1 files changed, 161 insertions, 2 deletions
diff --git a/llvm/test/CodeGen/X86/compress_expand.ll b/llvm/test/CodeGen/X86/compress_expand.ll
index 555d998eddb..1e34c1aa098 100644
--- a/llvm/test/CodeGen/X86/compress_expand.ll
+++ b/llvm/test/CodeGen/X86/compress_expand.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mattr=+avx512vl,+avx512dq,+avx512bw < %s | FileCheck %s --check-prefix=ALL --check-prefix=SKX
-; RUN: llc -mattr=+avx512f < %s | FileCheck %s --check-prefix=ALL --check-prefix=KNL
+; RUN: llc -mcpu=skylake-avx512 < %s | FileCheck %s --check-prefix=ALL --check-prefix=SKX
+; RUN: llc -mcpu=knl < %s | FileCheck %s --check-prefix=ALL --check-prefix=KNL
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
@@ -235,6 +235,157 @@ define void @test12(float* %base, <4 x float> %V, <4 x i1> %mask) {
   ret void
 }
 
+define <2 x float> @test13(float* %base, <2 x float> %src0, <2 x i32> %trigger) {
+; SKX-LABEL: test13:
+; SKX:       # BB#0:
+; SKX-NEXT:    vpxord %xmm2, %xmm2, %xmm2
+; SKX-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; SKX-NEXT:    vpcmpeqq %xmm2, %xmm1, %k0
+; SKX-NEXT:    kshiftlb $6, %k0, %k0
+; SKX-NEXT:    kshiftrb $6, %k0, %k1
+; SKX-NEXT:    vexpandps (%rdi), %xmm0 {%k1}
+; SKX-NEXT:    retq
+;
+; KNL-LABEL: test13:
+; KNL:       # BB#0:
+; KNL-NEXT:    # kill: %XMM0<def> %XMM0<kill> %ZMM0<def>
+; KNL-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; KNL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; KNL-NEXT:    vpcmpeqq %xmm2, %xmm1, %xmm1
+; KNL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; KNL-NEXT:    vmovq {{.*#+}} xmm1 = xmm1[0],zero
+; KNL-NEXT:    vpxord %zmm2, %zmm2, %zmm2
+; KNL-NEXT:    vinserti32x4 $0, %xmm1, %zmm2, %zmm1
+; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vexpandps (%rdi), %zmm0 {%k1}
+; KNL-NEXT:    # kill: %XMM0<def> %XMM0<kill> %ZMM0<kill>
+; KNL-NEXT:    retq
+  %mask = icmp eq <2 x i32> %trigger, zeroinitializer
+  %res = call <2 x float> @llvm.masked.expandload.v2f32(float* %base, <2 x i1> %mask, <2 x float> %src0)
+  ret <2 x float> %res
+}
+
+define void @test14(float* %base, <2 x float> %V, <2 x i32> %trigger) {
+; SKX-LABEL: test14:
+; SKX:       # BB#0:
+; SKX-NEXT:    vpxord %xmm2, %xmm2, %xmm2
+; SKX-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; SKX-NEXT:    vpcmpeqq %xmm2, %xmm1, %k0
+; SKX-NEXT:    kshiftlb $6, %k0, %k0
+; SKX-NEXT:    kshiftrb $6, %k0, %k1
+; SKX-NEXT:    vcompressps %xmm0, (%rdi) {%k1}
+; SKX-NEXT:    retq
+;
+; KNL-LABEL: test14:
+; KNL:       # BB#0:
+; KNL-NEXT:    # kill: %XMM0<def> %XMM0<kill> %ZMM0<def>
