[llvm] r363084 - [NFC][CodeGen] Add unary fneg tests to X86/fma-fneg-combine.ll
Cameron McInally via llvm-commits
llvm-commits at lists.llvm.org
Tue Jun 11 10:05:36 PDT 2019
Author: mcinally
Date: Tue Jun 11 10:05:36 2019
New Revision: 363084
URL: http://llvm.org/viewvc/llvm-project?rev=363084&view=rev
Log:
[NFC][CodeGen] Add unary fneg tests to X86/fma-fneg-combine.ll
Modified:
llvm/trunk/test/CodeGen/X86/fma-fneg-combine.ll
Modified: llvm/trunk/test/CodeGen/X86/fma-fneg-combine.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fma-fneg-combine.ll?rev=363084&r1=363083&r2=363084&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fma-fneg-combine.ll (original)
+++ llvm/trunk/test/CodeGen/X86/fma-fneg-combine.ll Tue Jun 11 10:05:36 2019
@@ -16,6 +16,17 @@ entry:
ret <16 x float> %0
}
+define <16 x float> @test1_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test1_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmsub213ps {{.*#+}} zmm0 = (zmm1 * zmm0) - zmm2
+; CHECK-NEXT: retq
+entry:
+ %neg.i = fneg <16 x float> %c
+ %0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %neg.i, i32 4) #2
+ ret <16 x float> %0
+}
+
declare <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float>, <16 x float>, <16 x float>, i32)
declare <16 x float> @llvm.x86.avx512.mask.vfnmadd.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32)
declare <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float>, <16 x float>, <16 x float>, i16, i32)
@@ -32,6 +43,17 @@ entry:
ret <16 x float> %sub.i
}
+define <16 x float> @test2_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test2_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfnmsub213ps {{.*#+}} zmm0 = -(zmm1 * zmm0) - zmm2
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i32 4) #2
+ %neg.i = fneg <16 x float> %0
+ ret <16 x float> %neg.i
+}
+
define <16 x float> @test3(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
; CHECK-LABEL: test3:
; CHECK: # %bb.0: # %entry
@@ -43,6 +65,17 @@ entry:
ret <16 x float> %sub.i
}
+define <16 x float> @test3_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test3_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmsub213ps {{.*#+}} zmm0 = (zmm1 * zmm0) - zmm2
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <16 x float> @llvm.x86.avx512.mask.vfnmadd.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 -1, i32 4) #2
+ %neg.i = fneg <16 x float> %0
+ ret <16 x float> %neg.i
+}
+
define <16 x float> @test4(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
; CHECK-LABEL: test4:
; CHECK: # %bb.0: # %entry
@@ -54,6 +87,17 @@ entry:
ret <16 x float> %sub.i
}
+define <16 x float> @test4_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test4_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmadd213ps {{.*#+}} zmm0 = (zmm1 * zmm0) + zmm2
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 -1, i32 4) #2
+ %neg.i = fneg <16 x float> %0
+ ret <16 x float> %neg.i
+}
+
define <16 x float> @test5(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
; CHECK-LABEL: test5:
; CHECK: # %bb.0: # %entry
@@ -65,6 +109,17 @@ entry:
ret <16 x float> %0
}
+define <16 x float> @test5_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test5_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmsub213ps {ru-sae}, %zmm2, %zmm1, %zmm0
+; CHECK-NEXT: retq
+entry:
+ %neg.i = fneg <16 x float> %c
+ %0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %neg.i, i32 10) #2
+ ret <16 x float> %0
+}
+
define <16 x float> @test6(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
; CHECK-LABEL: test6:
; CHECK: # %bb.0: # %entry
@@ -76,6 +131,16 @@ entry:
ret <16 x float> %sub.i
}
+define <16 x float> @test6_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test6_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmadd213ps {ru-sae}, %zmm2, %zmm1, %zmm0
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 -1, i32 10) #2
+ %neg.i = fneg <16 x float> %0
+ ret <16 x float> %neg.i
+}
define <8 x float> @test7(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-LABEL: test7:
@@ -88,6 +153,17 @@ entry:
ret <8 x float> %sub.i
}
