Boris Zbarsky
c57647d6f4
Bug 451605. Make sure to properly restore the thisp for the innermost frame when taking a nested side exit with a callDepth > 0. r=brendan,gal
2008-09-08 08:55:20 -04:00
Igor Bukanov
27d83904e5
bug 453133 - merging JSOP_DEFUN and JSOP_CLOSURE. r=brendan
2008-09-08 13:58:17 +02:00
Igor Bukanov
1694ed5ded
bug 454163 - fixing build problem with 64-bit debug builds.
2008-09-08 12:37:41 +02:00
Brendan Eich
082cb07c53
Make return at callDepth 0 terminate the loop, like break (454027, r=gal).
2008-09-08 00:31:18 -07:00
Brendan Eich
e697e5113d
Bug 454044 - TM: JSOP_NOT needs isPromoteInt/::demote help.
2008-09-07 01:40:24 -07:00
David Anderson
5018287de0
Trace through parseInt(double), bug 454037 r=gal+brendan.
2008-09-07 02:41:10 -05:00
Brendan Eich
ba17412f5e
Easy fixes to avoid aborting on V8/raytrace.js: MAX_CALLDEPTH doubled to 10; JSOP_NOT handles int and undefined.
2008-09-05 18:57:32 -07:00
Brendan Eich
623433f549
Merge.
2008-09-05 18:34:16 -07:00
Brendan Eich
fc936fa1da
Checkpoint arguments tour-de-force (453730).
2008-09-05 18:29:08 -07:00
Ginn Chen
2763e5d63e
Make TraceMonkey build on Solaris x86 with Sun Studio 12 (bug 452588, r=danderson).
2008-09-05 16:56:03 -07:00
David Anderson
9937dc1262
Trace through Math.log as a known native (bug 453868, r=brendan).
2008-09-05 14:25:01 -07:00
David Anderson
ba87213e86
Abort tracing if a callee encounters a primitive |this| and doesn't support it (bug 453701, r=brendan).
2008-09-05 13:55:03 -07:00
Blake Kaplan
b4d3674f58
Bug 453747 - Replace any holes on the stack with undefined. r=gal
2008-09-04 22:14:27 -07:00
Brendan Eich
ce684a0a6e
Bug 453747 - Avoid overuse of the fun_getProperty class getter to optimize getting and setting random properties on functions. r=mrbkap
2008-09-04 21:56:37 -07:00
David Anderson
3e31bc7fb2
Another fix for the patch to bug 452884 -- when ignoring switch guards, don't update the tracker stack.
2008-09-04 20:11:58 -07:00
David Anderson
8c610797f3
Merge.
2008-09-04 19:43:58 -07:00
David Anderson
05991c905d
Better fix for bug 452884 (changeset badf4c84665f regressed) - don't emit guards rather than bailing out of comparisons (r=gal).
2008-09-04 19:43:07 -07:00
Brendan Eich
ae2a22c019
Bug 453747 - Avoid overuse of the fun_getProperty class getter to optimize getting and setting random properties on functions. r=mrbkap
2008-09-04 19:30:58 -07:00
Brendan Eich
b35babfe1d
House style bracing for multiline then clause.
2008-09-04 18:53:31 -07:00
David Anderson
09123ebd63
Don't guard on constant decisions, which will always have the same path (bug 452884, r=gal).
2008-09-04 14:44:59 -07:00
David Anderson
ede63892f9
Disabled building AMD64 JIT in shell until regressions are tracked down. Fixed some compiling errors while I'm poking around here (bug 453674).
2008-09-04 10:58:35 -07:00
David Anderson
afe91e2757
Merge.
2008-09-03 23:28:27 -07:00
David Anderson
b791826dc5
Fixed JSOP_NEG with 0 being tracked as a promotable int, and added an equals-zero guard for the same opcode (bug 453049, r=gal).
2008-09-03 23:27:58 -07:00
Brendan Eich
9657023ef6
Missing guard for CallGetter builtin (453580, r=gal).
2008-09-03 22:32:13 -07:00
Blake Kaplan
d9be63b689
Handle an interpreted constructor returning an object. r=brendan
2008-09-03 18:21:15 -07:00
Andreas Gal
1d44de3cb6
Merge.
2008-09-03 01:22:07 -07:00
Andreas Gal
483b7fd5a1
Removed a bogus assert. Inner loops can exit on a goto in case we hit a break (453051).
