Tue, 23 Aug 2011 22:30:56 +0200 |
Artem Tikhomirov |
Use StringProxy when parsing manifest to minimize number of useless conversions and array instances
|
Tue, 23 Aug 2011 21:27:56 +0200 |
Artem Tikhomirov |
Revert manifest parser changes - no single string and back to HashMap Pool
|
Fri, 19 Aug 2011 04:59:32 +0200 |
Artem Tikhomirov |
Improve manifest parsing: reduce number of arrays instantiated for Nodeid
|
Fri, 19 Aug 2011 04:02:48 +0200 |
Artem Tikhomirov |
Generous defaults for SparseSet not to fail on big manifests
|
Fri, 19 Aug 2011 03:36:25 +0200 |
Artem Tikhomirov |
Optimize memory usage (reduce number of objects instantiated) when pooling file names and nodeids during manifest parsing
|
Thu, 18 Aug 2011 18:06:44 +0200 |
Artem Tikhomirov |
RevlogStream - reduce number of object instances, reuse when possible
|
Thu, 18 Aug 2011 03:46:36 +0200 |
Artem Tikhomirov |
Improve manifest parsing: decode bytes to chars once, minimize arraycopy on String instantiation, keep set of file revisions from previous manifest only
|
Wed, 17 Aug 2011 20:30:06 +0200 |
Artem Tikhomirov |
Avoid redundant calls to library when converting a char to hex digit
|
Wed, 17 Aug 2011 02:35:15 +0200 |
Artem Tikhomirov |
Experimental alternative to HashMap in Pool to reduce memory footprint
|
Tue, 16 Aug 2011 22:15:02 +0200 |
Artem Tikhomirov |
Avoid IAE when repository has removed tags. Do not duplicate pools that are already in ManifestParser
|
Tue, 16 Aug 2011 22:08:15 +0200 |
Artem Tikhomirov |
Reduce number of objects instantiated on revlog read
|
Tue, 16 Aug 2011 22:07:16 +0200 |
Artem Tikhomirov |
Handle revision constants correctly
|
Tue, 16 Aug 2011 15:45:47 +0200 |
Artem Tikhomirov |
Yet another slight perf/mem improvement for the tags per file sample
|
Tue, 16 Aug 2011 14:33:11 +0200 |
Artem Tikhomirov |
Improve memory footprint of tag collection (about 14 Mb saved for cpython repo without HashMap.Entry and Entry[])
|