diff options
author | Justin Holewinski <jholewinski@nvidia.com> | 2013-06-28 17:58:13 +0000 |
---|---|---|
committer | Justin Holewinski <jholewinski@nvidia.com> | 2013-06-28 17:58:13 +0000 |
commit | f51a22e815a7494cbe1aba25d275015a70b5c31d (patch) | |
tree | 1abd574a3cced1781dd5add580a196854a952d02 /test/CodeGen | |
parent | ac78a0645ddd2046fb66237ba4cfadffa2d367d7 (diff) | |
download | external_llvm-f51a22e815a7494cbe1aba25d275015a70b5c31d.zip external_llvm-f51a22e815a7494cbe1aba25d275015a70b5c31d.tar.gz external_llvm-f51a22e815a7494cbe1aba25d275015a70b5c31d.tar.bz2 |
[NVPTX] Add (1.0 / sqrt(x)) => rsqrt(x) generation when allowable by FP flags
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@185178 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen')
-rw-r--r-- | test/CodeGen/NVPTX/rsqrt.ll | 13 |
1 files changed, 13 insertions, 0 deletions
diff --git a/test/CodeGen/NVPTX/rsqrt.ll b/test/CodeGen/NVPTX/rsqrt.ll new file mode 100644 index 0000000..d49eebe --- /dev/null +++ b/test/CodeGen/NVPTX/rsqrt.ll @@ -0,0 +1,13 @@ +; RUN: llc < %s -march=nvptx -mcpu=sm_20 -nvptx-prec-divf32=1 -nvptx-prec-sqrtf32=0 | FileCheck %s + +target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64" + +declare float @llvm.nvvm.sqrt.f(float) + +define float @foo(float %a) { +; CHECK: rsqrt.approx.f32 + %val = tail call float @llvm.nvvm.sqrt.f(float %a) + %ret = fdiv float 1.0, %val + ret float %ret +} +
\ No newline at end of file |