3 files changed, 152 insertions, 0 deletions
diff --git a/test/CodeGen/X86/avx512-arith.ll b/test/CodeGen/X86/avx512-arith.ll
index 55ce9f9..d5af76f 100644
--- a/test/CodeGen/X86/avx512-arith.ll
+++ b/test/CodeGen/X86/avx512-arith.ll
@@ -186,6 +186,43 @@ define <16 x i32> @vpmulld_test(<16 x i32> %i, <16 x i32> %j) {
   ret <16 x i32> %x
 }
 
+; CHECK-LABEL: sqrtA
+; CHECK: vsqrtssz
+; CHECK: ret
+declare float @sqrtf(float) readnone
+define float @sqrtA(float %a) nounwind uwtable readnone ssp {
+entry:
+  %conv1 = tail call float @sqrtf(float %a) nounwind readnone
+  ret float %conv1
+}
+
+; CHECK-LABEL: sqrtB
+; CHECK: vsqrtsdz
+; CHECK: ret
+declare double @sqrt(double) readnone
+define double @sqrtB(double %a) nounwind uwtable readnone ssp {
+entry:
+  %call = tail call double @sqrt(double %a) nounwind readnone
+  ret double %call
+}
+
+; CHECK-LABEL: sqrtC
+; CHECK: vsqrtssz
+; CHECK: ret
+declare float @llvm.sqrt.f32(float)
+define float @sqrtC(float %a) nounwind {
+  %b = call float @llvm.sqrt.f32(float %a)
+  ret float %b
+}
+
+; CHECK-LABEL: fadd_broadcast
+; CHECK: LCP{{.*}}(%rip){1to16}, %zmm0, %zmm0
+; CHECK: ret
+define <16 x float> @fadd_broadcast(<16 x float> %a) nounwind {
+  %b = fadd <16 x float> %a, <float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000, float 0x3FB99999A0000000>
+  ret <16 x float> %b
+}
+
 ; CHECK-LABEL: addq_broadcast
 ; CHECK: vpaddq LCP{{.*}}(%rip){1to8}, %zmm0, %zmm0
 ; CHECK: ret
diff --git a/test/CodeGen/X86/avx512-cmp.ll b/test/CodeGen/X86/avx512-cmp.ll
new file mode 100644
index 0000000..ba52745
--- /dev/null
+++ b/test/CodeGen/X86/avx512-cmp.ll
@@ -0,0 +1,27 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+
+; CHECK: vucomisdz
+define double @test1(double %a, double %b) nounwind {
+  %tobool = fcmp une double %a, %b
+  br i1 %tobool, label %l1, label %l2
+
+l1:
+  %c = fsub double %a, %b
+  ret double %c
+l2:
+  %c1 = fadd double %a, %b
+  ret double %c1
+}
+
+; CHECK: vucomissz
+define float @test2(float %a, float %b) nounwind {
+  %tobool = fcmp olt float %a, %b
+  br i1 %tobool, label %l1, label %l2
+
+l1:
+  %c = fsub float %a, %b
+  ret float %c
+l2:
+  %c1 = fadd float %a, %b
+  ret float %c1
+}
diff --git a/test/CodeGen/X86/avx512-intrinsics.ll b/test/CodeGen/X86/avx512-intrinsics.ll
new file mode 100644
index 0000000..c0ac719
--- /dev/null
+++ b/test/CodeGen/X86/avx512-intrinsics.ll
@@ -0,0 +1,88 @@
+; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl | FileCheck %s
+
+declare i32 @llvm.x86.avx512.kortestz(i16, i16) nounwind readnone
+; CHECK: test_x86_avx3_kortestz
+; CHECK: kortestw
+; CHECK: sete
+define i32 @test_x86_avx3_kortestz(i16 %a0, i16 %a1) {
+  %res = call i32 @llvm.x86.avx512.kortestz(i16 %a0, i16 %a1) 
+  ret i32 %res
+}
+
+declare i32 @llvm.x86.avx512.kortestc(i16, i16) nounwind readnone
+; CHECK: test_x86_avx3_kortestc
+; CHECK: kortestw
+; CHECK: sbbl
+define i32 @test_x86_avx3_kortestc(i16 %a0, i16 %a1) {
