in Tamarin-Tracing, we experimented with an approach used in SpiderMonkey to improve array performance: assume most arrays are "mostly" dense, and allocate contiguous storage for them even if not 100% dense, falling back to a hashtable approach only if a certain threshold of sparseness is met. In most cases this is reputed to provide better performance (as well as better memory usage). See https://bugzilla.mozilla.org/show_bug.cgi?id=412813 for the original TT work.

This patch started out simply: in order to simply the "dense array" work, I wanted to remove the "friend" access that ArrayClass had, so that all knowledge of the innards of ArrayObject was kept internal. My change turned into serious scope creep, as I decided that the best way to do this was to change the affected methods (concat, splice, shift, reverse, pop) to be implemented as virtual methods based in ScriptObject, with overrides in ArrayObject... once I had done that I decided that it was more sensible to move everything else to mimic that as well, as several of them may also be more amenable to Array-specific optimization once true dense arrays are in place (eg indexOf), and this pattern will make that easier... One the one hand it's definitely a case of premature optimization, but on the other hand it's arguably cleaner code now. (Note that I left "sort" using the old pattern, though, as it would have required moving a huge chunk of code.)

Hm, I thought we already had an open bug on this topic... No matter. I will try to review the patch this week but other items are unfortuately higher priority.

(In reply to comment #2) > Hm, I thought we already had an open bug on this topic... I searched for one but could not find one that was relevant to TR. I may find a different reviewer if anyone else has substantially more bandwidth, but this is not a high-priority review.

I'd like to put a hold on this patch until the tamarin parts of the exactgc work land. I'm in continual rebasing hell because of the recent and sweeping list, array, and vector changes and I'd like to get out of it as soon as I can. (These deep changes are all good for the code base but they create a lot of trouble for me.)

(In reply to comment #4) > I'd like to put a hold on this patch until the tamarin parts of the exactgc > work land. Not a problem -- I'm going to continue to poke at it but won't plan on landing it anytime soon. (Review on your part would still be welcome, at your convenience...)

Here's a prototype patch that works and passes acceptance tests; unfortunately, it's not a performance win. The only places I'm seeing performances gains are benches that had known issue with growing dense arrays downward (a known limitation of the existing code); in particular, boidshack is +10% and gameoflife is %25%... but a lot of other benches are off in the -3% to -5% range. More investigation needed to figure out why. Archiving this patch here for now.

Comment on attachment 489264 [details] [diff] [review] Prototype of dense arrays (In reply to comment #6) > The only places I'm seeing performances gains are > benches that had known issue with growing dense arrays downward (a known > limitation of the existing code); in particular, boidshack is +10% and > gameoflife is %25%... That's really the only place I'd expect to see a win, so, great (and i hapilly disagree that the patch isn't a win). The regressions need looking into, assuming they're outside the noise range (obviously). The other thing to hand-check is to see if the common access paths are any shorter, or if they could be made any shorter within the scope of this patch. Ideally, the expression (x = a[i]) would only involve a null pointer check on a, a dense check, a range check on I, and a load. Writing down the access path by hand could uncover something. My favorite way is looking at assembly code in gdb, but YMMV. > if (m_type == kDense && isDenseIndex(index)) at the risk of cleverness, could the dense range encode "not dense" (say, m_denseUsed == 0), and avoid the m_type load-and-compare? (maybe m_denseUsed == -1 means kSealed, so we can tell the difference between sparse and sealed on the slow path?)

I found some subtle errors that are causing us to revert to to sparse arrays much more frequently, but there are still other regressions. I'll keep poking as I have time and post a new patch when I get more info.

another gotcha is that we were reverting to sparse if there were any non-numeric properties, meaning that all "arguments" were sparse due to the "callee" property.

(In reply to comment #7) > at the risk of cleverness, could the dense range encode "not dense" (say, > m_denseUsed == 0), and avoid the m_type load-and-compare? I like it, but m_denseStart is even better since we have to use it anyway... stand by for a new patch.

This uses m_denseStart as a sentinel field to avoid unnecessary dense-vs-sparse checks, plus some other streamlining, to eliminate the performance regressions... we still get the wins on grow-down cases, but many other benches are often now in the +3% or more range, with almost no regressions. (numbers to be posted subsequently) In fact, I think this is almost review-worthy; the main things that need fiddling with are: -- more testing, including memory load testing and perf testing on non-x86 -- integration with Flash/AIR to see if I changed anything they happened to rely on -- the trick used here relies on having dense arrays only when no numeric properties are > 0x7fffffff; I think this is correctly checked everywhere but there might be pathologies in concat, splice, etc that could allow for such a thing to happen; need to construct specific test cases to look for this

Results from tests on MacTel32 below. Contrary to my previous statement, most results are in the noise range (on my machine, anything +- 1% is noise). I'll run longer tests overnight. avm avm2 test best avg best avg %dBst %dAvg Metric: time Dir: misc/ boids 1951 1959 1840 1847.3 5.7 5.7 ++ boidshack 502 505.7 446 447.3 11.2 11.5 ++ gameoflife 2470 2475.3 1814 1817.3 26.6 26.6 ++ primes 5156 5174.7 5155 5164.3 0.0 0.2 Dir: scimark/ FFT 2280 2282.7 2312 2315 -1.4 -1.4 - LU 3085 3090.3 3029 3035.3 1.8 1.8 + MonteCarlo 2752 2757.7 2717 2733.7 1.3 0.9 + SOR 2531 2535.3 2568 2573.7 -1.5 -1.5 - SparseCompRow 102 102.3 99 99.3 2.9 2.9 + Dir: jsbench/ Crypt 3837 3847.7 3802 3805 0.9 1.1 + Euler 7433 7465.7 7210 7240.7 3.0 3.0 + FFT 7057 7168.3 7084 7103 -0.4 0.9 HeapSort 3537 3556.7 3585 3596.7 -1.4 -1.1 - LUFact 5869 5879 5791 5802 1.3 1.3 + Moldyn 9971 10004 10010 10039.3 -0.4 -0.4 RayTracer 7088 7110 7080 7103.3 0.1 0.1 SOR 32555 32752 32598 32665 -0.1 0.3 Series 8907 8917.7 9005 9047.3 -1.1 -1.5 - SparseMatmult 10411 10497 10322 10428.3 0.9 0.7 Dir: jsbench/typed/ Crypt 965 965.7 965 966.3 0 -0.1 Euler 9192 9247 9108 9153 0.9 1.0 + FFT 1577 1587 1626 1629.7 -3.1 -2.7 - HeapSort 1373 1378 1391 1393.7 -1.3 -1.1 - LUFact 1327 1335.7 1332 1339.3 -0.4 -0.3 Moldyn 4554 4559.3 4545 4560.3 0.2 -0.0 RayTracer 1348 1352.3 1351 1351.7 -0.2 0.0 SOR 4472 4476.7 4467 4487.3 0.1 -0.2 Series 7704 7730.7 7701 7717.7 0.0 0.2 SparseMatmult 2644 2671.7 2644 2664 0 0.3 Dir: v8/ crypto 448 448 444 442.7 -0.9 -1.2 - deltablue 1553 1551 1549 1531.7 -0.3 -1.2 earley-boyer 1009 1002 1005 989 -0.4 -1.3 raytrace 2903 2881 2891 2873 -0.4 -0.3 richards 1001 1000.3 973 972 -2.8 -2.8 - Dir: v8/typed/ crypto 466 463.3 455 452.3 -2.4 -2.4 - deltablue 2422 2420.7 2439 2429.3 0.7 0.4 + earley-boyer 1002 998 1008 1002 0.6 0.4 raytrace 6157 6148.3 6249 6173 1.5 0.4 + richards 1854 1847 1859 1847 0.3 0 Dir: v8.5/js/ crypto 397 392.3 394 390.7 -0.8 -0.4 deltablue 304 304 299 298.3 -1.6 -1.9 - earley-boyer 1005 998.7 1001 999.7 -0.4 0.1 raytrace 648 647 658 656.7 1.5 1.5 + regexp 87.3 87.1 88.2 88.1 1.0 1.1 + richards 244 242.3 243 240 -0.4 -1.0 splay 738 737 749 747.3 1.5 1.4 + Dir: v8.5/optimized/ crypto 3544 3536 3536 3531 -0.2 -0.1 deltablue 2991 2972.3 3116 3098.7 4.2 4.3 + earley-boyer 992 976 995 989.3 0.3 1.4 raytrace 7525 7505 7488 7376.7 -0.5 -1.7 regexp 87.5 86.8 88.5 88.5 1.1 1.9 + richards 3703 3693 3731 3698.3 0.8 0.1 splay 5443 5406 5520 5477.7 1.4 1.3 + Dir: v8.5/typed/ crypto 2736 2734 2752 2748.7 0.6 0.5 + deltablue 3265 3257.3 3214 3208.7 -1.6 -1.5 - earley-boyer 1003 993 1001 998 -0.2 0.5 raytrace 7525 7522.3 7466 7400 -0.8 -1.6 regexp 87.3 87.2 88.9 88.7 1.8 1.7 + richards 3707 3704.3 3739 3739 0.9 0.9 + splay 921 919 933 925.7 1.3 0.7 + Dir: v8.5/untyped/ crypto 439 439 436 435.3 -0.7 -0.8 - deltablue 1598 1597.7 1605 1604.7 0.4 0.4 + earley-boyer 1003 1000.3 994 986 -0.9 -1.4 raytrace 3109 3103 3118 3112 0.3 0.3 regexp 87.6 87.5 88.7 88.6 1.3 1.3 + richards 387 386.3 391 387.3 1.0 0.3 + splay 875 874 886 884 1.3 1.1 +

