llvm/test/CodeGen/NVPTX/fma-assoc.ll
Nicolai Haehnle ab43652716 [DAGCombine] require UnsafeFPMath for re-association of addition
Summary:
The affected transforms all implicitly use associativity of addition,
for which we usually require unsafe math to be enabled.

The "Aggressive" flag is only meant to convey information about the
performance of the fused ops relative to a fmul+fadd sequence.

Fixes Bug 31626.

Reviewers: spatel, hfinkel, mehdi_amini, arsenm, tstellarAMD

Subscribers: jholewinski, nemanjai, wdng, llvm-commits

Differential Revision: https://reviews.llvm.org/D28675

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@293635 91177308-0d34-0410-b5e6-96231b3b80d8
2017-01-31 14:35:37 +00:00

40 lines
1.3 KiB
LLVM

; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast | FileCheck %s -check-prefix=CHECK
; RUN: llc < %s -march=nvptx -mcpu=sm_20 -fp-contract=fast -enable-unsafe-fp-math | FileCheck %s -check-prefix=CHECK -check-prefix=CHECK-UNSAFE
define ptx_device float @t1_f32(float %x, float %y, float %z,
float %u, float %v) {
; CHECK-UNSAFE: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
; CHECK-UNSAFE: fma.rn.f32 %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}}, %f{{[0-9]+}};
; CHECK: ret;
%a = fmul float %x, %y
%b = fmul float %u, %v
%c = fadd float %a, %b
%d = fadd float %c, %z
ret float %d
}
define ptx_device double @t1_f64(double %x, double %y, double %z,
double %u, double %v) {
; CHECK-UNSAFE: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
; CHECK-UNSAFE: fma.rn.f64 %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}}, %fd{{[0-9]+}};
; CHECK: ret;
%a = fmul double %x, %y
%b = fmul double %u, %v
%c = fadd double %a, %b
%d = fadd double %c, %z
ret double %d
}
define double @two_choices(double %val1, double %val2) {
; CHECK-LABEL: two_choices(
; CHECK: mul.f64
; CHECK-NOT: mul.f64
; CHECK: fma.rn.f64
%1 = fmul double %val1, %val2
%2 = fmul double %1, %1
%3 = fadd double %1, %2
ret double %3
}