[llvm] c8bc89a - Regenerate FMA tests
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sat Feb 8 07:42:26 PST 2020
Author: Simon Pilgrim
Date: 2020-02-08T15:23:40Z
New Revision: c8bc89a933fc5e6fe160cffbd6e6a0aca0277cd9
URL: https://github.com/llvm/llvm-project/commit/c8bc89a933fc5e6fe160cffbd6e6a0aca0277cd9
DIFF: https://github.com/llvm/llvm-project/commit/c8bc89a933fc5e6fe160cffbd6e6a0aca0277cd9.diff
LOG: Regenerate FMA tests
Added:
Modified:
llvm/test/CodeGen/X86/extended-fma-contraction.ll
llvm/test/CodeGen/X86/fma-commute-x86.ll
llvm/test/CodeGen/X86/fma-intrinsics-phi-213-to-231.ll
llvm/test/CodeGen/X86/fmaddsub-combine.ll
llvm/test/CodeGen/X86/intersect-fma-fmf.ll
Removed:
################################################################################
diff --git a/llvm/test/CodeGen/X86/extended-fma-contraction.ll b/llvm/test/CodeGen/X86/extended-fma-contraction.ll
index 8ac47bd77f7e..c072d210750b 100644
--- a/llvm/test/CodeGen/X86/extended-fma-contraction.ll
+++ b/llvm/test/CodeGen/X86/extended-fma-contraction.ll
@@ -1,20 +1,18 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mcpu=bdver2 -mattr=-fma -mtriple=i686-apple-darwin < %s | FileCheck %s
; RUN: llc -mcpu=bdver2 -mattr=-fma,-fma4 -mtriple=i686-apple-darwin < %s | FileCheck %s --check-prefix=CHECK-NOFMA
-; CHECK-LABEL: fmafunc
define <3 x float> @fmafunc(<3 x float> %a, <3 x float> %b, <3 x float> %c) {
-
-; CHECK-NOT: vmulps
-; CHECK-NOT: vaddps
-; CHECK: vfmaddps
-; CHECK-NOT: vmulps
-; CHECK-NOT: vaddps
-
-; CHECK-NOFMA-NOT: calll
-; CHECK-NOFMA: vmulps
-; CHECK-NOFMA: vaddps
-; CHECK-NOFMA-NOT: calll
-
+; CHECK-LABEL: fmafunc:
+; CHECK: ## %bb.0:
+; CHECK-NEXT: vfmaddps %xmm2, %xmm1, %xmm0, %xmm0
+; CHECK-NEXT: retl
+;
+; CHECK-NOFMA-LABEL: fmafunc:
+; CHECK-NOFMA: ## %bb.0:
+; CHECK-NOFMA-NEXT: vmulps %xmm1, %xmm0, %xmm0
+; CHECK-NOFMA-NEXT: vaddps %xmm2, %xmm0, %xmm0
+; CHECK-NOFMA-NEXT: retl
%ret = tail call <3 x float> @llvm.fmuladd.v3f32(<3 x float> %a, <3 x float> %b, <3 x float> %c)
ret <3 x float> %ret
}
diff --git a/llvm/test/CodeGen/X86/fma-commute-x86.ll b/llvm/test/CodeGen/X86/fma-commute-x86.ll
index 64180f641107..21ff30f63727 100644
--- a/llvm/test/CodeGen/X86/fma-commute-x86.ll
+++ b/llvm/test/CodeGen/X86/fma-commute-x86.ll
@@ -21,7 +21,7 @@ define <4 x float> @test_x86_fmadd_aba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fmadd_aba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rcx), %xmm0
-; FMA-NEXT: vfmadd132ss (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfmadd132ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float> %a, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -31,7 +31,7 @@ define <4 x float> @test_x86_fmadd_bba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fmadd_bba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rdx), %xmm0
-; FMA-NEXT: vfmadd213ss (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfmadd213ss {{.*#+}} xmm0 = (xmm0 * xmm0) + mem
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfmadd.ss(<4 x float> %b, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -115,7 +115,7 @@ define <2 x double> @test_x86_fmadd_aba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fmadd_aba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rcx), %xmm0
-; FMA-NEXT: vfmadd132sd (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfmadd132sd {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double> %a, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -125,7 +125,7 @@ define <2 x double> @test_x86_fmadd_bba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fmadd_bba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rdx), %xmm0
-; FMA-NEXT: vfmadd213sd (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfmadd213sd {{.*#+}} xmm0 = (xmm0 * xmm0) + mem
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfmadd.sd(<2 x double> %b, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -210,7 +210,7 @@ define <4 x float> @test_x86_fnmadd_aba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fnmadd_aba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rcx), %xmm0
