70770 2011-10-24 16:02:59 -0700 op_new_array_buffer is not inlined correctly 2011-10-25 16:25:07 -0700 1 1 1 Unclassified WebKit JavaScriptCore 528+ (Nightly build) All All RESOLVED FIXED P2 Normal --- 70246 1 fpizlo webkit-unassigned oldest_to_newest 489742 0 fpizlo 2011-10-24 16:02:59 -0700 The constant buffers aren't moved from the inlinee to the caller, resulting in crashes and generally bizarre behavior. 489748 1 112270 fpizlo 2011-10-24 16:06:21 -0700 Created attachment 112270 the patch 489752 2 fpizlo 2011-10-24 16:11:36 -0700 This basically looks performance neutral. Benchmark report for SunSpider, V8, and Kraken. VMs tested: "TipOfTree" at /Volumes/Data/pizlo/OpenSource/WebKitBuild/Release/jsc "FixNewArrBuf" at /Volumes/Data/pizlo/secondary/OpenSource/WebKitBuild/Release/jsc Collected 12 samples per benchmark/VM, with 4 VM invocations per benchmark. Used 1 benchmark iteration per VM invocation for warm-up. Used the jsc-specific preciseTime() function to get microsecond-level timing. Reporting benchmark execution times with 95% confidence intervals in milliseconds. TipOfTree FixNewArrBuf SunSpider: 3d-cube 7.8879+-0.0339 7.8640+-0.0420 3d-morph 8.3969+-0.0325 8.3785+-0.0313 3d-raytrace 8.1488+-0.1055 ? 8.2466+-0.1088 ? might be 1.0120x slower access-binary-trees 1.7039+-0.0199 1.6854+-0.0067 might be 1.0110x faster access-fannkuch 7.7873+-0.0092 ^ 7.7302+-0.0129 ^ definitely 1.0074x faster access-nbody 4.5226+-0.0075 ? 4.5257+-0.0079 ? access-nsieve 3.1842+-0.0155 ? 3.1925+-0.0176 ? bitops-3bit-bits-in-byte 1.3212+-0.0146 1.3112+-0.0066 bitops-bits-in-byte 5.2840+-0.0199 5.2702+-0.0412 bitops-bitwise-and 3.4856+-0.0368 3.4834+-0.0350 bitops-nsieve-bits 5.6654+-0.0389 ? 5.6707+-0.0323 ? controlflow-recursive 2.3274+-0.0094 ? 2.3280+-0.0027 ? crypto-aes 7.5545+-0.0604 7.5486+-0.0644 crypto-md5 2.8674+-0.0214 ? 2.8790+-0.0331 ? crypto-sha1 2.6775+-0.0142 2.6437+-0.0209 might be 1.0128x faster date-format-tofte 10.7269+-0.1525 ? 10.7638+-0.1845 ? date-format-xparb 9.8674+-0.0846 9.7350+-0.0993 might be 1.0136x faster math-cordic 7.7788+-0.1623 7.7166+-0.2482 math-partial-sums 10.6469+-0.0392 10.5969+-0.0431 math-spectral-norm 2.8754+-0.0053 2.8750+-0.0045 regexp-dna 13.3414+-0.1880 ? 