[PATCH] D55234: Do not use a hash table to uniquify mergeable strings.

[George Rimar via Phabricator via llvm-commits]

grimar added a comment.

> Program         w/patch         w/o patch       Memory saving
>  chrome          1163088         1163572         0.04%

I guess that was not the debug chrome perhaps.
That might explain why it is 0.81% slower but saves an only minor amount of memory.

I am not sure what I feel about this patch. It can make the link faster in some cases,
but for regular use (linking not huge apps, like clang) it always seems to be up to a few percents slower.

Saving 4.18%, or 200Mb for Mozilla (4631424->4833940) is good, but is that so significant and worth doing?
200Mb is only 200Mb, while -2.77% of link time looks much more impressive to me and
unfortunately, the link time for most of the apps tested is slower with this patch.

Given that this patch is about a balance about finding the balance between speed and memory,
I would like to hear from other people on this.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D55234/new/

https://reviews.llvm.org/D55234