searching for reviewer(edwsmith)
3454779d0f771b40632d2f0a49504fbdb78e59c9: Bug 618215: callstatic ensure caller matches receiver (p=rickr, r=edwsmith, r=lhansen).
Felix Klock II <fklockii@adobe.com> - Wed, 11 Apr 2012 03:42:51 -0700 - rev 7347
Push 4223 by fklockii@adobe.com at Tue, 24 Apr 2012 00:43:32 +0000
Bug 618215: callstatic ensure caller matches receiver (p=rickr, r=edwsmith, r=lhansen). {{mercurial e5e6f50e4b6cb912ede9f8b67e96353a58b03dc4}} CL@1052248
f1b33c6cc29254407565e871be586a507d40deca: Bug 702552, 740903: VM stack pointer metrics and overflow recovery hooks (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Thu, 29 Mar 2012 20:56:35 -0700 - rev 7330
Push 4215 by fklockii@adobe.com at Fri, 30 Mar 2012 19:08:26 +0000
Bug 702552, 740903: VM stack pointer metrics and overflow recovery hooks (r=edwsmith) The patch addresses the handling of the StackOverflowError in AS3, and in native code to the extent that we are relying on the AS3 checks. This is principally a fix for WE 2949577, with changes to the Flash Player avmglue code. The avmplus enhancements consist of hooks to support better recovery by the host, and instrumentation that is useful both for testing and for tuning of allocation margins. On the player side, the patch was reviewed by Jeff Mott. Edwin reviewed only the Tamarin changes. CL@1048913
74ab7a8d7b7d7a48d78d20409c0553102ae5eb2c: Bug 620403: Remove 'pure' attribute from helper functions that may invoke coercions with side-effects (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Tue, 06 Mar 2012 11:46:04 -0800 - rev 7256
Push 4188 by wmaddox@adobe.com at Tue, 06 Mar 2012 19:47:06 +0000
Bug 620403: Remove 'pure' attribute from helper functions that may invoke coercions with side-effects (r=edwsmith)
f6f3c2a7ca125b2a8d4ebb419bc3e0aa52f651d8: Bug 620424: Order side-effects in OP_callstatic consistently between JIT and interpreter (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Wed, 29 Feb 2012 11:31:34 -0800 - rev 7237
Push 4174 by wmaddox@adobe.com at Wed, 29 Feb 2012 19:32:00 +0000
Bug 620424: Order side-effects in OP_callstatic consistently between JIT and interpreter (r=edwsmith)
42e0f73a6bf917a7cacca28f7267ca0cd88818d5: Bug 687338: Consistently coerce value prior to range check when setting Vector property (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Tue, 28 Feb 2012 15:11:06 -0800 - rev 7231
Push 4170 by wmaddox@adobe.com at Tue, 28 Feb 2012 23:11:19 +0000
Bug 687338: Consistently coerce value prior to range check when setting Vector property (r=edwsmith)
ba888f512b73d1d969eedee1f49bc8a7befbefdd: Bug 689882: Correct order in which coercions are applied to MOps store arguments (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Mon, 27 Feb 2012 14:24:45 -0800 - rev 7226
Push 4166 by wmaddox@adobe.com at Mon, 27 Feb 2012 22:44:53 +0000
Bug 689882: Correct order in which coercions are applied to MOps store arguments (r=edwsmith)
a59cc6992efaa640ca88408d69380f085c5ba197: Bug 700620: Handle integer wraparound correctly in increment_i and decrement_i (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Fri, 24 Feb 2012 21:06:01 -0800 - rev 7216
Push 4157 by wmaddox@adobe.com at Sat, 25 Feb 2012 07:06:18 +0000
Bug 700620: Handle integer wraparound correctly in increment_i and decrement_i (r=edwsmith) Also cleans up the code and tightens it up a bit for performance and clarity.
1fb053b49a9fb881d5390040f0f2c1197b5055e1: Bug 601426: Use range of 54-bit twos complement integer consistently as 64-bit intptr range (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Fri, 24 Feb 2012 14:55:43 -0800 - rev 7214
Push 4155 by wmaddox@adobe.com at Fri, 24 Feb 2012 22:59:40 +0000
Bug 601426: Use range of 54-bit twos complement integer consistently as 64-bit intptr range (r=edwsmith)
99987b9691556cc647756c0f994c485051e3daaa: Bug 723448: abcdump: undo in readUTFBytes leading UTF BOM compensation (r=edwsmith).
