3 files changed, 117 insertions, 3 deletions
diff --git a/test/CodeGen/X86/extended-fma-contraction.ll b/test/CodeGen/X86/extended-fma-contraction.ll
new file mode 100644
index 0000000..ef2c22b
--- /dev/null
+++ b/test/CodeGen/X86/extended-fma-contraction.ll
@@ -0,0 +1,22 @@
+; RUN: llc -march=x86 -mattr=+fma4 -mtriple=x86_64-apple-darwin < %s | FileCheck %s
+; RUN: llc -march=x86 -mtriple=x86_64-apple-darwin < %s | FileCheck %s --check-prefix=CHECK-NOFMA
+
+; CHECK: 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
+
+  %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
+}
+
+declare <3 x float> @llvm.fmuladd.v3f32(<3 x float>, <3 x float>, <3 x float>) nounwind readnone
diff --git a/test/CodeGen/X86/fma_patterns_wide.ll b/test/CodeGen/X86/fma_patterns_wide.ll
new file mode 100644
index 0000000..d84e5a0
--- /dev/null
+++ b/test/CodeGen/X86/fma_patterns_wide.ll
@@ -0,0 +1,84 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=core-avx2 -mattr=avx2,+fma -fp-contract=fast | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=bdver2 -mattr=-fma4 -fp-contract=fast | FileCheck %s
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mcpu=bdver1 -fp-contract=fast | FileCheck %s --check-prefix=CHECK_FMA4
+
+; CHECK: test_x86_fmadd_ps_y_wide
+; CHECK: vfmadd213ps
+; CHECK: vfmadd213ps
+; CHECK: ret
+; CHECK_FMA4: test_x86_fmadd_ps_y_wide
+; CHECK_FMA4: vfmaddps
+; CHECK_FMA4: vfmaddps
+; CHECK_FMA4: ret
+define <16 x float> @test_x86_fmadd_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) {
+  %x = fmul <16 x float> %a0, %a1
+  %res = fadd <16 x float> %x, %a2
+  ret <16 x float> %res
+}
+
+; CHECK: test_x86_fmsub_ps_y_wide
+; CHECK: vfmsub213ps
+; CHECK: vfmsub213ps
+; CHECK: ret
+; CHECK_FMA4: test_x86_fmsub_ps_y_wide
+; CHECK_FMA4: vfmsubps
+; CHECK_FMA4: vfmsubps
+; CHECK_FMA4: ret
+define <16 x float> @test_x86_fmsub_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) {
+  %x = fmul <16 x float> %a0, %a1
+  %res = fsub <16 x float> %x, %a2
+  ret <16 x float> %res
+}
+
+; CHECK: test_x86_fnmadd_ps_y_wide
+; CHECK: vfnmadd213ps
+; CHECK: vfnmadd213ps
+; CHECK: ret
+; CHECK_FMA4: test_x86_fnmadd_ps_y_wide
+; CHECK_FMA4: vfnmaddps
+; CHECK_FMA4: vfnmaddps
+; CHECK_FMA4: ret
+define <16 x float> @test_x86_fnmadd_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) {
+  %x = fmul <16 x float> %a0, %a1
+  %res = fsub <16 x float> %a2, %x
+  ret <16 x float> %res
+}
+
+; CHECK: test_x86_fnmsub_ps_y_wide
+; CHECK: vfnmsub213ps
+; CHECK: vfnmsub213ps
+; CHECK: ret
+define <16 x float> @test_x86_fnmsub_ps_y_wide(<16 x float> %a0, <16 x float> %a1, <16 x float> %a2) {
+  %x = fmul <16 x float> %a0, %a1
+  %y = fsub <16 x float> <float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, %x
+  %res = fsub <16 x float> %y, %a2
+  ret <16 x float> %res
+}
+
+; CHECK: test_x86_fmadd_pd_y_wide
+; CHECK: vfmadd213pd
+; CHECK: vfmadd213pd
+; CHECK: ret
+; CHECK_FMA4: test_x86_fmadd_pd_y_wide
+; CHECK_FMA4: vfmaddpd
+; CHECK_FMA4: vfmaddpd
+; CHECK_FMA4: ret
+define <8 x double> @test_x86_fmadd_pd_y_wide(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2) {
+  %x = fmul <8 x double> %a0, %a1
+  %res = fadd <8 x double> %x, %a2
+  ret <8 x double> %res
+}
+
+; CHECK: test_x86_fmsub_pd_y_wide
+; CHECK: vfmsub213pd
+; CHECK: vfmsub213pd
+; CHECK: ret
+; CHECK_FMA4: test_x86_fmsub_pd_y_wide
+; CHECK_FMA4: vfmsubpd
+; CHECK_FMA4: vfmsubpd
+; CHECK_FMA4: ret
+define <8 x double> @test_x86_fmsub_pd_y_wide(<8 x double> %a0, <8 x double> %a1, <8 x double> %a2) {
+  %x = fmul <8 x double> %a0, %a1
+  %res = fsub <8 x double> %x, %a2
+  ret <8 x double> %res
+}
diff --git a/test/CodeGen/X86/wide-fma-contraction.ll b/test/CodeGen/X86/wide-fma-contraction.ll
index d93f33b..7ee0fba 100644
--- a/test/CodeGen/X86/wide-fma-contraction.ll
+++ b/test/CodeGen/X86/wide-fma-contraction.ll
@@ -1,7 +1,9 @@
 ; RUN: llc -march=x86 -mattr=+fma4 -mtriple=x86_64-apple-darwin < %s | FileCheck %s
+; RUN: llc -march=x86 -mtriple=x86_64-apple-darwin < %s | FileCheck %s --check-prefix=CHECK-NOFMA
 
 ; CHECK: fmafunc
 define <16 x float> @fmafunc(<16 x float> %a, <16 x float> %b, <16 x float> %c) {
+
 ; CHECK-NOT: vmulps
 ; CHECK-NOT: vaddps
 ; CHECK: vfmaddps
@@ -10,11 +12,17 @@ define <16 x float> @fmafunc(<16 x float> %a, <16 x float> %b, <16 x float> %c)
 ; CHECK: vfmaddps
 ; CHECK-NOT: vmulps
 ; CHECK-NOT: vaddps
+
+; CHECK-NOFMA-NOT: calll
+; CHECK-NOFMA: vmulps
+; CHECK-NOFMA: vaddps
+; CHECK-NOFMA-NOT: calll
+; CHECK-NOFMA: vmulps
+; CHECK-NOFMA: vaddps
+; CHECK-NOFMA-NOT: calll
+
   %ret = tail call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %a, <16 x float> %b, <16 x float> %c)
   ret <16 x float> %ret
 }
 
 declare <16 x float> @llvm.fmuladd.v16f32(<16 x float>, <16 x float>, <16 x float>) nounwind readnone
-
-
-