Results from overnight run, 32 and 64 are below. On the whole, a few real positives, a lot of stuff in the noise range and a few meaningful regressions that need investigating. stevoquad:tl stejohns$ sudo nice -n -20 ../rperf-test.sh ../tr 5 Password: Tamarin tests started: 2010-11-10 18:16:46.748911 Executing 460 test(s) avm: /Users/stejohns/tamdev/tr/platform/mac/avmshell/build/Release/avm version: cyclone avm2: /Users/stejohns/tamdev/tl/platform/mac/avmshell/build/Release/avm version: cyclone iterations: 5 avm avm2 test best avg best avg %dBst %dAvg Metric: iterations/second Dir: asmicro/ alloc-1 37 37 37 37 0 0 alloc-10 13 13 14 13.8 7.7 6.2 ++ alloc-11 12 12 12 12 0 0 alloc-12 8 8 12 12 50 50 ++ alloc-13 84 83.4 83 83 -1.2 -0.5 - alloc-14 69 68.8 69 68.8 0 0 alloc-2 19 19 19 19 0 0 alloc-3 15 15 16 16 6.7 6.7 ++ alloc-4 42 42 42 42 0 0 alloc-5 29 28.2 29 29 0 2.8 alloc-6 67 66.6 67 67 0 0.6 alloc-7 38 38 38 38 0 0 alloc-8 15 15 15 15 0 0 alloc-9 15 15 15 15 0 0 arguments-1 676 674.2 675 674.4 -0.1 0.0 arguments-2 371 370.6 375 374 1.1 0.9 + arguments-3 19 19 19 18.8 0 -1.1 arguments-4 26 26 26 26 0 0 array-1 1473 1471.6 1611 1610.2 9.4 9.4 ++ array-2 500 497.4 532 531.4 6.4 6.8 ++ array-pop-1 304 301.2 286 283.8 -5.9 -5.8 -- array-push-1 251 250.2 254 254 1.2 1.5 + array-shift-1 131 130.6 156 154.6 19.1 18.4 ++ array-slice-1 17 17 17 16.4 0 -3.5 array-sort-1 25 25 24 24 -4 -4 - array-sort-2 2 2 2 2 0 0 array-sort-3 20 19.2 20 20 0 4.2 array-sort-4 9 9 9 9 0 0 array-unshift-1 147 146.6 153 152 4.1 3.7 + closedvar-read-1 3634 3626 3607 3605 -0.7 -0.6 - closedvar-write-1 4431 4391.4 4001 3982.8 -9.7 -9.3 -- closedvar-write-2 4743 4670 4557 4483.4 -3.9 -4.0 - do-1 3594 3590.6 3594 3591.6 0 0.0 for-1 3588 3585.4 3589 3585 0.0 -0.0 for-2 2189 2188.4 2191 2189.4 0.1 0.0 + for-3 2289 2287 2291 2288.8 0.1 0.1 for-in-1 339 339 326 326 -3.8 -3.8 - for-in-2 140 139.4 312 311.6 122.9 123.5 ++ funcall-1 321 320.2 321 320.4 0 0.1 funcall-2 210 210 210 210 0 0 funcall-3 296 295.4 300 300 1.4 1.6 + funcall-4 103 103 101 101 -1.9 -1.9 - globalvar-read-1 3615 3612.6 3602 3598.6 -0.4 -0.4 - globalvar-write-1 4685 4680.2 4313 4288 -7.9 -8.4 -- isNaN-1 3415 3409 3563 3559.2 4.3 4.4 + isNaN-2 3592 3583.8 3591 3588.2 -0.0 0.1 isNaN-3 3314 3277.2 3167 3165.2 -4.4 -3.4 - lookup-array-fetch-1 690 689.4 701 701 1.6 1.7 + lookup-array-in-1 1517 1513.2 1528 1521.8 0.7 0.6 + lookup-negindex-array-1 404 403.2 401 400.6 -0.7 -0.6 - lookup-negindex-array-2 348 347.8 344 343.2 -1.1 -1.3 - lookup-negindex-object-1 417 417 416 416 -0.2 -0.2 lookup-negindex-object-2 392 391.6 392 391.2 0 -0.1 lookup-object-fetch-1 787 786.8 788 787.4 0.1 0.1 + lookup-object-in-1 1358 1356.2 1358 1357 0 0.1 number-toString-1 5 5 5 5 0 0 number-toString-2 66 66 65 65 -1.5 -1.5 - oop-1 3 3 3 3 0 0 parseFloat-1 64 63.8 66 66 3.1 3.4 + parseInt-1 159 159 159 159 0 0 regex-exec-1 62 62 63 62.4 1.6 0.6 + regex-exec-2 73 72.4 74 74 1.4 2.2 + regex-exec-3 104 104 106 106 1.9 1.9 + regex-exec-4 284 283 285 284.4 0.4 0.5 restarg-1 676 675.2 675 674.4 -0.1 -0.1 - restarg-2 398 397.6 398 398 0 0.1 restarg-3 34 33.4 33 33 -2.9 -1.2 - restarg-4 26 26 26 26 0 0 string-casechange-1 23 23 23 22.2 0 -3.5 string-casechange-2 22 22 23 23 4.5 4.5 + string-charAt-1 1291 1290.4 1291 1290.6 0 0.0 string-charAt-2 75 75 74 74 -1.3 -1.3 - string-charCodeAt-1 1372 1371.4 1369 1368 -0.2 -0.2 - string-charCodeAt-2 1283 1282.6 1283 1281.8 0 -0.1 string-charCodeAt-3 1029 1027.8 1025 1024.6 -0.4 -0.3 - string-charCodeAt-4 1732 1730.2 1731 1728.6 -0.1 -0.1 string-charCodeAt-5 864 863.2 864 863.4 0 0.0 string-charCodeAt-6 1372 1371.2 1368 1367.4 -0.3 -0.3 - string-charCodeAt-7 1731 1730.6 1732 1730.4 0.1 -0.0 string-fromCharCode-1 288 288 287 286.6 -0.3 -0.5 - string-fromCharCode-2 59 59 57 57 -3.4 -3.4 - string-indexOf-1 202 201.2 191 190.6 -5.4 -5.3 -- string-indexOf-2 114 114 114 114 0 0 string-indexOf-3 103 103 103 103 0 0 string-lastIndexOf-1 461 461 461 461 0 0 string-lastIndexOf-2 148 147.4 148 145.6 0 -1.2 string-lastIndexOf-3 142.7 142.1 142.6 142.3 -0.1 0.1 string-slice-1 112 111.8 111 111 -0.9 -0.7 - string-split-1 8 8 8 8 0 0 string-split-2 8 8 8 8 0 0 string-substring-1 120 119.6 119 118.4 -0.8 -1.0 - switch-1 701 700.6 700 700 -0.1 -0.1 - switch-2 98 98 98 98 0 0 switch-3 172 172 172 172 0 0 try-1 219 219 219 219 0 0 try-2 14 14 14 14 0 0 try-3 46 46 45 45 -2.2 -2.2 - vector-push-1 40 40 39 39 -2.5 -2.5 - while-1 3590 3588.6 3586 3584 -0.1 -0.1 - Metric: time Dir: jsbench/ Crypt 3831 3833 3788 3793 1.1 1.0 + Euler 7410 7412.4 7178 7181.4 3.1 3.1 + FFT 7230 7245.8 7019 7085.4 2.9 2.2 + HeapSort 3532 3533 3574 3576.4 -1.2 -1.2 - LUFact 5884 5888.4 5742 5743.8 2.4 2.5 + Moldyn 10003 10018.4 9965 10018.6 0.4 -0.0 RayTracer 7071 7082 7090 7113.8 -0.3 -0.4 SOR 32615 32652.4 32178 32232.6 1.3 1.3 + Series 8900 8903.4 8933 8938.2 -0.4 -0.4 - SparseMatmult 10318 10330.6 10186 10242.6 1.3 0.9 + Dir: jsbench/typed/ Crypt 963 964.2 965 966 -0.2 -0.2 - Euler 9150 9158.8 9047 9062 1.1 1.1 + FFT 1563 1564.8 1558 1560 0.3 0.3 + HeapSort 1369 1369.6 1382 1383.6 -0.9 -1.0 - LUFact 1317 1318 1322 1323.6 -0.4 -0.4 - Moldyn 4519 4526.8 4499 4523.8 0.4 0.1 + RayTracer 1346 1346.6 1349 1349.4 -0.2 -0.2 - SOR 4445 4447.8 4442 4450.2 0.1 -0.1 Series 7688 7690 7680 7686 0.1 0.1 SparseMatmult 2640 2640.6 2640 2647.8 0 -0.3 Metric: iterations/second Dir: jsmicro/ alloc-1 37 36.4 36 36 -2.7 -1.1 - alloc-10 13 13 13 13 0 0 alloc-11 11 11 11 11 0 0 alloc-12 7 7 11 11 57.1 57.1 ++ alloc-13 71 71 72 72 1.4 1.4 + alloc-14 61 60.6 62 61.4 1.6 1.3 alloc-2 19 19 19 19 0 0 alloc-3 15 15 16 16 6.7 6.7 ++ alloc-4 42 42 41 41 -2.4 -2.4 - alloc-5 29 28.4 29 28.8 0 1.4 alloc-6 58 58 59 59 1.7 1.7 + alloc-7 37 37 37 37 0 0 alloc-8 14 14 14 14 0 0 alloc-9 14 14 14 14 0 0 arguments-1 147 147 147 147 0 0 arguments-2 104 104 104 104 0 0 arguments-3 17 17 17 17 0 0 array-1 372 372 382 381.8 2.7 2.6 + array-2 283 283 260 260 -8.1 -8.1 -- array-pop-1 61 60.8 64 63.2 4.9 3.9 + array-push-1 41 41 42 42 2.4 2.4 + array-shift-1 52 52 58 58 11.5 11.5 ++ array-slice-1 14 14 14 14 0 0 array-sort-1 23 23 23 23 0 0 array-sort-2 2 2 2 2 0 0 array-sort-3 20 19.6 20 19.8 0 1.0 array-sort-4 9 9 9 9 0 0 array-unshift-1 21 21 12 12 -42.9 -42.9 -- closedvar-read-1 603 602.2 612 610.8 1.5 1.4 + closedvar-write-1 443 442 443 442 0 0 closedvar-write-2 441 438.8 440 439.4 -0.2 0.1 do-1 609 609 610 609.4 0.2 0.1 + for-1 602 601.8 602 602 0 0.0 for-2 185 185 185 185 0 0 for-3 135 135 135 135 0 0 for-in-1 268 268 260 259.2 -3.0 -3.3 - for-in-2 129 128.6 251 251 94.6 95.2 ++ funcall-1 208 208 208 208 0 0 funcall-2 203 203 203 203 0 0 funcall-3 200 199.6 200 200 0 0.2 funcall-4 1332 1330.6 1331 1329.6 -0.1 -0.1 globalvar-read-1 600 599.6 600 599.4 0 -0.0 globalvar-write-1 430 429.8 430 429.4 0 -0.1 isNaN-1 518 518 518 517.6 0 -0.1 lookup-array-fetch-1 602 601.8 599 597.6 -0.5 -0.7 - lookup-array-in-1 990 989.2 984 982.4 -0.6 -0.7 - lookup-object-fetch-1 674 674 680 678 0.9 0.6 + lookup-object-in-1 887 886.2 882 882 -0.6 -0.5 - number-toString-1 5 5 5 5 0 0 number-toString-2 54 54 55 55 1.9 1.9 + oop-1 3 3 3 3 0 0 parseFloat-1 48 48 50 50 4.2 4.2 + parseInt-1 112 112 111 111 -0.9 -0.9 regex-exec-1 55 54.8 56 56 1.8 2.2 + regex-exec-2 68 67.2 69 69 1.5 2.7 + regex-exec-3 95 94.4 96 95.6 1.1 1.3 + regex-exec-4 235 233.8 239 238.4 1.7 2.0 + string-casechange-1 15 15 15 15 0 0 string-casechange-2 15 15 15 15 0 0 string-charAt-1 97 97 97 97 0 0 string-charAt-2 43 43 42 42 -2.3 -2.3 - string-charCodeAt-1 88 87.4 88 88 0 0.7 string-charCodeAt-2 89 89 88 87.6 -1.1 -1.6 - string-fromCharCode-1 74 74 73 73 -1.4 -1.4 - string-fromCharCode-2 37 37 36 36 -2.7 -2.7 - string-fromCharCode-3 66 65.8 65 64.4 -1.5 -2.1 - string-fromCharCode-4 69 69 68 68 -1.4 -1.4 - string-indexOf-1 79 79 78 78 -1.3 -1.3 - string-indexOf-2 51 51 51 51 0 0 string-indexOf-3 49 49 49 49 0 0 string-lastIndexOf-1 79 79 79 79 0 0 string-lastIndexOf-2 57 56.6 57 57 0 0.7 string-lastIndexOf-3 73 73 75 75 2.7 2.7 + string-slice-1 47 47 47 47 0 0 string-split-1 7 7 7 7 0 0 string-split-2 7 7 7 7 0 0 string-substring-1 50 50 50 49.4 0 -1.2 switch-1 109 109 109 109 0 0 switch-2 55 55 55 55 0 0 switch-3 68 67.2 68 67.2 0 0 try-1 144 143.2 143 143 -0.7 -0.1 - try-2 14 14 14 14 0 0 try-3 35 35 35 34.2 0 -2.3 while-1 602 602 603 602.2 0.2 0.0 + Metric: time Dir: language/describetype/ desctypeperf 603 603.6 593 593.8 1.7 1.6 + Dir: language/e4x/ addingToXMLList 18 18.8 19 19 -5.6 -1.1 -- appendChildAndString 53 53.4 53 53.6 0 -0.4 concatenatingStringsFromE4X 7 7.2 7 7.2 0 0 simpleStringConcatenation 1 1.8 1 1.8 0 0 usingAppendChildAndE4X 54 54.8 54 54.8 0 0 Dir: language/string/ append_concat 82 82 81 81 1.2 1.2 + append_equal_plus 67 68.2 66 66.4 1.5 2.6 append_plus_equal 67 67.6 66 66.6 1.5 1.5 charAt 174 174.6 174 174.2 0 0.2 charCodeAt 205 206.2 204 204.2 0.5 1.0 indexOf 302 302.8 303 303 -0.3 -0.1 - lastIndexOf 208 208 208 208 0 0 replace 518 518.8 533 539 -2.9 -3.9 - replace2 822 822.8 817 817.4 0.6 0.7 + search 35 35.6 35 35.2 0 1.1 slice 277 277.8 278 278.2 -0.4 -0.1 split 333 333.8 328 329.6 1.5 1.3 + static_ascii_array_100 976 976.4 251 251.8 74.3 74.2 ++ static_ascii_array_50 961 961.6 232 233.6 75.9 75.7 ++ static_latin1_array_100 1917 1922.6 473 474.8 75.3 75.3 ++ static_latin1_array_50 961 961 237 237.8 75.3 75.3 ++ substr 203 203.6 204 204.4 -0.5 -0.4 - substring 193 193.6 194 194.2 -0.5 -0.3 - Dir: language/string/typed/ append_concat 73 73.8 72 72.4 1.4 1.9 + append_equal_plus 61 61.4 61 61.4 0 0 append_plus_equal 62 62.2 61 61.4 1.6 1.3 + charAt 10 11 11 11.6 -10 -5.5 - charCodeAt 10 10.6 10 11 0 -3.8 indexOf 302 302.4 303 303 -0.3 -0.2 - lastIndexOf 207 207.8 208 208 -0.5 -0.1 - replace 517 518 519 525 -0.4 -1.4 replace2 824 824.4 819 820 0.6 0.5 + search 35 35.4 35 35.4 0 0 slice 182 182.2 183 183.6 -0.5 -0.8 - split 334 334.6 332 332.8 0.6 0.5 + substr 144 144 144 144.2 0 -0.1 substring 126 126.6 127 127.4 -0.8 -0.6 - Dir: misc/ boids 1938 1938.2 1833 1834.2 5.4 5.4 ++ boidshack 497 498.2 445 446 10.5 10.5 ++ gameoflife 2463 2464.2 1807 1808.2 26.6 26.6 ++ primes 5147 5148.2 5143 5145.4 0.1 0.1 + Dir: mmgc/ gcbench 2934 2938 2929 2939.8 0.2 -0.1 ofib-rc 269 269.8 269 269.4 0 0.1 ofib 1180 1200.8 1217 1232.6 -3.1 -2.6 - sfib 377 377 374 374.2 0.8 0.7 + Dir: scimark/ FFT 2270 2276 2304 2308.6 -1.5 -1.4 - LU 3064 3066 3019 3019.6 1.5 1.5 + MonteCarlo 2747 2748.2 2711 2715 1.3 1.2 + SOR 2523 2524.4 2558 2560.8 -1.4 -1.4 - SparseCompRow 101 101.8 99 99.2 2.0 2.6 + Dir: sunspider/ access-binary-trees 33 33.6 33 33.8 0 -0.6 access-fannkuch 87 87 82 82.4 5.7 5.3 ++ access-nbody 84 84 83 