diff options
| author | Joel Jones <joel_k_jones@apple.com> | 2012-07-18 00:02:16 +0000 |
|---|---|---|
| committer | Joel Jones <joel_k_jones@apple.com> | 2012-07-18 00:02:16 +0000 |
| commit | 7c82e6a32a84e238c9e4e57dd43eaba540a79ce1 (patch) | |
| tree | ee690b3fa0c29940515aa6a66214f8190b7fc180 /test | |
| parent | 0b5f1aca32076e9a29dc3d716945d1ec6003c96e (diff) | |
| download | external_llvm-7c82e6a32a84e238c9e4e57dd43eaba540a79ce1.zip external_llvm-7c82e6a32a84e238c9e4e57dd43eaba540a79ce1.tar.gz external_llvm-7c82e6a32a84e238c9e4e57dd43eaba540a79ce1.tar.bz2 | |
More replacing of target-dependent intrinsics with target-indepdent
intrinsics. The second instruction(s) to be handled are the vector versions
of count set bits (ctpop).
The changes here are to clang so that it generates a target independent
vector ctpop when it sees an ARM dependent vector bits set count. The changes
in llvm are to match the target independent vector ctpop and in
VMCore/AutoUpgrade.cpp to update any existing bc files containing ARM
dependent vector pop counts with target-independent ctpops. There are also
changes to an existing test case in llvm for ARM vector count instructions and
to a test for the bitcode upgrade.
<rdar://problem/11892519>
There is deliberately no test for the change to clang, as so far as I know, no
consensus has been reached regarding how to test neon instructions in clang;
q.v. <rdar://problem/8762292>
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@160410 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test')
| -rw-r--r-- | test/Bitcode/arm32_neon_vcnt_upgrade.ll | 11 | ||||
| -rw-r--r-- | test/CodeGen/ARM/vcnt.ll | 12 |
2 files changed, 16 insertions, 7 deletions
diff --git a/test/Bitcode/arm32_neon_vcnt_upgrade.ll b/test/Bitcode/arm32_neon_vcnt_upgrade.ll index b3f2f03..10b9284 100644 --- a/test/Bitcode/arm32_neon_vcnt_upgrade.ll +++ b/test/Bitcode/arm32_neon_vcnt_upgrade.ll @@ -1,5 +1,5 @@ ; RUN: llvm-as < %s | llvm-dis | FileCheck %s -; NB: currently tests only vclz, should also test vcnt and vcls +; Tests vclz and vcnt define <4 x i16> @vclz16(<4 x i16>* %A) nounwind { ;CHECK: @vclz16 @@ -9,4 +9,13 @@ define <4 x i16> @vclz16(<4 x i16>* %A) nounwind { ret <4 x i16> %tmp2 } +define <8 x i8> @vcnt8(<8 x i8>* %A) nounwind { +;CHECK: @vcnt8 + %tmp1 = load <8 x i8>* %A + %tmp2 = call <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8> %tmp1) +;CHECK: call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> + ret <8 x i8> %tmp2 +} + declare <4 x i16> @llvm.arm.neon.vclz.v4i16(<4 x i16>) nounwind readnone +declare <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8>) nounwind readnone diff --git a/test/CodeGen/ARM/vcnt.ll b/test/CodeGen/ARM/vcnt.ll index 7e54524..9f55c24 100644 --- a/test/CodeGen/ARM/vcnt.ll +++ b/test/CodeGen/ARM/vcnt.ll @@ -3,22 +3,22 @@ define <8 x i8> @vcnt8(<8 x i8>* %A) nounwind { ;CHECK: vcnt8: -;CHECK: vcnt.8 +;CHECK: vcnt.8 {{d[0-9]+}}, {{d[0-9]+}} %tmp1 = load <8 x i8>* %A - %tmp2 = call <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8> %tmp1) + %tmp2 = call <8 x i8> @llvm.ctpop.v8i8(<8 x i8> %tmp1) ret <8 x i8> %tmp2 } define <16 x i8> @vcntQ8(<16 x i8>* %A) nounwind { ;CHECK: vcntQ8: -;CHECK: vcnt.8 +;CHECK: vcnt.8 {{q[0-9]+}}, {{q[0-9]+}} %tmp1 = load <16 x i8>* %A - %tmp2 = call <16 x i8> @llvm.arm.neon.vcnt.v16i8(<16 x i8> %tmp1) + %tmp2 = call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %tmp1) ret <16 x i8> %tmp2 } -declare <8 x i8> @llvm.arm.neon.vcnt.v8i8(<8 x i8>) nounwind readnone -declare <16 x i8> @llvm.arm.neon.vcnt.v16i8(<16 x i8>) nounwind readnone +declare <8 x i8> @llvm.ctpop.v8i8(<8 x i8>) nounwind readnone +declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>) nounwind readnone define <8 x i8> @vclz8(<8 x i8>* %A) nounwind { ;CHECK: vclz8: |