+define <8 x float> @test7_unary_fneg(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: test7_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfnmadd213ps {{.*#+}} ymm0 = -(ymm1 * ymm0) + ymm2
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %c) #2
+ %neg.i = fneg <8 x float> %0
+ ret <8 x float> %neg.i
+}
+
define <8 x float> @test8(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
; CHECK-LABEL: test8:
; CHECK: # %bb.0: # %entry
@@ -99,6 +175,17 @@ entry:
ret <8 x float> %0
}
+define <8 x float> @test8_unary_fneg(<8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: test8_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmadd213ps {{.*#+}} ymm0 = (ymm1 * ymm0) + ymm2
+; CHECK-NEXT: retq
+entry:
+ %neg.c = fneg <8 x float> %c
+ %0 = tail call <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float> %a, <8 x float> %b, <8 x float> %neg.c) #2
+ ret <8 x float> %0
+}
+
declare <8 x float> @llvm.x86.fma.vfmsub.ps.256(<8 x float>, <8 x float>, <8 x float>)
@@ -113,6 +200,17 @@ entry:
ret <8 x double> %sub.i
}
+define <8 x double> @test9_unary_fneg(<8 x double> %a, <8 x double> %b, <8 x double> %c) {
+; CHECK-LABEL: test9_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfnmsub213pd {{.*#+}} zmm0 = -(zmm1 * zmm0) - zmm2
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <8 x double> @llvm.x86.avx512.vfmadd.pd.512(<8 x double> %a, <8 x double> %b, <8 x double> %c, i32 4) #2
+ %neg.i = fneg <8 x double> %0
+ ret <8 x double> %neg.i
+}
+
declare <8 x double> @llvm.x86.avx512.vfmadd.pd.512(<8 x double> %a, <8 x double> %b, <8 x double> %c, i32)
define <2 x double> @test10(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
@@ -127,6 +225,18 @@ entry:
ret <2 x double> %sub.i
}
+define <2 x double> @test10_unary_fneg(<2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: test10_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfmadd213sd {{.*#+}} xmm0 = (xmm1 * xmm0) + xmm2
+; CHECK-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm0
+; CHECK-NEXT: retq
+entry:
+ %0 = tail call <2 x double> @llvm.x86.avx512.mask.vfmadd.sd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 -1, i32 4) #2
+ %neg.i = fneg <2 x double> %0
+ ret <2 x double> %neg.i
+}
+
declare <2 x double> @llvm.x86.avx512.mask.vfmadd.sd(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8, i32)
define <4 x float> @test11(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
@@ -154,6 +264,31 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test11_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test11_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: vxorps {{.*}}(%rip){1to4}, %xmm2, %xmm3
+; SKX-NEXT: vfmsub213ss {{.*#+}} xmm0 = (xmm1 * xmm0) - xmm2
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vmovss %xmm0, %xmm3, %xmm3 {%k1}
+; SKX-NEXT: vmovaps %xmm3, %xmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test11_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: vbroadcastss {{.*#+}} xmm3 = [-0.0E+0,-0.0E+0,-0.0E+0,-0.0E+0]
+; KNL-NEXT: vxorps %xmm3, %xmm2, %xmm3
+; KNL-NEXT: vfmsub213ss {{.*#+}} xmm0 = (xmm1 * xmm0) - xmm2
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vmovss %xmm0, %xmm3, %xmm3 {%k1}
+; KNL-NEXT: vmovaps %xmm3, %xmm0
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %c
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float> %a, <4 x float> %b, <4 x float> %neg.i, i8 %mask, i32 4) #10
+ ret <4 x float> %0
+}
+
declare <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
define <4 x float> @test11b(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
@@ -174,6 +309,24 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test11b_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test11b_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfmsub213ss {{.*#+}} xmm0 = (xmm1 * xmm0) - xmm2
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test11b_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfmsub213ss {{.*#+}} xmm0 = (xmm1 * xmm0) - xmm2
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %c
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %b, <4 x float> %neg.i, i8 %mask, i32 4) #10
