[llvm] r362752 - [NFC][CodeGen] Add unary fneg tests to X86/fma4-intrinsics-x86.ll
Cameron McInally via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 6 14:49:59 PDT 2019
Author: mcinally
Date: Thu Jun 6 14:49:59 2019
New Revision: 362752
URL: http://llvm.org/viewvc/llvm-project?rev=362752&view=rev
Log:
[NFC][CodeGen] Add unary fneg tests to X86/fma4-intrinsics-x86.ll
Modified:
llvm/trunk/test/CodeGen/X86/fma4-intrinsics-x86.ll
Modified: llvm/trunk/test/CodeGen/X86/fma4-intrinsics-x86.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fma4-intrinsics-x86.ll?rev=362752&r1=362751&r2=362752&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fma4-intrinsics-x86.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fma4-intrinsics-x86.ll Thu Jun 6 14:49:59 2019
@@ -88,6 +88,16 @@ define <4 x float> @test_x86_fma_vfmsub_
ret <4 x float> %2
}
+define <4 x float> @test_x86_fma_vfmsub_ps_unary_fneg(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsub_ps_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubps %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x6c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x float> %a2
+ %2 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %1)
+ ret <4 x float> %2
+}
+
define <2 x double> @test_x86_fma_vfmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsub_pd:
; CHECK: # %bb.0:
@@ -98,6 +108,16 @@ define <2 x double> @test_x86_fma_vfmsub
ret <2 x double> %2
}
+define <2 x double> @test_x86_fma_vfmsub_pd_unary_fneg(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsub_pd_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubpd %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x6d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <2 x double> %a2
+ %2 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %1)
+ ret <2 x double> %2
+}
+
define <8 x float> @test_x86_fma_vfmsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsub_ps_256:
; CHECK: # %bb.0:
@@ -108,6 +128,16 @@ define <8 x float> @test_x86_fma_vfmsub_
ret <8 x float> %2
}
+define <8 x float> @test_x86_fma_vfmsub_ps_256_unary_fneg(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsub_ps_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubps %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x6c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <8 x float> %a2
+ %2 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %1)
+ ret <8 x float> %2
+}
+
define <4 x double> @test_x86_fma_vfmsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsub_pd_256:
; CHECK: # %bb.0:
@@ -118,6 +148,16 @@ define <4 x double> @test_x86_fma_vfmsub
ret <4 x double> %2
}
+define <4 x double> @test_x86_fma_vfmsub_pd_256_unary_fneg(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsub_pd_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubpd %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x6d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x double> %a2
+ %2 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %1)
+ ret <4 x double> %2
+}
+
; VFNMADD
define <4 x float> @test_x86_fma_vfnmadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmadd_ps:
@@ -129,6 +169,16 @@ define <4 x float> @test_x86_fma_vfnmadd
ret <4 x float> %2
}
+define <4 x float> @test_x86_fma_vfnmadd_ps_unary_fneg(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmadd_ps_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmaddps %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x78,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x float> %a0
+ %2 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> %a1, <4 x float> %a2)
+ ret <4 x float> %2
+}
+
define <2 x double> @test_x86_fma_vfnmadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmadd_pd:
; CHECK: # %bb.0:
@@ -139,6 +189,16 @@ define <2 x double> @test_x86_fma_vfnmad
ret <2 x double> %2
}
+define <2 x double> @test_x86_fma_vfnmadd_pd_unary_fneg(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmadd_pd_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmaddpd %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x79,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <2 x double> %a0
+ %2 = call <2 x double> @llvm.fma.v2f64(<2 x double> %1, <2 x double> %a1, <2 x double> %a2)
+ ret <2 x double> %2
+}
+
define <8 x float> @test_x86_fma_vfnmadd_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmadd_ps_256:
; CHECK: # %bb.0:
@@ -149,6 +209,16 @@ define <8 x float> @test_x86_fma_vfnmadd
ret <8 x float> %2
}
+define <8 x float> @test_x86_fma_vfnmadd_ps_256_unary_fneg(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmadd_ps_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmaddps %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x78,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <8 x float> %a0
+ %2 = call <8 x float> @llvm.fma.v8f32(<8 x float> %1, <8 x float> %a1, <8 x float> %a2)