Felix S Klock II <fklockii@adobe.com> - Thu, 09 Feb 2012 13:45:07 +0100 - rev 7193
Push 4135 by fklockii@adobe.com at Thu, 09 Feb 2012 12:45:28 +0000
Bug 723448: abcdump: undo in readUTFBytes leading UTF BOM compensation (r=edwsmith).
65e07c19b309f6dd435ead53c1bc6b1b0031c7a1: Bug 722688 - asm_store64 LIR_std2f cannot have scratch rt == value rv (r=edwsmith+)
Felix S Klock II <fklockii> - Tue, 31 Jan 2012 06:53:00 -0500 - rev 7175
Push 4118 by edwsmith@adobe.com at Tue, 31 Jan 2012 17:19:43 +0000
Bug 722688 - asm_store64 LIR_std2f cannot have scratch rt == value rv (r=edwsmith+)
29e3472aa9762dbf317f9a978a050c4cd3197322: Bug 688261 - nativegen.py should get the correct poolobject in case of Interface class generated methods (p=rulohani, r=wmaddox, sr=edwsmith)
Ruchi Lohani<rulohani@adobe.com> - Fri, 13 Jan 2012 16:03:45 -0800 - rev 7159
Push 4105 by rulohani@adobe.com at Sat, 14 Jan 2012 00:04:02 +0000
Bug 688261 - nativegen.py should get the correct poolobject in case of Interface class generated methods (p=rulohani, r=wmaddox, sr=edwsmith)
2e8247b57659bd80b5bda74ef87b358302704814: Bug 688261 - nativegen.py should get the correct poolobject in case of Interface class generated methods (p=rulohani, r=wmaddox, sr=edwsmith)
Ruchi Lohani<rulohani@adobe.com> - Fri, 13 Jan 2012 14:15:29 -0800 - rev 7157
Push 4103 by rulohani@adobe.com at Fri, 13 Jan 2012 22:16:09 +0000
Bug 688261 - nativegen.py should get the correct poolobject in case of Interface class generated methods (p=rulohani, r=wmaddox, sr=edwsmith)
96f8dd33480f33ed8c3cfad79a6cb7bbc9d22748: Bug 700613: handle large 64-bit atoms via general helper (r=edwsmith).
Felix S Klock II <fklockii@adobe.com> - Tue, 08 Nov 2011 17:04:17 +0100 - rev 6712
Push 4008 by fklockii@adobe.com at Tue, 08 Nov 2011 16:04:57 +0000
Bug 700613: handle large 64-bit atoms via general helper (r=edwsmith).
4bd85ff304e13e78ca5fed77250070155b27b4eb: Bug 697116: perf tests: increase fractional decimals when rendering small numbers (r=edwsmith).
Felix S Klock II <fklockii@adobe.com> - Wed, 26 Oct 2011 11:02:00 +0200 - rev 6693
Push 3991 by fklockii@adobe.com at Tue, 01 Nov 2011 11:18:48 +0000
Bug 697116: perf tests: increase fractional decimals when rendering small numbers (r=edwsmith).