84 1.2 0 access-nsieve 44 44.4 26 26 40.9 41.4 ++ bitops-3bit-bits-in-byte 10 10.2 10 10.4 0 -2.0 bitops-bits-in-byte 29 29.8 29 29.8 0 0 bitops-bitwise-and 210 210 209 209.8 0.5 0.1 + bitops-nsieve-bits 41 41.6 41 41 0 1.4 controlflow-recursive 15 15.2 15 15 0 1.3 crypto-aes 41 41.8 41 41.4 0 1.0 crypto-md5 20 20 19 19.4 5 3.0 + crypto-sha1 19 19.8 20 20 -5.3 -1.0 -- date-format-tofte 171 177.2 171 179.4 0 -1.2 math-cordic 54 54 54 54.4 0 -0.7 math-partial-sums 193 193 194 194.4 -0.5 -0.7 - math-spectral-norm 28 28 27 27.8 3.6 0.7 + s3d-cube 69 70 68 68.6 1.4 2.0 s3d-morph 41 41.2 41 41.4 0 -0.5 s3d-raytrace 79 79 78 78.8 1.3 0.3 + string-fasta 81 81.4 87 87 -7.4 -6.9 -- string-unpack-code 171 171 170 170 0.6 0.6 string-validate-input 43 43.4 43 43.6 0 -0.5 Dir: sunspider/as3/ access-binary-trees 9 9 9 9.2 0 -2.2 access-fannkuch 56 56 49 49.8 12.5 11.1 ++ access-nbody 6 6.4 6 6.4 0 0 access-nsieve 33 33 14 14 57.6 57.6 ++ bitops-3bit-bits-in-byte 5 5.8 5 5.8 0 0 bitops-bits-in-byte 9 9.2 9 9.2 0 0 bitops-bitwise-and 2 2 2 2.4 0 -20.0 bitops-nsieve-bits 27 27 27 27 0 0 controlflow-recursive 4 4 4 4 0 0 crypto-aes 31 31.4 31 31.8 0 -1.3 crypto-md5 24 24.2 23 23.6 4.2 2.5 + crypto-sha1 18 18.6 18 18.8 0 -1.1 date-format-tofte 164 168.2 163 168 0.6 0.1 math-cordic 14 14 13 13.6 7.1 2.9 + math-partial-sums 65 66 66 66.2 -1.5 -0.3 math-spectral-norm 5 5.8 5 5.4 0 6.9 s3d-cube 21 21 20 20.6 4.8 1.9 + s3d-morph 29 29.4 29 29 0 1.4 s3d-raytrace 32 32.8 32 32.6 0 0.6 string-fasta 40 40 40 40 0 0 string-unpack-code 169 169.2 169 169.2 0 0 string-validate-input 33 33 33 33 0 0 Dir: sunspider/as3vector/ access-fannkuch 22 22.6 22 22.2 0 1.8 access-nbody 6 6 6 6 0 0 access-nsieve 12 12 11 11.8 8.3 1.7 ++ bitops-nsieve-bits 8 8 8 8 0 0 math-cordic 14 14 14 14 0 0 math-spectral-norm 16 16.4 16 16.2 0 1.2 s3d-cube 16 16.8 16 16.6 0 1.2 s3d-morph 16 16.2 16 16.4 0 -1.2 string-fasta 43 43 42 42.2 2.3 1.9 + string-validate-input 35 35 35 35.2 0 -0.6 Dir: sunspider-0.9.1/js/ access-binary-trees 33 33 33 33 0 0 access-fannkuch 85 85.6 81 81.2 4.7 5.1 + access-nbody 78 78.2 78 78 0 0.3 access-nsieve 42 42.6 25 25 40.5 41.3 ++ bitops-3bit-bits-in-byte 8 8.4 9 9 -12.5 -7.1 -- bitops-bits-in-byte 29 29.2 29 29.4 0 -0.7 bitops-bitwise-and 210 210.6 209 209.8 0.5 0.4 + bitops-nsieve-bits 40 40 39 39.6 2.5 1.0 + controlflow-recursive 15 15 14 14.6 6.7 2.7 + crypto-aes 35 35.2 35 35.6 0 -1.1 crypto-md5 16 16 16 16 0 0 crypto-sha1 18 18 17 17.6 5.6 2.2 + math-cordic 52 52.6 52 52 0 1.1 math-partial-sums 194 194.6 196 196 -1.0 -0.7 - math-spectral-norm 27 27.8 27 27.8 0 0 regexp-dna 826 829.6 825 827.6 0.1 0.2 s3d-cube 62 62.4 61 61.8 1.6 1.0 + s3d-morph 57 57.4 57 57.6 0 -0.3 s3d-raytrace 72 72.2 71 71.4 1.4 1.1 + string-fasta 56 56.4 57 57.2 -1.8 -1.4 - string-unpack-code 4386 4402 4399 4433 -0.3 -0.7 string-validate-input 42 42 42 42 0 0 Dir: sunspider-0.9.1/typed/ access-binary-trees 8 8.6 8 8.2 0 4.7 access-fannkuch 57 57.8 53 53 7.0 8.3 ++ access-nbody 4 4 4 4.2 0 -5.0 access-nsieve 30 30 12 12 60 60 ++ bitops-3bit-bits-in-byte 5 5 5 5 0 0 bitops-bits-in-byte 8 8.8 8 8.8 0 0 bitops-bitwise-and 2 2 2 2 0 0 bitops-nsieve-bits 26 26 25 25.6 3.8 1.5 + controlflow-recursive 3 3.6 3 3.6 0 0 crypto-aes 26 26 26 26.2 0 -0.8 crypto-md5 4 4.4 4 4.2 0 4.5 crypto-sha1 4 4.4 4 4.2 0 4.5 math-cordic 12 12.6 12 12.4 0 1.6 math-partial-sums 11 11.4 11 11 0 3.5 math-spectral-norm 8 8 7 7.8 12.5 2.5 ++ regexp-dna 795 816.4 794 800 0.1 2.0 s3d-cube 17 17.6 17 17.2 0 2.3 s3d-morph 47 47.4 46 47 2.1 0.8 s3d-raytrace 8 8.6 8 8.6 0 0 string-fasta 31 31 31 31.4 0 -1.3 string-unpack-code 4373 4393.8 4380 4487 -0.2 -2.1 string-validate-input 26 26.2 26 26.6 0 -1.5 Metric: v8 custom v8 normalized metric (hardcoded in the test) Dir: v8/ crypto 450 449.6 445 444 -1.1 -1.2 - deltablue 1557 1555.8 1553 1550.4 -0.3 -0.3 - earley-boyer 1018 1011.8 1012 1007.4 -0.6 -0.4 raytrace 2909 2906 2900 2894.4 -0.3 -0.4 - richards 1002 1001.4 978 977.2 -2.4 -2.4 - Dir: v8/typed/ crypto 467 466.4 456 455.4 -2.4 -2.4 - deltablue 2434 2432.4 2439 2438.4 0.2 0.2 + earley-boyer 1019 1014.8 1012 1008.6 -0.7 -0.6 - raytrace 6224 6213 6317 6312 1.5 1.6 + richards 1858 1856.4 1866 1864.4 0.4 0.4 + Dir: v8.5/js/ crypto 399 399 398 397.8 -0.3 -0.3 - deltablue 305 304.6 300 299.2 -1.6 -1.8 - earley-boyer 1012 1011.2 1010 1004.2 -0.2 -0.7 raytrace 650 649.4 659 658.8 1.4 1.4 + regexp 87.3 87.2 88.5 88.2 1.4 1.2 + richards 245 245 244 244 -0.4 -0.4 splay 742 736.2 753 752.2 1.5 2.2 + Dir: v8.5/optimized/ crypto 3558 3556.8 3560 3558.2 0.1 0.0 + deltablue 3013 3010.6 3136 3134.8 4.1 4.1 + earley-boyer 1006 1001.2 1005 1001.2 -0.1 0 raytrace 7547 7535.2 7502 7492.2 -0.6 -0.6 - regexp 87.7 87.4 88.9 88.5 1.4 1.3 + richards 3715 3714.2 3739 3737.4 0.6 0.6 + splay 5522 5489.8 5532 5494.6 0.2 0.1 Dir: v8.5/typed/ crypto 2746 2744.8 2756 2754.2 0.4 0.3 + deltablue 3265 3265 3214 3212.2 -1.6 -1.6 - earley-boyer 1005 1002 1004 1002 -0.1 0 raytrace 7540 7527.8 7495 7481.6 -0.6 -0.6 - regexp 87.6 87.4 88.8 88.6 1.4 1.3 + richards 3719 3716.6 3747 3742.2 0.8 0.7 + splay 921 921 937 934.6 1.7 1.5 + Dir: v8.5/untyped/ crypto 440 439.2 437 436 -0.7 -0.7 - deltablue 1604 1601.4 1609 1607 0.3 0.3 + earley-boyer 1007 1001.8 1004 1000.2 -0.3 -0.2 raytrace 3112 3108.4 3124 3119.2 0.4 0.3 + regexp 87.8 87.7 89.1 88.9 1.5 1.5 + richards 387 385.6 391 388.4 1.0 0.7 + splay 877 871.6 888 879.4 1.3 0.9 Tamarin tests started: 2010-11-10 19:32:54.387210 Executing 460 test(s) avm: /Users/stejohns/tamdev/tr/platform/mac/avmshell/build/Release64/avm version: cyclone avm2: /Users/stejohns/tamdev/tl/platform/mac/avmshell/build/Release64/avm version: cyclone iterations: 5 avm avm2 test best avg best avg %dBst %dAvg Metric: iterations/second Dir: asmicro/ alloc-1 41 41 41 41 0 0 alloc-10 12 12 12 12 0 0 alloc-11 13 13 12 12 -7.7 -7.7 -- alloc-12 9 9 12 12 33.3 33.3 ++ alloc-13 100 100 101 101 1 1 alloc-14 78 78 78 78 0 0 alloc-2 22 22 23 22.8 4.5 3.6 + alloc-3 11 11 11 11 0 0 alloc-4 51 51 48 48 -5.9 -5.9 -- alloc-5 38 37.8 37 37 -2.6 -2.1 - alloc-6 77 77 78 78 1.3 1.3 + alloc-7 43 43 43 43 0 0 alloc-8 14 13.2 14 13.8 0 4.5 alloc-9 14 13.4 14 13.8 0 3.0 arguments-1 925 924 924 924 -0.1 0 - arguments-2 489 489 490 489.4 0.2 0.1 + arguments-3 22 22 22 22 0 0 arguments-4 30 30 30 30 0 0 array-1 2108 2106.6 2292 2290.8 8.7 8.7 ++ array-2 657 656.8 674 674 2.6 2.6 + array-pop-1 435 429.4 367 364 -15.6 -15.2 -- array-push-1 287 287 306 305.8 6.6 6.6 ++ array-shift-1 133 132.6 150 148.8 12.8 12.2 ++ array-slice-1 20 20 20 19.8 0 -1.0 array-sort-1 31 31 31 31 0 0 array-sort-2 2 2 2 2 0 0 array-sort-3 23 23 23 23 0 0 array-sort-4 11 11 11 11 0 0 array-unshift-1 140 140 143 142.4 2.1 1.7 + closedvar-read-1 4081 4079 4072 4069.8 -0.2 -0.2 - closedvar-write-1 4189 4185.8 4194 4190.2 0.1 0.1 + closedvar-write-2 4193 4192 4195 4191.2 0.0 -0.0 do-1 4193 4191.6 4195 4193.6 0.0 0.0 for-1 3595 3593 3598 3594.2 0.1 0.0 for-2 2291 2289.8 2292 2290.2 0.0 0.0 for-3 2291 2289.2 2292 2290.6 0.0 0.1 for-in-1 408 407.8 390 389.8 -4.4 -4.4 - for-in-2 188 188 362 362 92.6 92.6 ++ funcall-1 384 384 388 387.6 1.0 0.9 + funcall-2 290 289.2 283 283 -2.4 -2.1 - funcall-3 345 345 342 342 -0.9 -0.9 funcall-4 114 114 115 114.4 0.9 0.4 + globalvar-read-1 3602 3599.6 3600 3597.2 -0.1 -0.1 globalvar-write-1 4188 4175.2 4179 4146.8 -0.2 -0.7 isNaN-1 2987 2984.8 2987 2985.4 0 0.0 isNaN-2 3478 3475.6 3480 3477.8 0.1 0.1 isNaN-3 3578 3575 3573 3565.8 -0.1 -0.3 lookup-array-fetch-1 740 739.4 735 734.8 -0.7 -0.6 - lookup-array-in-1 1551 1550.4 1572 1570.4 1.4 1.3 + lookup-negindex-array-1 415 412.4 413 409.6 -0.5 -0.7 lookup-negindex-array-2 330 330 330 329.8 0 -0.1 lookup-negindex-object-1 424 421.8 423 420.8 -0.2 -0.2 lookup-negindex-object-2 364 363.8 366 365.2 0.5 0.4 + lookup-object-fetch-1 841 840.4 831 830.8 -1.2 -1.1 - lookup-object-in-1 1438 1437.4 1420 1419 -1.3 -1.3 - number-toString-1 7 7 7 7 0 0 number-toString-2 84 84 85 84.6 1.2 0.7 + oop-1 3 3 4 4 33.3 33.3 ++ parseFloat-1 77 77 77 77 0 0 parseInt-1 174 173.6 178 178 2.3 2.5 + regex-exec-1 77 77 70 69.6 -9.1 -9.6 -- regex-exec-2 94 94 89 89 -5.3 -5.3 -- regex-exec-3 122 122 119 118.2 -2.5 -3.1 - regex-exec-4 374 374 355 353.8 -5.1 -5.4 -- restarg-1 924 923.2 924 924 0 0.1 restarg-2 514 514 514 514 0 0 restarg-3 40 40 38 38 -5 -5 - restarg-4 31 30.2 30 30 -3.2 -0.7 - string-casechange-1 31 31 31 30.4 0 -1.9 string-casechange-2 31 31 30 30 -3.2 -3.2 - string-charAt-1 1524 1523 1525 1524.6 0.1 0.1 string-charAt-2 89 89 90 90 1.1 1.1 + string-charCodeAt-1 1901 1900.6 1901 1900.6 0 0 string-charCodeAt-2 1710 1704.8 1722 1707.2 0.7 0.1 string-charCodeAt-3 1248 1247.6 1249 1248 0.1 0.0 string-charCodeAt-4 2025 2024.4 2026 2025 0.0 0.0 string-charCodeAt-5 1134 1132.8 1134 1133.2 0 0.0 string-charCodeAt-6 1901 1899.6 1901 1900.6 0 0.1 string-charCodeAt-7 1979 1977.4 1980 1979.4 0.1 0.1 string-fromCharCode-1 276 266.2 356 342.6 29.0 28.7 ++ string-fromCharCode-2 68 68 68 68 0 0 string-indexOf-1 277 276.8 271 271 -2.2 -2.1 - string-indexOf-2 150 150 150 150 0 0 string-indexOf-3 108 108 108 108 0 0 string-lastIndexOf-1 682 681.2 682 681.6 0 0.1 string-lastIndexOf-2 152 151.8 152 151.6 0 -0.1 string-lastIndexOf-3 156.8 156.8 156.8 156.8 0 0 string-slice-1 135 135 135 135 0 0 string-split-1 8 8 8 8 0 0 string-split-2 8 8 8 8 0 0 string-substring-1 147 147 147 146.8 0 -0.1 switch-1 742 741.4 742 741.8 0 0.1 switch-2 118 117.2 117 117 -0.8 -0.2 - switch-3 221 221 221 221 0 0 try-1 284 284 285 285 0.4 0.4 try-2 14 14 14 14 0 0 try-3 59 59 59 59 0 0 vector-push-1 51 51 52 51.8 2.0 1.6 + while-1 3595 3593.6 3596 3592.6 0.0 -0.0 Metric: time Dir: jsbench/ Crypt 3264 3266.6 3233 3234.4 0.9 1.0 + Euler 5956 5971 5760 5772.6 3.3 3.3 + FFT 6697 6710.6 6633 6639.2 1.0 1.1 + HeapSort 2988 2989.2 2938 2939.6 1.7 1.7 + LUFact 5264 5274.8 5155 5164.6 2.1 2.1 + Moldyn 8875 8889.2 8879 8887.2 -0.0 0.0 RayTracer 5975 5979.4 5992 5997.2 -0.3 -0.3 - SOR 28363 28372.8 28249 28267.2 0.4 0.4 + Series 8159 8167.6 7889 7896.6 3.3 3.3 + SparseMatmult 9341 9371.4 9216 9242.6 1.3 1.4 + Dir: jsbench/typed/ Crypt 902 903.2 901 903.6 0.1 -0.0 Euler 6774 6792.4 6770 6818.6 0.1 -0.4 FFT 1450 1451.4 1356 1357.4 6.5 6.5 ++ HeapSort 1103 1110 1155 1199.2 -4.7 -8.0 LUFact 1046 1047.6 1049 1052.4 -0.3 -0.5 Moldyn 3552 3557 3538 3545.8 0.4 0.3 + RayTracer 1061 1062 1067 1068.2 -0.6 -0.6 - SOR 3492 3495 3480 3486.2 0.3 0.3 + Series 6614 6633.2 6362 6378.8 3.8 3.8 + SparseMatmult 2004 2005.4 2003 2008.6 0.0 -0.2 Metric: iterations/second Dir: jsmicro/ alloc-1 39 39 39 39 0 0 alloc-10 12 12 12 11.8 0 -1.7 alloc-11 12 12 12 11.6 0 -3.3 alloc-12 8 8 12 11.6 