[InstCombine][X86] Add PACKSS/PACKUS tests for truncation where saturation won't occur
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@359185 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/test/Transforms/InstCombine/X86/x86-pack.ll b/test/Transforms/InstCombine/X86/x86-pack.ll
index f3c41a8..f61cc3a 100644
--- a/test/Transforms/InstCombine/X86/x86-pack.ll
+++ b/test/Transforms/InstCombine/X86/x86-pack.ll
@@ -350,6 +350,166 @@
ret <64 x i8> %4
}
+;
+; Truncation (without Saturation)
+;
+
+define <8 x i16> @trunc_packssdw_128(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: @trunc_packssdw_128(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <4 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = and <4 x i32> [[A1:%.*]], <i32 15, i32 15, i32 15, i32 15>
+; CHECK-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> [[TMP1]], <4 x i32> [[TMP2]])
+; CHECK-NEXT: ret <8 x i16> [[TMP3]]
+;
+ %1 = ashr <4 x i32> %a0, <i32 17, i32 17, i32 17, i32 17>
+ %2 = and <4 x i32> %a1, <i32 15, i32 15, i32 15, i32 15>
+ %3 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %1, <4 x i32> %2)
+ ret <8 x i16> %3
+}
+
+define <8 x i16> @trunc_packusdw_128(<4 x i32> %a0, <4 x i32> %a1) {
+; CHECK-LABEL: @trunc_packusdw_128(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <4 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = and <4 x i32> [[A1:%.*]], <i32 15, i32 15, i32 15, i32 15>
+; CHECK-NEXT: [[TMP3:%.*]] = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> [[TMP1]], <4 x i32> [[TMP2]])
+; CHECK-NEXT: ret <8 x i16> [[TMP3]]
+;
+ %1 = lshr <4 x i32> %a0, <i32 17, i32 17, i32 17, i32 17>
+ %2 = and <4 x i32> %a1, <i32 15, i32 15, i32 15, i32 15>
+ %3 = call <8 x i16> @llvm.x86.sse2.packssdw.128(<4 x i32> %1, <4 x i32> %2)
+ ret <8 x i16> %3
+}
+
+define <16 x i8> @trunc_packsswb_128(<8 x i16> %a0, <8 x i16> %a1) {
+; CHECK-LABEL: @trunc_packsswb_128(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <8 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> [[TMP1]], <8 x i16> [[TMP2]])
+; CHECK-NEXT: ret <16 x i8> [[TMP3]]
+;
+ %1 = ashr <8 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <8 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <16 x i8> @llvm.x86.sse2.packsswb.128(<8 x i16> %1, <8 x i16> %2)
+ ret <16 x i8> %3
+}
+
+define <16 x i8> @trunc_packuswb_128(<8 x i16> %a0, <8 x i16> %a1) {
+; CHECK-LABEL: @trunc_packuswb_128(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <8 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> [[TMP1]], <8 x i16> [[TMP2]])
+; CHECK-NEXT: ret <16 x i8> [[TMP3]]
+;
+ %1 = lshr <8 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <8 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <16 x i8> @llvm.x86.sse2.packuswb.128(<8 x i16> %1, <8 x i16> %2)
+ ret <16 x i8> %3
+}
+
+define <16 x i16> @trunc_packssdw_256(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: @trunc_packssdw_256(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <8 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = ashr <8 x i32> [[A1:%.*]], <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> [[TMP1]], <8 x i32> [[TMP2]])
+; CHECK-NEXT: ret <16 x i16> [[TMP3]]
+;
+ %1 = ashr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+ %2 = ashr <8 x i32> %a1, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+ %3 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %1, <8 x i32> %2)
+ ret <16 x i16> %3
+}
+
+define <16 x i16> @trunc_packusdw_256(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: @trunc_packusdw_256(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <8 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = and <8 x i32> [[A1:%.*]], <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+; CHECK-NEXT: [[TMP3:%.*]] = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> [[TMP1]], <8 x i32> [[TMP2]])
+; CHECK-NEXT: ret <16 x i16> [[TMP3]]
+;
+ %1 = lshr <8 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+ %2 = and <8 x i32> %a1, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+ %3 = call <16 x i16> @llvm.x86.avx2.packssdw(<8 x i32> %1, <8 x i32> %2)
