[PATCH] D154854: Use empty symbol name for XCOFF text csect
Hubert Tong via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Jul 11 13:54:50 PDT 2023
hubert.reinterpretcast added inline comments.
================
Comment at: llvm/lib/MC/MCObjectFileInfo.cpp:923
// get placed into this csect. The choice of csect name is not a property of
- // the ABI or object file format. For example, the XL compiler uses an unnamed
- // csect for program code.
+ // the ABI or object file format.
TextSection = Ctx->getXCOFFSection(
----------------
Comment wording suggestion.
================
Comment at: llvm/test/CodeGen/PowerPC/aix-pgo-function-sections.ll:123
; OBJ64-NEXT: 0000000000000000 *UND* 0000000000000000 external_func
-; OBJ64-NEXT: 0000000000000000 l .text 0000000000000000 .text
+; OBJ64-NEXT: 0000000000000000 l .text 0000000000000000
; OBJ64-NEXT: 0000000000000000 g .text 0000000000000033 .func1
----------------
It is somewhat unfortunate that the default tool outputs for empty names is especially bad, but I don't want to hold up this patch on changing that.
================
Comment at: llvm/test/CodeGen/PowerPC/aix-text.ll:2
+; RUN: llc -filetype=obj --function-sections -mtriple powerpc-ibm-aix-xcoff -o %t.o < %s
+; RUN: llvm-objdump --syms %t.o | FileCheck %s
+; RUN: llc -filetype=obj --function-sections -mtriple powerpc64-ibm-aix-xcoff -o %t64.o < %s
----------------
Since this is a new test, I think we should use `--symbol-description` (same for other case below).
================
Comment at: llvm/test/CodeGen/PowerPC/aix-text.ll:40
+!2 = !{i32 7, !"frame-pointer", i32 2}
+!3 = !{!"IBM Open XL C/C++ for AIX 17.1.2 (5725-C72, 5765-J18), version 17.1.2.0, clang version 17.0.0"}
+
----------------
I think it should be possible to remove this (and the directive referring to this). Please try to remove the other metadata and attributes (and their uses) as well.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D154854/new/
https://reviews.llvm.org/D154854
More information about the llvm-commits
mailing list