Andreas Gal
8a7bb9c550
Smarter speculative demotion of numbers to integers and promotion of the trace seems to require actual doubles. If the number at entry looks like an int we make the slot an int and compile as such. If the loop-tail proves the slot to be a double, we recompile the trace. Currently such miss-speculation cannot be handled on secondary traces since we are currently unable to recompile the primary trace. Such secondary traces are blacklisted.
2008-07-27 02:15:17 -07:00
Andreas Gal
a4fc19765e
Don't try to demote slots on secondary traces (we have to recompile the primary trace as well for that, which we currently don't do.)
2008-07-25 19:30:33 -07:00
Andreas Gal
059abfa50f
Added sunspider tests to t/ for dvander.
2008-07-25 18:43:19 -07:00
Andreas Gal
812989bb8a
Eliminate EntryRegs. Its not safe to keep a reference to the entry SP since we might extend the tree from a different outer stack frame. Instead just store the entryStackDepth.
2008-07-25 18:22:15 -07:00
Andreas Gal
bbb1bc89c0
Print real recording point, not entryRegs and add an assert that makes sure the guard we come out of is associated with the tree we entered into.
2008-07-25 18:13:57 -07:00
Andreas Gal
24447ee8fb
Merge.
2008-07-25 16:54:14 -07:00
Andreas Gal
1b7930d39d
The VP engineering broke JSOP_NAME! We have to check whether the slot is actually interned and otherwise abort.
2008-07-25 16:51:42 -07:00
Brendan Eich
a8c6046d8c
Avoid JSOP_POPV in global scripts from load(), etc.
2008-07-26 01:23:12 +02:00
David Anderson
4560ed961f
Fixed entry typemap having wrong allocation size.
2008-07-25 13:30:52 -07:00
Andreas Gal
d952f192ba
Can't use lr->from. Seems to not get set in certain cases. Go figure. The nanojit tree code is really weird.
2008-07-25 02:29:36 -07:00
Andreas Gal
dee805421b
Factor out AttemptToGrowTree.
2008-07-25 02:00:02 -07:00
Andreas Gal
ecb79b1640
Determine ahead of time whether a side exit is a loop-terminating side exit instead of re-determining this at every side exit.
2008-07-25 01:44:40 -07:00
Andreas Gal
f3573b9940
Move trace activation code into js_ExecuteTree().
2008-07-25 01:20:40 -07:00
Andreas Gal
3999210c12
Make demotion threshold optional (0=off, 32=default). Turn off to debug the type assert bug.
2008-07-24 17:33:34 -07:00
Andreas Gal
e838b138b8
Don't demote slots in overly long traces (based on counting the number of guards, current threshold=32).
2008-07-24 16:25:18 -07:00
Andreas Gal
f0399ffea7
Cleanup iteration over all slots to use a single macro to avoid code duplication.
2008-07-24 14:51:14 -07:00
Andreas Gal
e7f92ecf31
Add shark support to md5.js directly, no longer around the recorder.
2008-07-24 11:08:37 -07:00
Andreas Gal
6df91a21a0
Make the interpreter notify us when frames are popped so we see he right cx->fp value. Fix return value tracking for JSOP_STOP.
2008-07-24 10:35:10 -07:00
Andreas Gal
5ae4ef819e
Backed out changeset a58e7ce6eb7f. stack(0) is wrong, if at all it should have been stack(-(1 + argc)). But anyway, its impossible to write through to the frame via set() at that point because leaveFrame() doesn't actually modify cx->fp. We need to wait for the interpreter to do so. Until then nativeFrameOffset produces incorrect values and the assert fails in set.
2008-07-24 10:09:51 -07:00
shaver@mozilla.org
b989e7b3e7
fix return-value handling for new stack layout
2008-07-24 00:31:37 -07:00
Andreas Gal
41e0593392
Maintain a separate frame for globals using InterpState->gp.
...
This will allow nested loops to access the globals interned by outer loops while at the same time sharing parts of the stack of the outer loop.
2008-07-23 23:18:02 -07:00
Andreas Gal
117bc4d9f9
Add a fast path for nativeFrameOffset that walks entire groups at a time, not slots. Check against the result calculated by the FORALL macro in DEBUG builds (for sanity).
2008-07-23 16:39:17 -07:00
Andreas Gal
4cc2061091
Merge. Make sure we clear out all slots from the tracker when leaving a frame.
2008-07-23 15:56:27 -07:00
Andreas Gal
df280d08d7
When addressing locations above the stack water level, use ->sp as base. Shaver accidently changed this to ->spbase when merging Igor's patch, so we were writing past the water mark and the stores were killed.
2008-07-23 15:48:15 -07:00
shaver@mozilla.org
c29aa4c3d2
tighten up some bounds, still not quite right
2008-07-23 11:42:52 -04:00
Andreas Gal
5dd9876515
Lazy fill the stackTracker as we lookup the nativeFrameOffset for values. Purge all entries related to the current frame when it is popped. This code could be further simplified if Tracker was a template.
2008-07-23 02:57:56 -07:00
shaver@mozilla.org
f0a625f5bc
merge from andreas
2008-07-23 01:08:23 -04:00
shaver@mozilla.org
6f5e58e443
merge (still not working, probably box/unbox dumbness)
2008-07-23 01:06:18 -04:00
shaver@mozilla.org
a93fbcc99d
merge (now compiling, untested)
2008-07-23 00:20:29 -04:00
Andreas Gal
c9cac5757f
Handle leaving of frames in one central place.
2008-07-22 20:16:12 -07:00
Andreas Gal
c6828d3504
Reduce HOTLOOP threshold to 2.