7c2ac104ff18a8573e6757451313a62ce4d05c64: Bug 679840 - Use "safe unlinking" in CodeAlloc class (r=nnethercote,edwsmith)
William Maddox <wmaddox@adobe.com> - Wed, 19 Oct 2011 21:09:45 -0700 - rev 6671
Push 3971 by wmaddox@adobe.com at Thu, 20 Oct 2011 07:29:59 +0000
Bug 679840 - Use "safe unlinking" in CodeAlloc class (r=nnethercote,edwsmith)
717e35d976340a251dbf0801f80ea3da829be8c8: Fix 695328 - Vector inlining code needs to properly interact with CSE in the JIT (f=wmaddox, r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Thu, 20 Oct 2011 08:41:56 +0200 - rev 6670
Push 3970 by lhansen@adobe.com at Thu, 20 Oct 2011 06:42:29 +0000
Fix 695328 - Vector inlining code needs to properly interact with CSE in the JIT (f=wmaddox, r=edwsmith)
6b57469a935c7fe772281a8f79351b518b186098: Bug 686698 - Inline fastpath for Atom to expected type in InvokerCompiler (r=wmaddox,sr=edwsmith)
Tommy Reilly <treilly@adobe.com> - Wed, 19 Oct 2011 10:59:36 -0400 - rev 6668
Push 3968 by treilly@adobe.com at Wed, 19 Oct 2011 14:59:57 +0000
Bug 686698 - Inline fastpath for Atom to expected type in InvokerCompiler (r=wmaddox,sr=edwsmith)
18834e3c7831b0d5afd3d05560a7197c568dbeba: Fix 682280 - Vector.length is slow (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Tue, 18 Oct 2011 19:58:45 +0200 - rev 6663
Push 3964 by lhansen@adobe.com at Tue, 18 Oct 2011 17:58:51 +0000
Fix 682280 - Vector.length is slow (r=edwsmith)
f25fa9e0eb3d4417280f298b840ed75345a1dda6: Bug 624164 - [NPOTB] complete PowerPC nanojit for Firefox (r=rreitmai sr=edwsmith)
Cameron Kaiser <spectre> - Thu, 13 Oct 2011 09:18:42 -0700 - rev 6642
Push 3949 by rreitmai@adobe.com at Thu, 13 Oct 2011 17:53:39 +0000
Bug 624164 - [NPOTB] complete PowerPC nanojit for Firefox (r=rreitmai sr=edwsmith) . . attachment 542715 - Part 2: PPC backend . User-Agent: Mozilla/5.0 (Macintosh; PPC Mac OS X 10.4; rv:2.0b8pre) Gecko/20110107 Firefox/4.0b8pre TenFourFox/Debugging Build Identifier: This adds code for Firefox nanojit support for PowerPC to the tree. This bug does NOT enable PPC nanojit in the build (thus NPOTB); that will be in a future patch. However, it is already enabled in TenFourFox and will be part of beta 9 of that browser. I discussed this work with Edwin Smith in E-mail already -- thanks for all your help, Ed! This patch does the following: - Adds nFragExit to NativePPC.cpp - Adds support for the overflow math instructions to asm_arith() in NativePPC.cpp - Completes nPatchBranch in NativePPC so that it can "demote" 14-bit and 24-bit branch displacements to CTR-based branches, and conversely "promote" CTR-branches to 14-bit displacements where possible, in NativePPC.cpp - Adds additional opcodes to NativePPC.h required by the above - Implements a basic instruction scheduler that hoists independent instructions up higher to facilitate better ILP. This is done by (ab)using the EMIT1 macro in NativePPC.h to do instruction swapping, and adding a new struct to Assembler that tracks instruction history. Special cases are added for certain common sequences that this scheduler does not catch, as well as macros for disabling the optimizer for sequences that must be emitted in strictly serialized order. Performance is mixed. It actually takes a non-trivial hit on SunSpider to use the JIT, although it does significantly better on many parts of Dromaeo. The net effect is overall positive for most tasks, but not nearly enough. Therefore, I'm posting this first working draft to get more eyes on it and suggestions for further optimization. Notes on stuff that came up during development and other related bugs Ed and Nick Nethercote pointed out in our E-mail conversation: - There are lots of stack stores in hot code related to side exits (an empty loop takes an unbelievable amount of time because of this). This hurts badly on POWER which can take a big stall hit for memory access. The suggested solution for this is, as I understand it, controversial (bug 537842). - Some of the load and store noise could be improved by bug 514102, and possibly further with bug 602793. - Bug 545406 "TM: loop invariant code motion (LICM)" will probably benefit PPC greatly. - For certain cases where we can statistically predict the PPC nanojit will take a bath with traced code, bug 597439 where we can let the compiler do our optimization (of the LIR interpreter itself) may also be useful. I have not made any tracer changes in this patch to that end, although I am exploring them. Reproducible: Always Steps to Reproduce: n/a Actual Results: Warp 3. Expected Results: Ludicrous speed.