+ ret <4 x float> %0
+}
+
declare <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float>, <4 x float>, <4 x float>, i8, i32)
define <8 x double> @test12(<8 x double> %a, <8 x double> %b, <8 x double> %c, i8 %mask) {
@@ -198,6 +351,28 @@ entry:
ret <8 x double> %sub.i
}
+define <8 x double> @test12_unary_fneg(<8 x double> %a, <8 x double> %b, <8 x double> %c, i8 %mask) {
+; SKX-LABEL: test12_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfmadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2
+; SKX-NEXT: vxorpd {{.*}}(%rip){1to8}, %zmm0, %zmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test12_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfmadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) + zmm2
+; KNL-NEXT: vpxorq {{.*}}(%rip){1to8}, %zmm0, %zmm0
+; KNL-NEXT: retq
+entry:
+ %0 = tail call <8 x double> @llvm.x86.avx512.vfmadd.pd.512(<8 x double> %a, <8 x double> %b, <8 x double> %c, i32 4) #2
+ %bc = bitcast i8 %mask to <8 x i1>
+ %sel = select <8 x i1> %bc, <8 x double> %0, <8 x double> %a
+ %neg.i = fneg <8 x double> %sel
+ ret <8 x double> %neg.i
+}
+
define <2 x double> @test13(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) {
; SKX-LABEL: test13:
; SKX: # %bb.0: # %entry
@@ -223,6 +398,31 @@ entry:
ret <2 x double> %0
}
+define <2 x double> @test13_unary_fneg(<2 x double> %a, <2 x double> %b, <2 x double> %c, i8 %mask) {
+; SKX-LABEL: test13_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm3
+; SKX-NEXT: vfnmadd213sd {{.*#+}} xmm1 = -(xmm0 * xmm1) + xmm2
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vmovsd %xmm1, %xmm3, %xmm3 {%k1}
+; SKX-NEXT: vmovapd %xmm3, %xmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test13_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: vxorpd {{.*}}(%rip), %xmm0, %xmm3
+; KNL-NEXT: vfnmadd213sd {{.*#+}} xmm1 = -(xmm0 * xmm1) + xmm2
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vmovsd %xmm1, %xmm3, %xmm3 {%k1}
+; KNL-NEXT: vmovapd %xmm3, %xmm0
+; KNL-NEXT: retq
+
+entry:
+ %neg.i = fneg <2 x double> %a
+ %0 = tail call <2 x double> @llvm.x86.avx512.mask.vfmadd.sd(<2 x double> %neg.i, <2 x double> %b, <2 x double> %c, i8 %mask, i32 4)
+ ret <2 x double> %0
+}
+
define <16 x float> @test14(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
; SKX-LABEL: test14:
; SKX: # %bb.0: # %entry
@@ -243,6 +443,26 @@ entry:
ret <16 x float> %sub.i
}
+define <16 x float> @test14_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
+; SKX-LABEL: test14_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmsub132ps {ru-sae}, %zmm1, %zmm2, %zmm0 {%k1}
+; SKX-NEXT: vxorps {{.*}}(%rip){1to16}, %zmm0, %zmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test14_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmsub132ps {ru-sae}, %zmm1, %zmm2, %zmm0 {%k1}
+; KNL-NEXT: vpxord {{.*}}(%rip){1to16}, %zmm0, %zmm0
+; KNL-NEXT: retq
+entry:
+ %0 = tail call <16 x float> @llvm.x86.avx512.mask.vfnmsub.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask, i32 10) #2
+ %neg.i = fneg <16 x float> %0
+ ret <16 x float> %neg.i
+}
+
define <16 x float> @test15(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
; SKX-LABEL: test15:
; SKX: # %bb.0: # %entry
@@ -273,6 +493,36 @@ entry:
ret <16 x float> %sel2
}
+define <16 x float> @test15_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
+; SKX-LABEL: test15_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vxorps {{.*}}(%rip){1to16}, %zmm0, %zmm3
+; SKX-NEXT: vfnmadd213ps {ru-sae}, %zmm2, %zmm0, %zmm1
+; SKX-NEXT: vmovaps %zmm1, %zmm3 {%k1}
+; SKX-NEXT: vfnmadd132ps {rd-sae}, %zmm0, %zmm2, %zmm3 {%k1}
+; SKX-NEXT: vmovaps %zmm3, %zmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test15_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vpxord {{.*}}(%rip){1to16}, %zmm0, %zmm3
+; KNL-NEXT: vfnmadd213ps {ru-sae}, %zmm2, %zmm0, %zmm1
+; KNL-NEXT: vmovaps %zmm1, %zmm3 {%k1}
+; KNL-NEXT: vfnmadd132ps {rd-sae}, %zmm0, %zmm2, %zmm3 {%k1}
+; KNL-NEXT: vmovaps %zmm3, %zmm0
+; KNL-NEXT: retq
+entry:
+ %bc = bitcast i16 %mask to <16 x i1>
+ %neg.i = fneg <16 x float> %a
