forked from OSchip/llvm-project
19 lines
598 B
LLVM
19 lines
598 B
LLVM
|
; RUN: llc -march=hexagon -mcpu=hexagonv5 < %s | FileCheck %s
|
||
|
|
; Check that we generate double precision floating point multiply in V5.
|
||
|
|
|
||
|
|
; CHECK: r{{[0-9]+}}:{{[0-9]+}} = dfmpy(r{{[0-9]+}}:{{[0-9]+}}, r{{[0-9]+}}:{{[0-9]+}})
|
||
|
|
|
||
|
|
define i32 @main() nounwind {
|
||
|
|
entry:
|
||
|
|
%a = alloca double, align 8
|
||
|
|
%b = alloca double, align 8
|
||
|
|
%c = alloca double, align 8
|
||
|
|
store double 1.540000e+01, double* %a, align 8
|
||
|
|
store double 9.100000e+00, double* %b, align 8
|
||
|
|
%0 = load double* %b, align 8
|
||
|
|
%1 = load double* %a, align 8
|
||
|
|
%mul = fmul double %0, %1
|
||
|
|
store double %mul, double* %c, align 8
|
||
|
|
ret i32 0
|
||
|
|
}
|