AVX128: More various instructions #3740
Conversation
|
I was going to flag the copypaste in UCOMIS, instead of doing that I wrote #3741 to use 😄 |
Cherry-picked and used. |
|
Wrote #3743 for all the Literals flying around here, but I don't mind merging without porting to that, I can do the (fully mechanical) changes after merge if you prefer. |
Yeesss, that would be good to do in one lump commit once the 66 remaining commits are merged. Since there will be a decent amount of them after the fact and it will be painful to go through each individual commit |
Sure, that's fine with me. Although https://github.com/torbiak/git-autofixup is basically tailormade for this problem - you do all the changes at the end, run (Not saying you should do this, just singing the praises of this script!) |
I told Ryan to change this for AVX, but it needs to be changed in the original to match! Signed-off-by: Alyssa Rosenzweig <[email protected]>
|
I see the fcmp funniness is an existing funniness. So I wrote you #3744 |
|
VPSRA/VPSLL/VPSRL look like candidates to use the lambda helper. If that doesn't work for some reason (I don't see such a reason but maybe there's a subtlelty I'm missing here), we can at least define a common implementation for the 3 and use DeriveOp on an argument. Likewise for VPSLLI/VPSRAI/VPSRLI and VPSRLDQ/VPSLLDQ. I guess I'll push a fixup. |
|
oops accidental close |
so AVX128 can reuse it. Signed-off-by: Alyssa Rosenzweig <[email protected]>
|
I just pushed a version with all my comments addressed. I'm happy to merge it in this state if you are. Diff between what I pushed and what you pushed rebased on #3743, for review convenience: https://rosenzweig.io/diff.txt |
Looks good to me |
These are very closely related to each other so it makes sense to implement the roughly three different families in one commit.
A smattering of scalar operations and vector shifts.