+ %0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %neg.i, <16 x float> %b, <16 x float> %c, i32 10)
+ %sel = select <16 x i1> %bc, <16 x float> %0, <16 x float> %neg.i
+ %1 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %sel, <16 x float> %neg.i, <16 x float> %c, i32 9)
+ %sel2 = select <16 x i1> %bc, <16 x float> %1, <16 x float> %sel
+ ret <16 x float> %sel2
+}
+
define <16 x float> @test16(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
; SKX-LABEL: test16:
; SKX: # %bb.0:
@@ -291,6 +541,26 @@ define <16 x float> @test16(<16 x float>
%sel = select <16 x i1> %bc, <16 x float> %res, <16 x float> %a
ret <16 x float> %sel
}
+
+define <16 x float> @test16_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c, i16 %mask) {
+; SKX-LABEL: test16_unary_fneg:
+; SKX: # %bb.0:
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfmsubadd132ps {rd-sae}, %zmm1, %zmm2, %zmm0 {%k1}
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test16_unary_fneg:
+; KNL: # %bb.0:
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfmsubadd132ps {rd-sae}, %zmm1, %zmm2, %zmm0 {%k1}
+; KNL-NEXT: retq
+ %neg.i = fneg <16 x float> %c
+ %res = call <16 x float> @llvm.x86.avx512.vfmaddsub.ps.512(<16 x float> %a, <16 x float> %b, <16 x float> %neg.i, i32 9)
+ %bc = bitcast i16 %mask to <16 x i1>
+ %sel = select <16 x i1> %bc, <16 x float> %res, <16 x float> %a
+ ret <16 x float> %sel
+}
+
declare <16 x float> @llvm.x86.avx512.vfmaddsub.ps.512(<16 x float>, <16 x float>, <16 x float>, i32)
define <8 x double> @test17(<8 x double> %a, <8 x double> %b, <8 x double> %c, i8 %mask) {
@@ -311,6 +581,26 @@ define <8 x double> @test17(<8 x double>
%sel = select <8 x i1> %bc, <8 x double> %res, <8 x double> %a
ret <8 x double> %sel
}
+
+define <8 x double> @test17_unary_fneg(<8 x double> %a, <8 x double> %b, <8 x double> %c, i8 %mask) {
+; SKX-LABEL: test17_unary_fneg:
+; SKX: # %bb.0:
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfmsubadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) -/+ zmm2
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test17_unary_fneg:
+; KNL: # %bb.0:
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfmsubadd132pd {{.*#+}} zmm0 = (zmm0 * zmm1) -/+ zmm2
+; KNL-NEXT: retq
+ %neg.i = fneg <8 x double> %c
+ %res = call <8 x double> @llvm.x86.avx512.vfmaddsub.pd.512(<8 x double> %a, <8 x double> %b, <8 x double> %neg.i, i32 4)
+ %bc = bitcast i8 %mask to <8 x i1>
+ %sel = select <8 x i1> %bc, <8 x double> %res, <8 x double> %a
+ ret <8 x double> %sel
+}
+
declare <8 x double> @llvm.x86.avx512.vfmaddsub.pd.512(<8 x double>, <8 x double>, <8 x double>, i32)
define <4 x float> @test18(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
@@ -331,6 +621,24 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test18_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test18_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test18_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmadd213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) + xmm2
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %c, i8 %mask, i32 4) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test19(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test19:
; SKX: # %bb.0: # %entry
@@ -350,6 +658,25 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test19_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test19_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmsub213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) - xmm2
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test19_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmsub213ss {{.*#+}} xmm0 = -(xmm1 * xmm0) - xmm2
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %neg.i.2 = fneg <4 x float> %c
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %neg.i.2, i8 %mask, i32 4) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test20(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test20:
; SKX: # %bb.0: # %entry
@@ -370,6 +697,26 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test20_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test20_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmadd231ss {{.*#+}} xmm2 = -(xmm0 * xmm1) + xmm2