50 45.0 ++ alloc-13 87 87 88 87.4 1.1 0.5 + alloc-14 69 69 70 70 1.4 1.4 + alloc-2 22 22 22 22 0 0 alloc-3 10 10 11 10.8 10 8.0 ++ alloc-4 49 49 46 46 -6.1 -6.1 -- alloc-5 37 37 36 36 -2.7 -2.7 - alloc-6 69 68.4 69 69 0 0.9 alloc-7 43 43 43 43 0 0 alloc-8 13 13 13 13 0 0 alloc-9 13 13 13 13 0 0 arguments-1 173 173 170 169.2 -1.7 -2.2 - arguments-2 100 100 102 102 2 2 + arguments-3 19 19 19 19 0 0 array-1 408 408 415 415 1.7 1.7 + array-2 286 286 291 290.8 1.7 1.7 + array-pop-1 73 72.2 79 78.4 8.2 8.6 ++ array-push-1 52 51.8 53 52.8 1.9 1.9 + array-shift-1 58 58 62 62 6.9 6.9 ++ array-slice-1 17 17 17 17 0 0 array-sort-1 30 29.4 29 29 -3.3 -1.4 - array-sort-2 2 2 2 2 0 0 array-sort-3 23 23 23 23 0 0 array-sort-4 11 11 11 11 0 0 array-unshift-1 23 23 14 14 -39.1 -39.1 -- closedvar-read-1 647 646.6 647 647 0 0.1 closedvar-write-1 476 475.8 476 476 0 0.0 closedvar-write-2 476 475.8 476 476 0 0.0 do-1 682 681.8 682 681.2 0 -0.1 for-1 664 663.8 664 664 0 0.0 for-2 192 192 202 201.6 5.2 5.0 ++ for-3 149 149 147 147 -1.3 -1.3 - for-in-1 299 299 287 286.8 -4.0 -4.1 - for-in-2 156 156 272 271.8 74.4 74.2 ++ funcall-1 266 266 265 264.8 -0.4 -0.5 - funcall-2 247 247 250 249.8 1.2 1.1 + funcall-3 235 234.6 238 237.2 1.3 1.1 + funcall-4 1648 1647.6 1643 1642 -0.3 -0.3 - globalvar-read-1 647 646.8 647 646.4 0 -0.1 globalvar-write-1 476 476 476 476 0 0 isNaN-1 476 476 476 476 0 0 lookup-array-fetch-1 650 649.2 647 647 -0.5 -0.3 - lookup-array-in-1 1032 1030.8 1029 1028 -0.3 -0.3 - lookup-object-fetch-1 726 726 716 715.4 -1.4 -1.5 - lookup-object-in-1 945 944.4 941 939.8 -0.4 -0.5 - number-toString-1 7 7 7 7 0 0 number-toString-2 66 66 66 66 0 0 oop-1 3 3 3 3 0 0 parseFloat-1 58 57.6 57 57 -1.7 -1.0 - parseInt-1 126 126 126 126 0 0 regex-exec-1 63 63 58 58 -7.9 -7.9 -- regex-exec-2 79 78.4 79 78.4 0 0 regex-exec-3 106 105.2 105 104.2 -0.9 -1.0 - regex-exec-4 277 276.4 276 274.8 -0.4 -0.6 string-casechange-1 20 20 20 20 0 0 string-casechange-2 20 20 20 20 0 0 string-charAt-1 116 116 114 113.2 -1.7 -2.4 - string-charAt-2 51 51 51 51 0 0 string-charCodeAt-1 111 111 106 106 -4.5 -4.5 - string-charCodeAt-2 108 107.8 103 103 -4.6 -4.5 - string-fromCharCode-1 91 91 91 91 0 0 string-fromCharCode-2 44 44 44 44 0 0 string-fromCharCode-3 81 81 81 81 0 0 string-fromCharCode-4 85 85 86 86 1.2 1.2 + string-indexOf-1 95 95 98 98 3.2 3.2 + string-indexOf-2 64 64 65 65 1.6 1.6 + string-indexOf-3 55 55 56 56 1.8 1.8 + string-lastIndexOf-1 97 97 99 99 2.1 2.1 + string-lastIndexOf-2 65 64.4 66 66 1.5 2.5 + string-lastIndexOf-3 92 92 94 94 2.2 2.2 + string-slice-1 60 60 60 60 0 0 string-split-1 8 8 8 8 0 0 string-split-2 8 8 8 8 0 0 string-substring-1 60 60 60 60 0 0 switch-1 116 116 116 116 0 0 switch-2 64 64 64 64 0 0 switch-3 89 89 89 89 0 0 try-1 187 182.2 195 195 4.3 7.0 + try-2 14 13.8 14 14 0 1.4 try-3 48 48 48 48 0 0 while-1 664 663.6 664 663.6 0 0 Metric: time Dir: language/describetype/ desctypeperf 569 572.6 571 572.4 -0.4 0.0 Dir: language/e4x/ addingToXMLList 16 16 17 17 -6.2 -6.2 -- appendChildAndString 45 45.6 46 46.2 -2.2 -1.3 - concatenatingStringsFromE4X 5 5.4 6 6 -20 -11.1 -- simpleStringConcatenation 1 1.4 1 1.6 0 -14.3 usingAppendChildAndE4X 47 47.8 47 47.6 0 0.4 Dir: language/string/ append_concat 69 69.2 68 68.2 1.4 1.4 + append_equal_plus 58 58.4 57 57.2 1.7 2.1 + append_plus_equal 58 58.2 57 57.6 1.7 1.0 + charAt 130 130.2 130 130.2 0 0 charCodeAt 149 149.4 149 149.4 0 0 indexOf 302 302 302 302.4 0 -0.1 lastIndexOf 207 208 208 208 -0.5 0 - replace 397 399.8 400 400.6 -0.8 -0.2 - replace2 745 746.4 847 849 -13.7 -13.7 -- search 35 35.6 35 35.4 0 0.6 slice 207 208.4 206 206.8 0.5 0.8 split 440 441.2 442 446 -0.5 -1.1 static_ascii_array_100 847 852.2 268 269.4 68.4 68.4 ++ static_ascii_array_50 832 832.8 249 250.6 70.1 69.9 ++ static_latin1_array_100 1676 1678.2 514 515.8 69.3 69.3 ++ static_latin1_array_50 838 839.2 257 257.8 69.3 69.3 ++ substr 154 154 153 153 0.6 0.6 substring 147 147.6 148 148 -0.7 -0.3 - Dir: language/string/typed/ append_concat 64 64.4 63 63.4 1.6 1.6 + append_equal_plus 52 53 52 52.8 0 0.4 append_plus_equal 53 53.2 52 52.6 1.9 1.1 charAt 11 11 11 11.4 0 -3.6 charCodeAt 11 11.4 11 11.4 0 0 indexOf 302 302 301 301.4 0.3 0.2 + lastIndexOf 207 207.6 207 207.6 0 0 replace 391 395.2 396 397 -1.3 -0.5 - replace2 747 750.2 847 847.8 -13.4 -13.0 -- search 35 35.6 35 35 0 1.7 slice 144 144.4 144 144.2 0 0.1 split 440 440.8 424 443 3.6 -0.5 + substr 105 105.8 105 105.8 0 0 substring 96 96 96 96.2 0 -0.2 Dir: misc/ boids 1503 1504 1517 1517.4 -0.9 -0.9 - boidshack 418 419.4 375 376.8 10.3 10.2 ++ gameoflife 1830 1831.6 1278 1280 30.2 30.1 ++ primes 4826 4827 4828 4829.6 -0.0 -0.1 Dir: mmgc/ gcbench 2532 2605 2638 2640.8 -4.2 -1.4 - ofib-rc 289 292.4 359 362 -24.2 -23.8 -- ofib 1191 1194.2 1178 1189.2 1.1 0.4 sfib 315 315.2 312 312.6 1.0 0.8 + Dir: scimark/ FFT 1654 1656.2 1639 1639.4 0.9 1.0 + LU 2546 2547 2505 2507.6 1.6 1.5 + MonteCarlo 1582 1597 1534 1548 3.0 3.1 + SOR 1832 1833.6 1839 1841.4 -0.4 -0.4 - SparseCompRow 88 88.6 84 84.4 4.5 4.7 + Dir: sunspider/ access-binary-trees 30 30 29 29.6 3.3 1.3 + access-fannkuch 72 72 66 66.4 8.3 7.8 ++ access-nbody 65 65.2 65 66.2 0 -1.5 access-nsieve 42 42.4 25 25.8 40.5 39.2 ++ bitops-3bit-bits-in-byte 8 8.2 8 8 0 2.4 bitops-bits-in-byte 22 22.6 23 24 -4.5 -6.2 bitops-bitwise-and 183 183.2 180 180.8 1.6 1.3 + bitops-nsieve-bits 26 26.6 26 26 0 2.3 controlflow-recursive 13 13 13 13.4 0 -3.1 crypto-aes 34 34 34 34 0 0 crypto-md5 13 13.4 13 13.6 0 -1.5 crypto-sha1 11 11.6 12 12 -9.1 -3.4 -- date-format-tofte 208 212.2 207 210.8 0.5 0.7 math-cordic 45 45.2 45 45.6 0 -0.9 math-partial-sums 172 172 174 174 -1.2 -1.2 - math-spectral-norm 22 22.4 22 22.2 0 0.9 s3d-cube 59 59.4 59 59.4 0 0 s3d-morph 38 38.6 37 37.8 2.6 2.1 + s3d-raytrace 65 65.8 65 65 0 1.2 string-fasta 78 78.8 79 79 -1.3 -0.3 - string-unpack-code 119 119.6 119 119 0 0.5 string-validate-input 39 39.4 38 38 2.6 3.6 + Dir: sunspider/as3/ access-binary-trees 7 7.8 8 8 -14.3 -2.6 -- access-fannkuch 43 43 37 37 14.0 14.0 ++ access-nbody 5 5.6 5 5.2 0 7.1 access-nsieve 31 31.8 15 15 51.6 52.8 ++ bitops-3bit-bits-in-byte 4 4.4 4 4.4 0 0 bitops-bits-in-byte 8 8.4 8 8.2 0 2.4 bitops-bitwise-and 2 2 2 2 0 0 bitops-nsieve-bits 13 13 12 12.8 7.7 1.5 ++ controlflow-recursive 3 3.4 3 3.4 0 0 crypto-aes 26 26.4 26 26.6 0 -0.8 crypto-md5 10 10.2 10 10.4 0 -2.0 crypto-sha1 9 9 8 8.8 11.1 2.2 ++ date-format-tofte 186 197.4 186 196.6 0 0.4 math-cordic 12 12.8 13 13 -8.3 -1.6 -- math-partial-sums 63 63.4 65 65.2 -3.2 -2.8 - math-spectral-norm 4 4.8 5 5 -25 -4.2 -- s3d-cube 17 17.8 17 17.6 0 1.1 s3d-morph 24 24.2 24 24 0 0.8 s3d-raytrace 26 26.2 26 26 0 0.8 string-fasta 33 33.2 32 33 3.0 0.6 string-unpack-code 117 117.2 117 117.4 0 -0.2 string-validate-input 31 31.8 31 31.8 0 0 Dir: sunspider/as3vector/ access-fannkuch 20 20 20 20 0 0 access-nbody 5 5 5 5.2 0 -4.0 access-nsieve 11 11.6 11 11.2 0 3.4 bitops-nsieve-bits 6 6.8 6 6.4 0 5.9 math-cordic 11 11.4 11 11.8 0 -3.5 math-spectral-norm 12 12.6 12 12.8 0 -1.6 s3d-cube 14 14 14 14.4 0 -2.9 s3d-morph 13 13.8 13 13.4 0 2.9 string-fasta 35 35.2 35 35 0 0.6 string-validate-input 31 31.8 32 32 -3.2 -0.6 - Dir: sunspider-0.9.1/js/ access-binary-trees 29 29.4 30 30.8 -3.4 -4.8 - access-fannkuch 70 70.6 65 65 7.1 7.9 ++ access-nbody 62 62.6 62 62 0 1.0 access-nsieve 39 39.2 21 21.6 46.2 44.9 ++ bitops-3bit-bits-in-byte 6 6.8 7 7 -16.7 -2.9 -- bitops-bits-in-byte 22 22.4 22 22.2 0 0.9 bitops-bitwise-and 183 183 179 179.4 2.2 2.0 + bitops-nsieve-bits 24 24.8 25 25 -4.2 -0.8 - controlflow-recursive 13 13 13 13 0 0 crypto-aes 28 28.8 29 29.6 -3.6 -2.8 crypto-md5 10 10.2 10 10.2 0 0 crypto-sha1 9 9.6 9 9.6 0 0 math-cordic 44 44.6 44 44.6 0 0 math-partial-sums 182 182.6 181 182 0.5 0.3 math-spectral-norm 21 21.4 21 21.8 0 -1.9 regexp-dna 433 433.8 482 482.2 -11.3 -11.2 -- s3d-cube 54 54 54 54 0 0 s3d-morph 50 50.6 50 50.4 0 0.4 s3d-raytrace 60 60.6 60 60.2 0 0.7 string-fasta 59 59 60 60.6 -1.7 -2.7 - string-unpack-code 2853 2881.4 2901 2929.4 -1.7 -1.7 string-validate-input 37 37.4 36 36.6 2.7 2.1 + Dir: sunspider-0.9.1/typed/ access-binary-trees 7 7 7 7 0 0 access-fannkuch 43 43 37 37.2 14.0 13.5 ++ access-nbody 3 3 3 3.2 0 -6.7 access-nsieve 27 27.8 12 12 55.6 56.8 ++ bitops-3bit-bits-in-byte 4 4 4 4 0 0 bitops-bits-in-byte 8 8.2 8 8.4 0 -2.4 bitops-bitwise-and 2 2.2 2 2 0 9.1 bitops-nsieve-bits 13 13 12 12.8 7.7 1.5 ++ controlflow-recursive 3 3 3 3 0 0 crypto-aes 21 21.2 21 21.2 0 0 crypto-md5 3 3.2 3 3.6 0 -12.5 crypto-sha1 2 2.6 3 3 -50 -15.4 -- math-cordic 12 12.2 11 11.8 8.3 3.3 + math-partial-sums 10 10 10 10 0 0 math-spectral-norm 6 6 6 6 0 0 regexp-dna 408 408.2 458 458 -12.3 -12.2 -- s3d-cube 14 14.8 14 14 0 5.4 s3d-morph 36 36.6 36 36.4 0 0.5 s3d-raytrace 8 8 8 8 0 0 string-fasta 25 25.4 25 25.2 0 0.8 string-unpack-code 2849 2860.4 2905 2920.6 -2.0 -2.1 - string-validate-input 24 24.2 24 24 0 0.8 Metric: v8 custom v8 normalized metric (hardcoded in the test) Dir: v8/ crypto 749 748.4 755 754.2 0.8 0.8 + deltablue 1966 1961.4 2005 2000.6 2.0 2.0 + earley-boyer 1041 1034.6 1043 1038.4 0.2 0.4 raytrace 3251 3244.8 3248 3244.4 -0.1 -0.0 richards 1252 1251.2 1253 1249.8 0.1 -0.1 Dir: v8/typed/ crypto 801 800.8 798 797 -0.4 -0.5 - deltablue 3114 3099.8 3087 3076.8 -0.9 -0.7 - earley-boyer 1036 1028.8 1046 1038 1.0 0.9 raytrace 7260 7243 7317 7291 0.8 0.7 + richards 2299 2298.2 2306 2303.2 0.3 0.2 + Dir: v8.5/js/ crypto 639 638.6 640 639.6 0.2 0.2 + deltablue 315 314.6 311 311 -1.3 -1.1 - earley-boyer 1068 1060.8 1076 1072.6 0.7 1.1 raytrace 742 740.4 743 743 0.1 0.4 regexp 132 132 122 122 -7.6 -7.6 -- richards 286 286 285 284.6 -0.3 -0.5 - splay 1018 866.2 1029 901.6 1.1 4.1 Dir: v8.5/optimized/ crypto 4251 4247.8 4195 4189.6 -1.3 -1.4 - deltablue 3503 3475.6 3648 3640.2 4.1 4.7 + earley-boyer 1062 1055.4 1079 1072.6 1.6 1.6 + raytrace 8843 8809.6 8733 8707 -1.2 -1.2 - regexp 133 133 123 123 -7.5 -7.5 -- richards 4361 4357.4 4339 4336 -0.5 -0.5 - splay 6177 6158.2 6256 6251.2 1.3 1.5 + Dir: v8.5/typed/ crypto 2991 2984.4 2591 2572 -13.4 -13.8 -- deltablue 3719 3690.4 3701 3663.4 -0.5 -0.7 earley-boyer 1069 1063 1077 1072 0.7 0.8 raytrace 8817 8799 8707 8696.4 -1.2 -1.2 - regexp 133 133 123 123 -7.5 -7.5 -- richards 4355 4355 4334 4331.6 -0.5 -0.5 - splay 1351 1349 1304 1301.6 -3.5 -3.5 - Dir: v8.5/untyped/ crypto 725 724.8 732 731.6 1.0 0.9 + deltablue 2039 2032.4 2062 2053.4 1.1 1.0 + earley-boyer 1068 1061.6 1070 1065.2 0.2 0.3 raytrace 3544 3531.2 3544 3534.2 0 0.1 regexp 132 132 122 122 -7.6 -7.6 -- richards 484 482.6 488 486.2 0.8 0.7 + splay 1280 1277 1250 1240 -2.3 -2.9 -

