Add tests for combining mask with shuffles; NFC

Differential Revision: https://reviews.llvm.org/D144144

diff --git a/llvm/test/CodeGen/X86/combine-mask-with-shuffle.ll b/llvm/test/CodeGen/X86/combine-mask-with-shuffle.ll
new file mode 100644 (file)
index 0000000..e105d4e
--- /dev/null
+++ b/llvm/test/CodeGen/X86/combine-mask-with-shuffle.ll
@@ -0,0 +1,229 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512vpopcntdq,+avx512f,+avx512vl | FileCheck %s --check-prefixes=CHECK
+
+
+define <16 x i32> @combine_mask_with_or(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_or:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $2570, %ax # imm = 0xA0A
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpord %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = or <16 x i32> %shuf0_0, %shuf0_1
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_mul(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_mul:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $2570, %ax # imm = 0xA0A
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpmulld %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = mul <16 x i32> %shuf0_0, %shuf0_1
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_abs(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_abs:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    vpabsd %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0_tmp0 = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> %shuf0_0, i1 true)
+  %op0_0_tmp1 = tail call <16 x i32> @llvm.abs.v16i32(<16 x i32> %shuf0_1, i1 true)
+  %op0_0 = shufflevector <16 x i32> %op0_0_tmp0, <16 x i32> %op0_0_tmp0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_umin(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_umin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $-21846, %ax # imm = 0xAAAA
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpminud %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = tail call <16 x i32> @llvm.umin.v16i32(<16 x i32> %shuf0_0, <16 x i32> %shuf0_1)
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_umax(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_umax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $-21846, %ax # imm = 0xAAAA
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpmaxud %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = tail call <16 x i32> @llvm.umax.v16i32(<16 x i32> %shuf0_0, <16 x i32> %shuf0_1)
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_smin(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_smin:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $-21846, %ax # imm = 0xAAAA
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpminsd %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = tail call <16 x i32> @llvm.smin.v16i32(<16 x i32> %shuf0_0, <16 x i32> %shuf0_1)
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_smax(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_smax:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $-21846, %ax # imm = 0xAAAA
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpmaxsd %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = tail call <16 x i32> @llvm.smax.v16i32(<16 x i32> %shuf0_0, <16 x i32> %shuf0_1)
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_shl(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_shl:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $2570, %ax # imm = 0xA0A
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpsllvd %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = shl <16 x i32> %shuf0_0, %shuf0_1
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_ashr(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_ashr:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $2570, %ax # imm = 0xA0A
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpsravd %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = ashr <16 x i32> %shuf0_0, %shuf0_1
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+define <16 x i32> @combine_mask_with_lshr(<16 x i32> %v0) {
+; CHECK-LABEL: combine_mask_with_lshr:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vpshufd {{.*#+}} zmm1 = zmm0[1,0,3,2,5,4,7,6,9,8,11,10,13,12,15,14]
+; CHECK-NEXT:    movw $2570, %ax # imm = 0xA0A
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vpblendmd %zmm1, %zmm0, %zmm2 {%k1}
+; CHECK-NEXT:    vpsrlvd %zmm2, %zmm1, %zmm1
+; CHECK-NEXT:    vpopcntd %zmm0, %zmm0
+; CHECK-NEXT:    movb $-52, %al
+; CHECK-NEXT:    kmovw %eax, %k1
+; CHECK-NEXT:    vmovdqa64 %zmm0, %zmm1 {%k1}
+; CHECK-NEXT:    vmovdqa64 %zmm1, %zmm0
+; CHECK-NEXT:    retq
+  %shuf0_0 = shufflevector <16 x i32> %v0, <16 x i32> poison, <16 x i32> <i32 1, i32 0, i32 3, i32 2, i32 5, i32 4, i32 7, i32 6, i32 9, i32 8, i32 11, i32 10, i32 13, i32 12, i32 15, i32 14>
+  %shuf0_1 = shufflevector <16 x i32> %v0, <16 x i32> %shuf0_0, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31>
+  %op0_0 = lshr <16 x i32> %shuf0_0, %shuf0_1
+  %op1_0 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %v0)
+  %r = shufflevector <16 x i32> %op0_0, <16 x i32> %op1_0, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31>
+  ret <16 x i32> %r
+}
+
+declare <16 x i32> @llvm.smin.v16i32(<16 x i32>, <16 x i32>)
+declare <16 x i32> @llvm.smax.v16i32(<16 x i32>, <16 x i32>)
+declare <16 x i32> @llvm.umin.v16i32(<16 x i32>, <16 x i32>)
+declare <16 x i32> @llvm.umax.v16i32(<16 x i32>, <16 x i32>)
+declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>)
+declare <16 x i32> @llvm.abs.v16i32(<16 x i32>, i1)