2008-07-22 18:31:44 -07:00
Andreas Gal
06365155c2
Avoid calculating native frame offset by tracking the last load/store for every slot and extracting the offset from there.
2008-07-22 17:54:04 -07:00
Andreas Gal
48b195e651
Avoid nativeFrameOffset calls in import().
2008-07-22 17:24:29 -07:00
Andreas Gal
14b0a0827d
Merge.
2008-07-21 19:38:03 -07:00
Andreas Gal
9b799905ed
Add a pattern to recognize safe_add and emit a simple add for it.
2008-07-21 19:37:43 -07:00
shaver@mozilla.org
e0b2e86c09
use nargs rather than argc, since bytecode only refers to [o-nargs) args
2008-07-21 22:31:08 -04:00
Andreas Gal
2d2bcf0fc8
Reverse the insanity that used to be tamarin's arg passing in ins2. Now it makes sense.
2008-07-21 17:59:42 -07:00
Andreas Gal
3b2a943332
Avoid overflow checks on integer adds when its safe to do so.
2008-07-21 17:50:17 -07:00
Andreas Gal
410c0b9a98
Eliminate entryFrame pointer (not safe to carry that around.)
2008-07-21 16:20:53 -07:00
Andreas Gal
b30f4c9be6
Use callDepth instead of entryFrame in the FORALL macro.
2008-07-21 16:19:38 -07:00
Andreas Gal
f1ef7f0175
Eliminate exit filter, write the type map directly in snapshot.
2008-07-21 15:09:19 -07:00
Andreas Gal
9495f9cb50
Sink stack-targeting type conversion in the set() path instead of in the ExitFilter.
2008-07-21 15:01:47 -07:00
Andreas Gal
a7cc4892f9
Use callDepth, not entryFrame in nativeFrameSlots().
2008-07-21 14:55:41 -07:00
Andreas Gal
bce8e13528
Remove excess parameters of verifyTypeStability.
2008-07-21 13:21:00 -07:00
Andreas Gal
a0cd046167
Track call depth in the trace recorder. This lays the groundwork to get rid of entryFrame.
2008-07-21 13:18:08 -07:00
shaver@mozilla.org
124244af4b
beginning of entryFrame removal
2008-07-21 15:57:02 -04:00
Andreas Gal
5067fc2638
Merge.
2008-07-21 12:44:03 -07:00
Andreas Gal
3f6f0d1ea5
Use ti instead of fi for TreeInfo pointers.
2008-07-21 12:43:51 -07:00
Andreas Gal
e9b2918575
Rename VMFragmentInfo to TreeInfo.
2008-07-21 12:41:43 -07:00
Andreas Gal
2f5e227963
Merge.
2008-07-20 17:36:47 -07:00
Andreas Gal
221f99adce
Style cleanup and eliminate redundant f2u(i2f|u2f) chains.
2008-07-20 17:36:11 -07:00
Brendan Eich
4a9828f126
Cite bug 445262 with a FIXME instead of omfgHack_ prefix ;-).
2008-07-20 16:21:45 -07:00
Brendan Eich
464b70d055
Tighten up js_IsLoopEdge and include JSOP_IFNEX.
2008-07-20 16:11:52 -07:00
Brendan Eich
c278ae56e0
Nit patrol.
2008-07-20 16:09:08 -07:00
Andreas Gal
e82ab0c21f
Simplify updating of the side exit target.
2008-07-20 14:56:04 -07:00
shaver@mozilla.org
368d1588bf
simpler loop exit check (backwards is a loop exit, forwards is not)
2008-07-20 17:54:16 -04:00
shaver@mozilla.org
7546f723f5
force Math to be interned, and defend against non-interned globals until upvar saves us all
2008-07-20 17:23:39 -04:00
Andreas Gal
0b74e54db8
Use exponential backoff when trying to extend trees.
2008-07-20 14:28:56 -07:00
Andreas Gal
a0e38ffd84
Merge.
2008-07-20 13:29:20 -07:00
Andreas Gal
d4d2117f7c
Don't emit boxed values, use the unboxed representation for constants.
2008-07-20 13:28:53 -07:00
Andreas Gal
71eb1da52f
Removed meaningless assert.
2008-07-20 13:25:36 -07:00
shaver@mozilla.org
cadc5891e5
don't pretend we can trace INCELEM and friends (fixes trace-test.js, ahem)
2008-07-20 16:21:43 -04:00
shaver@mozilla.org
ed1d0a8953
improve tracing/recording diagnostics
2008-07-20 15:42:19 -04:00
shaver@mozilla.org
d03348f92d
discard trace on global shape mismatch
2008-07-19 22:32:45 -07:00
Brendan Eich
5b009b1b49
Breathe, vertical-spacing, breathe!
2008-07-19 15:42:31 -07:00
shaver@mozilla.org
ef2e8305dc
Avoid tracking object properties, make incElem work.
2008-07-19 10:24:10 -07:00
shaver@mozilla.org
f34cd71b78
Implement DEFLOCALFUN for the fully-lightweight case.
...
Reset atoms base register on function entry and exit.
2008-07-19 06:29:56 -07:00
Brendan Eich
293e6bd4c4
Improve js_IsLoopExit (unannotated JSOP_GOTO is still ambiguous).
2008-07-19 10:52:24 -07:00
Brendan Eich
a26be1bd4b
Fix incProp to box and store (incElem still todo).
2008-07-19 10:24:28 -07:00
Andreas Gal
f2ad868fc0
Merge. hg sucks.
2008-07-19 00:40:54 -07:00
Andreas Gal
457c855d15
Don't demote fmul since it overflows in math-partial. We might try to demote it and then promote back if this happens (depends how aggressively we want so speculate). Speedup 6.6 for math-partial (since its sin/cos/pow heavy.)
