[llvm] r359435 - [InstCombine][X86] Add PACKSS tests for truncation of sign-extended comparisons
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Mon Apr 29 03:36:20 PDT 2019
Author: rksimon
Date: Mon Apr 29 03:36:20 2019
New Revision: 359435
URL: http://llvm.org/viewvc/llvm-project?rev=359435&view=rev
Log:
[InstCombine][X86] Add PACKSS tests for truncation of sign-extended comparisons
Modified:
llvm/trunk/test/Transforms/InstCombine/X86/x86-pack.ll
Modified: llvm/trunk/test/Transforms/InstCombine/X86/x86-pack.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/X86/x86-pack.ll?rev=359435&r1=359434&r2=359435&view=diff
==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/X86/x86-pack.ll (original)
+++ llvm/trunk/test/Transforms/InstCombine/X86/x86-pack.ll Mon Apr 29 03:36:20 2019
@@ -510,6 +510,112 @@ define <64 x i8> @trunc_packuswb_512(<32
ret <64 x i8> %3
}
+;
+; Signed Pack Comparison Results
+;
+
+define <8 x i16> @cmp_packssdw_128(<4 x i32> %a0, <4 x i32> %a1, <4 x i32> %a2, <4 x i32> %a3) {
+; CHECK-LABEL: @cmp_packssdw_128(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <4 x i32> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <4 x i32> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <4 x i1> [[TMP1]] to <4 x i32>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <4 x i1> [[TMP2]] to <4 x i32>
+; CHECK-NEXT: [[TMP5:%.*]] = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> [[TMP3]], <4 x i32> [[TMP4]])
+; CHECK-NEXT: ret <8 x i16> [[TMP5]]
+;
+ %1 = icmp eq <4 x i32> %a0, %a1
+ %2 = icmp eq <4 x i32> %a2, %a3
+ %3 = sext <4 x i1> %1 to <4 x i32>
+ %4 = sext <4 x i1> %2 to <4 x i32>
+ %5 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %3, <4 x i32> %4)
+ ret <8 x i16> %5
+}
+
+define <16 x i8> @cmp_packsswb_128(<8 x i16> %a0, <8 x i16> %a1, <8 x i16> %a2, <8 x i16> %a3) {
+; CHECK-LABEL: @cmp_packsswb_128(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <8 x i16> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <8 x i16> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <8 x i1> [[TMP1]] to <8 x i16>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <8 x i1> [[TMP2]] to <8 x i16>
+; CHECK-NEXT: [[TMP5:%.*]] = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> [[TMP3]], <8 x i16> [[TMP4]])
+; CHECK-NEXT: ret <16 x i8> [[TMP5]]
+;
+ %1 = icmp eq <8 x i16> %a0, %a1
+ %2 = icmp eq <8 x i16> %a2, %a3
+ %3 = sext <8 x i1> %1 to <8 x i16>
+ %4 = sext <8 x i1> %2 to <8 x i16>
+ %5 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %3, <8 x i16> %4)
+ ret <16 x i8> %5
+}
+
+define <16 x i16> @cmp_packssdw_256(<8 x i32> %a0, <8 x i32> %a1, <8 x i32> %a2, <8 x i32> %a3) {
+; CHECK-LABEL: @cmp_packssdw_256(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <8 x i32> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <8 x i32> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <8 x i1> [[TMP1]] to <8 x i32>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <8 x i1> [[TMP2]] to <8 x i32>
+; CHECK-NEXT: [[TMP5:%.*]] = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> [[TMP3]], <8 x i32> [[TMP4]])
+; CHECK-NEXT: ret <16 x i16> [[TMP5]]
+;
+ %1 = icmp eq <8 x i32> %a0, %a1
+ %2 = icmp eq <8 x i32> %a2, %a3
+ %3 = sext <8 x i1> %1 to <8 x i32>
+ %4 = sext <8 x i1> %2 to <8 x i32>
+ %5 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %3, <8 x i32> %4)
+ ret <16 x i16> %5
+}
+
+define <32 x i8> @cmp_packsswb_256(<16 x i16> %a0, <16 x i16> %a1, <16 x i16> %a2, <16 x i16> %a3) {
+; CHECK-LABEL: @cmp_packsswb_256(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <16 x i16> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <16 x i16> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <16 x i1> [[TMP1]] to <16 x i16>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <16 x i1> [[TMP2]] to <16 x i16>
+; CHECK-NEXT: [[TMP5:%.*]] = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> [[TMP3]], <16 x i16> [[TMP4]])
+; CHECK-NEXT: ret <32 x i8> [[TMP5]]
+;
+ %1 = icmp eq <16 x i16> %a0, %a1
+ %2 = icmp eq <16 x i16> %a2, %a3
+ %3 = sext <16 x i1> %1 to <16 x i16>
+ %4 = sext <16 x i1> %2 to <16 x i16>
+ %5 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %3, <16 x i16> %4)
+ ret <32 x i8> %5
+}
+
+define <32 x i16> @cmp_packssdw_512(<16 x i32> %a0, <16 x i32> %a1, <16 x i32> %a2, <16 x i32> %a3) {
+; CHECK-LABEL: @cmp_packssdw_512(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <16 x i32> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <16 x i32> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <16 x i1> [[TMP1]] to <16 x i32>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <16 x i1> [[TMP2]] to <16 x i32>
+; CHECK-NEXT: [[TMP5:%.*]] = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> [[TMP3]], <16 x i32> [[TMP4]])
+; CHECK-NEXT: ret <32 x i16> [[TMP5]]
+;
+ %1 = icmp eq <16 x i32> %a0, %a1
+ %2 = icmp eq <16 x i32> %a2, %a3
+ %3 = sext <16 x i1> %1 to <16 x i32>
+ %4 = sext <16 x i1> %2 to <16 x i32>
+ %5 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %3, <16 x i32> %4)
+ ret <32 x i16> %5
+}
+
+define <64 x i8> @cmp_packsswb_512(<32 x i16> %a0, <32 x i16> %a1, <32 x i16> %a2, <32 x i16> %a3) {
+; CHECK-LABEL: @cmp_packsswb_512(
+; CHECK-NEXT: [[TMP1:%.*]] = icmp eq <32 x i16> [[A0:%.*]], [[A1:%.*]]
+; CHECK-NEXT: [[TMP2:%.*]] = icmp eq <32 x i16> [[A2:%.*]], [[A3:%.*]]
+; CHECK-NEXT: [[TMP3:%.*]] = sext <32 x i1> [[TMP1]] to <32 x i16>
+; CHECK-NEXT: [[TMP4:%.*]] = sext <32 x i1> [[TMP2]] to <32 x i16>
+; CHECK-NEXT: [[TMP5:%.*]] = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> [[TMP3]], <32 x i16> [[TMP4]])
+; CHECK-NEXT: ret <64 x i8> [[TMP5]]
+;
+ %1 = icmp eq <32 x i16> %a0, %a1
+ %2 = icmp eq <32 x i16> %a2, %a3
+ %3 = sext <32 x i1> %1 to <32 x i16>
+ %4 = sext <32 x i1> %2 to <32 x i16>
+ %5 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %3, <32 x i16> %4)
+ ret <64 x i8> %5
+}
+
declare <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32>, <4 x i32>) nounwind readnone
declare <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16>, <8 x i16>) nounwind readnone
declare <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16>, <8 x i16>) nounwind readnone
More information about the llvm-commits
mailing list