Thread (55 messages) flat view 55 messages, 11 authors, 2016-06-15

Re: [PATCH 2/2] Implement a simple delta_base cache

From: Linus Torvalds <torvalds@linux-foundation.org>
Date: 2016-06-15 22:43:00


On Tue, 20 Mar 2007, David Brodsky wrote:
And (hopefully) permanent http://steamer.kajka.koleje.cuni.cz/Eclipse
Ok, thanks, downloaded. Although the pack-file is just 1.7GB for me, not 
3.7 like somebody said.

Anyway, doing a

	git blame --incremental HEAD -- org.eclipse.debug.ui/plugin.xml > /dev/null

on that thing (picked a random file that got modified in a recent commit) 
took something like 12 seconds, so this is certainly a perfectly fine 
test-case.

Sadly, "git-gui blame" doesn't work in a bare git repository, so I had to 
do an ugly

	ln -s . .git

to make git-gui happy, and that worked, and was pretty usable. Still, 
12seconds should be something we can improve on.

And yeah, the profile is pretty horrid:

	samples  %        app name                 symbol name
	70307    20.9412  libc-2.5.so              strlen
	50925    15.1682  libz.so.1.2.3            (no symbols)
	24295     7.2364  git                      tree_entry_interesting
	19816     5.9023  libc-2.5.so              memcpy
	19569     5.8287  git                      tree_entry_extract
	17693     5.2699  vmlinux                  memcpy_c
	17032     5.0730  git                      assign_blame
	16956     5.0504  git                      get_mode
	12401     3.6937  git                      get_origin
	11815     3.5191  git                      skip_uninteresting
	10449     3.1123  git                      update_tree_entry
	10359     3.0855  git                      find_pack_entry_one
	7946      2.3667  git                      cmp_suspect
	4572      1.3618  libc-2.5.so              strncmp
	...

so I guess we need to find some more strlen's to remove ;)

			Linus
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help