2008-07-19 00:27:45 -07:00
Brendan Eich
d1b171e5e2
First stab at incops, plus unbox_jsval tag-masking guard fixes, plus misc. cleanup.
2008-07-19 00:15:22 -07:00
Andreas Gal
5495ccf586
Abort trace is STOP is leaving the recording context.
2008-07-18 19:15:49 -07:00
Andreas Gal
47ae25da20
Make entryFrame/entryRegs relative to the tree entry since we only adjust ip/sp/rp when exiting from the tree, not in between fragments.
2008-07-18 16:21:22 -07:00
Andreas Gal
a8b127d5af
Adjust ip/sp/rp in the interpreter recovery code, not on the trace.
2008-07-18 15:35:59 -07:00
Andreas Gal
1c9c67b6a1
Write all the debug output to stdout (we can't easily make everything go to stderr, so this way at least it all goes into the same place.)
2008-07-18 14:13:31 -07:00
Andreas Gal
89b50ed124
Rework some of David's changes. Make sure we don't modify the tree until we are ready to attach the new fragment. Share lirbufs between tree branches.
2008-07-18 14:11:01 -07:00
David Anderson
4770d4b3e0
Got side exits seemingly working, added if.js to trace-test.js
2008-07-18 10:33:17 -07:00
shaver@mozilla.org
61d58bad87
Handle more operand types in ifop().
2008-07-18 08:01:51 -07:00
shaver@mozilla.org
983e9cdb83
Better diagnostic info for trace abort.
2008-07-18 07:59:37 -07:00
shaver@mozilla.org
3eaeae50b5
lookup/find, but do not fill prop cache from recorder
2008-07-18 00:46:18 -07:00
shaver@mozilla.org
7078a99d21
When PROPERTY_CACHE_TEST misses, fall back to doing our own lookup via js_GetPropertyHelper.
2008-07-17 23:57:56 -07:00
Brendan Eich
84c0998916
Warning elimination.
2008-07-17 18:30:54 -07:00
Andreas Gal
f94180fbe2
Make sure we compile again if we reject a trace by decrementing hits.
2008-07-17 18:15:55 -07:00
Andreas Gal
f58160134a
Fix hot loop triggering.
2008-07-17 18:10:47 -07:00
shaver@mozilla.org
859ec6fbb6
can't easily find real names for globals, so just number them
2008-07-17 17:33:18 -07:00
Andreas Gal
dded69f2a3
Merge.
2008-07-17 17:13:41 -07:00
Andreas Gal
19ed222624
Disable trees for the time being.
2008-07-17 17:12:28 -07:00
Andreas Gal
43fbfbfe1e
Add if.js example for trees.
2008-07-17 16:38:09 -07:00
Andreas Gal
dbf2dc54bd
Add the beginning of support for trees.
2008-07-17 16:20:13 -07:00
shaver@mozilla.org
3db81baa04
Label globals and locals with their source-names in trace.
...
Show line number on abort.
2008-07-17 15:09:48 -07:00
Andreas Gal
f91ca871ac
Add js_IsLoopExit that indicates whether a side exit is likely to continue the loop or not.
2008-07-17 14:58:09 -07:00
Andreas Gal
3df09fb85c
Cleanup recorder activation, use a single hot loop threshold (10).
2008-07-17 13:42:58 -07:00
Andreas Gal
954c5830e1
Added statistics and a js_DestroyJIT hook. Also use debug_only and not verbose_only as #ifdef DEBUG shortcut.
2008-07-17 10:22:40 -07:00
Andreas Gal
bfb3e457f1
Fix signature of insCall in the FuncFilter and remove f2i(i2f) elimination since we shouldn't really ever hit it.
2008-07-17 03:24:17 -07:00
Andreas Gal
e12047397c
Move up addName to make sure it gets inline in non-debug mode.
2008-07-17 02:02:48 -07:00
shaver
87eacb1139
Add setprop and tests for setprop and cleanup labling code for debug mode (m=gal).
2008-07-17 02:00:23 -07:00
Andreas Gal
9d6fa41738
Cleanup trace abort handling.
2008-07-17 01:29:41 -07:00
Andreas Gal
3fa6978605
Merge.
2008-07-17 00:48:30 -07:00
Andreas Gal
3cd2b84ec4
Fixed ASSERT JS_DOUBLE != JS_VOID issue. If the root fragment fails to compile, trash the type-map and re-capture it when we try to re-compile.
2008-07-17 00:48:10 -07:00
shaver@mozilla.org
acde4fceb0
JSOP_LENGTH
2008-07-16 23:53:14 -07:00
Andreas Gal
6cfd498b7c
Increase code cache size.
2008-07-16 23:47:57 -07:00
Andreas Gal
ad6b349616
Merge.
2008-07-16 23:39:12 -07:00
Andreas Gal
e2a0438e22
Fix f2i(i2f) filter (broken since the last TT sync, argument order changed from right to left to left to right).
2008-07-16 23:38:39 -07:00
Brendan Eich
13d6469f9b
Prefer getarg;length, etc. to getargprop.
2008-07-16 23:33:04 -07:00
Andreas Gal
83ac88d4c7
Don't clear out arguments, clear out variables at method entry.
2008-07-16 23:04:50 -07:00
Andreas Gal
b243a8350a
Better naming of the after_JSOP_CALL event (EnterFrame).
2008-07-16 23:00:59 -07:00
Andreas Gal
2442099090
Notify the recorder when a new frame was created by JSOP_CALL.
