160733 2016-08-10 00:45:51 -0700 [JSC] Speed up SparseCollection & related maps 2016-08-10 14:26:59 -0700 1 1 1 Unclassified WebKit New Bugs WebKit Nightly Build Unspecified Unspecified RESOLVED FIXED P2 Normal --- 1 benjamin benjamin commit-queue keith_miller mark.lam msaboff saam oldest_to_newest 1218988 0 benjamin 2016-08-10 00:45:51 -0700 [JSC] Speed up SparseCollection & related maps 1218991 1 285721 benjamin 2016-08-10 00:51:56 -0700 Created attachment 285721 Patch 1218993 2 benjamin 2016-08-10 00:54:28 -0700 Numbers on MBP below. I don't think this is relevant for MBP, I am more interested in the bots for dual-core configs. Conf#1 Conf#2 SunSpider: 3d-cube 4.7261+-0.0859 ? 4.7345+-0.0965 ? 3d-morph 4.9893+-0.0508 ? 5.0213+-0.0738 ? 3d-raytrace 4.9468+-0.0370 ? 5.0262+-0.0889 ? might be 1.0161x slower access-binary-trees 1.9674+-0.0196 ? 1.9767+-0.0544 ? access-fannkuch 5.6886+-0.1146 ! 5.9070+-0.0552 ! definitely 1.0384x slower access-nbody 2.4529+-0.0628 2.3820+-0.0253 might be 1.0298x faster access-nsieve 2.9535+-0.0570 2.9250+-0.0169 bitops-3bit-bits-in-byte 1.0879+-0.0260 ? 1.0904+-0.0313 ? bitops-bits-in-byte 2.6040+-0.0539 2.5926+-0.0471 bitops-bitwise-and 1.9998+-0.0345 1.9760+-0.0147 might be 1.0121x faster bitops-nsieve-bits 3.0860+-0.0283 ? 3.0919+-0.0343 ? controlflow-recursive 2.2840+-0.0195 2.2771+-0.0193 crypto-aes 4.3992+-0.0604 4.3843+-0.0637 crypto-md5 2.6474+-0.0244 2.6395+-0.0191 crypto-sha1 2.7947+-0.0699 2.7183+-0.0296 might be 1.0281x faster date-format-tofte 6.4832+-0.1073 6.4402+-0.0280 date-format-xparb 4.6997+-0.0619 ? 4.7147+-0.0674 ? math-cordic 2.7077+-0.0208 ? 2.7264+-0.0184 ? math-partial-sums 4.0195+-0.0813 3.9907+-0.0581 math-spectral-norm 2.0682+-0.0566 2.0334+-0.0395 might be 1.0172x faster regexp-dna 6.3291+-0.1092 ? 6.4171+-0.1728 ? might be 1.0139x slower string-base64 4.0062+-0.1028 ? 4.0149+-0.0945 ? string-fasta 5.5715+-0.1064 5.4995+-0.0511 might be 1.0131x faster string-tagcloud 8.5775+-0.2117 ^ 8.2740+-0.0916 ^ definitely 1.0367x faster string-unpack-code 18.2859+-0.3425 ? 18.4312+-0.4127 ? string-validate-input 4.0729+-0.0477 4.0387+-0.0552 <arithmetic> 4.4403+-0.0177 4.4355+-0.0213 might be 1.0011x faster Conf#1 Conf#2 Octane: encrypt 0.15507+-0.00154 ? 