summaryrefslogtreecommitdiffstats
path: root/chrome/test/ppapi/ppapi_test.h
diff options
context:
space:
mode:
authorpliard@chromium.org <pliard@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98>2013-12-26 18:12:22 +0000
committerpliard@chromium.org <pliard@chromium.org@0039d316-1c4b-4281-b951-d872f2087c98>2013-12-26 18:12:22 +0000
commit1f48f758c388b0f402e5c166a4d63b7460c7dfa4 (patch)
tree2519b5a544d771f78c5a5ba0ea329453b34e43f0 /chrome/test/ppapi/ppapi_test.h
parent7935926e17379c232cef3a30a14684c704dc55de (diff)
downloadchromium_src-1f48f758c388b0f402e5c166a4d63b7460c7dfa4.zip
chromium_src-1f48f758c388b0f402e5c166a4d63b7460c7dfa4.tar.gz
chromium_src-1f48f758c388b0f402e5c166a4d63b7460c7dfa4.tar.bz2
Always use the ashmem allocator when creating DiscardableMemory instances.
Experiments and measurements showed that the possibly accumulated discardable external fragmentation caused by the allocator's free chunks doesn't matter much in practice. The fact that (unpinned) free chunks could be contributing to private dirty was the main reason to use the allocator only as a fallback (when file descriptor usage gets too high). However experimentation showed that unpinned ashmem eviction happened without degrading user experience (i.e. without killing other apps nor causing jankiness). This significantly simplifies the DiscardableMemory creation path on Android and should provide a speed boost too (10x on Nexus10) in addition to the fact that ashmem creation + mmap() appeared to have very unpredictable performance. Review URL: https://codereview.chromium.org/98213004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@242544 0039d316-1c4b-4281-b951-d872f2087c98
Diffstat (limited to 'chrome/test/ppapi/ppapi_test.h')
0 files changed, 0 insertions, 0 deletions