2008-07-16 22:58:06 -07:00
Andreas Gal
2379d6dcde
Initialize arguments to void and write JSVAL_HOLE's value onto the native stack, not its boxed representation.
2008-07-16 22:25:56 -07:00
Andreas Gal
df5005a322
Hand in typemap the recorder is supposed to use (preparation for side exit compilation and trees).
2008-07-16 21:49:42 -07:00
Andreas Gal
def4afc614
Move the allocation of the LIR buffer and the FragmentInfo data structure creation (which includes typemap and interned global calculation) into loopedge.
2008-07-16 21:41:03 -07:00
Andreas Gal
af95344998
Make JSOP_CALLNAME tell us why it doesn't like to trace.
2008-07-16 19:00:14 -07:00
Andreas Gal
74f767e62b
Fix generation of VOID immediates and initialize rval after the new frame was setup by CALL.
2008-07-16 17:44:44 -07:00
Andreas Gal
db7e41ed40
Add generic recorder hooks that are called before and after all opcodes as we trace. We might want to instead just move to pre_OP and post_OP.
2008-07-16 17:29:08 -07:00
Andreas Gal
c041bad939
Don't intern global function objects onto the native frame since we rarely ever need them anyway (we call them via the property cache).
2008-07-16 15:48:51 -07:00
Andreas Gal
7f119dc35a
Enable CALLNAME and add test case for it so we can call global functions.
2008-07-16 15:40:35 -07:00
Andreas Gal
c7714a3ad5
Make sure this/argv[-1] is set fpr CALLVAR and CALLARG.
2008-07-16 15:13:27 -07:00
Andreas Gal
04ea9e0e09
Enable inlining for calls. Deep bailouts (from within side exits) are generated but are not allowed to be ever taken (for now).
2008-07-16 15:01:55 -07:00
Andreas Gal
43b7ba1be4
Don't use a builtin for this. For functions its interned in the native frame. For global this we read from fp->thisp. Restore jsinterp.cpp and jsinterp.h (no longer need COMPUTE_THIS exposed).
2008-07-16 14:36:50 -07:00
David Anderson
5dca2ad552
Synced nanojit with TT tip.
2008-07-16 14:21:31 -07:00
shaver@mozilla.org
ef9a5dc216
fix GETXPROP; thanks to brendan for playing chewbacca
2008-07-16 13:10:17 -04:00
shaver@mozilla.org
f8c80ec490
CALLARG, CALLVAR
2008-07-15 23:37:57 -04:00
shaver@mozilla.org
4937938676
fix scalpel left in nativeFrameSlots from aborted thisp addition to frame
2008-07-15 23:20:53 -04:00
shaver@mozilla.org
21a3ec3485
implement JSOP_MOD
2008-07-15 23:19:29 -04:00
shaver@mozilla.org
3a4c348313
Trace JSOP_THIS and JSOP_THISPROP.
...
Rename getprop to get_prop, just because.
Extract COMPUTE_THIS to jsinterp.h as JS_COMPUTE_THIS for reuse in tracer.
2008-07-15 21:37:00 -04:00
Andreas Gal
21490a6e41
Fix uncomplete range check for slot numbers of interned globals.
2008-07-15 17:10:52 -07:00
Andreas Gal
cd739bdb35
Strict aliasing cleanup.
2008-07-15 16:14:00 -07:00
Andreas Gal
3aeb53f580
Add a few consts to char* pointers to pacify gcc 4.2.
2008-07-15 16:04:08 -07:00
shaver@mozilla.org
3693365174
rename getpropfromval, it burns mine eyes
2008-07-15 18:29:42 -04:00
Andreas Gal
1b783425aa
Assign blame where blame is due.
2008-07-15 15:05:16 -07:00
Andreas Gal
cbf4813605
Merge.
2008-07-15 13:14:05 -07:00
Andreas Gal
a81b83090e
Sync with TT.
2008-07-15 13:06:05 -07:00
shaver@mozilla.org
919139d11b
initialize dslots_ins
2008-07-15 13:40:11 -04:00
Brendan Eich
3ccdb62685
Propagate error exceptions from TraceRecorder ctor; fiddle/trim space.
2008-07-15 10:26:15 -07:00
shaver@mozilla.org
7d7ece76a8
GETVARPROP, GETARGPROP, GETXPROP
2008-07-15 13:17:51 -04:00
shaver@mozilla.org
409bb7f29d
more ABORT_TRACE instrumentation (some should be asserts?)
2008-07-15 09:07:54 -04:00
Andreas Gal
5848fd6694
Generate a list of interned global slots (gslots) when we process the tree header. This list is then used whenever we iterate over the native frame. This is faster and safer than looking up properties in the global object every time.
2008-07-15 01:53:39 -07:00
Andreas Gal
84608819b9
Remove state exposing accessor functions from recorder and instead hand in that state via the constructor into ExitFilter.
2008-07-15 01:27:14 -07:00
Andreas Gal
1693476391
Guard in FragmentInfo on the shape of the global object. Don't check for the shape of the global object on the trace.
2008-07-15 01:08:13 -07:00
Andreas Gal
bac6ed12f4
Store list of interned global slots in struct FragmentInfo
2008-07-15 01:03:49 -07:00
Andreas Gal
9f761173f0
Added a callstack that will track the pc of the call that caused a function call to be inlined. This is necessary to recover from deep side exits inside inline functions. The callstack is subject to store elimination, so unnecessary stores to the stack will go dead automatically during compilation (i.e. if we inline a function that doesn't have side exits.)
2008-07-14 19:12:50 -07:00
Andreas Gal
df84a21add
Fix warnings.