+; KNL-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; KNL-NEXT:    vpblendd {{.*#+}} xmm1 = xmm1[0],xmm2[1],xmm1[2],xmm2[3]
+; KNL-NEXT:    vpcmpeqq %xmm2, %xmm1, %xmm1
+; KNL-NEXT:    vpshufd {{.*#+}} xmm1 = xmm1[0,2,2,3]
+; KNL-NEXT:    vmovq {{.*#+}} xmm1 = xmm1[0],zero
+; KNL-NEXT:    vpxord %zmm2, %zmm2, %zmm2
+; KNL-NEXT:    vinserti32x4 $0, %xmm1, %zmm2, %zmm1
+; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
+; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1
+; KNL-NEXT:    vcompressps %zmm0, (%rdi) {%k1}
+; KNL-NEXT:    retq
+  %mask = icmp eq <2 x i32> %trigger, zeroinitializer
+  call void @llvm.masked.compressstore.v2f32(<2 x float> %V, float* %base, <2 x i1> %mask)
+  ret void
+}
+
+define <32 x float> @test15(float* %base, <32 x float> %src0, <32 x i32> %trigger) {
+; ALL-LABEL: test15:
+; ALL:       # BB#0:
+; ALL-NEXT:    vpxord %zmm4, %zmm4, %zmm4
+; ALL-NEXT:    vpcmpeqd %zmm4, %zmm3, %k1
+; ALL-NEXT:    vpcmpeqd %zmm4, %zmm2, %k2
+; ALL-NEXT:    kmovw %k2, %eax
+; ALL-NEXT:    popcntl %eax, %eax
+; ALL-NEXT:    vexpandps (%rdi,%rax,4), %zmm1 {%k1}
+; ALL-NEXT:    vexpandps (%rdi), %zmm0 {%k2}
+; ALL-NEXT:    retq
+  %mask = icmp eq <32 x i32> %trigger, zeroinitializer
+  %res = call <32 x float> @llvm.masked.expandload.v32f32(float* %base, <32 x i1> %mask, <32 x float> %src0)
+  ret <32 x float> %res
+}
+
+define <16 x double> @test16(double* %base, <16 x double> %src0, <16 x i32> %trigger) {
+; SKX-LABEL: test16:
+; SKX:       # BB#0:
+; SKX-NEXT:    vextracti32x8 $1, %zmm2, %ymm3
+; SKX-NEXT:    vpxord %ymm4, %ymm4, %ymm4
+; SKX-NEXT:    vpcmpeqd %ymm4, %ymm3, %k1
+; SKX-NEXT:    vpcmpeqd %ymm4, %ymm2, %k2
+; SKX-NEXT:    kmovb %k2, %eax
+; SKX-NEXT:    popcntl %eax, %eax
+; SKX-NEXT:    vexpandpd (%rdi,%rax,8), %zmm1 {%k1}
+; SKX-NEXT:    vexpandpd (%rdi), %zmm0 {%k2}
+; SKX-NEXT:    retq
+;
+; KNL-LABEL: test16:
+; KNL:       # BB#0:
+; KNL-NEXT:    vpxor %ymm3, %ymm3, %ymm3
+; KNL-NEXT:    vextracti64x4 $1, %zmm2, %ymm4
+; KNL-NEXT:    vpcmpeqd %zmm3, %zmm4, %k1
+; KNL-NEXT:    vpcmpeqd %zmm3, %zmm2, %k2
+; KNL-NEXT:    vexpandpd (%rdi), %zmm0 {%k2}
+; KNL-NEXT:    kmovw %k2, %eax
+; KNL-NEXT:    movzbl %al, %eax
+; KNL-NEXT:    popcntl %eax, %eax
+; KNL-NEXT:    vexpandpd (%rdi,%rax,8), %zmm1 {%k1}
+; KNL-NEXT:    retq
+  %mask = icmp eq <16 x i32> %trigger, zeroinitializer
+  %res = call <16 x double> @llvm.masked.expandload.v16f64(double* %base, <16 x i1> %mask, <16 x double> %src0)
+  ret <16 x double> %res
+}
+
+define void @test17(float* %base, <32 x float> %V, <32 x i32> %trigger) {
+; ALL-LABEL: test17:
+; ALL:       # BB#0:
+; ALL-NEXT:    vpxord %zmm4, %zmm4, %zmm4
+; ALL-NEXT:    vpcmpeqd %zmm4, %zmm3, %k1
+; ALL-NEXT:    vpcmpeqd %zmm4, %zmm2, %k2
+; ALL-NEXT:    kmovw %k2, %eax
+; ALL-NEXT:    popcntl %eax, %eax
+; ALL-NEXT:    vcompressps %zmm1, (%rdi,%rax,4) {%k1}
+; ALL-NEXT:    vcompressps %zmm0, (%rdi) {%k2}
+; ALL-NEXT:    retq
+  %mask = icmp eq <32 x i32> %trigger, zeroinitializer
+  call void @llvm.masked.compressstore.v32f32(<32 x float> %V, float* %base, <32 x i1> %mask)
+  ret void
+}
+
+define void @test18(double* %base, <16 x double> %V, <16 x i1> %mask) {
+; SKX-LABEL: test18:
+; SKX:       # BB#0:
+; SKX-NEXT:    vpsllw $7, %xmm2, %xmm2
+; SKX-NEXT:    vpmovb2m %xmm2, %k1
+; SKX-NEXT:    kshiftrw $8, %k1, %k2
+; SKX-NEXT:    kmovb %k1, %eax
+; SKX-NEXT:    popcntl %eax, %eax
+; SKX-NEXT:    vcompresspd %zmm1, (%rdi,%rax,8) {%k2}
+; SKX-NEXT:    vcompresspd %zmm0, (%rdi) {%k1}
+; SKX-NEXT:    retq
+;
+; KNL-LABEL: test18:
+; KNL:       # BB#0:
+; KNL-NEXT:    vpmovsxbd %xmm2, %zmm2
+; KNL-NEXT:    vpslld $31, %zmm2, %zmm2
+; KNL-NEXT:    vptestmd %zmm2, %zmm2, %k1
+; KNL-NEXT:    kshiftrw $8, %k1, %k2
+; KNL-NEXT:    kmovw %k1, %eax
+; KNL-NEXT:    movzbl %al, %eax
+; KNL-NEXT:    popcntl %eax, %eax
+; KNL-NEXT:    vcompresspd %zmm1, (%rdi,%rax,8) {%k2}
+; KNL-NEXT:    vcompresspd %zmm0, (%rdi) {%k1}
+; KNL-NEXT:    retq
+  call void @llvm.masked.compressstore.v16f64(<16 x double> %V, double* %base, <16 x i1> %mask)
+  ret void
+}
+
 declare void @llvm.masked.compressstore.v16f32(<16 x float>, float* , <16 x i1>)
 declare void @llvm.masked.compressstore.v8f32(<8 x float>, float* , <8 x i1>)
 declare void @llvm.masked.compressstore.v8f64(<8 x double>, double* , <8 x i1>)
@@ -245,3 +396,11 @@ declare void @llvm.masked.compressstore.v4i32(<4 x i32>, i32* , <4 x i1>)
 declare void @llvm.masked.compressstore.v4f32(<4 x float>, float* , <4 x i1>)
 declare void @llvm.masked.compressstore.v4i64(<4 x i64>, i64* , <4 x i1>)
 declare void @llvm.masked.compressstore.v2i64(<2 x i64>, i64* , <2 x i1>)
+declare void @llvm.masked.compressstore.v2f32(<2 x float>, float* , <2 x i1>)
+declare void @llvm.masked.compressstore.v32f32(<32 x float>, float* , <32 x i1>)
+declare void @llvm.masked.compressstore.v16f64(<16 x double>, double* , <16 x i1>)
+declare void @llvm.masked.compressstore.v32f64(<32 x double>, double* , <32 x i1>)
+
+declare <2 x float> @llvm.masked.expandload.v2f32(float* , <2 x i1> , <2 x float> )
+declare <32 x float> @llvm.masked.expandload.v32f32(float* , <32 x i1> , <32 x float> )
+declare <16 x double> @llvm.masked.expandload.v16f64(double* , <16 x i1> , <16 x double> )