-; FMA-NEXT: vfnmadd132ss (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfnmadd132ss {{.*#+}} xmm0 = -(xmm0 * mem) + xmm0
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float> %a, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -220,7 +220,7 @@ define <4 x float> @test_x86_fnmadd_bba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fnmadd_bba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rdx), %xmm0
-; FMA-NEXT: vfnmadd213ss (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfnmadd213ss {{.*#+}} xmm0 = -(xmm0 * xmm0) + mem
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfnmadd.ss(<4 x float> %b, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -304,7 +304,7 @@ define <2 x double> @test_x86_fnmadd_aba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fnmadd_aba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rcx), %xmm0
-; FMA-NEXT: vfnmadd132sd (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfnmadd132sd {{.*#+}} xmm0 = -(xmm0 * mem) + xmm0
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double> %a, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -314,7 +314,7 @@ define <2 x double> @test_x86_fnmadd_bba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fnmadd_bba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rdx), %xmm0
-; FMA-NEXT: vfnmadd213sd (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfnmadd213sd {{.*#+}} xmm0 = -(xmm0 * xmm0) + mem
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfnmadd.sd(<2 x double> %b, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -398,7 +398,7 @@ define <4 x float> @test_x86_fmsub_aba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fmsub_aba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rcx), %xmm0
-; FMA-NEXT: vfmsub132ss (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfmsub132ss {{.*#+}} xmm0 = (xmm0 * mem) - xmm0
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float> %a, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -408,7 +408,7 @@ define <4 x float> @test_x86_fmsub_bba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fmsub_bba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rdx), %xmm0
-; FMA-NEXT: vfmsub213ss (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfmsub213ss {{.*#+}} xmm0 = (xmm0 * xmm0) - mem
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfmsub.ss(<4 x float> %b, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -492,7 +492,7 @@ define <2 x double> @test_x86_fmsub_aba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fmsub_aba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rcx), %xmm0
-; FMA-NEXT: vfmsub132sd (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfmsub132sd {{.*#+}} xmm0 = (xmm0 * mem) - xmm0
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double> %a, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -502,7 +502,7 @@ define <2 x double> @test_x86_fmsub_bba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fmsub_bba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rdx), %xmm0
-; FMA-NEXT: vfmsub213sd (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfmsub213sd {{.*#+}} xmm0 = (xmm0 * xmm0) - mem
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfmsub.sd(<2 x double> %b, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -587,7 +587,7 @@ define <4 x float> @test_x86_fnmsub_aba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fnmsub_aba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rcx), %xmm0
-; FMA-NEXT: vfnmsub132ss (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfnmsub132ss {{.*#+}} xmm0 = -(xmm0 * mem) - xmm0
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float> %a, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -597,7 +597,7 @@ define <4 x float> @test_x86_fnmsub_bba_ss(<4 x float> %a, <4 x float> %b) #0 {
; FMA-LABEL: test_x86_fnmsub_bba_ss:
; FMA: # %bb.0:
; FMA-NEXT: vmovaps (%rdx), %xmm0
-; FMA-NEXT: vfnmsub213ss (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfnmsub213ss {{.*#+}} xmm0 = -(xmm0 * xmm0) - mem
; FMA-NEXT: retq
