diff options
author | Chris Lattner <sabre@nondot.org> | 2009-02-03 18:15:05 +0000 |
---|---|---|
committer | Chris Lattner <sabre@nondot.org> | 2009-02-03 18:15:05 +0000 |
commit | 1a3257bbf53eff4c7cfcbef972dd382f7baa7592 (patch) | |
tree | 7d60231912299c206c4ec23545d341fc788b18d3 /test/Transforms/ScalarRepl | |
parent | 67e3ba3f1dda6a5e4a9c774fbf0f96ec18db111a (diff) | |
download | llvm-1a3257bbf53eff4c7cfcbef972dd382f7baa7592.tar.gz llvm-1a3257bbf53eff4c7cfcbef972dd382f7baa7592.tar.bz2 llvm-1a3257bbf53eff4c7cfcbef972dd382f7baa7592.tar.xz |
Make SROA produce a vector only when the alloca is actually
accessed at least once as a vector. This prevents it from
compiling the example in not-a-vector into:
define double @test(double %A, double %B) {
%tmp4 = insertelement <7 x double> undef, double %A, i32 0
%tmp = insertelement <7 x double> %tmp4, double %B, i32 4
%tmp2 = extractelement <7 x double> %tmp, i32 4
ret double %tmp2
}
instead, producing the integer code. Producing vectors when they
aren't otherwise in the program is dangerous because a lot of other
code treats them carefully and doesn't want to break them down.
OTOH, many things want to break down tasty i448's.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@63638 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/Transforms/ScalarRepl')
-rw-r--r-- | test/Transforms/ScalarRepl/not-a-vector.ll | 19 |
1 files changed, 19 insertions, 0 deletions
diff --git a/test/Transforms/ScalarRepl/not-a-vector.ll b/test/Transforms/ScalarRepl/not-a-vector.ll new file mode 100644 index 0000000000..e2111e7b31 --- /dev/null +++ b/test/Transforms/ScalarRepl/not-a-vector.ll @@ -0,0 +1,19 @@ +; RUN: llvm-as < %s | opt -scalarrepl | llvm-dis | not grep alloca +; RUN: llvm-as < %s | opt -scalarrepl | llvm-dis | not grep {7 x double} +; RUN: llvm-as < %s | opt -scalarrepl -instcombine | llvm-dis | grep {ret double %B} + +define double @test(double %A, double %B) { + %ARR = alloca [7 x i64] + %C = bitcast [7 x i64]* %ARR to double* + store double %A, double* %C + + %D = getelementptr [7 x i64]* %ARR, i32 0, i32 4 + %E = bitcast i64* %D to double* + store double %B, double* %E + + %F = getelementptr double* %C, i32 4 + %G = load double* %F + ret double %G +} + + |