13.3536+-0.1834 ? string-base64 4.4311+-0.0241 ? 4.4390+-0.0143 ? string-fasta 7.0924+-0.0211 ? 7.1354+-0.0458 ? string-tagcloud 13.2534+-0.1381 ? 13.3815+-0.1511 ? string-unpack-code 22.6295+-0.1511 ? 22.9132+-0.2867 ? might be 1.0125x slower string-validate-input 5.6016+-0.0239 ? 5.6099+-0.0405 ? <arithmetic> * 6.9638+-0.0185 ? 6.9722+-0.0248 ? <geometric> 5.6392+-0.0128 5.6350+-0.0145 <harmonic> 4.4687+-0.0130 4.4571+-0.0101 TipOfTree FixNewArrBuf V8: crypto 80.9993+-0.2219 ? 81.1349+-0.2958 ? deltablue 194.4328+-0.6984 ? 195.5024+-1.1510 ? earley-boyer 110.9332+-0.3148 ? 111.6831+-0.6856 ? raytrace 69.3179+-0.5469 ! 70.5104+-0.6422 ! definitely 1.0172x slower regexp 124.9975+-0.4080 ^ 123.4700+-0.4680 ^ definitely 1.0124x faster richards 143.2010+-0.1146 ! 145.1758+-0.2890 ! definitely 1.0138x slower splay 125.1640+-0.5155 124.4957+-0.4643 <arithmetic> 121.2923+-0.1544 ? 121.7103+-0.2651 ? <geometric> * 115.3238+-0.1639 ! 115.7682+-0.2355 ! definitely 1.0039x slower <harmonic> 109.5375+-0.2064 ! 110.0625+-0.2430 ! definitely 1.0048x slower TipOfTree FixNewArrBuf Kraken: ai-astar 825.4978+-11.2327 ? 835.0474+-2.1459 ? might be 1.0116x slower audio-beat-detection 210.7485+-1.2126 ! 213.1404+-1.0511 ! definitely 1.0113x slower audio-dft 261.5475+-2.8238 260.9661+-2.8364 audio-fft 135.4228+-0.5462 ^ 133.0906+-0.6482 ^ definitely 1.0175x faster audio-oscillator 290.9940+-1.1317 ? 292.0005+-1.1513 ? imaging-darkroom 444.8460+-3.7579 ? 447.3978+-2.4742 ? imaging-desaturate 237.8529+-0.0435 ^ 231.6030+-0.0502 ^ definitely 1.0270x faster imaging-gaussian-blur 621.9686+-0.9858 ? 622.3993+-2.0852 ? json-parse-financial 69.9952+-0.3373 ! 70.5335+-0.1825 ! definitely 1.0077x slower json-stringify-tinderbox 78.7417+-0.2044 ! 79.6831+-0.2539 ! definitely 1.0120x slower stanford-crypto-aes 152.7402+-1.4333 152.2338+-1.4184 stanford-crypto-ccm 115.8948+-0.4873 115.4034+-0.6088 stanford-crypto-pbkdf2 236.8705+-1.9291 236.6271+-1.5212 stanford-crypto-sha256-iterative 85.4106+-0.2098 ? 85.5004+-0.2730 ? <arithmetic> * 269.1808+-0.9075 ? 269.6876+-0.4099 ? <geometric> 205.0797+-0.4329 205.0601+-0.3080 <harmonic> 161.3684+-0.2685 ? 161.4633+-0.2603 ? TipOfTree FixNewArrBuf All benchmarks: <arithmetic> 102.0986+-0.2871 ? 