+ ret <16 x i16> %3
+}
+
+define <32 x i8> @trunc_packsswb_256(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: @trunc_packsswb_256(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <16 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <16 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> [[TMP1]], <16 x i16> [[TMP2]])
+; CHECK-NEXT: ret <32 x i8> [[TMP3]]
+;
+ %1 = ashr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <16 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <32 x i8> @llvm.x86.avx2.packsswb(<16 x i16> %1, <16 x i16> %2)
+ ret <32 x i8> %3
+}
+
+define <32 x i8> @trunc_packuswb_256(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: @trunc_packuswb_256(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <16 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <16 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> [[TMP1]], <16 x i16> [[TMP2]])
+; CHECK-NEXT: ret <32 x i8> [[TMP3]]
+;
+ %1 = lshr <16 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <16 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <32 x i8> @llvm.x86.avx2.packuswb(<16 x i16> %1, <16 x i16> %2)
+ ret <32 x i8> %3
+}
+
+define <32 x i16> @trunc_packssdw_512(<16 x i32> %a0, <16 x i32> %a1) {
+; CHECK-LABEL: @trunc_packssdw_512(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <16 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = ashr <16 x i32> [[A1:%.*]], <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+; CHECK-NEXT: [[TMP3:%.*]] = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> [[TMP1]], <16 x i32> [[TMP2]])
+; CHECK-NEXT: ret <32 x i16> [[TMP3]]
+;
+ %1 = ashr <16 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+ %2 = ashr <16 x i32> %a1, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+ %3 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %1, <16 x i32> %2)
+ ret <32 x i16> %3
+}
+
+define <32 x i16> @trunc_packusdw_512(<16 x i32> %a0, <16 x i32> %a1) {
+; CHECK-LABEL: @trunc_packusdw_512(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <16 x i32> [[A0:%.*]], <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+; CHECK-NEXT: [[TMP2:%.*]] = and <16 x i32> [[A1:%.*]], <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+; CHECK-NEXT: [[TMP3:%.*]] = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> [[TMP1]], <16 x i32> [[TMP2]])
+; CHECK-NEXT: ret <32 x i16> [[TMP3]]
+;
+ %1 = lshr <16 x i32> %a0, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+ %2 = and <16 x i32> %a1, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+ %3 = call <32 x i16> @llvm.x86.avx512.packssdw.512(<16 x i32> %1, <16 x i32> %2)
+ ret <32 x i16> %3
+}
+
+define <64 x i8> @trunc_packsswb_512(<32 x i16> %a0, <32 x i16> %a1) {
+; CHECK-LABEL: @trunc_packsswb_512(
+; CHECK-NEXT: [[TMP1:%.*]] = ashr <32 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <32 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> [[TMP1]], <32 x i16> [[TMP2]])
+; CHECK-NEXT: ret <64 x i8> [[TMP3]]
+;
+ %1 = ashr <32 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <32 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <64 x i8> @llvm.x86.avx512.packsswb.512(<32 x i16> %1, <32 x i16> %2)
+ ret <64 x i8> %3
+}
+
+define <64 x i8> @trunc_packuswb_512(<32 x i16> %a0, <32 x i16> %a1) {
+; CHECK-LABEL: @trunc_packuswb_512(
+; CHECK-NEXT: [[TMP1:%.*]] = lshr <32 x i16> [[A0:%.*]], <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+; CHECK-NEXT: [[TMP2:%.*]] = and <32 x i16> [[A1:%.*]], <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+; CHECK-NEXT: [[TMP3:%.*]] = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> [[TMP1]], <32 x i16> [[TMP2]])
+; CHECK-NEXT: ret <64 x i8> [[TMP3]]
+;
+ %1 = lshr <32 x i16> %a0, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+ %2 = and <32 x i16> %a1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+ %3 = call <64 x i8> @llvm.x86.avx512.packuswb.512(<32 x i16> %1, <32 x i16> %2)
+ ret <64 x i8> %3
+}
+
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