0.15536+-0.00134 ? decrypt 2.76505+-0.03807 2.75964+-0.01474 deltablue x2 0.13048+-0.00091 ? 0.13089+-0.00078 ? earley 0.28601+-0.00106 ^ 0.28399+-0.00067 ^ definitely 1.0071x faster boyer 4.94989+-0.04288 4.94168+-0.02736 navier-stokes x2 4.93369+-0.01234 ? 4.93813+-0.01067 ? raytrace x2 0.78915+-0.00300 ? 0.79218+-0.00258 ? richards x2 0.08257+-0.00082 0.08253+-0.00080 splay x2 0.33799+-0.00123 ? 0.33815+-0.00053 ? regexp x2 16.41046+-0.24608 ! 16.92794+-0.21083 ! definitely 1.0315x slower pdfjs x2 38.44409+-0.16747 38.13687+-0.16331 mandreel x2 42.61708+-0.11795 42.50511+-0.12022 gbemu x2 29.45155+-0.10237 29.40763+-0.08854 closure 0.48012+-0.00120 ? 0.48228+-0.00106 ? jquery 6.43811+-0.01698 6.41802+-0.01232 box2d x2 9.23391+-0.03311 ^ 9.14402+-0.01908 ^ definitely 1.0098x faster zlib x2 358.58399+-3.71226 ? 359.20175+-2.80684 ? typescript x2 598.18347+-3.63030 593.23887+-2.47101 <geometric> 4.99866+-0.00932 ? 5.00115+-0.00708 ? might be 1.0005x slower Conf#1 Conf#2 Kraken: ai-astar 86.581+-0.782 86.201+-0.720 audio-beat-detection 38.800+-0.074 ? 38.855+-0.119 ? audio-dft 99.401+-1.354 98.560+-1.191 audio-fft 30.193+-0.029 ? 30.355+-0.242 ? audio-oscillator 48.086+-0.170 ? 48.321+-0.499 ? imaging-darkroom 61.294+-0.406 ? 61.523+-0.640 ? imaging-desaturate 43.871+-0.379 ? 44.398+-0.834 ? might be 1.0120x slower imaging-gaussian-blur 60.382+-0.762 ? 60.957+-1.005 ? json-parse-financial 34.196+-0.553 ? 34.235+-0.351 ? json-stringify-tinderbox 24.080+-0.723 23.764+-0.568 might be 1.0133x faster stanford-crypto-aes 36.911+-0.159 ? 37.105+-0.267 ? stanford-crypto-ccm 34.079+-0.660 ? 34.460+-0.754 ? might be 1.0112x slower stanford-crypto-pbkdf2 92.815+-0.551 92.679+-0.377 stanford-crypto-sha256-iterative 30.898+-0.830 30.547+-0.160 might be 1.0115x faster <arithmetic> 51.542+-0.169 ? 51.569+-0.199 ? might be 1.0005x slower Conf#1 Conf#2 AsmBench: bigfib.cpp 423.1337+-1.1114 ? 423.5357+-1.7526 ? cray.c 385.6242+-1.6527 383.6395+-1.3958 dry.c 438.4418+-18.7959 ? 439.2329+-21.7475 ? FloatMM.c 732.6120+-12.5241 725.0609+-11.3600 might be 1.0104x faster gcc-loops.cpp 3596.2006+-11.3758 3593.8597+-12.0773 n-body.c 801.1809+-2.6053 798.4202+-1.0607 Quicksort.c 392.8017+-1.7963 ? 