2008-07-14 17:52:38 -07:00
Andreas Gal
e59427d0f7
Use JS_GetGlobalForObject to get the global object. Walking back the call chain is not equivalent and not safe.
2008-07-14 16:40:38 -07:00
shaver@mozilla.org
0180d69fd8
fix our aliasing idiocy by extending jsdpun, add Math.sqrt, add strict-aliasing to our Makefile.ref flags
2008-07-14 18:22:05 -04:00
Brendan Eich
815146e0e4
Non-null prop from js_LookupProperty means found, so must unlock obj2 (now pobj) in all such cases.
2008-07-14 00:13:31 -07:00
Andreas Gal
4a598274f5
Clean up shaver's cleanup.
2008-07-13 22:02:42 -07:00
Andreas Gal
28e3ed165f
Fixed printing of integer incoming values in DEBUG mode.
2008-07-13 21:53:35 -07:00
shaver@mozilla.org
0f64113822
merge, and fix locking and logic for FORALL_SLOTS
2008-07-14 00:51:43 -04:00
shaver@mozilla.org
4779e69591
make JSOP_CALL builtin specialization data-driven
2008-07-14 00:28:31 -04:00
Andreas Gal
8750e39409
Skip properties that were not found by LookupProperty.
2008-07-13 21:17:56 -07:00
Andreas Gal
30c7f0493e
Reserve space for every global property that the current script has an atom for instead of trying to rely on ngvars.
2008-07-13 21:14:34 -07:00
Brendan Eich
044d6f9c93
Merge.
2008-07-12 23:44:23 -07:00
shaver@mozilla.org
46b4c7def2
fix the secondary map-native guard
2008-07-13 00:09:24 -04:00
Brendan Eich
ba3c776fc2
js_ for library-extern names like math_sin.
2008-07-12 12:35:36 -07:00
shaver@mozilla.org
6b30961e58
Guard against subzero array indices
2008-07-12 09:04:58 -04:00
Andreas Gal
3ccfc13098
Demote fneg to neg if input is known to be an integer (untested).
2008-07-12 00:34:10 -07:00
Andreas Gal
223f4f23f7
Merge.
2008-07-12 00:28:28 -07:00
Andreas Gal
b5e654f120
5 hours of debugging, and 9 keystrokes to fix it. That was one expensive bug. shaver's reduced fannkuch example works now. I think independently of this one we don't check properly for index underflow in dense arrays. Shaver is going to have to take a look at that (this fix makes fannkuch indexes not become negative, but doesn't explain why we crash so hard if they do become negative).
2008-07-12 00:27:21 -07:00
shaver@mozilla.org
b7dac1100d
add JSOP_NEG and tests
2008-07-12 00:20:48 -04:00
shaver@mozilla.org
fd5ac5b844
add Math.cos and Math.pow to the specialized-call party, and add tests
2008-07-12 00:03:33 -04:00
shaver@mozilla.org
d180a6755d
I will remember that stacked values are not boxed.
...
I will remember that stacked values are not boxed.
I will remember that stacked values are not boxed.
I will remember that stacked values are not boxed.
2008-07-11 23:04:29 -04:00
shaver@mozilla.org
9c69964575
Specialized tracing of Math.sin, as a proof of concept. Doesn't quite work due to regalloc mismatch, but close!
2008-07-11 22:59:09 -04:00
shaver@mozilla.org
8ffe777b5a
Incomplete stab at CALLPROP, added ABORT_TRACE for better diagnostics, make math_sin non-static in preparation for specializing call.
2008-07-11 20:59:10 -04:00
Andreas Gal
3e39ec9d3b
Print meaningful filename/line-number info for trace entry/exit.
2008-07-11 13:57:16 -07:00
Andreas Gal
8c6dbda691
Report the source location when recording a trace.
2008-07-11 13:45:38 -07:00
Andreas Gal
b222836a73
Don't demote u2f conversions and sink the type cast into the side exit type map, because this loses the sign bit for unsigned values. We could fix this by adding an explicit unsigned type to the map, but for now I think we should stick to int/double only since there is the risk of fanning out trees. Crypto doesn't seem to use ush all that much so we should be ok performance-wise.
2008-07-11 13:18:29 -07:00
Andreas Gal
cd0b213cec
Poking around in the arm code, trying to make it not die miserabily with BUILD_OPT=1.
2008-07-11 00:46:44 -07:00
Andreas Gal
3f31810cee
Still trying to get the ARM register updating right.
2008-07-11 00:31:48 -07:00
Andreas Gal
5d93de0225
Trying to make arm work. Flying blind here.
2008-07-10 22:35:06 -07:00
Andreas Gal
bde4b2e588
Adjust sp/ip for ARM. Very useful when trying to run on ARM.
2008-07-10 22:29:06 -07:00
shaver@mozilla.org
6bb689e6c3
some ARM stuff
2008-07-11 01:27:06 -04:00
Andreas Gal
fe9354b377
Merge (no, really, can someone teach Mercurial to not do this?).
2008-07-10 22:12:04 -07:00
Andreas Gal
41ded9b128
Output the value if we can't enter a trace because of a type mismatch.
2008-07-10 22:10:02 -07:00
Brendan Eich
a17263a036
1. Fix !JS_THREADED_INTERP bugs in BRANCH and recording switch case generation.
...
2. Rename TraceRecorder op methods to record_JSOP_xxx to avoid shadowing JSOP_*
enumerators used inside jsopcode.h macros -- death to ::JSOP_*!
3. Added atoms TraceRecorder member for giant literal pool scripts.
4. Condition assert in TraceRecorder::test_property_cache_direct_slot to avoid
false positive when executing function code accessing a global.