99% of all Arrays are plain-vanilla arrays, but in a many contexts, we have to act as though the "length" getter/setter might be overridden (as in a subclass); this adds a couple of function pointers to allow us to take a much quicker path for the common case. Adding this optimization to the previous ones puts us at ahead-or-even (within noise) almost pretty much across the board (timings to follow), so I think this is pretty close to reviewable.

Timings with latest patch sequence. (Note that there are definite red herrings here; e.g., globalvar-write-1 claims to be +30% but the inner loop has nothing to do with Array; it's one that we have noticed inexplicable timing deltas on in the past. Ditto the various regexp regressions, which don't seem to have any significant time in Array code before or after. That said, overall trend is definite positive.) avm avm2 test best avg best avg %dBst %dAvg Metric: iterations/second Dir: asmicro/ alloc-1 37 37 38 38 2.7 2.7 + alloc-10 13 13 14 13.3 7.7 2.6 ++ alloc-11 12 12 12 12 0 0 alloc-12 8 8 12 12 50 50 ++ alloc-13 83 80.7 84 82.7 1.2 2.5 alloc-14 69 68 68 66.7 -1.4 -2.0 alloc-2 19 19 19 19 0 0 alloc-3 15 15 16 15.3 6.7 2.2 ++ alloc-4 41 40.7 43 42.3 4.9 4.1 + alloc-5 28 27.7 30 30 7.1 8.4 ++ alloc-6 66 65.7 66 65.3 0 -0.5 alloc-7 38 37.3 37 37 -2.6 -0.9 - alloc-8 14 14 14 14 0 0 alloc-9 15 14.3 15 14.7 0 2.3 arguments-1 674 672 671 669 -0.4 -0.4 arguments-2 369 367.3 368 367.3 -0.3 0 arguments-3 19 19 19 19 0 0 arguments-4 25 25 26 26 4 4 + array-1 1455 1451 1598 1595.3 9.8 9.9 ++ array-2 500 499 530 527.3 6 5.7 ++ array-pop-1 300 297.7 283 279.3 -5.7 -6.2 -- array-push-1 250 249 260 258.3 4 3.7 + array-shift-1 128 128 149 149 16.4 16.4 ++ array-slice-1 17 17 19 18.7 11.8 9.8 ++ array-sort-1 24 24 25 25 4.2 4.2 + array-sort-2 2 2 2 2 0 0 array-sort-3 19 19 19 19 0 0 array-sort-4 9 9 9 9 0 0 array-unshift-1 145 143.3 151 149.7 4.1 4.4 + closedvar-read-1 3633 3625.3 3602 3581.7 -0.9 -1.2 - closedvar-write-1 3910 3877.7 4464 4394.3 14.2 13.3 ++ closedvar-write-2 4228 4201.3 3668 3627 -13.2 -13.7 -- do-1 3589 3572.7 3560 3553.7 -0.8 -0.5 - for-1 3559 3546 3566 3536.7 0.2 -0.3 for-2 2186 2176.7 2189 2174.3 0.1 -0.1 for-3 2285 2279.3 2287 2271 0.1 -0.4 for-in-1 334 331.3 323 319.3 -3.3 -3.6 - for-in-2 137 135.7 307 304.7 124.1 124.6 ++ funcall-1 319 318 319 318.3 0 0.1 funcall-2 209 208.7 210 210 0.5 0.6 + funcall-3 294 294 295 293 0.3 -0.3 funcall-4 102 101 103 102.3 1.0 1.3 globalvar-read-1 3596 3563.7 3585 3557.7 -0.3 -0.2 globalvar-write-1 3584 3576.3 4844 4680.3 35.2 30.9 ++ isNaN-1 3403 3396.7 3414 3404.3 0.3 0.2 isNaN-2 3584 3576.3 3591 3575 0.2 -0.0 isNaN-3 3156 3151 3165 3160.3 0.3 0.3 + lookup-array-fetch-1 676 675.7 667 666.3 -1.3 -1.4 - lookup-array-in-1 1485 1482.3 1541 1539.3 3.8 3.8 + lookup-negindex-array-1 405 404.3 400 400 -1.2 -1.1 - lookup-negindex-array-2 347 343.3 341 339 -1.7 -1.3 lookup-negindex-object-1 413 407.3 414 412.3 0.2 1.2 lookup-negindex-object-2 382 380.7 385 380.3 0.8 -0.1 lookup-object-fetch-1 745 735.7 722 721.3 -3.1 -1.9 - lookup-object-in-1 1325 1311 1346 1333.3 1.6 1.7 number-toString-1 5 5 5 5 0 0 number-toString-2 66 65.3 66 65.7 0 0.5 oop-1 3 3 3 3 0 0 parseFloat-1 67 67 66 66 -1.5 -1.5 - parseInt-1 158 158 159 159 0.6 0.6 regex-exec-1 61 60.7 62 61.3 1.6 1.1 + regex-exec-2 73 72.7 73 73 0 0.5 regex-exec-3 101 101 103 103 2.0 2.0 + regex-exec-4 278 276.3 285 284.3 2.5 2.9 + restarg-1 674 672.7 675 671 0.1 -0.2 restarg-2 396 394.3 396 394.3 0 0 restarg-3 34 34 34 33.3 0 -2.0 restarg-4 25 25 26 26 4 4 + string-casechange-1 22 22 23 23 4.5 4.5 + string-casechange-2 23 22.7 23 22.7 0 0 string-charAt-1 1288 1286 1286 1285.7 -0.2 -0.0 string-charAt-2 75 74.3 75 74.7 0 0.4 string-charCodeAt-1 1359 1356.7 1367 1359 0.6 0.2 string-charCodeAt-2 1281 1279.3 1283 1281.7 0.2 0.2 string-charCodeAt-3 1020 1012.3 1026 1023.7 0.6 1.1 string-charCodeAt-4 1728 1726.3 1728 1727.3 0 0.1 string-charCodeAt-5 862 861.7 863 856.7 0.1 -0.6 string-charCodeAt-6 1371 1369 1368 1366.7 -0.2 -0.2 - string-charCodeAt-7 1730 1729 1728 1721 -0.1 -0.5 string-fromCharCode-1 288 286.7 288 286 0 -0.2 string-fromCharCode-2 59 59 60 60 1.7 1.7 + string-indexOf-1 191 191 191 190.3 0 -0.3 string-indexOf-2 114 114 145 145 27.2 27.2 ++ string-indexOf-3 103 102.7 103 102 0 -0.6 string-lastIndexOf-1 461 460.3 461 460.7 0 0.1 string-lastIndexOf-2 145 145 143 143 -1.4 -1.4 - string-lastIndexOf-3 141.9 141.7 140.9 140.8 -0.7 -0.7 string-slice-1 111 111 111 111 0 0 string-split-1 8 8 8 8 0 0 string-split-2 8 8 8 8 0 0 string-substring-1 118 118 119 119 0.8 0.8 switch-1 700 699.3 700 699 0 -0.0 switch-2 98 97.7 98 98 0 0.3 switch-3 172 172 172 172 0 0 try-1 218 218 219 218.7 0.5 0.3 + try-2 14 14 14 14 0 0 try-3 45 45 46 46 2.2 2.2 + vector-push-1 39 39 37 37 -5.1 -5.1 -- while-1 3583 3581.7 3583 3565.7 0 -0.4 Metric: time Dir: jsbench/ Crypt 3827 3838.3 3772 3776 1.4 1.6 + Euler 7426 7443 7205 7230 3.0 2.9 + FFT 7018 7055 7136 7152.7 -1.7 -1.4 - HeapSort 3559 3575.3 3535 3539.7 0.7 1.0 LUFact 5882 5885 5813 5816 1.2 1.2 + Moldyn 10067 10086.7 9979 10029.3 0.9 0.6 + RayTracer 7172 7194.3 7199 7230.3 -0.4 -0.5 SOR 32557 32571.7 32486 32503.3 0.2 0.2 + Series 8871 8879.7 8954 8978 -0.9 -1.1 - SparseMatmult 10494 10543.7 10248 10299.3 2.3 2.3 + Dir: jsbench/typed/ Crypt 965 966.3 965 965.7 0 0.1 Euler 9070 9085.7 9033 9082 0.4 0.0 FFT 1578 1581.7 1601 1602 -1.5 -1.3 - HeapSort 1353 1366.7 1336 1337 1.3 2.2 LUFact 1326 1327.7 1319 1319.3 0.5 0.6 + Moldyn 4506 4511.7 4508 4518.7 -0.0 -0.2 RayTracer 1347 1347.7 1342 1342.7 0.4 0.4 + SOR 4450 4465.7 4442 4455 0.2 0.2 Series 7710 7751.3 7665 7720.7 0.6 0.4 SparseMatmult 2642 2645.3 2657 2658.3 -0.6 -0.5 - Metric: iterations/second Dir: jsmicro/ alloc-1 37 36.7 37 37 0 0.9 alloc-10 13 13 13 13 0 0 alloc-11 11 11 11 11 0 0 alloc-12 7 7 11 11 57.1 57.1 ++ alloc-13 72 71.3 72 71 0 -0.5 alloc-14 61 60.7 62 61.7 1.6 1.6 + alloc-2 19 19 19 19 0 0 alloc-3 15 15 15 15 0 0 alloc-4 41 41 41 41 0 0 alloc-5 30 30 29 29 -3.3 -3.3 - alloc-6 58 58 58 58 0 0 alloc-7 37 37 37 37 0 0 alloc-8 14 14 14 14 0 0 alloc-9 14 14 14 14 0 0 arguments-1 146 145.7 145 144 -0.7 -1.1 arguments-2 104 104 104 104 0 0 arguments-3 17 17 17 17 0 0 array-1 371 371 378 377.3 1.9 1.7 + array-2 282 282 259 259 -8.2 -8.2 -- array-pop-1 61 60.3 64 64 4.9 6.1 + array-push-1 41 41 42 42 2.4 2.4 + array-shift-1 52 52 58 57.7 11.5 10.9 ++ array-slice-1 14 14 16 15.7 14.3 11.9 ++ array-sort-1 23 23 24 24 4.3 4.3 + array-sort-2 2 2 2 2 0 0 array-sort-3 19 19 19 19 0 0 array-sort-4 9 9 9 9 0 0 array-unshift-1 20 20 21 21 5 5 + closedvar-read-1 609 608 611 610 0.3 0.3 + closedvar-write-1 441 441 441 440 0 -0.2 closedvar-write-2 430 425 438 435 1.9 2.4 do-1 607 606 610 608.7 0.5 0.4 + for-1 604 603.3 601 600.3 -0.5 -0.5 - for-2 184 184 185 184.7 0.5 0.4 + for-3 135 134.7 134 134 -0.7 -0.5 - for-in-1 265 264.7 257 256.3 -3.0 -3.1 - for-in-2 128 127.7 250 249.7 95.3 95.6 ++ funcall-1 208 207.3 207 207 -0.5 -0.2 - funcall-2 203 203 202 202 -0.5 -0.5 funcall-3 200 200 199 198.3 -0.5 -0.8 - funcall-4 1324 1319 1328 1313 0.3 -0.5 globalvar-read-1 600 598.7 599 598.3 -0.2 -0.1 globalvar-write-1 433 431.7 429 428.7 -0.9 -0.7 - isNaN-1 516 515.3 519 517.7 0.6 0.5 + lookup-array-fetch-1 592 591.3 586 585.3 -1.0 -1.0 - lookup-array-in-1 980 979 984 983.3 0.4 0.4 + lookup-object-fetch-1 648 647.7 643 642.7 -0.8 -0.8 - lookup-object-in-1 879 877.3 884 883.3 0.6 0.7 + number-toString-1 5 5 5 5 0 0 number-toString-2 55 54 55 55 0 1.9 oop-1 3 3 3 3 0 0 parseFloat-1 50 50 50 49.3 0 -1.3 parseInt-1 111 111 114 114 2.7 2.7 + regex-exec-1 54 53.7 53 52.7 -1.9 -1.9 - regex-exec-2 64 64 65 65 1.6 1.6 + regex-exec-3 92 92 93 93 1.1 1.1 + regex-exec-4 235 234 234 233.3 -0.4 -0.3 string-casechange-1 15 15 15 15 0 0 string-casechange-2 15 15 15 15 0 0 string-charAt-1 97 96 97 96.3 0 0.3 string-charAt-2 42 42 43 43 2.4 2.4 + string-charCodeAt-1 87 87 88 87.3 1.1 0.4 + string-charCodeAt-2 88 88 88 88 0 0 string-fromCharCode-1 73 73 73 73 0 0 string-fromCharCode-2 37 36.3 37 37 0 1.8 string-fromCharCode-3 65 65 65 65 0 0 string-fromCharCode-4 69 69 69 69 0 0 string-indexOf-1 78 78 78 78 0 0 string-indexOf-2 51 51 56 56 