[llvm] [AArch64] Enable GlobalMerge on externals (PR #158592)

[Sjoerd Meijer via llvm-commits]

================
@@ -694,12 +694,6 @@ bool AArch64PassConfig::addPreISel() {
     // is disabled as we emit the .subsections_via_symbols directive which
     // means that merging extern globals is not safe.
     bool MergeExternalByDefault = !TM->getTargetTriple().isOSBinFormatMachO();
-
-    // FIXME: extern global merging is only enabled when we optimise for size
-    // because there are some regressions with it also enabled for performance.
----------------
sjoerdmeijer wrote:

I was going to ask about these "some regressions", but did a git blame first and noticed I contributed this, so yes, that's funny and I can't remember what they were. Does this ring a bell to you, @davemgreen ? If I had to take a guess, then I did some measurements on smaller A-cores, i.e. not the wide X or Neoverse cores we have today. You did mention that it didn't affect perf very much. I am fine with the change, but my only question out of curiousity is which cores you've tested.

https://github.com/llvm/llvm-project/pull/158592