+ ret <8 x float> %2
+}
+
define <4 x double> @test_x86_fma_vfnmadd_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmadd_pd_256:
; CHECK: # %bb.0:
@@ -159,6 +229,16 @@ define <4 x double> @test_x86_fma_vfnmad
ret <4 x double> %2
}
+define <4 x double> @test_x86_fma_vfnmadd_pd_256_unary_fneg(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmadd_pd_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmaddpd %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x79,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x double> %a0
+ %2 = call <4 x double> @llvm.fma.v4f64(<4 x double> %1, <4 x double> %a1, <4 x double> %a2)
+ ret <4 x double> %2
+}
+
; VFNMSUB
define <4 x float> @test_x86_fma_vfnmsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmsub_ps:
@@ -171,6 +251,17 @@ define <4 x float> @test_x86_fma_vfnmsub
ret <4 x float> %3
}
+define <4 x float> @test_x86_fma_vfnmsub_ps_unary_fneg(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmsub_ps_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmsubps %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x7c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x float> %a0
+ %2 = fneg <4 x float> %a2
+ %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %1, <4 x float> %a1, <4 x float> %2)
+ ret <4 x float> %3
+}
+
define <2 x double> @test_x86_fma_vfnmsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmsub_pd:
; CHECK: # %bb.0:
@@ -182,6 +273,17 @@ define <2 x double> @test_x86_fma_vfnmsu
ret <2 x double> %3
}
+define <2 x double> @test_x86_fma_vfnmsub_pd_unary_fneg(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmsub_pd_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmsubpd %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x7d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <2 x double> %a0
+ %2 = fneg <2 x double> %a2
+ %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %1, <2 x double> %a1, <2 x double> %2)
+ ret <2 x double> %3
+}
+
define <8 x float> @test_x86_fma_vfnmsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmsub_ps_256:
; CHECK: # %bb.0:
@@ -193,6 +295,17 @@ define <8 x float> @test_x86_fma_vfnmsub
ret <8 x float> %3
}
+define <8 x float> @test_x86_fma_vfnmsub_ps_256_unary_fneg(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmsub_ps_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmsubps %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x7c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <8 x float> %a0
+ %2 = fneg <8 x float> %a2
+ %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %1, <8 x float> %a1, <8 x float> %2)
+ ret <8 x float> %3
+}
+
define <4 x double> @test_x86_fma_vfnmsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfnmsub_pd_256:
; CHECK: # %bb.0:
@@ -204,6 +317,17 @@ define <4 x double> @test_x86_fma_vfnmsu
ret <4 x double> %3
}
+define <4 x double> @test_x86_fma_vfnmsub_pd_256_unary_fneg(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfnmsub_pd_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfnmsubpd %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x7d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = fneg <4 x double> %a0
+ %2 = fneg <4 x double> %a2
+ %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %1, <4 x double> %a1, <4 x double> %2)
+ ret <4 x double> %3
+}
+
; VFMADDSUB
define <4 x float> @test_x86_fma_vfmaddsub_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmaddsub_ps:
@@ -217,6 +341,18 @@ define <4 x float> @test_x86_fma_vfmadds
ret <4 x float> %4
}
+define <4 x float> @test_x86_fma_vfmaddsub_ps_unary_fneg(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmaddsub_ps_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmaddsubps %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x5c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
+ %2 = fneg <4 x float> %a2
+ %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %2)
+ %4 = shufflevector <4 x float> %3, <4 x float> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+ ret <4 x float> %4
+}
+
define <2 x double> @test_x86_fma_vfmaddsub_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmaddsub_pd:
; CHECK: # %bb.0:
@@ -229,6 +365,18 @@ define <2 x double> @test_x86_fma_vfmadd
ret <2 x double> %4
}
+define <2 x double> @test_x86_fma_vfmaddsub_pd_unary_fneg(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmaddsub_pd_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmaddsubpd %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x5d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
+ %2 = fneg <2 x double> %a2
+ %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %2)
+ %4 = shufflevector <2 x double> %3, <2 x double> %1, <2 x i32> <i32 0, i32 3>
+ ret <2 x double> %4
+}
+
define <8 x float> @test_x86_fma_vfmaddsub_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmaddsub_ps_256:
