[AVX-512] Add additional test cases for broadcast intrinsics that demonstates that we don't fold the loads to use a broadcast instruction.

llvm-svn: 292465

4 files changed, 158 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 3015a2b499f..2ed27dbff39 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -4155,6 +4155,18 @@ define <16 x float>@test_int_x86_avx512_mask_broadcastf32x4_512(<4 x float> %x0,
   ret <16 x float> %res5
 }
 
+define <16 x float>@test_int_x86_avx512_mask_broadcastf32x4_512_load(<4 x float>* %x0ptr, <16 x float> %x2, i16 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf32x4_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovaps (%rdi), %xmm1
+; CHECK-NEXT:    vshuff32x4 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq
+  %x0 = load <4 x float>, <4 x float>* %x0ptr
+  %res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x4.512(<4 x float> %x0, <16 x float> %x2, i16 %mask)
+  ret <16 x float> %res
+}
+
 declare <8 x double> @llvm.x86.avx512.mask.broadcastf64x4.512(<4 x double>, <8 x double>, i8)
 
 define <8 x double>@test_int_x86_avx512_mask_broadcastf64x4_512(<4 x double> %x0, <8 x double> %x2, i8 %mask) {
@@ -4177,6 +4189,19 @@ define <8 x double>@test_int_x86_avx512_mask_broadcastf64x4_512(<4 x double> %x0
   ret <8 x double> %res5
 }
 
+define <8 x double>@test_int_x86_avx512_mask_broadcastf64x4_512_load(<4 x double>* %x0ptr, <8 x double> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf64x4_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovapd (%rdi), %ymm1
+; CHECK-NEXT:    vshuff64x2 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq
+
+  %x0 = load <4 x double>, <4 x double>* %x0ptr
+  %res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x4.512(<4 x double> %x0, <8 x double> %x2, i8 %mask)
+  ret <8 x double> %res
+}
+
 declare <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32>, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x4_512(<4 x i32> %x0, <16 x i32> %x2, i16 %mask) {
@@ -4199,6 +4224,19 @@ define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x4_512(<4 x i32> %x0, <16
   ret <16 x i32> %res5
 }
 
+define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x4_512_load(<4 x i32>* %x0ptr, <16 x i32> %x2, i16 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti32x4_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovdqa (%rdi), %xmm1
+; CHECK-NEXT:    vshufi32x4 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,0,1,2,3,0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq
+
+  %x0 = load <4 x i32>, <4 x i32>* %x0ptr
+  %res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x4.512(<4 x i32> %x0, <16 x i32> %x2, i16 %mask)
+  ret <16 x i32> %res
+}
+
 declare <8 x i64> @llvm.x86.avx512.mask.broadcasti64x4.512(<4 x i64>, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x4_512(<4 x i64> %x0, <8 x i64> %x2, i8 %mask) {
@@ -4221,6 +4259,19 @@ define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x4_512(<4 x i64> %x0, <8 x
   ret <8 x i64> %res5
 }
 
+define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x4_512_load(<4 x i64>* %x0ptr, <8 x i64> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti64x4_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovdqa (%rdi), %ymm1
+; CHECK-NEXT:    vshufi64x2 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq
+
+  %x0 = load <4 x i64>, <4 x i64>* %x0ptr
+  %res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x4.512(<4 x i64> %x0, <8 x i64> %x2, i8 %mask)
+  ret <8 x i64> %res
+}
+
 declare <16 x i32> @llvm.x86.avx512.mask.prorv.d.512(<16 x i32>, <16 x i32>, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_prorv_d_512(<16 x i32> %x0, <16 x i32> %x1, <16 x i32> %x2, i16 %x3) {
diff --git a/llvm/test/CodeGen/X86/avx512dq-intrinsics.ll b/llvm/test/CodeGen/X86/avx512dq-intrinsics.ll
index 375d6326451..d355c92061b 100644
--- a/llvm/test/CodeGen/X86/avx512dq-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512dq-intrinsics.ll
@@ -515,6 +515,19 @@ define <16 x float>@test_int_x86_avx512_mask_broadcastf32x8_512(<8 x float> %x0,
   ret <16 x float> %res5
 }
 
+define <16 x float>@test_int_x86_avx512_mask_broadcastf32x8_512_load(<8 x float>* %x0ptr, <16 x float> %x2, i16 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf32x8_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovaps (%rdi), %ymm1
+; CHECK-NEXT:    vshuff32x4 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; CHECK-NEXT:    retq
+
+  %x0 = load <8 x float>, <8 x float>* %x0ptr
+  %res = call <16 x float> @llvm.x86.avx512.mask.broadcastf32x8.512(<8 x float> %x0, <16 x float> %x2, i16 %mask)
+  ret <16 x float> %res
+}
+
 declare <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double>, <8 x double>, i8)
 
 define <8 x double>@test_int_x86_avx512_mask_broadcastf64x2_512(<2 x double> %x0, <8 x double> %x2, i8 %mask) {
@@ -537,6 +550,19 @@ define <8 x double>@test_int_x86_avx512_mask_broadcastf64x2_512(<2 x double> %x0
   ret <8 x double> %res5
 }
 