395.6572+-2.4215 ? stepanov_container.cpp 3268.1336+-10.3936 ? 3275.3264+-10.6962 ? Towers.c 264.9018+-0.8062 264.8246+-0.6862 <geometric> 718.6115+-3.2411 717.9080+-3.5248 might be 1.0010x faster Conf#1 Conf#2 Geomean of preferred means: <scaled-result> 30.1106+-0.0484 30.1026+-0.0697 might be 1.0003x faster 1219030 3 285721 saam 2016-08-10 09:43:17 -0700 Comment on attachment 285721 Patch r=me 1219131 4 285721 benjamin 2016-08-10 14:26:55 -0700 Comment on attachment 285721 Patch Clearing flags on attachment: 285721 Committed r204355: <http://trac.webkit.org/changeset/204355> 1219132 5 benjamin 2016-08-10 14:26:59 -0700 All reviewed patches have been landed. Closing bug. 285721 2016-08-10 00:51:56 -0700 2016-08-10 14:26:55 -0700 Patch bug-160733-20160810005021.patch text/plain 4078 benjamin U3VidmVyc2lvbiBSZXZpc2lvbjogMjA0MjY2CmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlw dENvcmUvQ2hhbmdlTG9nIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwppbmRleCAw NzcxNGM0YjVjNmE4NTg2MzVjMTQ0ZmIyNDIxOGQ4NDU0ZWU3MTA1Li5kMGQ3NDc0OTVjYWMzZTk1 YzFhM2QyZGE0MzhjN2U2ODhmMzQ1NjU0IDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENv cmUvQ2hhbmdlTG9nCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKQEAgLTEs MyArMSwzNCBAQAorMjAxNi0wOC0xMCAgQmVuamFtaW4gUG91bGFpbiAgPGJwb3VsYWluQGFwcGxl LmNvbT4KKworICAgICAgICBbSlNDXSBTcGVlZCB1cCBTcGFyc2VDb2xsZWN0aW9uICYgcmVsYXRl ZCBtYXBzCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0x NjA3MzMKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICBP biBNQkEsIEdyYXBoOjphZGROb2RlKCkgc2hvd3MgdXAgaW4gcHJvZmlsZXMgZHVlIHRvIFNwYXJz ZUNvbGxlY3Rpb246OmFkZCgpLgorICAgICAgICBUaGlzIGlzIHVuZm9ydHVuYXRlLgorCisgICAg ICAgIFRoZSBmaXJzdCBpbXByb3ZlbWVudCBpcyB0byBidWlsZCB0aGUgbmV3IHVuaXF1ZV9wdHIg aW4gdGhlIGVtcHR5IHNsb3QKKyAgICAgICAgaW5zdGVhZCBvZiBtb3ZpbmcgYSBuZXcgdmFsdWUg aW50byBpdC4KKworICAgICAgICBQcmV2aW91c2x5LCB0aGUgY29kZSB3b3VsZCBsb2FkIHRoZSBw cmV2aW91cyB2YWx1ZSwgdGVzdCBpZiBpdCBpcyBudWxsCisgICAgICAgIHRoZW4gaW52b2tlIHRo ZSBkZXN0cnVjdG9yIGFuZCBmaW5hbGx5IGZhc3RGcmVlKCkuIFRoZSBpbml0aWFsIHRlc3QKKyAg ICAgICAgb2J2aW91c2x5IGZhaWxzIHNvIHRoYXQncyBhIHdob2xlIGJ1bmNoIG9mIGNvZGUgdGhh