- User Since
- May 5 2014, 7:26 AM (276 w, 1 d)
LGTM - cheers
This was handled in rL367282
OK, thanks for checking - LGTM
Updated patch to correctly respect NSZ flags. I've also improved the logic to ensure we negate the most profitable multiplication factor, not just the first one.
LGTM - you just need to clang-format X86OptimizeLEAs.cpp as the addition of "X86" has screwed up indentations.
LGTM - thanks!
@vporpo Are you still looking at this please?
@lkail Are you still looking at this please?
@craig.topper Is this still relevant?
Is it worth adding in ISD::ABS support here ?
LGTM with more minor I noticed
a couple of minor comments
Sun, Aug 18
Fri, Aug 16
please rebase on rL369126
Please raise a bug if you manage to get a repro
LGTM - do we have constant folding coverage btw?
Thu, Aug 15
Please can you commit mulfix_combine.ll to trunk with current (trunk) codegen, then rebase this patch so that it shows the diff
LGTM - I've been trying to think of a general helper function we could create for this as I can only imagine that we're going to need this more....
I'm happy to remove the push/pull if reviewers prefer - its just a personal style preference tbh
Adding some people interested in MSVC builds
Disable MSVC divide by zero warnings for the whole of AlignmentTest.cpp
LGTM - cheers
Wed, Aug 14
Please could you add a couple of partial reduction tests (e.g. 4f32 in 8f32)
Not sure we should consider this NFC any more
This isn't handling all the cases - its probably better to push at the top of the file (before namespace) and similarly pop at the end of the file
OK, let's just go for v8i8
LGTM - I agree the AVX1 256-bit result cases might need further tweaking (especially for extension ratio = 2 as we can cheaply use movx+unpckh), but that can wait for now.