[PATCH] D71144: [AIX] Use csect reference for function address constants

Jason Liu via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Tue Dec 17 13:56:35 PST 2019


jasonliu added inline comments.


================
Comment at: llvm/test/CodeGen/PowerPC/aix-reference-func-addr-const.ll:2
+; RUN: llc -verify-machineinstrs -mcpu=pwr7 -mtriple powerpc-ibm-aix-xcoff < %s | FileCheck %s
+; RUN: llc -verify-machineinstrs -mcpu=pwr7 -mtriple powerpc64-ibm-aix-xcoff < %s | FileCheck --check-prefix=CHECK64 %s
+
----------------
hubert.reinterpretcast wrote:
> @cebowleratibm, we're still wondering the reason behind using `pwr7` here. When do we choose to use `pwr4`?
Discussed offline today, I will try to summarize the discussion.
Setting it to pwr7 makes people wondering about the reason of why we are choosing pwr7 (which is not our base support on AIX). So for consistency, we should just use -mcpu=pwr4, the base cpu we choose to support instead. 
For testing that could be cpu level sensitive (ABI for example), we should test both the base, and the cpu level that's in interest. 
In the future, if we decide to move up the default cpu level support on AIX, we should modify all the test case that intend to test the default cpu level to that higher new default cpu level.
Also part of the reason we have this discussion right now is that our default cpu level for AIX is not correct right now. (if it is, then we can just omit the -mcpu option all together.) So we should consider fix this issue as well. 


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71144/new/

https://reviews.llvm.org/D71144





More information about the llvm-commits mailing list