2008-09-03 01:21:21 -07:00
Brendan Eich
4fa7c5195d
Put the trace-recording/executing flag in the trace monitor to handle many contexts per thread (451666, r=gal).
2008-09-03 00:38:24 -07:00
Andreas Gal
c466258c18
Limit tree growth to 16 traces per tree to avoid code explosion due to tail duplication (452869, r=danderson).
2008-09-02 23:52:11 -07:00
Brendan Eich
3cd36a3b5f
Fix JS_THREADSAFE build, pick extern nit.
2008-09-02 23:00:23 -07:00
Vladimir Vukicevic
dfd7f72c0a
Fix up some ifdefs and debug printfs
2008-09-02 22:42:25 -07:00
Vladimir Vukicevic
a4ec50d569
Add VFP for floating point ops to nanojit ARM backend.
2008-09-02 22:29:23 -07:00
Vladimir Vukicevic
420e72ed85
b=449526, TM: fix up ARM code generation / softfloat
2008-09-02 22:29:23 -07:00
Brendan Eich
e3916b316d
Teach js_IsLoopExit about JSOP_AND, JSOP_OR, etc. and separate extended jump cases so they can get the extended offsets.
2008-09-02 22:29:06 -07:00
Blake Kaplan
3d517dff43
Remove outdated comment. r=gal
2008-09-02 22:18:57 -07:00
Andreas Gal
f91071f139
Allocate the oracle dynamically instead of making it a global object to avoid atexit C compatibility issues (453388, r=danderson).
2008-09-02 21:56:22 -07:00
Blake Kaplan
8f80eec070
Don't flush anything if we're not even enabled. r=gal
2008-09-02 14:02:13 -07:00
Ginn Chen
76f8dc0daf
On x86 compilers without fastcall, simulate it when invoking traces and un-simulate it when invoking builtins (bug 452390, r=dvander).
2008-09-02 11:43:55 -07:00
Andreas Gal
1632042ed8
Merge.
2008-09-01 18:37:27 -07:00
Andreas Gal
4a36a6b709
Backed out changeset 25f856429db3. Wrong fix. Looks like jstracer.h is not included. We in fact can call methods of TraceRecorder directly. Sigh. I wish we had a try server.
2008-09-01 18:37:04 -07:00
Andreas Gal
d85543063d
Merge.
2008-09-01 18:33:05 -07:00
Andreas Gal
fe70cfea9b
Introduce js_DeepAbort to be used from jsinterp.cpp, instead of trying to call TraceRecorder->deepAbort directly (453235).
2008-09-01 18:32:41 -07:00
Brendan Eich
b3042b928d
Holey single-element dense array has null dslots, requires nulld defense (453173).
2008-09-01 13:59:30 -07:00
Brendan Eich
cda5a77a1b
Fast followup fix for unintended change to 'f' prefix handling in TraceRecorder::record_JSOP_NEW (452878).
2008-09-01 01:44:26 -07:00
Brendan Eich
5a8cb06838
Kind of an Array initialiser tour-de-force for bug 452878:
...
1. Split FastNewArray from FastNewObject built-in for greater speed/specialization and further splitting into Array_1str, etc.
2. Add Array_1str, Array_2obj, and Array_3num builtins for benchmarked new Array(...) constructions.
3. Export ARRAY_SET_DENSE_LENGTH and ARRAY_GROWBY via jsarray.h to jstracer.cpp.
4. Tweaked SetArrayElement to make common/best case code be the predicted/prefetched path.
5. js_MakeArraySlow now preserves the pre-slow length in JSSLOT_ARRAY_COUTN as a jsval-tagged int if possible -- this will help the tracer avoid aborting on dense arrays that turned slow but not sparse by addition of a named property.
6. Export js_fun_apply and js_Object from their respective .cpp files, in these cases just to jstracer.cpp via local prototypes (no .h files involved).
7. More INS_CONSTPTR and INS_CONST macrology for better names in trace debug spew.
8. Fix TraceRecorder::test_property_cache to avoid aborting on JSOP_SETNAME that creates a new global, by setting it to undefined so it can be lazily imported. This helps 3d-raytrace.js, which has an unintended global loop control variable in a function.
9. JSTraceableNative loses its premature-deadwood tclasp member (my bad).
10. TraceRecorder::record_JSOP_NEW() handles 'new Object' now along with the 'new Array' variations. I also cut down the copy-paste code from JSOP_CALL's record method to mostly what is needed now.
11. Add KNOWN_NATIVE_DECL macro for concise prototype of library-private js_* native functions, and alphabetized the lists (too long for any other order to be winning).
