_LOAD1:%.*]], label [[ELSE2:%.*]] ; CHECK: cond.load1: @@ -20,9 +20,8 @@ define <2 x i64> @scalarize_v2i64(<2 x i64*> %p, <2 x i1> %mask, <2 x i64> %pass ; CHECK-NEXT: [[RES1:%.*]] = insertelement <2 x i64> [[RES_PHI_ELSE]], i64 [[LOAD1]], i32 1 ; CHECK-NEXT: br label [[ELSE2]] ; CHECK: else2: -; CHECK-NEXT: [[RES_PHI_SELECT:%.*]] = phi <2 x i64> [ [[RES1]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ] -; CHECK-NEXT: [[TMP1:%.*]] = select <2 x i1> [[MASK]], <2 x i64> [[RES_PHI_SELECT]], <2 x i64> [[PASSTHRU:%.*]] -; CHECK-NEXT: ret <2 x i64> [[TMP1]] +; CHECK-NEXT: [[RES_PHI_ELSE3:%.*]] = phi <2 x i64> [ [[RES1]], [[COND_LOAD1]] ], [ [[RES_PHI_ELSE]], [[ELSE]] ] +; CHECK-NEXT: ret <2 x i64> [[RES_PHI_ELSE3]] ; %ret = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %p, i32 8, <2 x i1> %mask, <2 x i64> %passthru) ret <2 x i64> %ret @@ -32,12 +31,11 @@ define <2 x i64> @scalarize_v2i64_ones_mask(<2 x i64*> %p, <2 x i64> %passthru) ; CHECK-LABEL: @scalarize_v2i64_ones_mask( ; CHECK-NEXT: [[PTR0:%.*]] = extractelement <2 x i64*> [[P:%.*]], i32 0 ; CHECK-NEXT: [[LOAD0:%.*]] = load i64, i64* [[PTR0]], align 8 -; CHECK-NEXT: [[RES0:%.*]] = insertelement <2 x i64> undef, i64 [[LOAD0]], i32 0 +; CHECK-NEXT: [[RES0:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[LOAD0]], i32 0 ; CHECK-NEXT: [[PTR1:%.*]] = extractelement <2 x i64*> [[P]], i32 1 ; CHECK-NEXT: [[LOAD1:%.*]] = load i64, i64* [[PTR1]], align 8 ; CHECK-NEXT: [[RES1:%.*]] = insertelement <2 x i64> [[RES0]], i64 [[LOAD1]], i32 1 -; CHECK-NEXT: [[TMP1:%.*]] = select <2 x i1> <i1 true, i1 true>, <2 x i64> [[RES1]], <2 x i64> [[PASSTHRU:%.*]] -; CHECK-NEXT: ret <2 x i64> [[TMP1]] +; CHECK-NEXT: ret <2 x i64> [[RES1]] ; %ret = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %p, i32 8, <2 x i1> <i1 true, i1 true>, <2 x i64> %passthru) ret <2 x i64> %ret @@ -45,8 +43,7 @@ define <2 x i64> @scalarize_v2i64_ones_mask(<2 x i64*> %p, <2 x i64> %passthru) define <2 x i64> @scalarize_v2i64_zero_mask(<2 x i64*> %p, <2 x i64> %passthru) { ; CHECK-LABEL: @scalarize_v2i64_zero_mask( -; CHECK-NEXT: [[TMP1:%.*]] = select <2 x i1> zeroinitializer, <2 x i64> undef, <2 x i64> [[PASSTHRU:%.*]] -; CHECK-NEXT: ret <2 x i64> [[TMP1]] +; CHECK-NEXT: ret <2 x i64> [[PASSTHRU:%.*]] ; %ret = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %p, i32 8, <2 x i1> <i1 false, i1 false>, <2 x i64> %passthru) ret <2 x i64> %ret @@ -56,9 +53,8 @@ define <2 x i64> @scalarize_v2i64_const_mask(<2 x i64*> %p, <2 x i64> %passthru) ; CHECK-LABEL: @scalarize_v2i64_const_mask( ; CHECK-NEXT: [[PTR1:%.*]] = extractelement <2 x i64*> [[P:%.*]], i32 1 ; CHECK-NEXT: [[LOAD1:%.*]] = load i64, i64* [[PTR1]], align 8 -; CHECK-NEXT: [[RES1:%.*]] = insertelement <2 x i64> undef, i64 [[LOAD1]], i32 1 -; CHECK-NEXT: [[TMP1:%.*]] = select <2 x i1> <i1 false, i1 true>, <2 x i64> [[RES1]], <2 x i64> [[PASSTHRU:%.*]] -; CHECK-NEXT: ret <2 x i64> [[TMP1]] +; CHECK-NEXT: [[RES1:%.*]] = insertelement <2 x i64> [[PASSTHRU:%.*]], i64 [[LOAD1]], i32 1 +; CHECK-NEXT: ret <2 x i64> [[RES1]] ; %ret = call <2 x i64> @llvm.masked.gather.v2i64.v2p0i64(<2 x i64*> %p, i32 8, <2 x i1> <i1 false, i1 true>, <2 x i64> %passthru) ret <2 x i64> %ret |