102.3165+-0.1368 ? <geometric> 25.7820+-0.0427 ? 25.7856+-0.0486 ? <harmonic> 7.8741+-0.0225 7.8547+-0.0176 TipOfTree FixNewArrBuf Geomean of preferred means: <scaled-result> 60.0162+-0.1069 ? 60.1551+-0.1138 ? 489763 3 fpizlo 2011-10-24 16:21:59 -0700 Landed in http://trac.webkit.org/changeset/98296 112270 2011-10-24 16:06:21 -0700 2011-10-24 16:11:00 -0700 the patch fixinlinenewarraybuffer_patch_1.diff text/plain 4744 fpizlo SW5kZXg6IFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9DaGFuZ2VMb2cKPT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gU291 cmNlL0phdmFTY3JpcHRDb3JlL0NoYW5nZUxvZwkocmV2aXNpb24gOTgyOTUpCisrKyBTb3VyY2Uv SmF2YVNjcmlwdENvcmUvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBAIC0xLDMgKzEsMTUgQEAK KzIwMTEtMTAtMjQgIEZpbGlwIFBpemxvICA8ZnBpemxvQGFwcGxlLmNvbT4KKworICAgICAgICBv cF9uZXdfYXJyYXlfYnVmZmVyIGlzIG5vdCBpbmxpbmVkIGNvcnJlY3RseQorICAgICAgICBodHRw czovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9NzA3NzAKKworICAgICAgICBSZXZp ZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKyAgICAgICAgCisgICAgICAgIERpc2FibGVkIGlubGlu aW5nIG9mIG9wX25ld19hcnJheV9idWZmZXIsIGZvciBub3cuCisKKyAgICAgICAgKiBkZmcvREZH Q2FwYWJpbGl0aWVzLmg6CisgICAgICAgIChKU0M6OkRGRzo6Y2FuSW5saW5lT3Bjb2RlKToKKwog MjAxMS0xMC0yNCAgWXVxaWFuZyBYaWFuICA8eXVxaWFuZy54aWFuQGludGVsLmNvbT4KIAogICAg ICAgICBBZGQgYm9vbGVhbiBzcGVjdWxhdGlvbnMgdG8gREZHIEpJVCAzMl82NApJbmRleDogU291 cmNlL0phdmFTY3JpcHRDb3JlL2RmZy9ERkdDYXBhYmlsaXRpZXMuaAo9PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBT b3VyY2UvSmF2YVNjcmlwdENvcmUvZGZnL0RGR0NhcGFiaWxpdGllcy5oCShyZXZpc2lvbiA5ODI5 MikKKysrIFNvdXJjZS9KYXZhU2NyaXB0Q29yZS9kZmcvREZHQ2FwYWJpbGl0aWVzLmgJKHdvcmtp bmcgY29weSkKQEAgLTE4MSw2ICsxODEsMTAgQEAgaW5saW5lIGJvb2wgY2FuSW5saW5lT3Bjb2Rl KE9wY29kZUlEIG9wYwogICAgIGNhc2Ugb3BfcmVzb2x2ZV9iYXNlOgogICAgIGNhc2Ugb3BfcmVz b2x2ZV9nbG9iYWw6CiAgICAgICAgIAorICAgIC8vIENvbnN0YW50IGJ1ZmZlcnMgYXJlbid0IGNv cGllZCBjb3JyZWN0bHkuIFRoaXMgaXMgZWFzeSB0byBmaXgsIGJ1dCBmb3IKKyAgICAvLyBub3cg d2UganVzdCBkaXNhYmxlIGlubGluaW5nIGZvciBmdW5jdGlvbnMgdGhhdCB1c2UgdGhlbS4KKyAg ICBjYXNlIG9wX25ld19hcnJheV9idWZmZXI6CisgICAgICAgIAogICAgIC8vIElubGluaW5nIGRv ZXNuJ3QgY29ycmVjdGx5IHJlbWFwIHJlZ3VsYXIgZXhwcmVzc2lvbiBvcGVyYW5kcy4KICAgICBj YXNlIG9wX25ld19yZWdleHA6CiAgICAgICAgIHJldHVybiBmYWxzZTsKSW5kZXg6IExheW91dFRl