cfallin edited PR #4785 from aarch64-isle-loads
to main
:
This retains
lower_amode
in the handwritten code (@akirilov-arm
reports that there is an upcoming patch to port this), but tweaks it
slightly to take aValue
rather than anInst
.
cfallin updated PR #4785 from aarch64-isle-loads
to main
.
elliottt submitted PR review.
elliottt created PR review comment:
Did you mean to remove the
Sextend
case?
elliottt submitted PR review.
cfallin submitted PR review.
cfallin created PR review comment:
Argh, no, I was playing with refactoring this a bit (splitting by uextend/sextend toplevel) and didn't revert completely. Thanks for catching!
cfallin updated PR #4785 from aarch64-isle-loads
to main
.
cfallin has enabled auto merge for PR #4785.
cfallin has disabled auto merge for PR #4785.
cfallin created PR review comment:
Also, now that I'm looking at this again, I had a momentary panic that we had another sign-extend bug (déjà-vu!); the reason that this isn't, for anyone watching, is that for narrower-than-32 extends we push the extended value (the output of the extend op) onto
addends64
, rather than the input to the extend. The 32-bit case is handled above specially with the extend incorporated into the amode if possible.
cfallin submitted PR review.
cfallin updated PR #4785 from aarch64-isle-loads
to main
.
cfallin requested elliottt for a review on PR #4785.
cfallin updated PR #4785 from aarch64-isle-loads
to main
.
elliottt submitted PR review.
cfallin merged PR #4785.
Last updated: Dec 23 2024 at 12:05 UTC