225bfd318eb701fd54ae3d07a5e7bd1d67df8399: Bug 616169 - Update the LIR opcode table file format to make adding opcodes easier (r=edwsmith,nnethercote)
Virgil Palanciuc <virgilp@adobe.com> - Thu, 13 Oct 2011 09:18:42 -0700 - rev 6640
Push 3949 by rreitmai@adobe.com at Thu, 13 Oct 2011 17:53:39 +0000
Bug 616169 - Update the LIR opcode table file format to make adding opcodes easier (r=edwsmith,nnethercote)
dc04812ad0325c4a63c3d7bb8257d1ddb0becf2f: For 685441 - The crash happens on ARM due to unaligned access in mop_sf64: Clean up the documentation for AVMSYSTEM_UNALIGNED_x_ACCESS (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Wed, 12 Oct 2011 17:07:57 +0200 - rev 6633
Push 3944 by lhansen@adobe.com at Wed, 12 Oct 2011 15:11:00 +0000
For 685441 - The crash happens on ARM due to unaligned access in mop_sf64: Clean up the documentation for AVMSYSTEM_UNALIGNED_x_ACCESS (r=edwsmith)
aef1458fd5564c3be2a8f640117393f8557ea739: Bug 676585 - fix JS compilation for Darwin/ARM. r=edwsmith.
Nicholas Nethercote <nnethercote@mozilla.com> - Wed, 07 Sep 2011 21:31:33 -0700 - rev 6626
Push 3940 by wmaddox@adobe.com at Tue, 11 Oct 2011 00:35:51 +0000
Bug 676585 - fix JS compilation for Darwin/ARM. r=edwsmith.
27b7095a62c6aedfc4771017da8d5dd955a18e4c: Bug 684799 - Measure slop in more JS memory reporters (NJ-specific part). r=edwsmith.
Nicholas Nethercote <nnethercote@mozilla.com> - Wed, 07 Sep 2011 18:15:04 -0700 - rev 6625
Push 3940 by wmaddox@adobe.com at Tue, 11 Oct 2011 00:35:51 +0000
Bug 684799 - Measure slop in more JS memory reporters (NJ-specific part). r=edwsmith.
b5877177462f094471455340f866079cde6c8904: Bug 685660 - Add mozilla tri-license to avmspec source files (pushing for Andre) [r=edwsmith]
Andre Kruetzfeldt <akruetzf> - Thu, 08 Sep 2011 13:16:00 -0400 - rev 6600
Push 3917 by edwsmith@adobe.com at Tue, 27 Sep 2011 16:12:53 +0000
Bug 685660 - Add mozilla tri-license to avmspec source files (pushing for Andre) [r=edwsmith]
ecae7c4f1a85807c3d6fc2c3ddf8fb731345056b: Fix 646541 - OP_applytype is not documented (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Wed, 14 Sep 2011 16:23:56 -0400 - rev 6582
Push 3900 by lhansen@adobe.com at Wed, 14 Sep 2011 20:24:01 +0000
Fix 646541 - OP_applytype is not documented (r=edwsmith)
b1b236ae99fc6c8ae4d039b114fa3515d49ee650: Fix 601817 - Specialized write path for Vector.<T> where T is known RCObject (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Fri, 09 Sep 2011 12:15:33 +0200 - rev 6568
Push 3887 by lhansen@adobe.com at Fri, 09 Sep 2011 10:18:55 +0000
Fix 601817 - Specialized write path for Vector.<T> where T is known RCObject (r=edwsmith)
db7dd23cd076a8f73224e469fc8798dd5d0c6dde: Fix 683507 - assertion fired when constructing a Vector with the wrong number of parameters (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Wed, 07 Sep 2011 09:04:04 +0200 - rev 6564
Push 3883 by lhansen@adobe.com at Wed, 07 Sep 2011 07:04:11 +0000
Fix 683507 - assertion fired when constructing a Vector with the wrong number of parameters (r=edwsmith)
62283c51bdf286fb51b15dc442eea42b5f7b9c48: For 599099 - investigate inline Vector.getNative*Property/setNative*Property calls: test cases (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Fri, 26 Aug 2011 10:17:39 +0200 - rev 6542
Push 3863 by lhansen@adobe.com at Fri, 26 Aug 2011 08:17:49 +0000
For 599099 - investigate inline Vector.getNative*Property/setNative*Property calls: test cases (r=edwsmith)
439a570dc1fea72990a2a31e3a23d72e55740cc0: Fix 678969 - Improve Vector range checking (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Fri, 19 Aug 2011 11:26:05 +0200 - rev 6533
