[x86] Introduce tests covering the gamut of 256-bit vector shuffling.
authorChandler Carruth <chandlerc@gmail.com>
Mon, 22 Sep 2014 20:25:08 +0000 (20:25 +0000)
committerChandler Carruth <chandlerc@gmail.com>
Mon, 22 Sep 2014 20:25:08 +0000 (20:25 +0000)
commit56c7cfe41f0364a5ac15db17ec4876b09acc012f
tree4272fe2bcf446bbc0a2a8046a17410b4387ada73
parenta956b28d32a23fa321020fbc94b49d61164004d8
[x86] Introduce tests covering the gamut of 256-bit vector shuffling.

These are just test cases, no actual code yet. This establishes the
baseline fallback strategy we're starting from on AVX2 and the expected
lowering we use on AVX1.

Also, these test cases are very much generated. I've manually crafted
the specific pattern set that I'm hoping will be useful at exercising
the lowering code, but I've not (and could not) manually verify *all* of
these. I've spot checked and they seem legit to me.

As with the rest of vector shuffling, at a certain point the only really
useful way to check the correctness of this stuff is through fuzz
testing.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@218267 91177308-0d34-0410-b5e6-96231b3b80d8
test/CodeGen/X86/vector-shuffle-256-v16.ll [new file with mode: 0644]
test/CodeGen/X86/vector-shuffle-256-v32.ll [new file with mode: 0644]
test/CodeGen/X86/vector-shuffle-256-v8.ll