mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-23 11:13:28 +01:00
f3ab4849fb
Now that we have fneg, prefer using it over "fsub -0.0, ...". This helps in particular with strict FP tests, as fneg does not raise any exceptions.
39 lines
1.4 KiB
LLVM
39 lines
1.4 KiB
LLVM
; Test strict vector multiply-and-add on z14.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z14 | FileCheck %s
|
|
|
|
declare <4 x float> @llvm.experimental.constrained.fma.v4f32(<4 x float>, <4 x float>, <4 x float>, metadata, metadata)
|
|
|
|
; Test a v4f32 multiply-and-add.
|
|
define <4 x float> @f1(<4 x float> %dummy, <4 x float> %val1,
|
|
<4 x float> %val2, <4 x float> %val3) #0 {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: vfmasb %v24, %v26, %v28, %v30
|
|
; CHECK: br %r14
|
|
%ret = call <4 x float> @llvm.experimental.constrained.fma.v4f32 (
|
|
<4 x float> %val1,
|
|
<4 x float> %val2,
|
|
<4 x float> %val3,
|
|
metadata !"round.dynamic",
|
|
metadata !"fpexcept.strict") #0
|
|
ret <4 x float> %ret
|
|
}
|
|
|
|
; Test a v4f32 multiply-and-subtract.
|
|
define <4 x float> @f2(<4 x float> %dummy, <4 x float> %val1,
|
|
<4 x float> %val2, <4 x float> %val3) #0 {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: vfmssb %v24, %v26, %v28, %v30
|
|
; CHECK: br %r14
|
|
%negval3 = fneg <4 x float> %val3
|
|
%ret = call <4 x float> @llvm.experimental.constrained.fma.v4f32 (
|
|
<4 x float> %val1,
|
|
<4 x float> %val2,
|
|
<4 x float> %negval3,
|
|
metadata !"round.dynamic",
|
|
metadata !"fpexcept.strict") #0
|
|
ret <4 x float> %ret
|
|
}
|
|
|
|
attributes #0 = { strictfp }
|