commit | 09f86d98995373b97d746d1c489cf097f84fe1ea | [log] [tgz] |
---|---|---|
author | Ryan Prichard <rprichard@google.com> | Tue May 07 21:31:53 2019 -0700 |
committer | Ryan Prichard <rprichard@google.com> | Mon May 13 20:26:56 2019 -0700 |
tree | 699f5c8473bfae8ed5ced8d3540a98fad886c6fa | |
parent | ee41accb399c458226bc2902ecbe529aabbfcc17 [diff] |
Replace the +1 dep_index bias with ssize_t This seems less likely to cause confusion. I measured the run-time of initializing dep_index with both ordinary stores and relaxed stores. On a 580MB deps log, the ordinary stores (seq_cst) took about 4ms, while relaxed stores took 0.5ms. Test: ninja_test Change-Id: Id1a2cc026220e33b25c87dba4b5f612850e42bd0