+define <8 x double>@test_int_x86_avx512_mask_broadcastf64x2_512_load(<2 x double>* %x0ptr, <8 x double> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf64x2_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovb %esi, %k1
+; CHECK-NEXT:    vmovapd (%rdi), %xmm1
+; CHECK-NEXT:    vshuff64x2 {{.*#+}} zmm0 {%k1} = zmm1[0,1,0,1,0,1,0,1]
+; CHECK-NEXT:    retq
+
+  %x0 = load <2 x double>, <2 x double>* %x0ptr
+  %res = call <8 x double> @llvm.x86.avx512.mask.broadcastf64x2.512(<2 x double> %x0, <8 x double> %x2, i8 %mask)
+  ret <8 x double> %res
+}
+
 declare <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32>, <16 x i32>, i16)
 
 define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x8_512(<8 x i32> %x0, <16 x i32> %x2, i16 %mask) {
@@ -559,6 +585,19 @@ define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x8_512(<8 x i32> %x0, <16
   ret <16 x i32> %res5
 }
 
+define <16 x i32>@test_int_x86_avx512_mask_broadcasti32x8_512_load(<8 x i32>* %x0ptr, <16 x i32> %x2, i16 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti32x8_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1
+; CHECK-NEXT:    vmovdqa (%rdi), %ymm1
+; CHECK-NEXT:    vshufi32x4 {{.*#+}} zmm0 {%k1} = zmm1[0,1,2,3,4,5,6,7,0,1,2,3,4,5,6,7]
+; CHECK-NEXT:    retq
+
+  %x0 = load <8 x i32>, <8 x i32>* %x0ptr
+  %res = call <16 x i32> @llvm.x86.avx512.mask.broadcasti32x8.512(<8 x i32> %x0, <16 x i32> %x2, i16 %mask)
+  ret <16 x i32> %res
+}
+
 declare <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64>, <8 x i64>, i8)
 
 define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x2_512(<2 x i64> %x0, <8 x i64> %x2, i8 %mask) {
@@ -580,3 +619,16 @@ define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x2_512(<2 x i64> %x0, <8 x
   %res5 = add <8 x i64> %res3, %res4
   ret <8 x i64> %res5
 }
+
+define <8 x i64>@test_int_x86_avx512_mask_broadcasti64x2_512_load(<2 x i64>* %x0ptr, <8 x i64> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti64x2_512_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovb %esi, %k1
+; CHECK-NEXT:    vmovdqa (%rdi), %xmm1
+; CHECK-NEXT:    vshufi64x2 {{.*#+}} zmm0 {%k1} = zmm1[0,1,0,1,0,1,0,1]
+; CHECK-NEXT:    retq
+
+  %x0 = load <2 x i64>, <2 x i64>* %x0ptr
+  %res = call <8 x i64> @llvm.x86.avx512.mask.broadcasti64x2.512(<2 x i64> %x0, <8 x i64> %x2, i8 %mask)
+  ret <8 x i64> %res
+}
diff --git a/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
index 3430c571537..0bf7f7bf3d3 100644
--- a/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512dqvl-intrinsics.ll
@@ -785,6 +785,7 @@ define <4 x i64>@test_int_x86_avx512_cvtmask2q_256(i8 %x0) {
   %res = call <4 x i64> @llvm.x86.avx512.cvtmask2q.256(i8 %x0)
   ret <4 x i64> %res
 }
+
 declare <4 x double> @llvm.x86.avx512.mask.broadcastf64x2.256(<2 x double>, <4 x double>, i8)
 
 define <4 x double>@test_int_x86_avx512_mask_broadcastf64x2_256(<2 x double> %x0, <4 x double> %x2, i8 %mask) {
@@ -810,6 +811,20 @@ define <4 x double>@test_int_x86_avx512_mask_broadcastf64x2_256(<2 x double> %x0
   ret <4 x double> %res5
 }
 