12. Big honking special case for foo.apply(obj, [str]), which we can generalize as needed. Helps string-tagcloud.js. What's cool is how tracing allows us to rewrite this to foo(str) with this set to obj, eliminating the Function.prototype.apply. This requires some rewriting in JSOP_ENDINIT's record method.
2008-09-01 01:24:58 -07:00
Jesse Ruderman
1b5148d940
Bug 452905 - Make js_RecordTree printfs only appear with TRACEMONKEY=verbose. r=gal
2008-08-30 02:21:00 -07:00
Brendan Eich
c7604aac4c
- Support JSOP_CONDSWITCH's JSOP_CASE and extended-jump-offset JSOP_CASEX variants properly.
...
- Fix sleep-deprived constant conditions due to failure to test flags & CMP_TRY_BRANCH_AFTER_COND!
- Use new TraceRecorder::fuseIf that Andreas added in a few more places where the previous try-branch-after-cond logic was open-coded.
2008-08-30 00:01:59 -07:00
Blake Kaplan
2782b3de85
Fix gmail crash by not tracing switches. r=brendan
2008-08-29 19:01:00 -07:00
Blake Kaplan
219161a0aa
Abort (deeply) if we re-enter js_Interpret in the middle of a trace. r=brendan
2008-08-29 18:59:21 -07:00
Andreas Gal
ea15fc19ef
Fixed an signed/unsigned comparison warning in an assert in jstracer.cpp.
2008-08-29 17:54:39 -07:00
Andreas Gal
33c4a8e742
Merge.
2008-08-29 17:35:34 -07:00
Andreas Gal
113ad8babe
Add builtins for toLowerCase, toUpperCase and replace(str,str) (452885, r=mrbkap).
2008-08-29 17:35:00 -07:00
Brendan Eich
45ea54d2d1
Don't let GC run when recording.
2008-08-29 16:04:54 -07:00
Blake Kaplan
48fb3e7f94
Bug 452875 - Ensure we intern the id for JSOP_IN. r=brendan
2008-08-29 15:59:09 -07:00
David Anderson
7515a0583e
Merge.
2008-08-29 15:38:05 -07:00
David Anderson
f720beb914
Fixed double-as-integer check not handling negative zero (bug 452170, r=gal).
2008-08-29 15:37:37 -07:00
Andreas Gal
f29178e950
Track number of breaks we generate loop exits for.
2008-08-29 15:29:55 -07:00
Andreas Gal
1e41ca88af
Long-form conditional branches (IFEQX, IFNEX) are never fused, so don't check for them in fuseIf.
2008-08-29 15:20:36 -07:00
Andreas Gal
57a8fae3d1
Merge.
2008-08-29 15:17:20 -07:00
Andreas Gal
9e7246a2b9
Track locations of control-flow merges in cfgMerges during recording (452869).
2008-08-29 15:12:17 -07:00
David Anderson
a2fdc866ff
Fixed assumptions that nanojit's insCall() would not clobber the input argument array (bug 452853, r=gal).
2008-08-29 14:22:21 -07:00
David Anderson
f10bd19c8e
Abort recording on invalid string indexes for JSOP_GETELEM (bug 452713, r=brendan).
2008-08-29 13:05:41 -07:00
Brendan Eich
f92de94117
Fix bogus JOF_VARPROP test; fix uninitialized id in JSOP_IN recorder, should have caught it when I reviewed danderson's patch.
2008-08-29 00:58:10 -07:00
Brendan Eich
655888f31e
Consolidate equal and cmp code harder, trace switch ops, use INS_CONST more (bug to be filed -- bugzilla down atm).
2008-08-28 23:50:48 -07:00
Andreas Gal
1115d0162b
Unroll loops we were not able to call once, but only if we don't have a tree available for that loop header.
2008-08-28 23:19:23 -07:00
Andreas Gal
535bdef3ee
Merge.
2008-08-28 22:34:14 -07:00
Andreas Gal
c263b7a218
Merge.
2008-08-28 22:33:45 -07:00
David Anderson
951c95658a
Fixed trying to record JSOP_IN with some unusuable left-hand values (bug 452724, r=brendan).
2008-08-28 22:33:32 -07:00
Andreas Gal
4fb4d1fdc5
Monitor downward branches only during recording (452709).
2008-08-28 22:33:22 -07:00
Brendan Eich
1f07d79c75
Abort trace if forInLoop values are not stable; currently, not strings (bug 452703).