dCBpcyBuZXZlciBleGVjdXRlZC4KKworICAgICAgICBXaXRoIHRoZSBuZXcgY29kZSwgd2UganVz dCBoYXZlIGEgc3RvcmUuCisKKyAgICAgICAgSSBhbHNvIHJlbW92ZWQgdGhlIGJvdW5kcyBjaGVj a2luZyBvbiBvdXIgbWFwcyBiYXNlZCBvbiBub2RlIGluZGV4LgorICAgICAgICBUaG9zZSBib3Vu ZHMgY2hlY2tzIGFyZSBuZXZlciBlbGltaW5hdGVkIGJ5IGNsYW5nIGJlY2F1c2UgdGhlIGluZGV4 CisgICAgICAgIGlzIGFsd2F5cyBsb2FkZWQgZnJvbSBtZW1vcnkgaW5zdGVhZCBvZiBiZWluZyBj b21wdXRlZC4KKyAgICAgICAgVGhlcmUgYXJlIHVuZm9ydHVuYXRlbHkgdG9vIG1hbnkgbm9kZXMg cHJvY2Vzc2VkIGFuZCB0aGUgYm91bmRzIGNoZWNrcworICAgICAgICBnZXQgY29zdGx5LgorCisg ICAgICAgICogYjMvQjNTcGFyc2VDb2xsZWN0aW9uLmg6CisgICAgICAgIChKU0M6OkIzOjpTcGFy c2VDb2xsZWN0aW9uOjphZGQpOgorICAgICAgICAqIGRmZy9ERkdHcmFwaC5oOgorICAgICAgICAo SlNDOjpERkc6OkdyYXBoOjphYnN0cmFjdFZhbHVlc0NhY2hlKToKKyAgICAgICAgKiBkZmcvREZH SW5QbGFjZUFic3RyYWN0U3RhdGUuaDoKKwogMjAxNi0wOC0wOCAgTWFyayBMYW0gIDxtYXJrLmxh bUBhcHBsZS5jb20+CiAKICAgICAgICAgQVNTRVJUSU9OIEZBSUxFRDogaGFzSW5saW5lU3RvcmFn ZSgpIGluIEpTRmluYWxPYmplY3Q6OnZpc2l0Q2hpbGRyZW4oKS4KZGlmZiAtLWdpdCBhL1NvdXJj ZS9KYXZhU2NyaXB0Q29yZS9iMy9CM1NwYXJzZUNvbGxlY3Rpb24uaCBiL1NvdXJjZS9KYXZhU2Ny aXB0Q29yZS9iMy9CM1NwYXJzZUNvbGxlY3Rpb24uaAppbmRleCA5YmMyM2IyMWFmMDMwODYwZjZj MTExMGYxYmYyZTU4NDM4NDNlMTE4Li4zNGU2MWRhMTMzZjRmYjkzZTk3OGM5MTBjMTdjYjkxYzE1 OWJlYmZiIDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvYjMvQjNTcGFyc2VDb2xs ZWN0aW9uLmgKKysrIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2IzL0IzU3BhcnNlQ29sbGVjdGlv bi5oCkBAIC01NSw4ICs1NSw4IEBAIHB1YmxpYzoKICAgICAgICAgICAgIGluZGV4ID0gbV9pbmRl eEZyZWVMaXN0LnRha2VMYXN0KCk7CiAKICAgICAgICAgdmFsdWUtPm1faW5kZXggPSBpbmRleDsK LQotICAgICAgICBtX3ZlY3RvcltpbmRleF0gPSBXVEZNb3ZlKHZhbHVlKTsKKyAgICAgICAgQVNT RVJUKCFtX3ZlY3RvcltpbmRleF0pOworICAgICAgICBuZXcgKE5vdE51bGwsICZtX3ZlY3Rvcltp bmRleF0pIHN0ZDo6dW5pcXVlX3B0cjxUPihXVEZNb3ZlKHZhbHVlKSk7CiAKICAgICAgICAgcmV0 dXJuIHJlc3VsdDsKICAgICB9CkBAIC0xNjgsOCArMTY4LDggQEAgcHVibGljOgogICAgIGl0ZXJh dG9yIGVuZCgpIGNvbnN0IHsgcmV0dXJuIGl0ZXJhdG9yKCp0aGlzLCBzaXplKCkpOyB9CiAKIHBy aXZhdGU6Ci0gICAgVmVjdG9yPHN0ZDo6dW5pcXVlX3B0cjxUPj4gbV92ZWN0b3I7Ci0gICAgVmVj dG9yPHNpemVfdD4gbV9pbmRleEZyZWVMaXN0OworICAgIFZlY3RvcjxzdGQ6OnVuaXF1ZV9wdHI8 VD4sIDAsIFVuc2FmZVZlY3Rvck92ZXJmbG93PiBtX3ZlY3RvcjsKKyAgICBWZWN0b3I8c2l6ZV90 