+; SKX-NEXT: vmovaps %xmm2, %xmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test20_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmadd231ss {{.*#+}} xmm2 = -(xmm0 * xmm1) + xmm2
+; KNL-NEXT: vmovaps %xmm2, %xmm0
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %c, i8 %mask, i32 4) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test21(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test21:
; SKX: # %bb.0: # %entry
@@ -388,6 +735,24 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test21_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test21_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmadd213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test21_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmadd213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %c, i8 %mask, i32 8) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test22(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test22:
; SKX: # %bb.0: # %entry
@@ -407,6 +772,25 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test22_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test22_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmsub213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test22_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmsub213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %neg.i.2 = fneg <4 x float> %c
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %neg.i.2, i8 %mask, i32 8) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test23(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test23:
; SKX: # %bb.0: # %entry
@@ -427,6 +811,26 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test23_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test23_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfnmadd231ss {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
+; SKX-NEXT: vmovaps %xmm2, %xmm0
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test23_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfnmadd231ss {rn-sae}, %xmm1, %xmm0, %xmm2 {%k1}
+; KNL-NEXT: vmovaps %xmm2, %xmm0
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %b
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask3.vfmadd.ss(<4 x float> %a, <4 x float> %neg.i, <4 x float> %c, i8 %mask, i32 8) #10
+ ret <4 x float> %0
+}
+
define <4 x float> @test24(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
; SKX-LABEL: test24:
; SKX: # %bb.0: # %entry
@@ -445,6 +849,24 @@ entry:
ret <4 x float> %0
}
+define <4 x float> @test24_unary_fneg(<4 x float> %a, <4 x float> %b, <4 x float> %c, i8 zeroext %mask) local_unnamed_addr #0 {
+; SKX-LABEL: test24_unary_fneg:
+; SKX: # %bb.0: # %entry
+; SKX-NEXT: kmovd %edi, %k1
+; SKX-NEXT: vfmsub213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; SKX-NEXT: retq
+;
+; KNL-LABEL: test24_unary_fneg:
+; KNL: # %bb.0: # %entry
+; KNL-NEXT: kmovw %edi, %k1
+; KNL-NEXT: vfmsub213ss {rn-sae}, %xmm2, %xmm1, %xmm0 {%k1}
+; KNL-NEXT: retq
+entry:
+ %neg.i = fneg <4 x float> %c
+ %0 = tail call <4 x float> @llvm.x86.avx512.mask.vfmadd.ss(<4 x float> %a, <4 x float> %b, <4 x float> %neg.i, i8 %mask, i32 8) #10
+ ret <4 x float> %0
+}
+
define <16 x float> @test25(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
; CHECK-LABEL: test25:
; CHECK: # %bb.0: # %entry
@@ -456,3 +878,15 @@ entry:
%0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %a, <16 x float> %sub.i, <16 x float> %sub.i.2, i32 8) #2
ret <16 x float> %0
}
+
+define <16 x float> @test25_unary_fneg(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+; CHECK-LABEL: test25_unary_fneg:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: vfnmsub213ps {rn-sae}, %zmm2, %zmm1, %zmm0
+; CHECK-NEXT: retq
+entry:
+ %neg.i = fneg <16 x float> %b
+ %neg.i.2 = fneg <16 x float> %c
+ %0 = tail call <16 x float> @llvm.x86.avx512.vfmadd.ps.512(<16 x float> %a, <16 x float> %neg.i, <16 x float> %neg.i.2, i32 8) #2
+ ret <16 x float> %0
+}
More information about the llvm-commits
mailing list