c3RzL0NoYW5nZUxvZwo9PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09Ci0tLSBMYXlvdXRUZXN0cy9DaGFuZ2VMb2cJKHJldmlz aW9uIDk4Mjk1KQorKysgTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCSh3b3JraW5nIGNvcHkpCkBAIC0x LDMgKzEsMTYgQEAKKzIwMTEtMTAtMjQgIEZpbGlwIFBpemxvICA8ZnBpemxvQGFwcGxlLmNvbT4K KworICAgICAgICBvcF9uZXdfYXJyYXlfYnVmZmVyIGlzIG5vdCBpbmxpbmVkIGNvcnJlY3RseQor ICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9NzA3NzAKKwor ICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAqIGZhc3QvanMv ZGZnLWlubGluZS1uZXctYXJyYXktYnVmZmVyLWV4cGVjdGVkLnR4dDogQWRkZWQuCisgICAgICAg ICogZmFzdC9qcy9kZmctaW5saW5lLW5ldy1hcnJheS1idWZmZXIuaHRtbDogQWRkZWQuCisgICAg ICAgICogZmFzdC9qcy9zY3JpcHQtdGVzdHMvZGZnLWlubGluZS1uZXctYXJyYXktYnVmZmVyLmpz OiBBZGRlZC4KKyAgICAgICAgKGZvbyk6CisgICAgICAgIChiYXIpOgorCiAyMDExLTEwLTI0ICBU aW0gSG9ydG9uICA8dGltb3RoeV9ob3J0b25AYXBwbGUuY29tPgogCiAgICAgICAgIGZlQ29sb3JN YXRyaXggZG9lc24ndCB1c2UgdGhlIGNvcnJlY3QgZGVmYXVsdCAibWF0cml4IiB0eXBlIHdoZW4g bm8gdHlwZSBhdHRyaWJ1dGUgaXMgc3BlY2lmaWVkCkluZGV4OiBMYXlvdXRUZXN0cy9mYXN0L2pz L2RmZy1pbmxpbmUtbmV3LWFycmF5LWJ1ZmZlci1leHBlY3RlZC50eHQKPT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0g TGF5b3V0VGVzdHMvZmFzdC9qcy9kZmctaW5saW5lLW5ldy1hcnJheS1idWZmZXItZXhwZWN0ZWQu dHh0CShyZXZpc2lvbiAwKQorKysgTGF5b3V0VGVzdHMvZmFzdC9qcy9kZmctaW5saW5lLW5ldy1h cnJheS1idWZmZXItZXhwZWN0ZWQudHh0CShyZXZpc2lvbiAwKQpAQCAtMCwwICsxLDQ5IEBACitU aGlzIHRlc3RzIHRoYXQgaW5saW5pbmcgY29ycmVjdGx5IGhhbmRsZXMgY29uc3RhbnQgYnVmZmVy cy4KKworT24gc3VjY2VzcywgeW91IHdpbGwgc2VlIGEgc2VyaWVzIG9mICJQQVNTIiBtZXNzYWdl cywgZm9sbG93ZWQgYnkgIlRFU1QgQ09NUExFVEUiLgorCisKK1BBU1MgYmFyKClbMF0gaXMgMQor UEFTUyBiYXIoKVsxXSBpcyAyCitQQVNTIGJhcigpWzJdIGlzIDMKK1BBU1MgYmFyKClbM10gaXMg NAorUEFTUyBiYXIoKVswXSBpcyAxCitQQVNTIGJhcigpWzFdIGlzIDIKK1BBU1MgYmFyKClbMl0g aXMgMworUEFTUyBiYXIoKVszXSBpcyA0CitQQVNTIGJhcigpWzBdIGlzIDEKK1BBU1MgYmFyKClb MV0gaXMgMgorUEFTUyBiYXIoKVsyXSBpcyAzCitQQVNTIGJhcigpWzNdIGlzIDQKK1BBU1MgYmFy KClbMF0gaXMgMQorUEFTUyBiYXIoKVsxXSBpcyAyCitQQVNTIGJhcigpWzJdIGlzIDMKK1BBU1Mg YmFyKClbM10gaXMgNAorUEFTUyBiYXIoKVswXSBpcyAxCitQQVNTIGJhcigpWzFdIGlzIDIKK1BB U1MgYmFyKClbMl0gaXMgMworUEFTUyBiYXIoKVszXSBpcyA0CitQQVNTIGJhcigpWzBdIGlzIDEK K1BBU1MgYmFyKClbMV0gaXMgMgorUEFTUyBiYXIoKVsyXSBpcyAzCitQQVNTIGJhcigpWzNdIGlz