Push 3855 by lhansen@adobe.com at Fri, 19 Aug 2011 09:26:13 +0000
Fix 678969 - Improve Vector range checking (r=edwsmith)
6f72616eadd7c517968331ded510b2d5f6f9255d: Fix 678952 - Operations on Vector.<C> do not make use of or provide information about C (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Fri, 19 Aug 2011 11:23:46 +0200 - rev 6532
Push 3855 by lhansen@adobe.com at Fri, 19 Aug 2011 09:26:13 +0000
Fix 678952 - Operations on Vector.<C> do not make use of or provide information about C (r=edwsmith)
6b300f7377e53335bebb7616d1b4e3f925610701: Fix 484509: - abcasm: need support for a[i] style setproperty (r=edwsmith, f=tharwood)
Lars T Hansen <lhansen@adobe.com> - Fri, 19 Aug 2011 10:55:45 +0200 - rev 6530
Push 3854 by lhansen@adobe.com at Fri, 19 Aug 2011 09:12:09 +0000
Fix 484509: - abcasm: need support for a[i] style setproperty (r=edwsmith, f=tharwood)
ecc1ce2a93efcd9e5753f7d2bd4e14bd4c946b18: Bug 679397 - X64 branch patch code seems to be wrong for jmp 64bit, but is actually fine: comment needed. (r=edwsmith)
Edwin Smith <edwsmith@adobe.com> - Wed, 17 Aug 2011 16:11:36 -0400 - rev 6523
Push 3851 by wmaddox@adobe.com at Thu, 18 Aug 2011 00:16:02 +0000
Bug 679397 - X64 branch patch code seems to be wrong for jmp 64bit, but is actually fine: comment needed. (r=edwsmith)
8de850f0d909c046af6f2f18f844bed101457b55: Bug 620860: clear specialized-function map when a label is encountered (r=wmaddox,sr=edwsmith)
Steven Johnson <stejohns@adobe.com> - Wed, 27 Jul 2011 13:00:13 -0700 - rev 6490
Push 3826 by stejohns@adobe.com at Wed, 27 Jul 2011 20:00:30 +0000
Bug 620860: clear specialized-function map when a label is encountered (r=wmaddox,sr=edwsmith)
bf9d9cfd04d4baff22e2b80528eb8d3e5475f259: Bug 576247 - asm_stkarg not implemented for x86_64. r=edwsmith
Makoto Kato <m_kato@ga2.so-net.ne.jp> - Wed, 25 May 2011 17:37:27 +0900 - rev 6474
Push 3815 by stejohns@adobe.com at Thu, 21 Jul 2011 00:23:51 +0000
Bug 576247 - asm_stkarg not implemented for x86_64. r=edwsmith
4d448a44169ab1858aeb8da417bbd7483b46250d: Bug 662851 - Remove redundant cast found during post push review (r=edwsmith)
Tommy Reilly <treilly@adobe.com> - Wed, 29 Jun 2011 12:47:38 -0400 - rev 6437
Push 3787 by treilly@adobe.com at Wed, 29 Jun 2011 16:47:52 +0000
Bug 662851 - Remove redundant cast found during post push review (r=edwsmith)
457203bf8d0d91277aecc586050bb6aec12b41e1: Fix 615181 - Extend SlotStorageType enum to allow more types (patch=virgilp, r=edwsmith, f=stejohns, push=lhansen)
Lars T Hansen <lhansen@adobe.com> - Fri, 17 Jun 2011 16:41:54 +0200 - rev 6396
Push 3753 by lhansen@adobe.com at Fri, 17 Jun 2011 14:42:15 +0000
Fix 615181 - Extend SlotStorageType enum to allow more types (patch=virgilp, r=edwsmith, f=stejohns, push=lhansen)
11b703b874a03ba6fd25a3b01238abc800e471a5: Bug 521245 - Use platform sin/cos/tan on Win32 (r=rreitmai,edwsmith)
William Maddox <wmaddox@adobe.com> - Fri, 27 May 2011 19:03:20 -0700 - rev 6361
Push 3719 by wmaddox@adobe.com at Sat, 28 May 2011 02:04:14 +0000
Bug 521245 - Use platform sin/cos/tan on Win32 (r=rreitmai,edwsmith)
206038c93867474903d00dc2af48473d2add75c7: Bug 655390 - argcOk should probably do a runtime check for argc<0 (r=rreitmai,sr=edwsmith)
Steven Johnson <stejohns@adobe.com> - Fri, 27 May 2011 11:11:35 -0700 - rev 6359
Push 3717 by stejohns@adobe.com at Fri, 27 May 2011 18:11:48 +0000
Bug 655390 - argcOk should probably do a runtime check for argc<0 (r=rreitmai,sr=edwsmith)
787e07e4fb6bfa1a82b5d3a591ed1d852c26578e: Bug 648862 - nanojit: Remove WINCE code. r=edwsmith,Jacob.Bramley.