; CHECK: # %bb.0:
@@ -241,6 +389,18 @@ define <8 x float> @test_x86_fma_vfmadds
ret <8 x float> %4
}
+define <8 x float> @test_x86_fma_vfmaddsub_ps_256_unary_fneg(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmaddsub_ps_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmaddsubps %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x5c,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
+ %2 = fneg <8 x float> %a2
+ %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %2)
+ %4 = shufflevector <8 x float> %3, <8 x float> %1, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
+ ret <8 x float> %4
+}
+
define <4 x double> @test_x86_fma_vfmaddsub_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmaddsub_pd_256:
; CHECK: # %bb.0:
@@ -253,6 +413,18 @@ define <4 x double> @test_x86_fma_vfmadd
ret <4 x double> %4
}
+define <4 x double> @test_x86_fma_vfmaddsub_pd_256_unary_fneg(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmaddsub_pd_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmaddsubpd %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x5d,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
+ %2 = fneg <4 x double> %a2
+ %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %2)
+ %4 = shufflevector <4 x double> %3, <4 x double> %1, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+ ret <4 x double> %4
+}
+
; VFMSUBADD
define <4 x float> @test_x86_fma_vfmsubadd_ps(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsubadd_ps:
@@ -266,6 +438,18 @@ define <4 x float> @test_x86_fma_vfmsuba
ret <4 x float> %4
}
+define <4 x float> @test_x86_fma_vfmsubadd_ps_unary_fneg(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsubadd_ps_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubaddps %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x5e,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %a2)
+ %2 = fneg <4 x float> %a2
+ %3 = call <4 x float> @llvm.fma.v4f32(<4 x float> %a0, <4 x float> %a1, <4 x float> %2)
+ %4 = shufflevector <4 x float> %1, <4 x float> %3, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+ ret <4 x float> %4
+}
+
define <2 x double> @test_x86_fma_vfmsubadd_pd(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsubadd_pd:
; CHECK: # %bb.0:
@@ -278,6 +462,19 @@ define <2 x double> @test_x86_fma_vfmsub
ret <2 x double> %4
}
+define <2 x double> @test_x86_fma_vfmsubadd_pd_unary_fneg(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsubadd_pd_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubaddpd %xmm2, %xmm1, %xmm0, %xmm0 # encoding: [0xc4,0xe3,0xf9,0x5f,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %a2)
+ %2 = fneg <2 x double> %a2
+ %3 = call <2 x double> @llvm.fma.v2f64(<2 x double> %a0, <2 x double> %a1, <2 x double> %2)
+ %4 = shufflevector <2 x double> %1, <2 x double> %3, <2 x i32> <i32 0, i32 3>
+ ret <2 x double> %4
+}
+
+
define <8 x float> @test_x86_fma_vfmsubadd_ps_256(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsubadd_ps_256:
; CHECK: # %bb.0:
@@ -290,6 +487,18 @@ define <8 x float> @test_x86_fma_vfmsuba
ret <8 x float> %4
}
+define <8 x float> @test_x86_fma_vfmsubadd_ps_256_unary_fneg(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsubadd_ps_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubaddps %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x5e,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %a2)
+ %2 = fneg <8 x float> %a2
+ %3 = call <8 x float> @llvm.fma.v8f32(<8 x float> %a0, <8 x float> %a1, <8 x float> %2)
+ %4 = shufflevector <8 x float> %1, <8 x float> %3, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15>
+ ret <8 x float> %4
+}
+
define <4 x double> @test_x86_fma_vfmsubadd_pd_256(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
; CHECK-LABEL: test_x86_fma_vfmsubadd_pd_256:
; CHECK: # %bb.0:
@@ -300,6 +509,18 @@ define <4 x double> @test_x86_fma_vfmsub
%3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %2)
%4 = shufflevector <4 x double> %1, <4 x double> %3, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
ret <4 x double> %4
+}
+
+define <4 x double> @test_x86_fma_vfmsubadd_pd_256_unary_fneg(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2) #0 {
+; CHECK-LABEL: test_x86_fma_vfmsubadd_pd_256_unary_fneg:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vfmsubaddpd %ymm2, %ymm1, %ymm0, %ymm0 # encoding: [0xc4,0xe3,0xfd,0x5f,0xc2,0x10]
+; CHECK-NEXT: retq # encoding: [0xc3]
+ %1 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %a2)
+ %2 = fneg <4 x double> %a2
+ %3 = call <4 x double> @llvm.fma.v4f64(<4 x double> %a0, <4 x double> %a1, <4 x double> %2)
+ %4 = shufflevector <4 x double> %1, <4 x double> %3, <4 x i32> <i32 0, i32 5, i32 2, i32 7>
+ ret <4 x double> %4
}
declare <4 x float> @llvm.fma.v4f32(<4 x float>, <4 x float>, <4 x float>) #2
More information about the llvm-commits
mailing list