%res = call <4 x float> @llvm.x86.fma.vfnmsub.ss(<4 x float> %b, <4 x float> %b, <4 x float> %a) nounwind
ret <4 x float> %res
@@ -681,7 +681,7 @@ define <2 x double> @test_x86_fnmsub_aba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fnmsub_aba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rcx), %xmm0
-; FMA-NEXT: vfnmsub132sd (%rdx), %xmm0, %xmm0
+; FMA-NEXT: vfnmsub132sd {{.*#+}} xmm0 = -(xmm0 * mem) - xmm0
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double> %a, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
@@ -691,7 +691,7 @@ define <2 x double> @test_x86_fnmsub_bba_sd(<2 x double> %a, <2 x double> %b) #0
; FMA-LABEL: test_x86_fnmsub_bba_sd:
; FMA: # %bb.0:
; FMA-NEXT: vmovapd (%rdx), %xmm0
-; FMA-NEXT: vfnmsub213sd (%rcx), %xmm0, %xmm0
+; FMA-NEXT: vfnmsub213sd {{.*#+}} xmm0 = -(xmm0 * xmm0) - mem
; FMA-NEXT: retq
%res = call <2 x double> @llvm.x86.fma.vfnmsub.sd(<2 x double> %b, <2 x double> %b, <2 x double> %a) nounwind
ret <2 x double> %res
diff --git a/llvm/test/CodeGen/X86/fma-intrinsics-phi-213-to-231.ll b/llvm/test/CodeGen/X86/fma-intrinsics-phi-213-to-231.ll
index 78e7471b8866..a509e93ef3f8 100644
--- a/llvm/test/CodeGen/X86/fma-intrinsics-phi-213-to-231.ll
+++ b/llvm/test/CodeGen/X86/fma-intrinsics-phi-213-to-231.ll
@@ -1,10 +1,22 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2,+fma | FileCheck %s
-; CHECK-LABEL: fmaddsubpd_loop_128:
-; CHECK: vfmaddsub231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fmaddsubpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fmaddsubpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB0_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB0_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmaddsub231pd {{.*#+}} xmm2 = (xmm0 * xmm1) +/- xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB0_2
+; CHECK-NEXT: .LBB0_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -26,11 +38,22 @@ for.end:
ret <2 x double> %c.addr.0
}
-; CHECK-LABEL: fmsubaddpd_loop_128:
-; CHECK: vfmsubadd231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fmsubaddpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fmsubaddpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB1_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB1_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsubadd231pd {{.*#+}} xmm2 = (xmm0 * xmm1) -/+ xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB1_2
+; CHECK-NEXT: .LBB1_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -52,11 +75,22 @@ for.end:
ret <2 x double> %c.addr.0
}
-; CHECK-LABEL: fmaddpd_loop_128:
-; CHECK: vfmadd231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fmaddpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fmaddpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB2_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB2_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmadd231pd {{.*#+}} xmm2 = (xmm0 * xmm1) + xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB2_2
+; CHECK-NEXT: .LBB2_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -78,11 +112,22 @@ for.end:
ret <2 x double> %c.addr.0
}
-; CHECK-LABEL: fmsubpd_loop_128:
-; CHECK: vfmsub231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fmsubpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fmsubpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB3_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB3_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsub231pd {{.*#+}} xmm2 = (xmm0 * xmm1) - xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB3_2
+; CHECK-NEXT: .LBB3_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -104,11 +149,22 @@ for.end:
ret <2 x double> %c.addr.0
}
-; CHECK-LABEL: fnmaddpd_loop_128:
-; CHECK: vfnmadd231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fnmaddpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fnmaddpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB4_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB4_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmadd231pd {{.*#+}} xmm2 = -(xmm0 * xmm1) + xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB4_2