IDQKK1BBU1MgYmFyKClbMF0gaXMgMQorUEFTUyBiYXIoKVsxXSBpcyAyCitQQVNTIGJhcigpWzJd IGlzIDMKK1BBU1MgYmFyKClbM10gaXMgNAorUEFTUyBiYXIoKVswXSBpcyAxCitQQVNTIGJhcigp WzFdIGlzIDIKK1BBU1MgYmFyKClbMl0gaXMgMworUEFTUyBiYXIoKVszXSBpcyA0CitQQVNTIGJh cigpWzBdIGlzIDEKK1BBU1MgYmFyKClbMV0gaXMgMgorUEFTUyBiYXIoKVsyXSBpcyAzCitQQVNT IGJhcigpWzNdIGlzIDQKK1BBU1MgYmFyKClbMF0gaXMgMQorUEFTUyBiYXIoKVsxXSBpcyAyCitQ QVNTIGJhcigpWzJdIGlzIDMKK1BBU1MgYmFyKClbM10gaXMgNAorUEFTUyBzdWNjZXNzZnVsbHlQ YXJzZWQgaXMgdHJ1ZQorCitURVNUIENPTVBMRVRFCisKSW5kZXg6IExheW91dFRlc3RzL2Zhc3Qv anMvZGZnLWlubGluZS1uZXctYXJyYXktYnVmZmVyLmh0bWwKPT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PQotLS0gTGF5b3V0 VGVzdHMvZmFzdC9qcy9kZmctaW5saW5lLW5ldy1hcnJheS1idWZmZXIuaHRtbAkocmV2aXNpb24g MCkKKysrIExheW91dFRlc3RzL2Zhc3QvanMvZGZnLWlubGluZS1uZXctYXJyYXktYnVmZmVyLmh0 bWwJKHJldmlzaW9uIDApCkBAIC0wLDAgKzEsMTAgQEAKKzwhRE9DVFlQRSBIVE1MIFBVQkxJQyAi LS8vSUVURi8vRFREIEhUTUwvL0VOIj4KKzxodG1sPgorPGhlYWQ+Cis8c2NyaXB0IHNyYz0icmVz b3VyY2VzL2pzLXRlc3QtcHJlLmpzIj48L3NjcmlwdD4KKzwvaGVhZD4KKzxib2R5PgorPHNjcmlw dCBzcmM9InNjcmlwdC10ZXN0cy9kZmctaW5saW5lLW5ldy1hcnJheS1idWZmZXIuanMiPjwvc2Ny aXB0PgorPHNjcmlwdCBzcmM9InJlc291cmNlcy9qcy10ZXN0LXBvc3QuanMiPjwvc2NyaXB0Pgor PC9ib2R5PgorPC9odG1sPgpJbmRleDogTGF5b3V0VGVzdHMvZmFzdC9qcy9zY3JpcHQtdGVzdHMv ZGZnLWlubGluZS1uZXctYXJyYXktYnVmZmVyLmpzCj09PT09PT09PT09PT09PT09PT09PT09PT09 PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT09PT0KLS0tIExheW91dFRlc3Rz L2Zhc3QvanMvc2NyaXB0LXRlc3RzL2RmZy1pbmxpbmUtbmV3LWFycmF5LWJ1ZmZlci5qcwkocmV2 aXNpb24gMCkKKysrIExheW91dFRlc3RzL2Zhc3QvanMvc2NyaXB0LXRlc3RzL2RmZy1pbmxpbmUt bmV3LWFycmF5LWJ1ZmZlci5qcwkocmV2aXNpb24gMCkKQEAgLTAsMCArMSwyNCBAQAorZGVzY3Jp cHRpb24oCisiVGhpcyB0ZXN0cyB0aGF0IGlubGluaW5nIGNvcnJlY3RseSBoYW5kbGVzIGNvbnN0 YW50IGJ1ZmZlcnMuIgorKTsKKworZnVuY3Rpb24gZm9vKCkgeworICAgIHJldHVybiBbMSwgMiwg MywgNF07Cit9CisKK2Z1bmN0aW9uIGJhcigpIHsKKyAgICByZXR1cm4gZm9vKCk7Cit9CisKK2Zv ciAodmFyIGkgPSAwOyBpIDwgMTAwMDsgKytpKSB7CisgICAgYmFyKCk7Cit9CisKK2ZvciAodmFy IGkgPSAwOyBpIDwgMTA7ICsraSkgeworICAgIHNob3VsZEJlKCJiYXIoKVswXSIsICIxIikKKyAg ICBzaG91bGRCZSgiYmFyKClbMV0iLCAiMiIpCisgICAgc2hvdWxkQmUoImJhcigpWzJdIiwgIjMi KQorICAgIHNob3VsZEJlKCJiYXIoKVszXSIsICI0IikKK30KKwordmFyIHN1Y2Nlc3NmdWxseVBh cnNlZCA9IHRydWU7Cg==