Ed Morley <bmo@edmorley.co.uk> - Wed, 18 May 2011 10:38:17 +1000 - rev 6332
Push 3697 by wmaddox@adobe.com at Tue, 24 May 2011 07:40:12 +0000
Bug 648862 - nanojit: Remove WINCE code. r=edwsmith,Jacob.Bramley.
cb0301d0f80de32ab4a1a45712b94907c49f5292: Bug 633653 - revamp about:memory (NJ-specific part). The original patch was landed on mozilla-central. But because NJ-to-TM merges clobber any NJ changes made in TM, I need to land the NJ-specific directly onto nanojit-central. r=edwsmith.
Nicholas Nethercote <nnethercote@mozilla.com> - Wed, 11 May 2011 16:53:11 -0700 - rev 6331
Push 3697 by wmaddox@adobe.com at Tue, 24 May 2011 07:40:12 +0000
Bug 633653 - revamp about:memory (NJ-specific part). The original patch was landed on mozilla-central. But because NJ-to-TM merges clobber any NJ changes made in TM, I need to land the NJ-specific directly onto nanojit-central. r=edwsmith.
9b955600941df3281e849d36e84f6ef98f6fb4a6: Bug 653244 - Error and Math builtins should be eagerly inited (r=lhansen,sr=edwsmith)
Steven Johnson <stejohns@adobe.com> - Thu, 28 Apr 2011 14:08:34 -0700 - rev 6242
Push 3619 by stejohns@adobe.com at Thu, 28 Apr 2011 21:12:27 +0000
Bug 653244 - Error and Math builtins should be eagerly inited (r=lhansen,sr=edwsmith)
0c16b8944742b1ded97069b62df0afa13b61949f: Bug 644900 - Generate probes for stack expansion when allocating large frames on Windows (r=edwsmith)
William Maddox <wmaddox@adobe.com> - Wed, 30 Mar 2011 21:45:21 -0700 - rev 6227
Push 3609 by stejohns@adobe.com at Wed, 27 Apr 2011 17:21:13 +0000
Bug 644900 - Generate probes for stack expansion when allocating large frames on Windows (r=edwsmith) Also extend lirasm with --stkskip option to exercise the fix.
ff7c627718e86676fa189e2583c219045548dce7: Bug 539094 - On-stack replacement, port from tr-spicy (r=edwsmith,rreitmai)
William Maddox <wmaddox@adobe.com> - Wed, 20 Apr 2011 20:08:36 -0700 - rev 6198
Push 3583 by wmaddox@adobe.com at Thu, 21 Apr 2011 03:11:30 +0000
Bug 539094 - On-stack replacement, port from tr-spicy (r=edwsmith,rreitmai) This is a cumulative patch containing the OSR work done in the tr-spicy branch. The original work was reviewed piecemeal by edwsmith, and the rebased cumulative patch by rreitmai. OSR may be enabled with the --enable-osr configuration switch. It is currently disabled by default. While it has been used on a limited basis in production, unresolved interpreter/jit semantic divergences make general deployment inadvisable at this time. Nonetheless, we are required to carry this functionality forward at this point in support of hosts that use it.