2008-08-28 21:41:15 -07:00
Brendan Eich
25a591f870
Fixed bogus assertion in test_property_cache (bug 452693, r=dvander).
2008-08-28 21:15:39 -07:00
Andreas Gal
f10a08c384
Cleanup detection of gotos that were emitted as a result of a BREAK statement.
2008-08-28 17:46:25 -07:00
Andreas Gal
65bc53f864
Merge.
2008-08-28 17:01:22 -07:00
Andreas Gal
c0e0852952
Notify the monitor of all branches, not just backwards branches.
2008-08-28 17:00:52 -07:00
Blake Kaplan
977123733e
Bug 452338 - Abort on the weird case where we find a global name on the prototype of the scope chain. r=brendan
2008-08-28 16:37:03 -07:00
Andreas Gal
c274daed5c
Merge.
2008-08-28 14:51:16 -07:00
David Anderson
1b50da7117
Abort recording on JSOP_INCPROP with an invalid slot (bug 452336, r=brendan).
2008-08-28 14:43:44 -07:00
Andreas Gal
453ea3339f
If the inner tree cannot be adjusted to match the call site of the outer tree because it uses an int where the outer tree uses a double, trash the inner tree, not the outer one. In CallTree, return the innermost guard we return from, not the outermost one. Jump over at most 1 loop edge that doesn't go back to our own header, not an arbitrary amount.
2008-08-28 14:24:58 -07:00
David Anderson
48fe80af99
Merge.
2008-08-28 13:15:34 -07:00
David Anderson
eb051d8c73
Better fix for bug 451666 - only track new rval for primitives (r=brendan).
2008-08-28 13:14:59 -07:00
Blake Kaplan
1006bbf99c
Bug 452333 - Don't push tagged jsvals on the stack. r=brendan
2008-08-28 12:40:48 -07:00
David Anderson
d7c4af1e0c
Merge.
2008-08-28 11:56:45 -07:00
David Anderson
d7b5791be9
Track new rval from JSOP_RETURN when constructing inline frames (same as JSOP_STOP).
2008-08-28 11:55:58 -07:00
Blake Kaplan
2bcc0a39b5
Bug 452573 - The interpreter's JSOP_VOID doesn't push anything, so we shouldn't expect it to. r=brendan
2008-08-28 11:55:27 -07:00
Peter Weilbacher
046dd17991
[OS/2] Bug 452630: Build break in jstracer.cpp (r=gal)
2008-08-28 18:28:52 +03:00
Jesse Ruderman
df0d2c793b
Remove extra space in tracing statistics message (bug 452571). r=gal
2008-08-28 00:16:50 -07:00
Brendan Eich
5d132dba83
Record JSOP_IN (452563, r=gal).
2008-08-27 23:00:43 -07:00
Andreas Gal
2316d281a0
Keep unrolling inner loops as long we are not hitting the same inner loop edge twice in a row (452362).
2008-08-27 17:48:15 -07:00
Andreas Gal
4b9cda096c
Merge.
2008-08-27 17:26:36 -07:00
Andreas Gal
cff18f06cd
Fix trashing of inner trees. Keep track of trees that call to a tree and flush them as well since they directly embed the code address. Since flushing an inner tree might invalidate the tree we are currently compiling, trees are now trashed in the destructor of TraceRecorder.
2008-08-27 17:25:56 -07:00
David Anderson
ccc26331ca
Removed INS_CONSTPTR use with JSVAL_TAG to discourage future 64-bit problems. Fixed 32-bit build.
2008-08-27 16:32:11 -07:00
David Anderson
98306ccb74
Merge.
2008-08-27 16:09:17 -07:00
David Anderson
d0d64bbeff
Added LIR_qior/pior. Fixed more AMD64 page jumping bugs. Fixed some other AMD64 usage bugs in jstracer.
2008-08-27 16:08:59 -07:00
Andreas Gal
e8fcb3bd37
Merge.
2008-08-27 14:48:59 -07:00
Andreas Gal
3fb237f38c
Loops with a single iteration start tracing, but never complete since we immediately walk past the loop edge after the first iteration. At least try to complete the loop around it by walking past such thin inner loops when compiling the outer loop (452362).
2008-08-27 14:46:57 -07:00
Brendan Eich
4842b435c5
Oops, not enough copying...
2008-08-27 13:54:05 -07:00
Brendan Eich
510ec8bdce
Fix a bug Jesse reported: insufficient copy/paste from record_JSOP_CALL to record_JSOP_NEW.