2008-07-10 21:55:09 -07:00
Andreas Gal
cc38ec5e21
Don't crash when expecting an int32 as double box in unbox but getting something else (and better debug output).
2008-07-10 21:23:32 -07:00
Andreas Gal
4cba5853d4
Merge.
2008-07-10 20:35:40 -07:00
Andreas Gal
4e34cfc559
IFEQ and IFNE are identical for us. We just expect the same boolean on the stack and side exit if not.
2008-07-10 20:35:19 -07:00
shaver@mozilla.org
66b22a063c
use a type that windows knows about for offset computation
2008-07-10 22:17:41 -04:00
shaver@mozilla.org
795a6bbef1
try to get alloca on Windows; remind me to strip these flailing commits before we merge
2008-07-10 22:10:48 -04:00
shaver@mozilla.org
03ad7cdff2
my turn to merge; had to happen eventually
2008-07-10 21:44:41 -04:00
Andreas Gal
ca4e6419ee
shaver's favorite operating system of choice doesn't like templates, so de-template tracker since we use it with LInsp only anyway.
2008-07-10 18:42:04 -07:00
shaver@mozilla.org
0695011b26
try to make alloca work for Windows
2008-07-10 21:37:49 -04:00
David Anderson
1bfa0a2b0e
Actually fixed ifeq/ifne fusion now.
2008-07-10 17:54:19 -07:00
Andreas Gal
4a0541e2ae
nanojit doesn't support loads with non-constant offsets so don't do that
2008-07-10 17:41:24 -07:00
Andreas Gal
16eda7e83e
Merge.
2008-07-10 17:05:13 -07:00
Andreas Gal
e6cb4750e2
Steal the reference to cx from the BoxDouble call instead of observing the load.
2008-07-10 17:03:59 -07:00
Andreas Gal
4ad95932a2
Fixed the setelem a[i] bug.
2008-07-10 16:55:37 -07:00
shaver@mozilla.org
60c7caba17
Unlike software developers, g++ doesn't like variable-sized arrays. Have some alloca!
2008-07-10 16:48:42 -04:00
David Anderson
8282c44200
Fixed guarding of eq+ifeq/ifne fusions and enabled tracing JSOP_GOTO (no-op)
2008-07-10 13:35:17 -07:00
Andreas Gal
1b8705e646
Root all strings and objects first when unboxing. Then box values that might trigger the GC (doubles/ints). This probably needs some performance tuning over time.
2008-07-10 13:24:49 -07:00
Andreas Gal
222a958b75
Merge.
2008-07-10 09:24:59 -07:00
Andreas Gal
ed0cfa6ce8
Don't concede an inch to ISO C++. Substract the size of array[1] from the overall struct size when allocating.
2008-07-10 09:22:01 -07:00
shaver@mozilla.org
0b4c4a7782
fix compilation, but possibly not logic, of shared-object defense
2008-07-10 08:05:25 -04:00
Andreas Gal
bdeea1b3cb
Prime the page cache during VM startup. This makes us eat the page cache allocation overhead there instead of during the first use. This is just a hotfix. We still need a rewrite of the page cache.
2008-07-09 23:47:17 -07:00
Andreas Gal
0a93e4dd69
Assert if no gvar is allocated for an undeclared global.
2008-07-09 18:51:08 -07:00
Andreas Gal
40a15bccc7
Merge.
2008-07-09 18:25:36 -07:00
Andreas Gal
ef76c5d316
varobj is not passed through the chain. Make sure to use global->varobj for gvar access.
2008-07-09 18:25:10 -07:00
Brendan Eich
31b11b4e90
Merge.
2008-07-09 17:58:19 -07:00
Andreas Gal
4617cc628c
Fixed name/setname code to just track the value move instead of touching memory since we have global variables in our native frame now. We have to fix the page cache issues before we can benchmark this.
2008-07-09 17:15:27 -07:00
Brendan Eich
8d88185086
Condition fp->arg*/*vars usage on fp->callee, not fp->down.
2008-07-09 17:10:42 -07:00
Andreas Gal
9a72a1d41d
Add verbose native stack frame printing.
2008-07-09 17:09:22 -07:00
Brendan Eich
ab59004c84
Memoize implicit gvars in the interpreter, on assignment (JSOP_BINDNAME/JSOP_SETNAME); fix recoder l/r operand order bug.
2008-07-09 16:37:31 -07:00
Brendan Eich
20a5a9ef31
1. Fix inc to address the right result stack slot; 2. Require via assertions that the interpreter memoize implicit gvars (patch to do that next; bitwise-and will assert until that lands).
2008-07-09 15:15:32 -07:00
Brendan Eich
2210ed4a21
Remove unnecessary JSOp cast.
2008-07-09 13:46:46 -07:00
Brendan Eich
d8cf234b33
Memoize global name gets and sets as if they refer to a declared gvar.
2008-07-09 11:42:31 -07:00
Andreas Gal
626026b70a
Merge.
2008-07-09 10:06:24 -07:00
Andreas Gal
202ec610da
Removed assert that compares nativeFrameSlots to nativeFrameOffset since nativeFrameOffset is undefined in case sp is at sp+depth, so the assert sometimes randomly bites. Also fixed typo in cmp(). trace-test.js compiles now.
2008-07-09 10:05:59 -07:00
Brendan Eich
cffa8d2279
Tweak a few interval tests, eliminate casts, space patrol.
2008-07-09 09:59:51 -07:00
Andreas Gal
05554d05b3
Stricter typing of the tracker code path which now only accepts jsval as suggested by Brendan.
2008-07-09 00:29:23 -07:00
Andreas Gal
19a2dcb6f5
Fixed bug in nativeFrameSlots and use an assert to compare the result with nativeFrameOffset (which is slower, but more precise).
