[X86] Add test cases and command lines demonstrating how we accidentally select vrangeps/vrangepd from vrangess/vrangesd instrinsics when the rounding mode is CUR_DIRECTION
llvm-svn: 317998
This commit is contained in:
Craig Topper
parent
1382932c12
commit
6b53c4a982
|
|
@ -1,5 +1,6 @@
|
|||
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
|
||||
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512dq | FileCheck %s
|
||||
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512dq | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512DQ
|
||||
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512dq,avx512vl | FileCheck %s --check-prefix=CHECK --check-prefix=AVX512DQVL
|
||||
|
||||
declare <8 x i64> @llvm.x86.avx512.mask.cvtpd2qq.512(<8 x double>, <8 x i64>, i8, i32)
|
||||
|
||||
|
|
@ -276,17 +277,31 @@ define <4 x float>@test_int_x86_avx512_mask_reduce_ss(<4 x float> %x0, <4 x floa
|
|||
declare <4 x float> @llvm.x86.avx512.mask.range.ss(<4 x float>, <4 x float>,<4 x float>, i8, i32, i32)
|
||||
|
||||
define <4 x float>@test_int_x86_avx512_mask_range_ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x3, i8 %x4) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_range_ss:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; CHECK-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm0
|
||||
; CHECK-NEXT: vaddps %xmm0, %xmm2, %xmm0
|
||||
; CHECK-NEXT: retq
|
||||
; AVX512DQ-LABEL: test_int_x86_avx512_mask_range_ss:
|
||||
; AVX512DQ: ## BB#0:
|
||||
; AVX512DQ-NEXT: kmovw %edi, %k1
|
||||
; AVX512DQ-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; AVX512DQ-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm3
|
||||
; AVX512DQ-NEXT: vrangess $4, %xmm1, %xmm0, %xmm0
|
||||
; AVX512DQ-NEXT: vaddps %xmm3, %xmm2, %xmm1
|
||||
; AVX512DQ-NEXT: vaddps %xmm1, %xmm0, %xmm0
|
||||
; AVX512DQ-NEXT: retq
|
||||
;
|
||||
; AVX512DQVL-LABEL: test_int_x86_avx512_mask_range_ss:
|
||||
; AVX512DQVL: ## BB#0:
|
||||
; AVX512DQVL-NEXT: kmovw %edi, %k1
|
||||
; AVX512DQVL-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; AVX512DQVL-NEXT: vrangess $4, {sae}, %xmm1, %xmm0, %xmm3
|
||||
; AVX512DQVL-NEXT: vaddps %xmm3, %xmm2, %xmm2
|
||||
; AVX512DQVL-NEXT: vrangeps $4, %xmm1, %xmm0, %xmm0
|
||||
; AVX512DQVL-NEXT: vaddps %xmm2, %xmm0, %xmm0
|
||||
; AVX512DQVL-NEXT: retq
|
||||
%res = call <4 x float> @llvm.x86.avx512.mask.range.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x3, i8 %x4, i32 4, i32 8)
|
||||
%res1 = call <4 x float> @llvm.x86.avx512.mask.range.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x3, i8 -1, i32 4, i32 8)
|
||||
%res2 = fadd <4 x float> %res, %res1
|
||||
ret <4 x float> %res2
|
||||
%res2 = call <4 x float> @llvm.x86.avx512.mask.range.ss(<4 x float> %x0, <4 x float> %x1, <4 x float> %x3, i8 -1, i32 4, i32 4)
|
||||
%res3 = fadd <4 x float> %res, %res1
|
||||
%res4 = fadd <4 x float> %res2, %res3
|
||||
ret <4 x float> %res4
|
||||
}
|
||||
|
||||
declare <2 x double> @llvm.x86.avx512.mask.reduce.sd(<2 x double>, <2 x double>,<2 x double>, i8, i32, i32)
|
||||
|
|
@ -308,17 +323,31 @@ define <2 x double>@test_int_x86_avx512_mask_reduce_sd(<2 x double> %x0, <2 x do
|
|||
declare <2 x double> @llvm.x86.avx512.mask.range.sd(<2 x double>, <2 x double>,<2 x double>, i8, i32, i32)
|
||||
|
||||
define <2 x double>@test_int_x86_avx512_mask_range_sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x3, i8 %x4) {
|
||||
; CHECK-LABEL: test_int_x86_avx512_mask_range_sd:
|
||||
; CHECK: ## BB#0:
|
||||
; CHECK-NEXT: kmovw %edi, %k1
|
||||
; CHECK-NEXT: vrangesd $4, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; CHECK-NEXT: vrangesd $4, {sae}, %xmm1, %xmm0, %xmm0
|
||||
; CHECK-NEXT: vaddpd %xmm0, %xmm2, %xmm0
|
||||
; CHECK-NEXT: retq
|
||||
; AVX512DQ-LABEL: test_int_x86_avx512_mask_range_sd:
|
||||
; AVX512DQ: ## BB#0:
|
||||
; AVX512DQ-NEXT: vrangesd $4, %xmm1, %xmm0, %xmm3
|
||||
; AVX512DQ-NEXT: kmovw %edi, %k1
|
||||
; AVX512DQ-NEXT: vrangesd $4, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; AVX512DQ-NEXT: vrangesd $4, {sae}, %xmm1, %xmm0, %xmm0
|
||||
; AVX512DQ-NEXT: vaddpd %xmm0, %xmm2, %xmm0
|
||||
; AVX512DQ-NEXT: vaddpd %xmm0, %xmm3, %xmm0
|
||||
; AVX512DQ-NEXT: retq
|
||||
;
|
||||
; AVX512DQVL-LABEL: test_int_x86_avx512_mask_range_sd:
|
||||
; AVX512DQVL: ## BB#0:
|
||||
; AVX512DQVL-NEXT: vrangepd $4, %xmm1, %xmm0, %xmm3
|
||||
; AVX512DQVL-NEXT: kmovw %edi, %k1
|
||||
; AVX512DQVL-NEXT: vrangesd $4, %xmm1, %xmm0, %xmm2 {%k1}
|
||||
; AVX512DQVL-NEXT: vrangesd $4, {sae}, %xmm1, %xmm0, %xmm0
|
||||
; AVX512DQVL-NEXT: vaddpd %xmm0, %xmm2, %xmm0
|
||||
; AVX512DQVL-NEXT: vaddpd %xmm0, %xmm3, %xmm0
|
||||
; AVX512DQVL-NEXT: retq
|
||||
%res = call <2 x double> @llvm.x86.avx512.mask.range.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x3, i8 %x4, i32 4, i32 4)
|
||||
%res1 = call <2 x double> @llvm.x86.avx512.mask.range.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x3, i8 -1, i32 4, i32 8)
|
||||
%res2 = fadd <2 x double> %res, %res1
|
||||
ret <2 x double> %res2
|
||||
%res2 = call <2 x double> @llvm.x86.avx512.mask.range.sd(<2 x double> %x0, <2 x double> %x1, <2 x double> %x3, i8 -1, i32 4, i32 4)
|
||||
%res3 = fadd <2 x double> %res, %res1
|
||||
%res4 = fadd <2 x double> %res2, %res3
|
||||
ret <2 x double> %res4
|
||||
}
|
||||
|
||||
declare i8 @llvm.x86.avx512.mask.fpclass.pd.512(<8 x double>, i32, i8)
|
||||
|
|
|
|||
Loading…
Reference in New Issue