9.8 9.8 ++ string-indexOf-3 48 47.7 49 49 2.1 2.8 + string-lastIndexOf-1 78 78 78 78 0 0 string-lastIndexOf-2 56 56 57 56.3 1.8 0.6 + string-lastIndexOf-3 73 72.3 74 73.7 1.4 1.8 + string-slice-1 47 47 46 46 -2.1 -2.1 - string-split-1 7 7 7 7 0 0 string-split-2 7 7 7 7 0 0 string-substring-1 50 50 49 49 -2 -2 - switch-1 109 109 108 108 -0.9 -0.9 switch-2 55 54.3 55 55 0 1.2 switch-3 69 68 70 69.7 1.4 2.5 try-1 143 143 145 145 1.4 1.4 + try-2 13 13 13 13 0 0 try-3 34 34 35 35 2.9 2.9 + while-1 605 604 602 600.7 -0.5 -0.6 - Metric: time Dir: language/describetype/ desctypeperf 599 599.3 600 602.3 -0.2 -0.5 Dir: language/e4x/ addingToXMLList 19 19 18 18.3 5.3 3.5 ++ appendChildAndString 53 53.7 53 53.7 0 0 concatenatingStringsFromE4X 7 7.3 7 7 0 4.5 simpleStringConcatenation 1 1.3 1 1.3 0 0 usingAppendChildAndE4X 55 56 54 54.3 1.8 3.0 Dir: language/string/ append_concat 81 82 81 81 0 1.2 append_equal_plus 67 67.7 66 66.7 1.5 1.5 append_plus_equal 67 68 66 66.3 1.5 2.5 charAt 171 173.7 172 172.3 -0.6 0.8 charCodeAt 204 204 204 204.7 0 -0.3 indexOf 303 303.3 298 298.3 1.7 1.6 + lastIndexOf 208 208 208 208.7 0 -0.3 replace 520 523.3 518 518.7 0.4 0.9 replace2 823 823.7 870 870.3 -5.7 -5.7 -- search 36 36.3 35 35.7 2.8 1.8 + slice 278 278 276 277 0.7 0.4 + split 338 346 339 341.7 -0.3 1.3 static_ascii_array_100 978 979.3 253 253.7 74.1 74.1 ++ static_ascii_array_50 965 966.7 235 235 75.6 75.7 ++ static_latin1_array_100 1920 1924 478 480 75.1 75.1 ++ static_latin1_array_50 963 963.7 239 239 75.2 75.2 ++ substr 204 204.3 204 204.3 0 0 substring 194 195 194 194 0 0.5 Dir: language/string/typed/ append_concat 73 73.7 75 75.3 -2.7 -2.3 - append_equal_plus 61 62 61 61.7 0 0.5 append_plus_equal 61 61.7 61 61.3 0 0.5 charAt 11 11.7 11 11.7 0 0 charCodeAt 12 12 10 11 16.7 8.3 ++ indexOf 303 303 297 297.7 2.0 1.8 + lastIndexOf 208 208.3 209 209 -0.5 -0.3 - replace 522 525.3 518 518.7 0.8 1.3 + replace2 825 827 870 872 -5.5 -5.4 -- search 36 36 35 35.7 2.8 0.9 + slice 181 182 182 182.3 -0.6 -0.2 split 334 335.3 340 340.3 -1.8 -1.5 - substr 144 144.3 143 143.7 0.7 0.5 + substring 126 126.7 127 127 -0.8 -0.3 - Dir: misc/ boids 1881 1883.7 1810 1810.3 3.8 3.9 + boidshack 500 501.3 443 445.3 11.4 11.2 ++ gameoflife 2474 2479 1816 1826.7 26.6 26.3 ++ primes 5149 5155.3 5148 5155.7 0.0 -0.0 Dir: mmgc/ gcbench 2917 2924.3 2912 2922.3 0.2 0.1 ofib-rc 270 275 269 270 0.4 1.8 ofib 1197 1218.7 1210 1217 -1.1 0.1 sfib 375 376 376 377 -0.3 -0.3 Dir: scimark/ FFT 2288 2292.7 2309 2309.3 -0.9 -0.7 - LU 3066 3068 3032 3036.7 1.1 1.0 + MonteCarlo 2746 2749.7 2722 2733 0.9 0.6 + SOR 2528 2533 2558 2563 -1.2 -1.2 - SparseCompRow 102 102.3 99 99.7 2.9 2.6 + Dir: sunspider/ access-binary-trees 33 33.3 34 34 -3.0 -2.0 - access-fannkuch 86 86.3 82 82.3 4.7 4.6 + access-nbody 84 84.3 83 83.7 1.2 0.8 + access-nsieve 44 45 26 26 40.9 42.2 ++ bitops-3bit-bits-in-byte 10 10.3 10 10.3 0 0 bitops-bits-in-byte 29 29 30 30 -3.4 -3.4 - bitops-bitwise-and 212 212.3 212 212 0 0.2 bitops-nsieve-bits 42 42.3 41 41 2.4 3.1 + controlflow-recursive 15 15.7 15 15.7 0 0 crypto-aes 42 42 42 42 0 0 crypto-md5 19 19.7 20 20 -5.3 -1.7 -- crypto-sha1 20 20 19 19.7 5 1.7 + date-format-tofte 179 181.3 179 180.7 0 0.4 math-cordic 54 54.3 54 54.3 0 0 math-partial-sums 185 185.3 185 185 0 0.2 math-spectral-norm 28 28.3 28 28 0 1.2 s3d-cube 70 70.7 69 69 1.4 2.4 + s3d-morph 41 41 41 41.3 0 -0.8 s3d-raytrace 78 78.3 77 77.7 1.3 0.9 + string-fasta 81 82 82 82 -1.2 0 - string-unpack-code 173 174.7 173 173.7 0 0.6 string-validate-input 43 43.3 43 43.7 0 -0.8 Dir: sunspider/as3/ access-binary-trees 9 9.3 9 9.3 0 0 access-fannkuch 55 55.7 50 50.7 9.1 9.0 ++ access-nbody 6 6.3 6 6.3 0 0 access-nsieve 33 33 14 14.3 57.6 56.6 ++ bitops-3bit-bits-in-byte 6 6 5 5.7 16.7 5.6 ++ bitops-bits-in-byte 9 9.3 9 9.3 0 0 bitops-bitwise-and 2 2 2 2 0 0 bitops-nsieve-bits 27 27.3 27 27 0 1.2 controlflow-recursive 4 4.3 4 4 0 7.7 crypto-aes 31 31 32 32 -3.2 -3.2 - crypto-md5 24 24 23 24 4.2 0 + crypto-sha1 19 19 18 18.3 5.3 3.5 ++ date-format-tofte 172 172.7 171 173.7 0.6 -0.6 math-cordic 14 14 13 13.3 7.1 4.8 ++ math-partial-sums 64 64.7 65 65 -1.6 -0.5 - math-spectral-norm 5 5.3 5 5.7 0 -6.3 s3d-cube 21 21.7 20 21 4.8 3.1 s3d-morph 29 29.3 29 30 0 -2.3 s3d-raytrace 33 33 33 33.3 0 -1.0 string-fasta 40 40 40 40 0 0 string-unpack-code 172 172 172 172.7 0 -0.4 string-validate-input 33 33 33 33.3 0 -1.0 Dir: sunspider/as3vector/ access-fannkuch 22 22 22 22.3 0 -1.5 access-nbody 6 6.3 6 6.7 0 -5.3 access-nsieve 12 12 12 12 0 0 bitops-nsieve-bits 8 8 8 8 0 0 math-cordic 14 14 14 14.3 0 -2.4 math-spectral-norm 16 16.7 16 16.7 0 0 s3d-cube 16 16.7 16 16.3 0 2.0 s3d-morph 16 16.7 16 16 0 4.0 string-fasta 42 42.7 42 42.3 0 0.8 string-validate-input 35 35.3 35 35.3 0 0 Dir: sunspider-0.9.1/js/ access-binary-trees 32 32.7 33 33.3 -3.1 -2.0 - access-fannkuch 85 85.3 81 81 4.7 5.1 + access-nbody 78 79 78 78.7 0 0.4 access-nsieve 42 42.3 25 25 40.5 40.9 ++ bitops-3bit-bits-in-byte 9 9 8 8.7 11.1 3.7 ++ bitops-bits-in-byte 29 29.7 29 29.3 0 1.1 bitops-bitwise-and 210 210.7 212 212 -1.0 -0.6 - bitops-nsieve-bits 40 41 40 40.3 0 1.6 controlflow-recursive 14 14.7 15 15 -7.1 -2.3 -- crypto-aes 35 35.3 35 35.3 0 0 crypto-md5 16 16.3 17 17 -6.2 -4.1 -- crypto-sha1 18 18 18 18 0 0 math-cordic 52 52.3 52 52 0 0.6 math-partial-sums 187 187.7 188 188 -0.5 -0.2 - math-spectral-norm 27 27.7 27 27.7 0 0 regexp-dna 829 839.3 858 874.3 -3.5 -4.2 - s3d-cube 63 64.7 62 63 1.6 2.6 s3d-morph 58 60.3 58 58.7 0 2.8 s3d-raytrace 71 71.3 71 71.7 0 -0.5 string-fasta 57 57 57 57.3 0 -0.6 string-unpack-code 4456 4480.7 4353 4442.3 2.3 0.9 + string-validate-input 42 42 42 42 0 0 Dir: sunspider-0.9.1/typed/ access-binary-trees 8 8 8 8 0 0 access-fannkuch 57 57.7 53 53 7.0 8.1 ++ access-nbody 4 4 4 4 0 0 access-nsieve 30 31.3 13 13 56.7 58.5 ++ bitops-3bit-bits-in-byte 5 5 4 4.3 20 13.3 + bitops-bits-in-byte 9 9 9 9 0 0 bitops-bitwise-and 2 2 2 2 0 0 bitops-nsieve-bits 27 27.3 26 26.7 3.7 2.4 + controlflow-recursive 3 3.3 4 4 -33.3 -20.0 -- crypto-aes 26 26.7 26 26.7 0 0 crypto-md5 4 4 4 4.3 0 -8.3 crypto-sha1 5 5 5 5 0 0 math-cordic 12 12.7 12 12.7 0 0 math-partial-sums 11 11 11 11.3 0 -3.0 math-spectral-norm 8 8 7 7.7 12.5 4.2 ++ regexp-dna 798 799.7 826 844.7 -3.5 -5.6 - s3d-cube 18 18.3 17 17 5.6 7.3 ++ s3d-morph 47 48 49 49 -4.3 -2.1 - s3d-raytrace 8 8.7 9 9.3 -12.5 -7.7 - string-fasta 31 32 31 31.7 0 1.0 string-unpack-code 4412 4483 4469 4491 -1.3 -0.2 string-validate-input 26 27 27 27 -3.8 0 - Metric: v8 custom v8 normalized metric (hardcoded in the test) Dir: v8/ crypto 448 447.3 446 445.3 -0.4 -0.4 - deltablue 1550 1549.3 1541 1541 -0.6 -0.5 - earley-boyer 1008 1006.3 1008 1004 0 -0.2 raytrace 2874 2866.3 2888 2882.7 0.5 0.6 + richards 1014 1010.3 990 988 -2.4 -2.2 - Dir: v8/typed/ crypto 462 460 458 457 -0.9 -0.7 - deltablue 2420 2417.7 2422 2420.7 0.1 0.1 earley-boyer 1015 1010 1009 1007 -0.6 -0.3 raytrace 6298 6291.7 6304 6291.7 0.1 0 richards 1912 1910 1865 1863.7 -2.5 -2.4 - Dir: v8.5/js/ crypto 398 398 397 396.3 -0.3 -0.4 - deltablue 299 298.3 299 299 0 0.2 earley-boyer 1009 998 1007 1003 -0.2 0.5 raytrace 639 638.7 647 644.7 1.3 0.9 + regexp 86.5 86.3 85.1 84.9 -1.6 -1.5 - richards 244 243.7 241 240 -1.2 -1.5 - splay 732 689.3 737 696 0.7 1.0 Dir: v8.5/optimized/ crypto 3551 3547 3521 3498 -0.8 -1.4 deltablue 3019 3018.3 3162 3162 4.7 4.8 + earley-boyer 1006 1002.3 999 997.7 -0.7 -0.5 - raytrace 7510 7507.3 7618 7607.7 1.4 1.3 + regexp 86.9 86.7 85.6 85.4 -1.5 -1.5 - richards 3707 3702 3623 3620.3 -2.3 -2.2 - splay 5564 5487.7 5293 5288.3 -4.9 -3.6 - Dir: v8.5/typed/ crypto 2736 2734 2740 2724.3 0.1 -0.4 deltablue 3262 3255.3 3236 3233.7 -0.8 -0.7 - earley-boyer 1007 1004 997 994.3 -1.0 -1.0 - raytrace 7540 7530 7602 7595 0.8 0.9 + regexp 86.8 86.7 85.6 85.4 -1.4 -1.5 - richards 3715 3706 3630 3625.3 -2.3 -2.2 - splay 931 930.7 937 931.7 0.6 0.1 Dir: v8.5/untyped/ crypto 436 434 434 433 -0.5 -0.2 deltablue 1597 1592.7 1598 1597.7 0.1 0.3 earley-boyer 1012 1009.3 1001 1000 -1.1 -0.9 - raytrace 3078 3072 3091 3089 0.4 0.6 + regexp 87.8 87.7 85.9 85.8 -2.2 -2.1 - richards 411 409.7 411 410.7 0 0.2 splay 888 872 887 878.3 -0.1 0.7

