code review 4383047: ld: 25% faster

ld: 25% faster

The ld time was dominated by symbol table processing, so
  * increase hash table size
  * emit fewer symbols in gc (just 1 per string, 1 per type)
  * add read-only lookup to avoid creating spurious symbols
  * add linked list to speed whole-table traversals

Breaks dwarf generator (no idea why), so disable dwarf.

Reduces time for 6l to link godoc by 25%.

[rsc, 2011-04-09 13:44:19]

Hello ken2 (cc: golang-dev@googlegroups.com),

I'd like you to review this change to
https://go.googlecode.com/hg/

[rsc, 2011-04-09 13:44:27]

*** Submitted as http://code.google.com/p/go/source/detail?r=9a64273f9d68 ***

ld: 25% faster

The ld time was dominated by symbol table processing, so
  * increase hash table size
  * emit fewer symbols in gc (just 1 per string, 1 per type)
  * add read-only lookup to avoid creating spurious symbols
  * add linked list to speed whole-table traversals

Breaks dwarf generator (no idea why), so disable dwarf.

Reduces time for 6l to link godoc by 25%.

R=ken2
CC=golang-dev
http://codereview.appspot.com/4383047

[ken3, 2011-04-10 22:04:40]

lgtm

i think that you can still save time in the
lookups across the whole sym table by
putting a test before the call to strncmp.
example:

for(s = allsym; s != S; s = s->allsym) {
      if(s->name[0] == 't')
      if(strncmp(s->name, "type.", 5) == 0) {


On Sat, Apr 9, 2011 at 6:44 AM,  <rsc@golang.org> wrote:
> *** Submitted as
> http://code.google.com/p/go/source/detail?r=9a64273f9d68 ***
>
> ld: 25% faster
>
> The ld time was dominated by symbol table processing, so
>  * increase hash table size
>  * emit fewer symbols in gc (just 1 per string, 1 per type)
>  * add read-only lookup to avoid creating spurious symbols
>  * add linked list to speed whole-table traversals
>
> Breaks dwarf generator (no idea why), so disable dwarf.
>
> Reduces time for 6l to link godoc by 25%.
>
> R=ken2
> CC=golang-dev
> http://codereview.appspot.com/4383047
>
>
> http://codereview.appspot.com/4383047/
>