+; CHECK-NEXT: .LBB4_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -130,11 +186,22 @@ for.end:
ret <2 x double> %c.addr.0
}
-; CHECK-LABEL: fnmsubpd_loop_128:
-; CHECK: vfnmsub231pd %xmm1, %xmm0, %xmm2
-; CHECK: vmovapd %xmm2, %xmm0
-; CHECK-NEXT: retq
define <2 x double> @fnmsubpd_loop_128(i32 %iter, <2 x double> %a, <2 x double> %b, <2 x double> %c) {
+; CHECK-LABEL: fnmsubpd_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB5_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB5_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmsub231pd {{.*#+}} xmm2 = -(xmm0 * xmm1) - xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB5_2
+; CHECK-NEXT: .LBB5_3: # %for.end
+; CHECK-NEXT: vmovapd %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -163,12 +230,22 @@ declare <2 x double> @llvm.x86.fma.vfmsub.pd(<2 x double>, <2 x double>, <2 x do
declare <2 x double> @llvm.x86.fma.vfnmadd.pd(<2 x double>, <2 x double>, <2 x double>)
declare <2 x double> @llvm.x86.fma.vfnmsub.pd(<2 x double>, <2 x double>, <2 x double>)
-
-; CHECK-LABEL: fmaddsubps_loop_128:
-; CHECK: vfmaddsub231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fmaddsubps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fmaddsubps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB6_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB6_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmaddsub231ps {{.*#+}} xmm2 = (xmm0 * xmm1) +/- xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB6_2
+; CHECK-NEXT: .LBB6_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -190,11 +267,22 @@ for.end:
ret <4 x float> %c.addr.0
}
-; CHECK-LABEL: fmsubaddps_loop_128:
-; CHECK: vfmsubadd231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fmsubaddps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fmsubaddps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB7_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB7_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsubadd231ps {{.*#+}} xmm2 = (xmm0 * xmm1) -/+ xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB7_2
+; CHECK-NEXT: .LBB7_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -216,11 +304,22 @@ for.end:
ret <4 x float> %c.addr.0
}
-; CHECK-LABEL: fmaddps_loop_128:
-; CHECK: vfmadd231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fmaddps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fmaddps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB8_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB8_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmadd231ps {{.*#+}} xmm2 = (xmm0 * xmm1) + xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB8_2
+; CHECK-NEXT: .LBB8_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -242,11 +341,22 @@ for.end:
ret <4 x float> %c.addr.0
}
-; CHECK-LABEL: fmsubps_loop_128:
-; CHECK: vfmsub231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fmsubps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fmsubps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB9_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB9_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsub231ps {{.*#+}} xmm2 = (xmm0 * xmm1) - xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB9_2
+; CHECK-NEXT: .LBB9_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -268,11 +378,22 @@ for.end:
ret <4 x float> %c.addr.0
}
-; CHECK-LABEL: fnmaddps_loop_128:
-; CHECK: vfnmadd231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fnmaddps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fnmaddps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB10_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB10_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmadd231ps {{.*#+}} xmm2 = -(xmm0 * xmm1) + xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB10_2
+; CHECK-NEXT: .LBB10_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -294,11 +415,22 @@ for.end:
ret <4 x float> %c.addr.0
}
-; CHECK-LABEL: fnmsubps_loop_128:
-; CHECK: vfnmsub231ps %xmm1, %xmm0, %xmm2
-; CHECK: vmovaps %xmm2, %xmm0