2008-07-09 00:26:01 -07:00
Andreas Gal
1c24e264ad
Removed numMapEntries since typeMap is not really part of the SideExit struct yet anyway.
2008-07-09 00:13:35 -07:00
Andreas Gal
d6b834b98a
Move builtin[] table in jsbuiltin.cpp
2008-07-08 23:48:06 -07:00
Andreas Gal
5ee8b4abd7
Housekeeping. Remove a few warnings during BUILT_OPT and spelling in comments.
2008-07-08 22:40:07 -07:00
Andreas Gal
216f33a0bb
sp_adj is now maintained in bytes, not words, so +8 is now the magic value to point to the top of the stack. All stores above that watermark are eliminated by StackFilter in nanojit.
2008-07-08 21:03:30 -07:00
Andreas Gal
c9ee018202
Merge.
2008-07-08 20:20:24 -07:00
shaver@mozilla.org
815f787a8a
merge
2008-07-08 23:14:47 -04:00
Andreas Gal
ade00cfeb0
Removed bogus debug printfs.
2008-07-08 20:09:15 -07:00
Andreas Gal
795edf3334
Fixed frame walking FORALL_PENDING_FRAME_SLOTS and add 4 to sp in getTop (hotfix, not the right way to do it.)
2008-07-08 20:06:54 -07:00
shaver@mozilla.org
4fcad3cb1e
rval tracking and JSOP_POPV
2008-07-08 22:21:25 -04:00
Andreas Gal
4d84965bb8
Merge.
2008-07-08 19:17:15 -07:00
Andreas Gal
8e439abde3
Use more efficient address mode for LIR_load if possible and removed bogus printf.
2008-07-08 19:16:56 -07:00
shaver@mozilla.org
ff2b728cfd
protect printf with DEBUG
2008-07-08 21:22:46 -04:00
Brendan Eich
6a2a47ede9
Merge.
2008-07-08 18:14:14 -07:00
Brendan Eich
654d46135f
Single-ended interval tests, plus space patrol.
2008-07-08 18:12:22 -07:00
shaver@mozilla.org
46bdd14d57
Implement JSOP_SETNAME for globals.
2008-07-08 21:07:27 -04:00
David Anderson
daa8ef943b
Merge.
2008-07-08 17:20:11 -07:00
David Anderson
fcf8410125
Builds against latest nanojit; merged VMSideExitInfo into SideExit
2008-07-08 17:16:51 -07:00
Andreas Gal
444ffb8230
Fixed argv[n] and vars[n], should be [0] of course.
2008-07-08 17:14:01 -07:00
Andreas Gal
27628a1abf
Add dummy vpname/vpnum parameters when not compiling in DEBUG mode.
2008-07-08 16:46:54 -07:00
Andreas Gal
4550dd77f0
Introduce a generic stack frame walking macro. This eliminates the 6 (hg pull) redundant blocks of code that all walked the stack in 6 different ways, each with their individuals quirks.
2008-07-08 16:29:23 -07:00
Andreas Gal
9a4e0a6467
Don't check the types for invalid global slots in checkType. Proper gvar lookup in boxing path. We have to unify the stack frame traversal and enumeration code. I fixed this bug before, just in a different place. 11x speedup for gvar.js
2008-07-08 01:30:29 -07:00
Andreas Gal
c1491b01c2
Properly display the 'any' type in the side exit map printout.
2008-07-08 01:14:02 -07:00
Andreas Gal
0afa601949
Cleanup definition of tracker (don't use LInsp, use T, its a template now).
2008-07-08 00:46:53 -07:00
Andreas Gal
ac0b138197
Hack: demotable stores have to be flagged as int in the exit typemap. This needs cleanup.
2008-07-07 22:41:58 -07:00
Andreas Gal
64bf4baa1c
Don't expect args and vars in a top-level frame that is mapped to the native frame.
2008-07-07 22:20:27 -07:00
Andreas Gal
1e0a9a8363
Merge.
2008-07-07 22:00:25 -07:00
Andreas Gal
a6d30ed7c2
Fixed native frame offset calculation for globals.
2008-07-07 22:00:07 -07:00
shaver@mozilla.org
8fa1b5215f
fix type checking of gvars
2008-07-08 00:19:20 -04:00
shaver@mozilla.org
3f51971710
more detailed type-instability diagnostics
2008-07-07 23:51:56 -04:00
Andreas Gal
205ab22204
We urgently need shaver's magic native stack/typeframe iterator. Fixed exit map builder to skip argv/vars for global scope.
2008-07-07 19:29:54 -07:00
Andreas Gal
2f939b51d1
Fixed increment for DECGVAR.
2008-07-07 19:27:36 -07:00
Andreas Gal
a0c960adee
Bugfixes to the native stack and typemap handling code.
2008-07-07 19:20:40 -07:00
Andreas Gal
56397d3c0e
Attempt at cleaning up the typemap/stack frame iteration code.
2008-07-07 18:35:38 -07:00
Andreas Gal
6aaa63a172
ngvars and the slot index are not related, so don't rely on ngvars < JS_INITIAL_NSLOTS to decide whether dslots exists
2008-07-07 18:08:28 -07:00
Andreas Gal
0fe5682c00
Make gvar ops use the global varobj values we unbox onto the native stack. The stack offset handling needs more work.
2008-07-07 17:21:54 -07:00
Andreas Gal
a4f454ae32
We shouldn't read globals onto the stack if the global object is used by another context so we just refuse to execute the trace in this case.