nice work, an idea late to the party and maybe you've already thought of it , but what about waiting until the first 'set' to decide whether or not its sparse? The justification being that a 'low' valued int index might be a good indicator for keeping it sparse. On the other hand seeing an object being used as an index might bias your decision. Likewise, with this technique you could introduce thresholds for transitioning out of sparse. I.e. integer based access might have a higher threshold , whereas string based could be lower.

If I understand the patches correctly, the design is "dense until proven otherwise". So an empty array is dense, and a 1-element array is dense, and you can't be sparse until you have 2 elements that are far enough away from each other. Rick: Given this, are you proposing that we add new code to support sparse->dense conversion? Because as things stand, the only transition is dense->sparse, and that can't happen until after 2 sets. (Steven: jump in and correct me any time).

Comment on attachment 489711 [details] [diff] [review] Faster prototype cool. I didn't quite grok why the upper limit of dense values needed to be 0x7fffffff. It's probably fine, but worth 10 minutes to see if we can think of a different clever encoding that would allow the full range. Would it be enough to just have the [start,end) range be empty?

Comment on attachment 489851 [details] [diff] [review] Add dense-array specialization for unshift looks straightforward. i didn't review the new code for correctness though.

Comment on attachment 489959 [details] [diff] [review] Add special-casing for length accessors > + /*static*/ uint32_t ArrayObject::LengthGetter_Undef(ArrayObject* self) ... > + typedef uint32_t (*LengthGetter)(ArrayObject*); Ow, my eyes! my eyes! InitialCapsAnd_Underscores? phulleeese. The blob of code that checks to see if length() is overridden should be factored as an API on Traits, probably. It sure doesn't belong in the Array object. In all this seems like a heavy and intrusive optimization, for what it's providing. Could you isolate the the tests it helps and explain better why we'd want the complexity? (virtual method which call a function pointer, etc)

(In reply to comment #18) > If I understand the patches correctly, the design is "dense until proven > otherwise". That's correct. > So an empty array is dense, and a 1-element array is dense, and > you can't be sparse until you have 2 elements that are far enough away from > each other. Yep; also there's a min-size threshold (currently 32) under which we assume dense even if there's "too few" items. > Rick: Given this, are you proposing that we add new code to support > sparse->dense conversion? Because as things stand, the only transition is > dense->sparse, and that can't happen until after 2 sets. (Steven: jump in and > correct me any time). Correct. In reply to comment #19) > cool. I didn't quite grok why the upper limit of dense values needed to be > 0x7fffffff. It's probably fine, but worth 10 minutes to see if we can think of > a different clever encoding that would allow the full range. Would it be > enough to just have the [start,end) range be empty? Main reason is that allows us to use m_denseStart as a flag for "dense/sparse/sealed"; by keeping indices <= 0x7fffffff we can ensure that "index - densestart < denselength" is a valid test for a dense index. (In reply to comment #21) > Ow, my eyes! my eyes! InitialCapsAnd_Underscores? phulleeese. Guilty as charged. > The blob of code that checks to see if length() is overridden should be > factored as an API on Traits, probably. It sure doesn't belong in the Array > object. Fair enough. > In all this seems like a heavy and intrusive optimization, for what it's > providing. Could you isolate the the tests it helps and explain better why > we'd want the complexity? (virtual method which call a function pointer, etc) Will do.

(In reply to comment #18) > Because as things stand, the only transition is dense->sparse, and that can't happen until after 2 sets. Perfect, thanks for clarifying and a threshold limit of 32 sounds more than generous.

(In reply to comment #22) > > In all this seems like a heavy and intrusive optimization, for what it's > > providing. Could you isolate the the tests it helps and explain better why > > we'd want the complexity? (virtual method which call a function pointer, etc) After looking a bit further, I think most of the improvement I was seeing is of the inexplicable-voodoo variety. I'm going to withdraw this part of the patch.