79e463a7dbfc2d7237ad1ad6d849d79c382f2396: Bug 643144 - GCC internal compiler error (r=edwsmith)
Tommy Reilly <treilly@adobe.com> - Mon, 21 Mar 2011 12:59:54 -0400 - rev 6118
Push 3511 by treilly@adobe.com at Mon, 21 Mar 2011 17:00:08 +0000
Bug 643144 - GCC internal compiler error (r=edwsmith)
3b45d4b7918354a07560344cd141542d4729b0db: Bug 642806 - List add crashes with GCInlineObject inheritance (r=stejohns,sr=edwsmith)
Tommy Reilly <treilly@adobe.com> - Mon, 21 Mar 2011 12:59:12 -0400 - rev 6117
Push 3510 by treilly@adobe.com at Mon, 21 Mar 2011 16:59:29 +0000
Bug 642806 - List add crashes with GCInlineObject inheritance (r=stejohns,sr=edwsmith)
152c1dddb0763df5ba45dd2856e5de196d4cdb6d: Fix 642392 - Reorganize AvmCore fields to make exact tracing tractable (r=edwsmith)
Lars T Hansen <lhansen@adobe.com> - Thu, 17 Mar 2011 15:51:50 +0100 - rev 6105
Push 3500 by lhansen@adobe.com at Thu, 17 Mar 2011 14:51:57 +0000
Fix 642392 - Reorganize AvmCore fields to make exact tracing tractable (r=edwsmith)
82b4008b9b3208a4ff73d030925b234ed12bc769: Bug 638233 - SO::constructProperty is never overridden, anywhere (r=edwsmith)
Steven Johnson <stejohns@adobe.com> - Wed, 16 Mar 2011 12:24:45 -0700 - rev 6101
Push 3497 by stejohns@adobe.com at Wed, 16 Mar 2011 19:24:48 +0000
Bug 638233 - SO::constructProperty is never overridden, anywhere (r=edwsmith)
5536a71f53b929ef821702e2722cc46839618585: Bug 607816 - nanojit: Mechanism to specify calls that never return (r=edwsmith)
Rick Reitmaier <rreitmai> - Mon, 14 Mar 2011 11:46:00 -0700 - rev 6100
Push 3496 by rreitmai@adobe.com at Wed, 16 Mar 2011 18:19:42 +0000
Bug 607816 - nanojit: Mechanism to specify calls that never return (r=edwsmith) . . attachment 519196 - insert LIR_ret after calls that never return (alternate) . None
fc8730a8d886dc0e845feaa0a68dca884df00e28: Bug 617244 - nanojit: remove AvmCore (r=edwsmith)
Nicholas Nethercote <nnethercote> - Mon, 06 Dec 2010 20:05:00 -0800 - rev 6086
Push 3483 by rreitmai@adobe.com at Mon, 14 Mar 2011 23:13:48 +0000
Bug 617244 - nanojit: remove AvmCore (r=edwsmith) . . attachment 495753 - TR patch (against TR 5632:00830e7731c4) . Created attachment 495751 NJ patch (against TM 58520:8921e3faccd2) Something I came across while working on bug 529430: there's this AvmCore class that is used throughout Nanojit. It's a Tamarin-ism, AFAICT, and isn't needed, and it's quite confusing. All that *is* needed is the Config struct that AvmCore (statically) contains. So I decided to clean this up. The NJ patch: - Removes AvmCore from the nanojit/avmplus.h file that TM and lirasm use (but TR doesn't). Removes all traces of AvmCore elsewhere. - Removes some other dead stuff from avmplus.h. - nInit() is now passed no arguments; none of the back-ends need an argument. - Removes 'has_cmov' from the Sparc back-end; it is dead. - Moves use_cmov() from AvmCore to Config. - Updates lirasm for all the above changes, by adding a nanojit::Config to class Lirasm.
bf20a8332a900edfdf4cfa857ef044167c37ad04: nanojit: remove LIR_xtbl. r=edwsmith.
Nicholas Nethercote <nnethercote@mozilla.com> - Thu, 03 Mar 2011 14:20:55 -0800 - rev 6080
Push 3483 by rreitmai@adobe.com at Mon, 14 Mar 2011 23:13:48 +0000
nanojit: remove LIR_xtbl. r=edwsmith.