2008-08-27 13:52:46 -07:00
Brendan Eich
9cb246a8fc
Fix another broken assert (452372).
2008-08-27 01:10:44 -07:00
Andreas Gal
7df899a9f5
Merge.
2008-08-26 21:46:21 -07:00
Andreas Gal
418b15eadd
Flag a side exit as loop exit only if the underlying loop condition actually targets the current loop header, otherwise treat it as a branch exit. This is required for inner loops where we unroll the first iteration but the loop condition immediately fails. If we report such exits as loop exits, no branch gets ever attached to them and we keep side-exiting over and over.
2008-08-26 21:43:59 -07:00
Brendan Eich
ae9605be38
Correct (but sadly a bit slower) property cache guarding (452140, r=shaver).
2008-08-26 21:26:10 -07:00
Brendan Eich
e326b78b55
Added knownNative for num_toString (bug 452340, r=dvander).
2008-08-26 20:00:11 -07:00
Andreas Gal
40796f4154
For global code assume that no slots are below the current frame.
2008-08-26 13:25:40 -07:00
Andreas Gal
795f7ae426
Merge.
2008-08-26 01:01:24 -07:00
Andreas Gal
eb0c261390
When extending an outer tree from its nesting guard, the guard we actually exited on determines the typemap for the current frame and all inlined frames, whereas the nesting guard is consulted for all type information frames below the current frame.
2008-08-26 01:00:53 -07:00
Brendan Eich
a7cdb3ddb2
Fix warning.
2008-08-25 23:22:20 -07:00
Andreas Gal
4bb68b852c
Rename lr->anchor in AttemptToExtendTree()
2008-08-25 23:19:00 -07:00
Andreas Gal
24b663fd61
Determine the tree we have to grow from the side exit we attach to in AttemptToExtendTree() instead of passing the parameter in.
2008-08-25 23:17:38 -07:00
Brendan Eich
43b25b3c09
Rename gcDontBlock to runningJittedCode, and assert it's false on entry to js_Interpret (for 451657, but not the fix, just prolog).
2008-08-25 23:08:21 -07:00
Andreas Gal
e86ad8730d
Merge.
2008-08-25 22:11:23 -07:00
Andreas Gal
d00742c962
Add an assert checking that celldepth is zero upon loop exit side exits and better verbose diagnostic for trace exits.
2008-08-25 22:11:04 -07:00
Andreas Gal
c86c7d09fb
Generate an always taken LOOP_EXIT guard when an inner loop encounters a break statement. When returning from such an inner tree, make sure the recorder resumes at the target address of the goto the break was emitted as, otherwise the outer tree confuses the location with a break in its own loop. Disabled outerlining (this patch subsumes it, but we might enable it later since it might handle some cases more efficiently, currently off for debugging through). fasta currently asserts, and fannkuch hits a perf regression. These are probably bugs this change exposed.
2008-08-25 17:24:49 -07:00
Blake Kaplan
48d275806e
Bug 451785 - Trace == and != involving booleans. r=gal
2008-08-25 16:58:18 -07:00
Blake Kaplan
298caaa80f
Bug 451782 - Trace Math.ceil. r=brendan
2008-08-25 16:54:02 -07:00
David Anderson
e4ef2c3451
Merge.
2008-08-25 16:06:17 -07:00
Brendan Eich
b385e456d9
Fixed cases where argc varied between recording time and execution time (bug 451727).
2008-08-25 16:04:10 -07:00
Andreas Gal
fc5667ef13
Split closeLoop() and compile() and add endLoop(), which omits an always-exit guard (will be used for break statements.)
2008-08-25 15:17:46 -07:00
Brendan Eich
eb8bbd11d2
Cope with sub-optimal JSOP_FORNAME instead of JSOP_FORVAR inside a with, or similar (eval-created locals; 451806).
2008-08-24 08:18:17 -07:00
Andreas Gal
e944751f2e
Merge.
2008-08-23 15:13:55 -07:00
Andreas Gal
71b12b53fe
Enable outlining. If a loop doesn't connect back to its own header (i.e. break inside the loop), see if the outer loop path can be attached to the inner loop (outerlining). Added test case (failure mode is high number of activated traces in trace-test.js). This test exposes a regression in access-fannkuch. Committing so we can diagnose that separately.
2008-08-23 15:11:38 -07:00
David Anderson
04ab3326c8
Merge.
2008-08-22 17:36:37 -07:00
David Anderson
e88386ec38
Fixed bogus assertion in debug mode.