-; CHECK-NEXT: retq
define <4 x float> @fnmsubps_loop_128(i32 %iter, <4 x float> %a, <4 x float> %b, <4 x float> %c) {
+; CHECK-LABEL: fnmsubps_loop_128:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB11_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB11_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmsub231ps {{.*#+}} xmm2 = -(xmm0 * xmm1) - xmm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB11_2
+; CHECK-NEXT: .LBB11_3: # %for.end
+; CHECK-NEXT: vmovaps %xmm2, %xmm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -327,11 +459,22 @@ declare <4 x float> @llvm.x86.fma.vfmsub.ps(<4 x float>, <4 x float>, <4 x float
declare <4 x float> @llvm.x86.fma.vfnmadd.ps(<4 x float>, <4 x float>, <4 x float>)
declare <4 x float> @llvm.x86.fma.vfnmsub.ps(<4 x float>, <4 x float>, <4 x float>)
-; CHECK-LABEL: fmaddsubpd_loop_256:
-; CHECK: vfmaddsub231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fmaddsubpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fmaddsubpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB12_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB12_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmaddsub231pd {{.*#+}} ymm2 = (ymm0 * ymm1) +/- ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB12_2
+; CHECK-NEXT: .LBB12_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -353,11 +496,22 @@ for.end:
ret <4 x double> %c.addr.0
}
-; CHECK-LABEL: fmsubaddpd_loop_256:
-; CHECK: vfmsubadd231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fmsubaddpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fmsubaddpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB13_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB13_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsubadd231pd {{.*#+}} ymm2 = (ymm0 * ymm1) -/+ ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB13_2
+; CHECK-NEXT: .LBB13_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -379,11 +533,22 @@ for.end:
ret <4 x double> %c.addr.0
}
-; CHECK-LABEL: fmaddpd_loop_256:
-; CHECK: vfmadd231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fmaddpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fmaddpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB14_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB14_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmadd231pd {{.*#+}} ymm2 = (ymm0 * ymm1) + ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB14_2
+; CHECK-NEXT: .LBB14_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -405,11 +570,22 @@ for.end:
ret <4 x double> %c.addr.0
}
-; CHECK-LABEL: fmsubpd_loop_256:
-; CHECK: vfmsub231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fmsubpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fmsubpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB15_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB15_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsub231pd {{.*#+}} ymm2 = (ymm0 * ymm1) - ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB15_2
+; CHECK-NEXT: .LBB15_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -431,11 +607,22 @@ for.end:
ret <4 x double> %c.addr.0
}
-; CHECK-LABEL: fnmaddpd_loop_256:
-; CHECK: vfnmadd231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fnmaddpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fnmaddpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB16_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB16_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmadd231pd {{.*#+}} ymm2 = -(ymm0 * ymm1) + ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB16_2
+; CHECK-NEXT: .LBB16_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -457,11 +644,22 @@ for.end:
ret <4 x double> %c.addr.0
}
-; CHECK-LABEL: fnmsubpd_loop_256:
-; CHECK: vfnmsub231pd %ymm1, %ymm0, %ymm2
-; CHECK: vmovapd %ymm2, %ymm0
-; CHECK-NEXT: retq
define <4 x double> @fnmsubpd_loop_256(i32 %iter, <4 x double> %a, <4 x double> %b, <4 x double> %c) {