LCAwLCBVbnNhZmVWZWN0b3JPdmVyZmxvdz4gbV9pbmRleEZyZWVMaXN0OwogfTsKIAogfSB9IC8v IG5hbWVzcGFjZSBKU0M6OkIzCmRpZmYgLS1naXQgYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZn L0RGR0dyYXBoLmggYi9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0dyYXBoLmgKaW5kZXgg YzEzN2U3ZjJmMzI1YTJjZDU5OTZhNTk4NDY1ZjU3MmUxODVkNjVlNC4uNTJiNGZmYTZmNGYxZDA0 MDZiNmI1NTY0MzlmNmQyZTJhN2QwMTg1MCAxMDA2NDQKLS0tIGEvU291cmNlL0phdmFTY3JpcHRD b3JlL2RmZy9ERkdHcmFwaC5oCisrKyBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHR3Jh cGguaApAQCAtMTk5LDcgKzE5OSw3IEBAIHB1YmxpYzoKICAgICBOb2RlKiBub2RlQXQodW5zaWdu ZWQgaW5kZXgpIGNvbnN0IHsgcmV0dXJuIG1fbm9kZXNbaW5kZXhdOyB9CiAgICAgdm9pZCBwYWNr Tm9kZUluZGljZXMoKTsKIAotICAgIFZlY3RvcjxBYnN0cmFjdFZhbHVlPiYgYWJzdHJhY3RWYWx1 ZXNDYWNoZSgpIHsgcmV0dXJuIG1fYWJzdHJhY3RWYWx1ZXNDYWNoZTsgfQorICAgIFZlY3RvcjxB YnN0cmFjdFZhbHVlLCAwLCBVbnNhZmVWZWN0b3JPdmVyZmxvdz4mIGFic3RyYWN0VmFsdWVzQ2Fj aGUoKSB7IHJldHVybiBtX2Fic3RyYWN0VmFsdWVzQ2FjaGU7IH0KIAogICAgIHZvaWQgZGV0aHJl YWQoKTsKICAgICAKQEAgLTk1Nyw3ICs5NTcsNyBAQCBwcml2YXRlOgogICAgIH0KIAogICAgIEIz OjpTcGFyc2VDb2xsZWN0aW9uPE5vZGU+IG1fbm9kZXM7Ci0gICAgVmVjdG9yPEFic3RyYWN0VmFs dWU+IG1fYWJzdHJhY3RWYWx1ZXNDYWNoZTsKKyAgICBWZWN0b3I8QWJzdHJhY3RWYWx1ZSwgMCwg VW5zYWZlVmVjdG9yT3ZlcmZsb3c+IG1fYWJzdHJhY3RWYWx1ZXNDYWNoZTsKIH07CiAKIH0gfSAv LyBuYW1lc3BhY2UgSlNDOjpERkcKZGlmZiAtLWdpdCBhL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9k ZmcvREZHSW5QbGFjZUFic3RyYWN0U3RhdGUuaCBiL1NvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcv REZHSW5QbGFjZUFic3RyYWN0U3RhdGUuaAppbmRleCA5NDU5YzdkYTJmYzI3NzVjZWFhMGRmZTNm ZTQ0NTY5Mzg4ZWNkOWI1Li42MmQ4YzNmNTBmMmYzZmU2ODlhNGZhNzg3OGZkNWU5NzhmZjBhMDA2 IDEwMDY0NAotLS0gYS9Tb3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0luUGxhY2VBYnN0cmFj dFN0YXRlLmgKKysrIGIvU291cmNlL0phdmFTY3JpcHRDb3JlL2RmZy9ERkdJblBsYWNlQWJzdHJh Y3RTdGF0ZS5oCkBAIC0xMzMsNyArMTMzLDcgQEAgcHJpdmF0ZToKICAgICAKICAgICBHcmFwaCYg bV9ncmFwaDsKIAotICAgIFZlY3RvcjxBYnN0cmFjdFZhbHVlPiYgbV9hYnN0cmFjdFZhbHVlczsK KyAgICBWZWN0b3I8QWJzdHJhY3RWYWx1ZSwgMCwgVW5zYWZlVmVjdG9yT3ZlcmZsb3c+JiBtX2Fi c3RyYWN0VmFsdWVzOwogICAgIE9wZXJhbmRzPEFic3RyYWN0VmFsdWU+IG1fdmFyaWFibGVzOwog ICAgIEJhc2ljQmxvY2sqIG1fYmxvY2s7CiAgICAgCg==