diff options
| author | Arnold Schwaighofer <aschwaighofer@apple.com> | 2013-02-20 21:33:32 +0000 |
|---|---|---|
| committer | Arnold Schwaighofer <aschwaighofer@apple.com> | 2013-02-20 21:33:32 +0000 |
| commit | 3f9568e921d706b0755b5edb2f0649ef8485ef17 (patch) | |
| tree | 898743473e4b0ec73b5d4df6fa24bd9675f3a87b /llvm/test/CodeGen | |
| parent | 64c2e69809bc09db413abc740c0bc881ade2d39e (diff) | |
| download | bcm5719-llvm-3f9568e921d706b0755b5edb2f0649ef8485ef17.tar.gz bcm5719-llvm-3f9568e921d706b0755b5edb2f0649ef8485ef17.zip | |
DAGCombiner: Fold pointless truncate, bitcast, buildvector series
(2xi32) (truncate ((2xi64) bitcast (buildvector i32 a, i32 x, i32 b, i32 y)))
can be folded into a (2xi32) (buildvector i32 a, i32 b).
Such a DAG would cause uneccessary vdup instructions followed by vmovn
instructions.
We generate this code on ARM NEON for a setcc olt, 2xf64, 2xf64. For example, in
the vectorized version of the code below.
double A[N];
double B[N];
void test_double_compare_to_double() {
int i;
for(i=0;i<N;i++)
A[i] = (double)(A[i] < B[i]);
}
radar://13191881
Fixes bug 15283.
llvm-svn: 175670
Diffstat (limited to 'llvm/test/CodeGen')
| -rw-r--r-- | llvm/test/CodeGen/ARM/neon_cmp.ll | 15 |
1 files changed, 15 insertions, 0 deletions
diff --git a/llvm/test/CodeGen/ARM/neon_cmp.ll b/llvm/test/CodeGen/ARM/neon_cmp.ll new file mode 100644 index 00000000000..046b5da2289 --- /dev/null +++ b/llvm/test/CodeGen/ARM/neon_cmp.ll @@ -0,0 +1,15 @@ +; RUN: llc < %s -march=arm -mcpu=cortex-a9 | FileCheck %s +; bug 15283 +; radar://13191881 +; CHECK: vfcmp +define void @vfcmp(<2 x double>* %a, <2 x double>* %b) { + %wide.load = load <2 x double>* %a, align 4 + %wide.load2 = load <2 x double>* %b, align 4 +; CHECK-NOT: vdup.32 +; CHECK-NOT: vmovn.i64 + %v1 = fcmp olt <2 x double> %wide.load, %wide.load2 + %v2 = zext <2 x i1> %v1 to <2 x i32> + %v3 = sitofp <2 x i32> %v2 to <2 x double> + store <2 x double> %v3, <2 x double>* %b, align 4 + ret void +} |