+; CHECK-LABEL: fnmsubpd_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB17_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB17_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmsub231pd {{.*#+}} ymm2 = -(ymm0 * ymm1) - ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB17_2
+; CHECK-NEXT: .LBB17_3: # %for.end
+; CHECK-NEXT: vmovapd %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -490,12 +688,22 @@ declare <4 x double> @llvm.x86.fma.vfmsub.pd.256(<4 x double>, <4 x double>, <4
declare <4 x double> @llvm.x86.fma.vfnmadd.pd.256(<4 x double>, <4 x double>, <4 x double>)
declare <4 x double> @llvm.x86.fma.vfnmsub.pd.256(<4 x double>, <4 x double>, <4 x double>)
-
-; CHECK-LABEL: fmaddsubps_loop_256:
-; CHECK: vfmaddsub231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fmaddsubps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fmaddsubps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB18_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB18_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmaddsub231ps {{.*#+}} ymm2 = (ymm0 * ymm1) +/- ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB18_2
+; CHECK-NEXT: .LBB18_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -517,11 +725,22 @@ for.end:
ret <8 x float> %c.addr.0
}
-; CHECK-LABEL: fmsubaddps_loop_256:
-; CHECK: vfmsubadd231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fmsubaddps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fmsubaddps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB19_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB19_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsubadd231ps {{.*#+}} ymm2 = (ymm0 * ymm1) -/+ ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB19_2
+; CHECK-NEXT: .LBB19_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -543,11 +762,22 @@ for.end:
ret <8 x float> %c.addr.0
}
-; CHECK-LABEL: fmaddps_loop_256:
-; CHECK: vfmadd231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fmaddps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fmaddps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB20_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB20_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmadd231ps {{.*#+}} ymm2 = (ymm0 * ymm1) + ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB20_2
+; CHECK-NEXT: .LBB20_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -569,11 +799,22 @@ for.end:
ret <8 x float> %c.addr.0
}
-; CHECK-LABEL: fmsubps_loop_256:
-; CHECK: vfmsub231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fmsubps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fmsubps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB21_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB21_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfmsub231ps {{.*#+}} ymm2 = (ymm0 * ymm1) - ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB21_2
+; CHECK-NEXT: .LBB21_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -595,11 +836,22 @@ for.end:
ret <8 x float> %c.addr.0
}
-; CHECK-LABEL: fnmaddps_loop_256:
-; CHECK: vfnmadd231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fnmaddps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fnmaddps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB22_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB22_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmadd231ps {{.*#+}} ymm2 = -(ymm0 * ymm1) + ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB22_2
+; CHECK-NEXT: .LBB22_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
@@ -621,11 +873,22 @@ for.end:
ret <8 x float> %c.addr.0
}
-; CHECK-LABEL: fnmsubps_loop_256:
-; CHECK: vfnmsub231ps %ymm1, %ymm0, %ymm2
-; CHECK: vmovaps %ymm2, %ymm0
-; CHECK-NEXT: retq
define <8 x float> @fnmsubps_loop_256(i32 %iter, <8 x float> %a, <8 x float> %b, <8 x float> %c) {
+; CHECK-LABEL: fnmsubps_loop_256:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: xorl %eax, %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jge .LBB23_3
+; CHECK-NEXT: .p2align 4, 0x90
+; CHECK-NEXT: .LBB23_2: # %for.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: vfnmsub231ps {{.*#+}} ymm2 = -(ymm0 * ymm1) - ymm2
+; CHECK-NEXT: incl %eax
+; CHECK-NEXT: cmpl %edi, %eax
