diff options
| author | Chris Lattner <sabre@nondot.org> | 2010-01-01 03:32:16 +0000 |
|---|---|---|
| committer | Chris Lattner <sabre@nondot.org> | 2010-01-01 03:32:16 +0000 |
| commit | f031e8ad011e9ad95d7c965936da07e3a9c42add (patch) | |
| tree | f548e6e940434f362f095323d22b87f189df7433 /test/CodeGen/X86/powi.ll | |
| parent | 0fba8cf9ffbb4397d3ddbb51469c8819d0faf3cc (diff) | |
| download | llvm-f031e8ad011e9ad95d7c965936da07e3a9c42add.tar.gz llvm-f031e8ad011e9ad95d7c965936da07e3a9c42add.tar.bz2 llvm-f031e8ad011e9ad95d7c965936da07e3a9c42add.tar.xz | |
Teach codegen to lower llvm.powi to an efficient (but not optimal)
multiply sequence when the power is a constant integer. Before, our
codegen for std::pow(.., int) always turned into a libcall, which was
really inefficient.
This should also make many gfortran programs happier I'd imagine.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@92388 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/X86/powi.ll')
| -rw-r--r-- | test/CodeGen/X86/powi.ll | 11 |
1 files changed, 11 insertions, 0 deletions
diff --git a/test/CodeGen/X86/powi.ll b/test/CodeGen/X86/powi.ll new file mode 100644 index 0000000000..c3d68312ce --- /dev/null +++ b/test/CodeGen/X86/powi.ll @@ -0,0 +1,11 @@ +; RUN: llc %s -march=x86 -mcpu=yonah -o - | grep mulsd | count 6 +; Ideally this would compile to 5 multiplies. + +define double @_Z3f10d(double %a) nounwind readonly ssp noredzone { +entry: + %0 = tail call double @llvm.powi.f64(double %a, i32 15) nounwind ; <double> [#uses=1] + ret double %0 +} + +declare double @llvm.powi.f64(double, i32) nounwind readonly + |