+define <4 x double>@test_int_x86_avx512_mask_broadcastf64x2_256_load(<2 x double>* %x0ptr, <4 x double> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf64x2_256_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovb %esi, %k1 ## encoding: [0xc5,0xf9,0x92,0xce]
+; CHECK-NEXT:    vmovapd (%rdi), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x28,0x0f]
+; CHECK-NEXT:    vshuff64x2 $0, %ymm1, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x23,0xc1,0x00]
+; CHECK-NEXT:    ## ymm0 {%k1} = ymm1[0,1,0,1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+
+  %x0 = load <2 x double>, <2 x double>* %x0ptr
+  %res = call <4 x double> @llvm.x86.avx512.mask.broadcastf64x2.256(<2 x double> %x0, <4 x double> %x2, i8 %mask)
+  ret <4 x double> %res
+}
+
 declare <4 x i64> @llvm.x86.avx512.mask.broadcasti64x2.256(<2 x i64>, <4 x i64>, i8)
 
 define <4 x i64>@test_int_x86_avx512_mask_broadcasti64x2_256(<2 x i64> %x0, <4 x i64> %x2, i8 %mask) {
@@ -834,3 +849,17 @@ define <4 x i64>@test_int_x86_avx512_mask_broadcasti64x2_256(<2 x i64> %x0, <4 x
   %res5 = add <4 x i64> %res3, %res4
   ret <4 x i64> %res5
 }
+
+define <4 x i64>@test_int_x86_avx512_mask_broadcasti64x2_256_load(<2 x i64>* %x0ptr, <4 x i64> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti64x2_256_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovb %esi, %k1 ## encoding: [0xc5,0xf9,0x92,0xce]
+; CHECK-NEXT:    vmovdqa (%rdi), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x0f]
+; CHECK-NEXT:    vshufi64x2 $0, %ymm1, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf3,0xf5,0x29,0x43,0xc1,0x00]
+; CHECK-NEXT:    ## ymm0 {%k1} = ymm1[0,1,0,1]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+
+  %x0 = load <2 x i64>, <2 x i64>* %x0ptr
+  %res = call <4 x i64> @llvm.x86.avx512.mask.broadcasti64x2.256(<2 x i64> %x0, <4 x i64> %x2, i8 %mask)
+  ret <4 x i64> %res
+}
diff --git a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
index 82014283246..8a2daf86859 100644
--- a/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512vl-intrinsics.ll
@@ -4206,6 +4206,19 @@ define <8 x float>@test_int_x86_avx512_mask_broadcastf32x4_256(<4 x float> %x0,
   ret <8 x float> %res5
 }
 
+define <8 x float>@test_int_x86_avx512_mask_broadcastf32x4_256_load(<4 x float>* %x0ptr, <8 x float> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcastf32x4_256_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vmovaps (%rdi), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf8,0x28,0x0f]
+; CHECK-NEXT:    vshuff32x4 $0, %ymm1, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x23,0xc1,0x00]
+; CHECK-NEXT:    ## ymm0 {%k1} = ymm1[0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %x0 = load <4 x float>, <4 x float>* %x0ptr
+  %res = call <8 x float> @llvm.x86.avx512.mask.broadcastf32x4.256(<4 x float> %x0, <8 x float> %x2, i8 %mask)
+  ret <8 x float> %res
+}
+
 declare <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32>, <8 x i32>, i8)
 
 define <8 x i32>@test_int_x86_avx512_mask_broadcasti32x4_256(<4 x i32> %x0, <8 x i32> %x2, i8 %mask) {
@@ -4230,6 +4243,19 @@ define <8 x i32>@test_int_x86_avx512_mask_broadcasti32x4_256(<4 x i32> %x0, <8 x
   ret <8 x i32> %res5
 }
 
+define <8 x i32>@test_int_x86_avx512_mask_broadcasti32x4_256_load(<4 x i32>* %x0ptr, <8 x i32> %x2, i8 %mask) {
+; CHECK-LABEL: test_int_x86_avx512_mask_broadcasti32x4_256_load:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    kmovw %esi, %k1 ## encoding: [0xc5,0xf8,0x92,0xce]
+; CHECK-NEXT:    vmovdqa (%rdi), %xmm1 ## EVEX TO VEX Compression encoding: [0xc5,0xf9,0x6f,0x0f]
+; CHECK-NEXT:    vshufi32x4 $0, %ymm1, %ymm1, %ymm0 {%k1} ## encoding: [0x62,0xf3,0x75,0x29,0x43,0xc1,0x00]
+; CHECK-NEXT:    ## ymm0 {%k1} = ymm1[0,1,2,3,0,1,2,3]
+; CHECK-NEXT:    retq ## encoding: [0xc3]
+  %x0 = load <4 x i32>, <4 x i32>* %x0ptr
+  %res = call <8 x i32> @llvm.x86.avx512.mask.broadcasti32x4.256(<4 x i32> %x0, <8 x i32> %x2, i8 %mask)
+  ret <8 x i32> %res
+}
+
 declare <4 x i32> @llvm.x86.avx512.mask.prorv.d.128(<4 x i32>, <4 x i32>, <4 x i32>, i8)
 
 define <4 x i32>@test_int_x86_avx512_mask_prorv_d_128(<4 x i32> %x0, <4 x i32> %x1, <4 x i32> %x2, i8 %x3) {