2008-08-22 17:32:45 -07:00
Andreas Gal
f5462e1c0a
Merge.
2008-08-22 17:32:01 -07:00
Andreas Gal
454b252c63
If the outer loop has a value in a float register but the inner tree expects an int, tell the oracle that we want doubles in that slot in the inner tree and trash the inner tree so it gets re-compiled with a matching layout (451673).
2008-08-22 17:31:23 -07:00
Andreas Gal
2fb04231fe
Allocate nanojit code cache incrementally (bug 443111).
2008-08-22 16:37:37 -07:00
David Anderson
36ea645fbe
Bug 451743 - On x86 processors, only use SSE2 if the processor supports it, otherwise default back to x87 FPU (r=gal, sr=mrbkap)
2008-08-22 12:21:27 -07:00
Andreas Gal
80088cd777
Fix warning about signed/unsigned comparison.
2008-08-21 11:40:56 -07:00
Andreas Gal
ee5f8e441f
Merge.
2008-08-21 05:21:37 -07:00
Andreas Gal
8f6bbaaa5f
Abort recording when an error occurs.
2008-08-21 05:21:01 -07:00
Brendan Eich
fd148818e1
Add split built-in, also sort knownNatives for my own sanity (should sort builtins.tbl and jsbuiltins.cpp by function while at it, but too tired).
2008-08-21 04:13:07 -07:00
Brendan Eich
147018b013
Merge, plus restore TraceRecorder::activeCallOrGlobalSlot.
2008-08-21 03:51:37 -07:00
Brendan Eich
9cd8e5a539
Upvar, v0.1 (limited to looking up from eval in a function, to the function's args and vars).
2008-08-21 03:47:33 -07:00
shaver@mozilla.org
35915d94ea
disable tracing with active call object pending perf fix
2008-08-21 03:19:17 -07:00
Andreas Gal
d9acfb10f0
Remove FastEval, broken by design (r=brendan).
2008-08-21 03:13:31 -07:00
Brendan Eich
324b9c7f4e
Remove lambda-replace built-in, it's not safe to record into a nested js_Interpret.
2008-08-21 01:39:16 -07:00
Andreas Gal
2e767b124c
Convert undefined to NULL on tree entry if the tree wants an object type.
2008-08-21 01:13:07 -07:00
shaver@mozilla.org
ed7fab1440
trace |new Array|, via the magic of copy and paste
2008-08-21 00:50:20 -07:00
Brendan Eich
321ae383ba
Merge.
2008-08-21 00:47:51 -07:00
Andreas Gal
fcd8670fec
If we expect a concrete value but see an undefined come into the trace, convert undefined to the concerete value.
2008-08-21 00:43:45 -07:00
Brendan Eich
9b56ca4984
Tidy up globalShape mismatch reporting.
2008-08-21 00:19:04 -07:00
Andreas Gal
e26e3234f4
Merge.
2008-08-20 23:40:24 -07:00
Andreas Gal
12fcdd141f
Remove bogus fadd +1 check at the end of the loop. We have long switched over to starting slots as int and hence n++ will be emitted as integer add with i2f, so checking for i2f only is sufficient.
2008-08-20 23:39:57 -07:00
Andreas Gal
ddf445130e
Increase HOTEXIT to 1 (from 0).
2008-08-20 23:35:00 -07:00
Andreas Gal
1261e82b42
Eliminate redundant BoxDouble(UnboxDouble) chains.
2008-08-20 23:19:48 -07:00
shaver@mozilla.org
6870d297ef
say what we're returning from from EnterFrame
2008-08-20 23:12:28 -07:00
Brendan Eich
9e4119701e
Merge.
2008-08-20 22:44:16 -07:00
Brendan Eich
e9b635b38a
1. Add activeCallOrGlobalSlot, used by JSOP_NAME, JSOP_CALLNAME, etc. recorders to cope with Call objects on the scope chain, if they represent still-active frames covered by callDepth.
...
2. Add builtin for eval, to be optimized further, to handle very short eval'ed programs as found in popular benchmarks.
3. Add common scopeChain TraceRecorder helper, to generate the LIR to load cx->fp->scopeChain.
4. Add ABORT_TRACE to LeaveFrame for clarity.
2008-08-20 22:40:39 -07:00
Andreas Gal
8468ac7444
Support multiple returns from called trees by continueing the outer tree with a new nested guard.
