From 1d8fbacd7d05fa650d9dc9522647cee488babd83 Mon Sep 17 00:00:00 2001 From: Craig Topper Date: Mon, 13 Nov 2017 06:46:46 +0000 Subject: [PATCH] [X86] Add tests for missed opportunities to fold a 128-bit vector load into vfpclassss and vpfpclasssd. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@318018 91177308-0d34-0410-b5e6-96231b3b80d8 --- test/CodeGen/X86/avx512dq-intrinsics.ll | 26 ++++++++++++++++++++++++++ 1 file changed, 26 insertions(+) diff --git a/test/CodeGen/X86/avx512dq-intrinsics.ll b/test/CodeGen/X86/avx512dq-intrinsics.ll index e4f854912ac..49f439f8e27 100644 --- a/test/CodeGen/X86/avx512dq-intrinsics.ll +++ b/test/CodeGen/X86/avx512dq-intrinsics.ll @@ -396,6 +396,19 @@ define i8 @test_int_x86_avx512_mask_fpclass_sd(<2 x double> %x0, i8 %x1) { ret i8 %res2 } +define i8 @test_int_x86_avx512_mask_fpclass_sd_load(<2 x double>* %x0ptr) { +; CHECK-LABEL: test_int_x86_avx512_mask_fpclass_sd_load: +; CHECK: ## BB#0: +; CHECK-NEXT: vmovapd (%rdi), %xmm0 +; CHECK-NEXT: vfpclasssd $4, %xmm0, %k0 +; CHECK-NEXT: kmovw %k0, %eax +; CHECK-NEXT: ## kill: %AL %AL %EAX +; CHECK-NEXT: retq + %x0 = load <2 x double>, <2 x double>* %x0ptr + %res = call i8 @llvm.x86.avx512.mask.fpclass.sd(<2 x double> %x0, i32 4, i8 -1) + ret i8 %res +} + declare i8 @llvm.x86.avx512.mask.fpclass.ss(<4 x float>, i32, i8) define i8 @test_int_x86_avx512_mask_fpclass_ss(<4 x float> %x0, i8 %x1) { @@ -415,6 +428,19 @@ define i8 @test_int_x86_avx512_mask_fpclass_ss(<4 x float> %x0, i8 %x1) { ret i8 %res2 } +define i8 @test_int_x86_avx512_mask_fpclass_ss_load(<4 x float>* %x0ptr, i8 %x1) { +; CHECK-LABEL: test_int_x86_avx512_mask_fpclass_ss_load: +; CHECK: ## BB#0: +; CHECK-NEXT: vmovaps (%rdi), %xmm0 +; CHECK-NEXT: vfpclassss $4, %xmm0, %k0 +; CHECK-NEXT: kmovw %k0, %eax +; CHECK-NEXT: ## kill: %AL %AL %EAX +; CHECK-NEXT: retq + %x0 = load <4 x float>, <4 x float>* %x0ptr + %res = call i8 @llvm.x86.avx512.mask.fpclass.ss(<4 x float> %x0, i32 4, i8 -1) + ret i8 %res +} + declare i16 @llvm.x86.avx512.cvtd2mask.512(<16 x i32>) define i16@test_int_x86_avx512_cvtd2mask_512(<16 x i32> %x0) { -- 2.11.0