+  %res = call i32 @llvm.x86.avx512.kortestc(i16 %a0, i16 %a1) 
+  ret i32 %res
+}
+
+define <16 x float> @test_x86_avx3_rcp_ps_512(<16 x float> %a0) {
+  ; CHECK: vrcp14ps
+  %res = call <16 x float> @llvm.x86.avx512.rcp14.ps.512(<16 x float> %a0) ; <<16 x float>> [#uses=1]
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.rcp14.ps.512(<16 x float>) nounwind readnone
+
+define <8 x double> @test_x86_avx3_rcp_pd_512(<8 x double> %a0) {
+  ; CHECK: vrcp14pd
+  %res = call <8 x double> @llvm.x86.avx512.rcp14.pd.512(<8 x double> %a0) ; <<8 x double>> [#uses=1]
+  ret <8 x double> %res
+}
+declare <8 x double> @llvm.x86.avx512.rcp14.pd.512(<8 x double>) nounwind readnone
+
+
+define <8 x double> @test_x86_avx3_rndscale_pd_512(<8 x double> %a0) {
+  ; CHECK: vrndscale
+  %res = call <8 x double> @llvm.x86.avx512.rndscale.pd.512(<8 x double> %a0, i32 7) ; <<8 x double>> [#uses=1]
+  ret <8 x double> %res
+}
+declare <8 x double> @llvm.x86.avx512.rndscale.pd.512(<8 x double>, i32) nounwind readnone
+
+
+define <16 x float> @test_x86_avx3_rndscale_ps_512(<16 x float> %a0) {
+  ; CHECK: vrndscale
+  %res = call <16 x float> @llvm.x86.avx512.rndscale.ps.512(<16 x float> %a0, i32 7) ; <<16 x float>> [#uses=1]
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.rndscale.ps.512(<16 x float>, i32) nounwind readnone
+
+
+define <16 x float> @test_x86_avx3_rsqrt_ps_512(<16 x float> %a0) {
+  ; CHECK: vrsqrt14ps
+  %res = call <16 x float> @llvm.x86.avx512.rsqrt14.ps.512(<16 x float> %a0) ; <<16 x float>> [#uses=1]
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.rsqrt14.ps.512(<16 x float>) nounwind readnone
+
+
+define <8 x double> @test_x86_avx3_sqrt_pd_512(<8 x double> %a0) {
+  ; CHECK: vsqrtpd
+  %res = call <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double> %a0) ; <<8 x double>> [#uses=1]
+  ret <8 x double> %res
+}
+declare <8 x double> @llvm.x86.avx512.sqrt.pd.512(<8 x double>) nounwind readnone
+
+
+define <16 x float> @test_x86_avx3_sqrt_ps_512(<16 x float> %a0) {
+  ; CHECK: vsqrtps
+  %res = call <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float> %a0) ; <<16 x float>> [#uses=1]
+  ret <16 x float> %res
+}
+declare <16 x float> @llvm.x86.avx512.sqrt.ps.512(<16 x float>) nounwind readnone
+
+define <4 x float> @test_x86_avx3_sqrt_ss(<4 x float> %a0, <4 x float> %a1) {
+  ; CHECK: vsqrtssz
+  %res = call <4 x float> @llvm.x86.avx512.sqrt.ss(<4 x float> %a0, <4 x float> %a1) ; <<4 x float>> [#uses=1]
+  ret <4 x float> %res
+}
+declare <4 x float> @llvm.x86.avx512.sqrt.ss(<4 x float>, <4 x float>) nounwind readnone
+
+define <2 x double> @test_x86_avx3_sqrt_sd(<2 x double> %a0, <2 x double> %a1) {
+  ; CHECK: vsqrtsdz
+  %res = call <2 x double> @llvm.x86.avx512.sqrt.sd(<2 x double> %a0, <2 x double> %a1) ; <<2 x double>> [#uses=1]
+  ret <2 x double> %res
+}
+declare <2 x double> @llvm.x86.avx512.sqrt.sd(<2 x double>, <2 x double>) nounwind readnone
+