FWIW, I think the remaining slowdown in a few cases is actually due to the fastpath in setuintproperty having to read the existing slot (to see if it had a hole and thus needs to adjust m_denseUsed). I'll ponder if there's any way to finess that to avoid it for the common case (which tends to be 100% dense).

If the arrays in the slowed-down cases are 100% dense, then a) that's surprising since dense arrays should be faster than sparse ones, and b) the simple experiment for just those tests is to only support 100% denseness, and skip the "hole" check.

Comment on attachment 489959 [details] [diff] [review] Add special-casing for length accessors Withdrawing, dubious usefulness

(In reply to comment #26) > If the arrays in the slowed-down cases are 100% dense, then a) that's > surprising since dense arrays should be faster than sparse ones They are; the issue is that the "old" dense array code didn't have to check for holes in the dense array, while the new one does. In a few situations (mainly synthetic benchmarks that do nothing but call setuintproperty 10000000 times) this adds up.

Mostly small effects, but add up for the microbenchmarks: -- in _setUintProperty, don't check for DENSE_ARRAY_HOLE unless denseUsed < denseLen -- add a List::replace() function; this is like List::set() but doesn't check for list lengthening. (Useful when we know the index is valid, avoid redundant runtime checks.) -- remove runtime "isEmpty" checks from List::removeFirst and List::removeLast (we already were asserting that the list wasn't empty, so the check was redundant) -- Specialize List::removeLast rather than calling removeAt(len-1). -- minor optimizing of code in List::removeAt.

Results from latest patch series (ignoring noise): Again, some of these are sensible but some are inexplicable... eg, several regressions in regexp-related tests, none of which have hotspots anywhere near array code. Even weirder is isNaN-3, which is a loop that calls isNaN a jillion times... hard to figure why array changes could affect is plus or minus. avm avm2 test best avg best avg %dBst %dAvg Metric: iterations/second Dir: asmicro/ alloc-12 8 8 12 12 50 50 ++ array-1 1472 1465 1606 1604.7 9.1 9.5 ++ array-shift-1 128 128 151 150.7 18.0 17.7 ++ for-in-2 139 139 308 307.7 121.6 121.3 ++ globalvar-write-1 4619 4495.7 4900 4874.3 6.1 8.4 + isNaN-1 3508 3467.7 3374 3332.7 -3.8 -3.9 - isNaN-3 3535 3510 3184 3177.7 -9.9 -9.5 -- parseFloat-1 64 62.7 65 65 1.6 3.7 + regex-exec-1 59 58.7 60 60 1.7 2.3 + regex-exec-3 102 101.7 104 103.7 2.0 2.0 + restarg-3 33 33 34 34 3.0 3.0 + restarg-4 26 25.7 26 26 0 1.3 string-casechange-1 23 22.7 22 21.3 -4.3 -5.9 - string-indexOf-1 199 198.7 190 189.7 -4.5 -4.5 - string-indexOf-2 145 144.7 114 114 -21.4 -21.2 -- try-1 212 211 217 216.7 2.4 2.7 + try-2 14 13.3 14 14 0 5.0 try-3 42 42 45 45 7.1 7.1 ++ Metric: time Dir: jsbench/ Euler 7485 7524.3 7240 7244.7 3.3 3.7 + FFT 7193 7206.3 7036 7085 2.2 1.7 + LUFact 5959 6006.3 5812 5852.7 2.5 2.6 + Moldyn 10215 10305.3 10074 10149.7 1.4 1.5 Dir: jsbench/typed/ FFT 1626 1638.3 1574 1599.3 3.2 2.4 + Metric: iterations/second Dir: jsmicro/ alloc-12 7 7 11 11 57.1 57.1 ++ alloc-2 18 18 19 19 5.6 5.6 ++ alloc-3 15 15 16 16 6.7 6.7 ++ array-2 277.7 277.0 263.7 263.6 -5.0 -4.8 -- array-pop-1 61.2 61.1 63.7 63.5 4.0 4.0 + array-push-1 41.5 41.4 42.3 42.2 2.0 2.0 + array-shift-1 52.9 52.8 57.2 57.1 8.0 8.1 ++ array-slice-1 14.8 14.8 14.3 14.3 -3.5 -3.6 - array-sort-4 9.8 9.7 9.9 9.9 1.6 2.3 + array-unshift-1 21.6 21.5 16.0 16.0 -25.9 -25.7 -- for-in-2 129 128.7 250 249.3 93.8 93.8 ++ globalvar-write-1 427 426.3 435 434 1.9 1.8 + lookup-array-fetch-1 586 586 593 593 1.2 1.2 + string-indexOf-2 57 57 52 52 -8.8 -8.8 -- string-lastIndexOf-1 75 74.7 78 77.7 4 4.0 + Metric: time Dir: language/string/ replace2 823 824 872 873 -6.0 -5.9 -- static_ascii_array_100 969 970 261 262 73.1 73.0 ++ static_ascii_array_50 956 958 243 246.3 74.6 74.3 ++ static_latin1_array_100 1908 1908 492 494.7 74.2 74.1 ++ static_latin1_array_50 955 956 247 247 74.1 74.2 ++ Dir: language/string/typed/ append_plus_equal 63 63 61 61.7 3.2 2.1 + replace2 827 828 874 875.3 -5.7 -5.7 -- Dir: misc/ boids 1878 1879.7 1852 1853.3 1.4 1.4 + boidshack 500 502 454 454.7 9.2 9.4 ++ gameoflife 2463 2465.3 1823 1823.3 26.0 26.0 ++ Dir: mmgc/ ofib 1238 1257 1182 1201 4.5 4.5 + Dir: scimark/ SparseCompRow 109 109 100 100 8.3 8.3 ++ Dir: sunspider/ access-fannkuch 86 86.3 83 83.3 3.5 3.5 + access-nsieve 44 44.3 27 27 38.6 39.1 ++ Dir: sunspider/as3/ access-fannkuch 55 55 49 49 10.9 10.9 ++ access-nsieve 33 33 14 14 57.6 57.6 ++ Dir: sunspider-0.9.1/js/ access-nsieve 43 43 25 25.7 41.9 40.3 ++ regexp-dna 841 843 884 885.7 -5.1 -5.1 -- Dir: sunspider-0.9.1/typed/ access-fannkuch 57 57 52 52.3 8.8 8.2 ++ access-nsieve 30 30.3 12 12.3 60 59.3 ++ regexp-dna 793 795 826 827.7 -4.2 -4.1 - Metric: v8 custom v8 normalized metric (hardcoded in the test) Dir: v8/typed/ richards 1929 1927.3 1838 1836 -4.7 -4.7 - Dir: v8.5/js/ regexp 88.1 87.9 84.7 84.6 -3.9 -3.8 - Dir: v8.5/optimized/ deltablue 3031 3024.3 3171 3156.3 4.6 4.4 + regexp 89 88.8 84.9 84.7 -4.6 -4.6 - Dir: v8.5/typed/ regexp 88.9 88.5 85 84.6 -4.4 -4.4 - Dir: v8.5/untyped/ regexp 88.6 88.4 84.9 84.7 -4.2 -4.1 -

Comment on attachment 491383 [details] [diff] [review] A few more optimizations I would consider making the replace() methods private and friending the trusted callers of it (Array, maybe Vector, nothing else).

(In reply to comment #31) > I would consider making the replace() methods private and friending the trusted > callers of it (Array, maybe Vector, nothing else). Seems reasonable, I'll do that. This entire patch queue is pretty close to landing-worthy IMHO, but (aside from the lack of formal R+) Lars has asked me to hold off pending some exactGC work; I'll wait for his say-so before landing, but in the meantime I'm going to ask for the "F+" to be upgraded to R+...

Comment on attachment 488618 [details] [diff] [review] Refactor "generic" array methods in ArrayClass to remove friend access Documentation on ScriptObject::toArrayObject is incorrect, it says it returns "true if this is an ArrayObject" but it returns a pointer. (Also "return" should be "returns".) The generic implementation of ScriptObject::array_concat appears incorrect, because it transfers 'undefined' values where the source array has a hole (missing property). Note that E262-3 says that the 'this' object of array_concat need not be an Array object. Same goes for array_reverse and array_slice. (I stopped looking at this point; I have not reviewed the rest of ScriptObject.cpp, and pretty much none of the changes in ArrayClass.cpp.) I have a hard time seeing the utility of the refactoring. The behavior of the generic methods is invariant with respect to the type of the 'this' object. The implementation varies because Array and Vector have other representations, but Vector already has its own implementations for the most part; it would have seemed more surgical to me to simply make the generic_X methods deal with the two different implementations on two different code paths. Minor issues found while reviewing: Documentation on ScriptObject::getLengthProperty pointlessly brings up ArrayObject, when it should just state that the method reads the "length" property of any object. However, the API does not make sense for that since the "length" property may be undefined on some objects, so something needs to be said about that. About setLengthProperty, it's not clear what happens if "length" is a typed property that does not admit uint32_t - is the value coerced or is it an error? Documentation for array methods in ScriptObject.h needs to be capitalized.

(In reply to comment #33) > I have a hard time seeing the utility of the refactoring. The behavior of the > generic methods is invariant with respect to the type of the 'this' object. > The implementation varies because Array and Vector have other representations, > but Vector already has its own implementations for the most part; it would have > seemed more surgical to me to simply make the generic_X methods deal with the > two different implementations on two different code paths. This was definitely scope creep, but the initial motivation for this was that the ArrayClass implementations had special-casing for ArrayObjects in order to deal with "dense arrays", but this approach required a lot of special knowledge to be promulgated outside ArrayObject proper. Granted, ArrayClass is fairly friendly but nevertheless it seemed that such friendliness could be minimized further (which this patch accomplished). The defects and nits you point out can obviously be corrected, but I'd like guidance as to whether you oppose the refactoring in general or just question whether such a drastic change is necessary... no point in me expending effort on improving a patch if there's no chance of it being approved.

(In reply to comment #34) > > The defects and nits you point out can obviously be corrected, but I'd like > guidance as to whether you oppose the refactoring in general or just question > whether such a drastic change is necessary... no point in me expending effort > on improving a patch if there's no chance of it being approved. I don't oppose it as such but I definitely question the drastic change, as you say. I don't understand why the operations have to be virtual and I'm pretty sure I would not have done it that way - the array operations are specified one way for all data types, and need to have at most two implementation in the system at this time; I don't think subclasses ought to be able to override anything. (And of course I oppose what I believe is incorrect implementations of the generic Array operations in ScriptObject, but that's somewhat orthogonal maybe.)

(In reply to comment #35) > I don't oppose it as such but I definitely question the drastic change, as you > say. OK, I'll revisit this and see if a less drastic change can be made to co-exist.

Combined & revised to use a generally more surgical approach to the generics; otherwise mostly a rebased version of the previous patch series. BUT: cleanup this time revealed that generic_shift wasn't honoring spoofed length property properly; we had an acceptance test for it, but said test was wrong. (This will probably require a followup bug to unfix the old behavior in a versioned way; I'd prefer to land this with the "proper" behavior rather than mix in BugCompatibility changes.) (Note, no security risk here, as we never attempted to use the spoofed value.)

Comment on attachment 498475 [details] [diff] [review] Revised & Combined Patch removeAt() should have an assertion for !isEmpty(). (I know the callers do too but the invariant is actually important only in removeAt.) The use of the word "enable" in the comment on the definition of NO_DEBUG_ARRAY_VERIFY is very strange given the negative name. I think simply commenting out the line instead of prefixing the name with "NO_" would make the point much better, it's what we do everywhere else I think. Accident or deliberate? The definition of DENSE_ARRAY_HOLE is reasonable but it's not reasonable to have it here without additional documentation on the Atom documentation that undefined values cannot carry additional payload (which is what allows you to assume that here). Also it would be most excellent to add documentation there that you're using that facility here, so that we won't have to hunt for it when we redo the atom representations eventually, as we must. Of course the implementations of "generic" operations (eg reverse, concat) are completely wrong according to the ES spec - they do not preserve holes - and this is only made worse by the correct documentation being present in our source code. But it appears you make them no worse than they were, so what can I say? I'll file a bug if one is not filed already...

(In reply to comment #38) > Of course the implementations of "generic" operations (eg reverse, concat) are > completely wrong according to the ES spec - they do not preserve holes - and > this is only made worse by the correct documentation being present in our > source code. But it appears you make them no worse than they were, so what can > I say? I'll file a bug if one is not filed already... To be sure I understand: the generic operations are incorrect according to ES spec? As long as my changes preserve this incorrect behavior, then we're all good. (Fixing the incorrect behavior will have to be versioned, of course, and I'm not aware that there is a bug on this.)

(In reply to comment #39) > > To be sure I understand: the generic operations are incorrect according to ES > spec? At least some of them, but I have not looked at all. > As long as my changes preserve this incorrect behavior, then we're all > good. (Fixing the incorrect behavior will have to be versioned, of course, and > I'm not aware that there is a bug on this.) Yes. I filed bug #620651 to track the problem.

pushed with the relevant nits fixed: 5713:9c0aeb405ad3