2008-08-20 22:37:00 -07:00
Andreas Gal
36ddb45efb
Restore state.sp and state.rp before the nested guard exit point so we can have a sequence of them and still have rp/sp adjusted correctly.
2008-08-20 22:23:17 -07:00
Andreas Gal
e9f3ccf839
Push actual guard that failed in a nested call, not the expected.
2008-08-20 20:12:19 -07:00
Andreas Gal
87295d3d95
Merge.
2008-08-20 19:25:26 -07:00
Andreas Gal
1b99fe0e1e
Don't emit a guard at the end of CALLNAME and friends since the values CALLNAME puts on the stack (callee, this) are not seen by the interpreter yet, and thus get store killed. Instead, emit the guard in CALL, at which point the values are properly stacked.
2008-08-20 19:24:57 -07:00
shaver@mozilla.org
9817a0cd97
fix js_IsLoopExit to better handle for-in exits, r=mrbkap
2008-08-20 18:50:49 -07:00
Andreas Gal
49f0b4f6cb
Add a switch to disable the oracle, in which case we never demote any slots.
2008-08-20 16:10:56 -07:00
Andreas Gal
fdb661bd5e
Clear the oracle on every GC.
2008-08-20 16:01:56 -07:00
Andreas Gal
3e55729492
If at loop entry we expect a double, but at the loop tail we store an int because we suck the i2f into the side exit, we have to explicitly cast back up to double using i2f otherwise we won't be able to make sense of the value when we load it at the top of the next iteration.
2008-08-20 15:43:26 -07:00
Andreas Gal
98f67af6d0
Add a way to clear the oracle.
2008-08-20 15:18:43 -07:00
Andreas Gal
ccfca5d79f
Clear global slots and global type map when clearing the code cache.
2008-08-20 15:03:51 -07:00
Andreas Gal
f3972bfb84
Backed out changeset be63a51a0a3b. Breaks tofte if run from ./time.sh.
2008-08-20 14:57:49 -07:00
Brendan Eich
6bc330ef33
Don't disable property caching in eval (this requires GC'ing eval scripts). Do tolerate active Call object at head of scope chain in record_JSOP_CALLNAME, and go straight to the stack slots.
2008-08-20 14:04:37 -07:00
Andreas Gal
8df2fc18c8
When adjusting integers to doubles in tree calls start with the current stack frame only.
2008-08-20 12:47:39 -07:00
David Anderson
d6aef7cf26
Fixed some valgrind whinings (not actual bug fixes).
2008-08-20 12:17:31 -07:00
David Anderson
54cfb5e424
Fixed typo that broke the tree. Sorry!
2008-08-20 10:38:50 -07:00
shaver@mozilla.org
d1f7543e1b
default to non-verbose for DEBUG, set TRACEMONKEY=verbose in env to go verbose
2008-08-20 10:15:42 -07:00
David Anderson
738ce385fa
Merge.
2008-08-20 09:50:41 -07:00
Brendan Eich
deca6a9d50
Add str + obj concatenation for 3d-raytrace.js.
2008-08-20 02:35:17 -07:00
Andreas Gal
3a4d105a0f
If the outer tree has a slot in an integer register, but the inner tree expects a double, make sure to promote the value before calling the tree, otherwise the inner tree will read the value from the stack incorrectly.
2008-08-20 00:51:56 -07:00
Andreas Gal
fb3cf056d9
Correctly calculate tree call stack adjustment, even if we don't have any arguments in the current frame and hand pick some cleanups from the backed-out changeset.
2008-08-19 23:31:33 -07:00
Andreas Gal
39d9d53d83
Back out multi-trees. There is absolutely no way this will be debugged in time for tomorrow. Need a bit more gdb man-power.
2008-08-19 23:28:13 -07:00
Andreas Gal
55c1e3d15d
Always blacklist the first fragment in the peer list, and blacklist if we see a loop that is not type-stable.
2008-08-19 23:10:28 -07:00
Andreas Gal
25bf562fe6
Don't build trees inside tree calls if no suitable tree can be found (for now.)
2008-08-19 23:00:19 -07:00
Andreas Gal
4ee5ead209
Improved debug output.
2008-08-19 22:59:08 -07:00
David Anderson
69c4638355
Merge.
2008-08-19 22:33:43 -07:00
Andreas Gal
8569359e49
Add debugging facilities for typemaps and fix sp_adj calculation for tree calls.
2008-08-19 22:29:02 -07:00
David Anderson
9fa5be54af
Merge.
2008-08-19 21:32:22 -07:00
Andreas Gal
454c89d72b
Support multiple fragments with different entry maps per PC location.
