25656 2009-05-08 16:56:46 -0700 Reduce the size of CSSParserValueList 2009-05-08 17:53:42 -0700 1 1 1 Unclassified WebKit CSS 528+ (Nightly build) Mac OS X 10.5 RESOLVED FIXED P2 Normal --- 1 simon.fraser simon.fraser hyatt oldest_to_newest 120490 0 simon.fraser 2009-05-08 16:56:46 -0700 CSSParserValueList is 544 bytes in 64-bit, and accounts for about 2.4MB of churn when running the PLTs. I think the vector of m_values can uses a smaller initial buffer size. 120491 1 simon.fraser 2009-05-08 17:03:51 -0700 I browsed some common sites, and the frequency distribution of the high water mark of the m_values size is: 1 20717 2 864 3 1166 4 1380 5 293 6 121 7 26 8 8 9 9 10 2 11 2 12 7 13 1 So a size of 4 will include 98% of cases. 120494 2 30149 simon.fraser 2009-05-08 17:20:13 -0700 Created attachment 30149 Patch, changelog I ran the PLTs with this patch, and was not able to detect any performance decrease. 120495 3 simon.fraser 2009-05-08 17:53:42 -0700 http://trac.webkit.org/changeset/43423 30149 2009-05-08 17:20:13 -0700 2009-05-08 17:49:14 -0700 Patch, changelog 25656_patch.txt text/plain 1020 simon.fraser ZGlmZiAtLWdpdCBhL1dlYkNvcmUvQ2hhbmdlTG9nIGIvV2ViQ29yZS9DaGFuZ2VMb2cKaW5kZXgg ZDk5YTY1Yy4uYzMzY2RiZCAxMDA2NDQKLS0tIGEvV2ViQ29yZS9DaGFuZ2VMb2cKKysrIGIvV2Vi Q29yZS9DaGFuZ2VMb2cKQEAgLTEsMyArMSwxNCBAQAorMjAwOS0wNS0wOCAgU2ltb24gRnJhc2Vy ICA8c2ltb24uZnJhc2VyQGFwcGxlLmNvbT4KKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkg KE9PUFMhKS4KKyAgICAgICAgCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3df YnVnLmNnaT9pZD0yNTY1NgorICAgICAgICAKKyAgICAgICAgUmVkdWNlIHRoZSBpbmxpbmUgY2Fw YWNpdHkgb2YgQ1NTUGFyc2VyVmFsdWVMaXN0J3MgbV92YWx1ZXMKKyAgICAgICAgdmVjdG9yIHRv IHJlZHVjZSB0aGUgc2l6ZSBvZiBDU1NQYXJzZXJWYWx1ZUxpc3QgZnJvbSA1NDQgdG8gMTYwIGJ5 dGVzIGluIDY0LWJpdC4KKworICAgICAgICAqIGNzcy9DU1NQYXJzZXJWYWx1ZXMuaDoKKwogMjAw OS0wNS0wOCAgRGFuIEJlcm5zdGVpbiAgPG1pdHpAYXBwbGUuY29tPgogCiAgICAgICAgIFJldmll d2VkIGJ5IE1hY2llaiBTdGFjaG93aWFrLgpkaWZmIC0tZ2l0IGEvV2ViQ29yZS9jc3MvQ1NTUGFy c2VyVmFsdWVzLmggYi9XZWJDb3JlL2Nzcy9DU1NQYXJzZXJWYWx1ZXMuaAppbmRleCA0ZTBhMjgw Li4yNGJkOWI3IDEwMDY0NAotLS0gYS9XZWJDb3JlL2Nzcy9DU1NQYXJzZXJWYWx1ZXMuaAorKysg Yi9XZWJDb3JlL2Nzcy9DU1NQYXJzZXJWYWx1ZXMuaApAQCAtODMsNyArODMsNyBAQCBwdWJsaWM6 CiAgICAgYm9vbCBjb250YWluc1ZhcmlhYmxlcygpIGNvbnN0IHsgcmV0dXJuIG1fdmFyaWFibGVz Q291bnQ7IH0KIAogcHJpdmF0ZToKLSAgICBWZWN0b3I8Q1NTUGFyc2VyVmFsdWUsIDE2PiBtX3Zh bHVlczsKKyAgICBWZWN0b3I8Q1NTUGFyc2VyVmFsdWUsIDQ+IG1fdmFsdWVzOwogICAgIHVuc2ln bmVkIG1fY3VycmVudDsKICAgICB1bnNpZ25lZCBtX3ZhcmlhYmxlc0NvdW50OwogfTsK