[llvm-commits] [PATCH] AArch64 backend: LLVM changes

[Renato Golin Linaro]

Hi Tim,

I had a look and I'm afraid I haven't seen 0.1% of it. I don't think anyone
will have time to review every line of it and not end up in a mental
hospital, so I had a look at a few files, mostly the modified ones, and
ignored the new ones.

Overall, the code seems fairly orthogonal to the current back-ends, and
there's a large number of tests, which will help keeping it in shape in the
near future.

A few questions:

Have you run the build and check-all on an ARM32 system? I say that because
currently the ARM buildbots have at least three different types of failure,
not to mention a "make clean" error, so adding a whole new back-end that
has its own failures on ARM (especially because you probably develop on a
64-bit machine), won't help matters.

Now to code specific questions:

+// Vector Multiply-subtract (Floating Point)
+def int_aarch64_neon_vfms : Neon_3Arg_Intrinsic;
(...)
+// Vector Multiply Extended (Floating Point)
+def int_aarch64_neon_vmulx : Neon_2Arg_Intrinsic;

Can't these be patterns?


-      APFloat apf(APInt::getNullValue(VT.getSizeInBits()));
+      APFloat apf(APInt::getNullValue(VT.getSizeInBits()), /*isIEEE=*/
true);

I thought you had changed APFloat...


-        ? dwarf::DW_EH_PE_udata4 : dwarf::DW_EH_PE_absptr;
+         ? dwarf::DW_EH_PE_udata4 : dwarf::DW_EH_PE_absptr;

unnecessary space change, might confuse history


That's all for now. I might have another look further down, but not sure
that would actually help anything... ;)

cheers,
--renato
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20130112/6364ed03/attachment.html>