+; CHECK-NEXT: jl .LBB23_2
+; CHECK-NEXT: .LBB23_3: # %for.end
+; CHECK-NEXT: vmovaps %ymm2, %ymm0
+; CHECK-NEXT: retq
entry:
br label %for.cond
diff --git a/llvm/test/CodeGen/X86/fmaddsub-combine.ll b/llvm/test/CodeGen/X86/fmaddsub-combine.ll
index 941f6ba204d3..320f2575e97c 100644
--- a/llvm/test/CodeGen/X86/fmaddsub-combine.ll
+++ b/llvm/test/CodeGen/X86/fmaddsub-combine.ll
@@ -1,14 +1,14 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma | FileCheck -check-prefix=FMA3 -check-prefix=FMA3_256 %s
-; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma,+avx512f | FileCheck -check-prefix=FMA3 -check-prefix=FMA3_512 %s
-; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma4 | FileCheck -check-prefix=FMA4 %s
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma | FileCheck %s -check-prefixes=FMA3,FMA3_256
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma,+avx512f | FileCheck %s -check-prefixes=FMA3,FMA3_512
+; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu -mattr=+fma4 | FileCheck %s -check-prefixes=FMA4
; This test checks the fusing of MUL + ADDSUB to FMADDSUB.
define <2 x double> @mul_addsub_pd128(<2 x double> %A, <2 x double> %B, <2 x double> %C) #0 {
; FMA3-LABEL: mul_addsub_pd128:
; FMA3: # %bb.0: # %entry
-; FMA3-NEXT: vfmaddsub213pd %xmm2, %xmm1, %xmm0
+; FMA3-NEXT: vfmaddsub213pd {{.*#+}} xmm0 = (xmm1 * xmm0) +/- xmm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: mul_addsub_pd128:
@@ -26,7 +26,7 @@ entry:
define <4 x float> @mul_addsub_ps128(<4 x float> %A, <4 x float> %B, <4 x float> %C) #0 {
; FMA3-LABEL: mul_addsub_ps128:
; FMA3: # %bb.0: # %entry
-; FMA3-NEXT: vfmaddsub213ps %xmm2, %xmm1, %xmm0
+; FMA3-NEXT: vfmaddsub213ps {{.*#+}} xmm0 = (xmm1 * xmm0) +/- xmm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: mul_addsub_ps128:
@@ -44,7 +44,7 @@ entry:
define <4 x double> @mul_addsub_pd256(<4 x double> %A, <4 x double> %B, <4 x double> %C) #0 {
; FMA3-LABEL: mul_addsub_pd256:
; FMA3: # %bb.0: # %entry
-; FMA3-NEXT: vfmaddsub213pd %ymm2, %ymm1, %ymm0
+; FMA3-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- ymm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: mul_addsub_pd256:
@@ -62,7 +62,7 @@ entry:
define <8 x float> @mul_addsub_ps256(<8 x float> %A, <8 x float> %B, <8 x float> %C) #0 {
; FMA3-LABEL: mul_addsub_ps256:
; FMA3: # %bb.0: # %entry
-; FMA3-NEXT: vfmaddsub213ps %ymm2, %ymm1, %ymm0
+; FMA3-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- ymm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: mul_addsub_ps256:
@@ -80,13 +80,13 @@ entry:
define <8 x double> @mul_addsub_pd512(<8 x double> %A, <8 x double> %B, <8 x double> %C) #0 {
; FMA3_256-LABEL: mul_addsub_pd512:
; FMA3_256: # %bb.0: # %entry
-; FMA3_256-NEXT: vfmaddsub213pd %ymm4, %ymm2, %ymm0
-; FMA3_256-NEXT: vfmaddsub213pd %ymm5, %ymm3, %ymm1
+; FMA3_256-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm2 * ymm0) +/- ymm4
+; FMA3_256-NEXT: vfmaddsub213pd {{.*#+}} ymm1 = (ymm3 * ymm1) +/- ymm5
; FMA3_256-NEXT: retq
;
; FMA3_512-LABEL: mul_addsub_pd512:
; FMA3_512: # %bb.0: # %entry
-; FMA3_512-NEXT: vfmaddsub213pd %zmm2, %zmm1, %zmm0
+; FMA3_512-NEXT: vfmaddsub213pd {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2
; FMA3_512-NEXT: retq
;
; FMA4-LABEL: mul_addsub_pd512:
@@ -105,13 +105,13 @@ entry:
define <16 x float> @mul_addsub_ps512(<16 x float> %A, <16 x float> %B, <16 x float> %C) #0 {
; FMA3_256-LABEL: mul_addsub_ps512:
; FMA3_256: # %bb.0: # %entry
-; FMA3_256-NEXT: vfmaddsub213ps %ymm4, %ymm2, %ymm0
-; FMA3_256-NEXT: vfmaddsub213ps %ymm5, %ymm3, %ymm1
+; FMA3_256-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm2 * ymm0) +/- ymm4
+; FMA3_256-NEXT: vfmaddsub213ps {{.*#+}} ymm1 = (ymm3 * ymm1) +/- ymm5
; FMA3_256-NEXT: retq
;
; FMA3_512-LABEL: mul_addsub_ps512:
; FMA3_512: # %bb.0: # %entry
-; FMA3_512-NEXT: vfmaddsub213ps %zmm2, %zmm1, %zmm0
+; FMA3_512-NEXT: vfmaddsub213ps {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2
; FMA3_512-NEXT: retq
;
; FMA4-LABEL: mul_addsub_ps512:
@@ -130,7 +130,7 @@ entry:
define <4 x float> @buildvector_mul_addsub_ps128(<4 x float> %C, <4 x float> %D, <4 x float> %B) #0 {
; FMA3-LABEL: buildvector_mul_addsub_ps128:
; FMA3: # %bb.0: # %bb
-; FMA3-NEXT: vfmaddsub213ps %xmm2, %xmm1, %xmm0
+; FMA3-NEXT: vfmaddsub213ps {{.*#+}} xmm0 = (xmm1 * xmm0) +/- xmm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_ps128:
@@ -161,7 +161,7 @@ bb:
define <2 x double> @buildvector_mul_addsub_pd128(<2 x double> %C, <2 x double> %D, <2 x double> %B) #0 {
; FMA3-LABEL: buildvector_mul_addsub_pd128:
; FMA3: # %bb.0: # %bb
-; FMA3-NEXT: vfmaddsub213pd %xmm2, %xmm1, %xmm0
+; FMA3-NEXT: vfmaddsub213pd {{.*#+}} xmm0 = (xmm1 * xmm0) +/- xmm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_pd128:
@@ -184,7 +184,7 @@ bb:
define <8 x float> @buildvector_mul_addsub_ps256(<8 x float> %C, <8 x float> %D, <8 x float> %B) #0 {
; FMA3-LABEL: buildvector_mul_addsub_ps256:
; FMA3: # %bb.0: # %bb
-; FMA3-NEXT: vfmaddsub213ps %ymm2, %ymm1, %ymm0
+; FMA3-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm1 * ymm0) +/- ymm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_ps256:
@@ -231,7 +231,7 @@ bb:
define <4 x double> @buildvector_mul_addsub_pd256(<4 x double> %C, <4 x double> %D, <4 x double> %B) #0 {
; FMA3-LABEL: buildvector_mul_addsub_pd256:
; FMA3: # %bb.0: # %bb
-; FMA3-NEXT: vfmaddsub213pd %ymm2, %ymm1, %ymm0
+; FMA3-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm1 * ymm0) +/- ymm2
; FMA3-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_pd256:
@@ -262,13 +262,13 @@ bb:
define <16 x float> @buildvector_mul_addsub_ps512(<16 x float> %C, <16 x float> %D, <16 x float> %B) #0 {
; FMA3_256-LABEL: buildvector_mul_addsub_ps512:
; FMA3_256: # %bb.0: # %bb
-; FMA3_256-NEXT: vfmaddsub213ps %ymm4, %ymm2, %ymm0
-; FMA3_256-NEXT: vfmaddsub213ps %ymm5, %ymm3, %ymm1
+; FMA3_256-NEXT: vfmaddsub213ps {{.*#+}} ymm0 = (ymm2 * ymm0) +/- ymm4
+; FMA3_256-NEXT: vfmaddsub213ps {{.*#+}} ymm1 = (ymm3 * ymm1) +/- ymm5
; FMA3_256-NEXT: retq
;
; FMA3_512-LABEL: buildvector_mul_addsub_ps512:
; FMA3_512: # %bb.0: # %bb
-; FMA3_512-NEXT: vfmaddsub213ps %zmm2, %zmm1, %zmm0
+; FMA3_512-NEXT: vfmaddsub213ps {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2
; FMA3_512-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_ps512:
@@ -348,13 +348,13 @@ bb:
define <8 x double> @buildvector_mul_addsub_pd512(<8 x double> %C, <8 x double> %D, <8 x double> %B) #0 {
; FMA3_256-LABEL: buildvector_mul_addsub_pd512:
; FMA3_256: # %bb.0: # %bb
-; FMA3_256-NEXT: vfmaddsub213pd %ymm4, %ymm2, %ymm0
-; FMA3_256-NEXT: vfmaddsub213pd %ymm5, %ymm3, %ymm1
+; FMA3_256-NEXT: vfmaddsub213pd {{.*#+}} ymm0 = (ymm2 * ymm0) +/- ymm4
+; FMA3_256-NEXT: vfmaddsub213pd {{.*#+}} ymm1 = (ymm3 * ymm1) +/- ymm5
; FMA3_256-NEXT: retq
;
; FMA3_512-LABEL: buildvector_mul_addsub_pd512:
; FMA3_512: # %bb.0: # %bb
-; FMA3_512-NEXT: vfmaddsub213pd %zmm2, %zmm1, %zmm0
+; FMA3_512-NEXT: vfmaddsub213pd {{.*#+}} zmm0 = (zmm1 * zmm0) +/- zmm2
; FMA3_512-NEXT: retq
;
; FMA4-LABEL: buildvector_mul_addsub_pd512:
diff --git a/llvm/test/CodeGen/X86/intersect-fma-fmf.ll b/llvm/test/CodeGen/X86/intersect-fma-fmf.ll
index 3c2a3d2c0257..601b4c454187 100644
--- a/llvm/test/CodeGen/X86/intersect-fma-fmf.ll
+++ b/llvm/test/CodeGen/X86/intersect-fma-fmf.ll
@@ -1,10 +1,16 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=i686-unknown-unknown -mattr=+avx512f < %s | FileCheck %s
-define float @test_x86_fma_intersection_fmf(float %a, float %b) {
+define float @test_x86_fma_intersection_fmf(float %a, float %b) nounwind {
; CHECK-LABEL: test_x86_fma_intersection_fmf:
-; CHECK: # %bb.0:
-; CHECK: vfmadd132ss {{[0-9]+}}(%esp), %xmm0, %xmm0
-; CHECK: retl
+; CHECK: # %bb.0:
+; CHECK-NEXT: pushl %eax
+; CHECK-NEXT: vmovss {{.*#+}} xmm0 = mem[0],zero,zero,zero
+; CHECK-NEXT: vfmadd132ss {{.*#+}} xmm0 = (xmm0 * mem) + xmm0
+; CHECK-NEXT: vmovss %xmm0, (%esp)
+; CHECK-NEXT: flds (%esp)
+; CHECK-NEXT: popl %eax
+; CHECK-NEXT: retl
%tmp8 = fmul fast float %a, %b
%tmp9 = fadd fast float %tmp8, %b
%tmp10 = insertelement <4 x float> undef, float %tmp9, i32 1
More information about the llvm-commits
mailing list