2008-08-19 20:30:16 -07:00
David Anderson
9a5b3b4307
AMD64 trace-tests.js passes now (hacked in LIR_qcmov opcode, corrected builtin return types).
2008-08-19 17:19:19 -07:00
Andreas Gal
b3efe4fbeb
Clear out the list of global slots when the shape changes.
2008-08-19 13:34:20 -07:00
Andreas Gal
1ee261ca2e
Emit the tree call stack setup code before executing the inner tree, otherwise we might see stale references to instructions prior to the call that no longer guarantee certain types (since the inner tree overwrite them.)
2008-08-19 13:11:18 -07:00
Andreas Gal
1ded8ec0ad
Merge.
2008-08-19 11:32:52 -07:00
Andreas Gal
e79f9ed689
Intern globals per-context, not per tree, which allows transitioning between trees that use globals. The type of all global slots compiled code uses is expected to remain stable, otherwise the entire code cache has to be flushed. Changes to the shape of the global object also flush the code cache. Working for trace-tests.js, but fails math-partial-sum.js
2008-08-19 11:32:36 -07:00
David Anderson
ee4a887045
Merge.
2008-08-19 11:02:57 -07:00
Brendan Eich
c2229dbd5e
Back out changeset 1d0574db8320 (for 450997), it was based on a bum steer.
2008-08-18 22:18:33 -07:00
Andreas Gal
dfb94ccec1
Add a global type map to the monitor and rename slotList to globalSlots.
2008-08-18 21:08:35 -07:00
Brendan Eich
76380c598b
Add a final js_ReconstructStackDepth assert after setting fp->regs->sp in js_ExecuteTree's bail-out epilog.
2008-08-18 19:26:47 -07:00
Brendan Eich
87a6e416ef
1. Fix json2.js (it crashes later, another bug) to restore a non-empty stack for a call with missing args after a nested bail-out. This involves changing js_SynthesizeFrame to return the number of stack slots it reconstructed that map to native slots. It also means not counting the missing args in the spdist uint16 pushed onto the state.rp stack (this was the root of all over-counting evil, for this bug).
...
2. Use if (C) { do {...} while (C); } instead of if (C) { while (C) {...} } and other style fixage.
2008-08-18 18:36:05 -07:00
David Anderson
e503629dbe
Fixed AMD64 accidentally treating the shape as a 64-bit integer.
2008-08-18 17:36:24 -07:00
David Anderson
908183cf6f
Merge.
2008-08-18 16:38:32 -07:00
David Anderson
0f46c5d568
Fixed some bugs in the AMD64 port, still crash in 5 sunspider tests.
2008-08-18 16:23:37 -07:00
Brendan Eich
21b6e353e2
Cosmetic cleanup.
2008-08-18 15:32:17 -07:00
Andreas Gal
ad4bfa3b5b
Add a global slot list to the trace monitor.
2008-08-18 13:12:19 -07:00
David Anderson
751ef6d2af
Improved AMD64 nanojit compatibility.
2008-08-18 12:54:04 -07:00
Andreas Gal
afc782cbe2
Make sure to trasm the vmprivate of dependent trees as well.
2008-08-18 12:53:38 -07:00
Andreas Gal
f7aac4002e
Tighter fencepost in getTop, don't overshoot the stack (450997).
2008-08-18 10:09:24 -07:00
Andreas Gal
202a9de38b
Doh. Deep bailots are pretty rare and js_SynthesizeFrame already uses the fairly slow js_ReconstructStackDepth helper internally, so anotherone to avoid counting stack slots via spbase can't hurt.
2008-08-18 10:04:03 -07:00
Andreas Gal
e7645176d5
Merge.
2008-08-18 08:03:51 -07:00
Andreas Gal
535584eecb
Add a method to type map that captures missing slots (which have been lazily added to the slot list since the type map was created.)
2008-08-18 07:32:42 -07:00
Brendan Eich
4d8c1fda9b
Fix missing paren in comment; also s/scope/frame/ in same comment.
2008-08-17 12:04:33 -07:00
Andreas Gal
9f71dd226f
Extend the globalTypeMap import(), not in the caller, since we missed extending it in case of re-reading registers after a tree call.
2008-08-17 11:38:49 -07:00
Andreas Gal
f08e18fc14
If RecordTree is triggered, create a new peer fragment (or recycle and unused one) if we already have a tree for this location.
2008-08-17 00:49:59 -07:00
