Page MenuHomeFreeBSD

arm atomic-v6.h: Use eoreqs rather than eorseq to appease GNU as.
AcceptedPublic

Authored by jhb on Sep 27 2022, 5:42 PM.
Tags
None
Referenced Files
Unknown Object (File)
Tue, Jun 4, 7:57 AM
Unknown Object (File)
May 18 2024, 11:49 AM
Unknown Object (File)
May 17 2024, 4:30 PM
Unknown Object (File)
May 17 2024, 1:05 AM
Unknown Object (File)
May 16 2024, 2:38 PM
Unknown Object (File)
Apr 8 2024, 5:55 PM
Unknown Object (File)
Apr 1 2024, 1:13 PM
Unknown Object (File)
Mar 18 2024, 6:37 AM
Subscribers

Details

Reviewers
imp
manu
andrew
Summary

GNU as seems to require the 's' flag after the condition code
rather than before it which seems consistent with the instruction
description for EOR in the ARM ARM. Curiously, however, objdump -d
from binutils displays the instruction as eorseq.

Diff Detail

Repository
rG FreeBSD src repository
Lint
Lint Skipped
Unit
Tests Skipped
Build Status
Buildable 47569
Build 44456: arc lint + arc unit

Event Timeline

jhb requested review of this revision.Sep 27 2022, 5:42 PM

If it build with both clang and gcc, this is fine

This revision is now accepted and ready to land.Sep 27 2022, 5:43 PM

eorseq is correct as it's created from the eors instruction and eq condition code.

My copy of the ARM ARM (for Armv8/9) has it as EORS{<c>}{<q>} where <c> is the condition code. The <q> is unimportant here as it's to tell the assembler to use a 16 or 32 bit Thumb instruction.

I guess GCC isn't emitting .syntax unified to enable Unified Assembler Language.

Hmm, shouldn't the -masm-syntax-unified be added to gcc CFLAGS? (just guessing, untested)

Hmm, shouldn't the -masm-syntax-unified be added to gcc CFLAGS? (just guessing, untested)

Yes, I just haven't gotten back to testing the right CFLAGS fix.