2008-07-07 15:23:48 -07:00
Andreas Gal
1d36b97516
Unbox global variables onto the native stack and re-box them. This has certain synchronization implications but will make top-level scripts quite a bit faster.
2008-07-07 15:15:56 -07:00
Andreas Gal
4da1b65be4
Don't try to demote a slot we already decided we can't demote. Also, make sure we only demote additions with constants where the constant fits into an int.
2008-07-07 12:58:28 -07:00
Brendan Eich
affcda00d2
Merge.
2008-07-07 02:56:52 -07:00
Brendan Eich
175ce14f5b
Indentation and comment wrapping (at 79, 99 looks too long and most comments still avoid going past 80+/-).
2008-07-07 02:55:03 -07:00
Andreas Gal
ef4a0f7533
Merge.
2008-07-07 02:51:36 -07:00
Andreas Gal
e4f8fa2f97
Sync nanojit with tamarin-tracing tip.
2008-07-07 02:47:40 -07:00
Brendan Eich
6d65efee60
Use jsopcode.tbl for tracer JSOP_* method decls; trim trailing whitespace.
2008-07-07 02:21:04 -07:00
Andreas Gal
e15355bc04
Cleanup the native frame -> interpreter frame boxing code. We sometimes end up with numbers on our stack in double format that are really ints. We have to detect those and properly store them as in on the interpreter stack.
2008-07-07 01:37:40 -07:00
Andreas Gal
a38e8a96c3
Add support to demote stores of constants that are currently represented as float but are really integers. bitwise.js is now compiled complete fp-casts free.
2008-07-07 01:05:53 -07:00
Andreas Gal
efa3309778
Demote floating point comparisons even if one side is constant (bug in nanojit, filed as #443884 against tamarin). Also demote add/sub/mul using the integer overflow detection side exit code that Ed adopted upstream. Tight loops (bitwise) are not emitted completely as integer code after the initial compilation using doubles triggered a speculative demotion of the context slots.
2008-07-07 00:43:40 -07:00
Andreas Gal
8ce0de1ad7
Sink type conversions into the side exit by updating the map and seeing though the cast in the store (in ExitFilter). Add a whole bunch of asserts on the varios speculative type states to make sure we get the state machine right. Import speculated integer values as integers into the trace (indicate in entry map).
2008-07-07 00:10:22 -07:00
Andreas Gal
34c2283213
Added loop-tail driven context slot type demotion. Without all the buzzwords, what this essentially means is that we detect if the last value that the trace leaves in a slot (which is the value that flows along the backedge back to the loop header) is known to originate from an integer value via i2f, we flag the slot as type integer and recompile the trace. We do this where type stability is certain (result of an and, i.e.) or where its very likely (++operator). If the speculation fails, the same analysis flags the slot as blocked, which means it will always be double. The hope is that this analysis converges quickly (1-2 recompilations tops).
2008-07-06 22:35:19 -07:00
Andreas Gal
d6172e525e
Make sure we don't get the argument order in BoxDouble wrong again.
2008-07-06 20:27:50 -07:00
Andreas Gal
efee2381b5
Make typemap uint8_t and fix order of arguments in call to BoxDouble.
2008-07-06 20:25:49 -07:00
shaver@mozilla.org
9ed160ad12
remove outdated guard (now inferred via filter as needed)
2008-07-06 23:11:59 -04:00
shaver@mozilla.org
2c69d44f6f
we only speak double around here now
2008-07-06 23:01:58 -04:00
shaver@mozilla.org
10d122ba09
remove bogus NOT_REACHED
2008-07-06 22:39:02 -04:00
Andreas Gal
312d176937
Strength reduce i2f(doubleToInt32(x)) and u2f(doubleToUint32(x)) to x. This eliminate most of the on-trace overhead, but we still need type peeling of loop variable into int to win back the performance loss casting introduced.
2008-07-06 17:43:08 -07:00
Andreas Gal
bd6deef4fc
Reduce redundant doubleToInt32(i2f(x)) chains to simply x.
2008-07-06 17:32:21 -07:00
Andreas Gal
5ec69f76a7
Strength reduce BoxDouble(i2f(x)) to BoxInt32(x). Make sure loop indexes are actually integers. This guard will be eliminated further down in the pipeline if we decided to peel the type of the loop variable (index) down to integer.
2008-07-06 16:38:54 -07:00
Andreas Gal
a990d4a2e6
Since numbers are now always represented by doubles at the recorder/type level, using BoxDouble and UnboxDouble only when moving numbers from or to memory. A filter will then turn f2i(UnboxDouble) into UnboxInt where appropriate.
2008-07-06 16:18:45 -07:00
Andreas Gal
aeeff83beb
Move the type level from int/double to number. All traces start out as double in all slots, and denote and promote to/from int as needed. The FuncFilter optimizes on-trace casting and elininates redundant f->i-> chains. More optimization needed on this of course, and this code is now a bit slower than the previous integer-register use. However, this does solve the q += 2.5 issues. The heap access code does not properly cast yet and is likely unstable.
2008-07-06 15:55:04 -07:00
shaver@mozilla.org
b28afaa80d
Trace GETGVAR and INCGVAR, though not yet correctly.
2008-07-06 16:59:59 -04:00
Andreas Gal
d30b7d1c8a
Add unary and binary helpers that automatically demote and promote when dealing with integer operations (not used yet, need loop typemap peeling in place first.)
2008-07-06 13:16:34 -07:00
Andreas Gal
4e1aa1d977
JSVAL_IS_BOOLEAN does what isTrueOrFalse was trying to do.
2008-07-06 12:16:55 -07:00
Andreas Gal
5816dee707
Fix template, should use typename, not class.
2008-07-06 12:03:51 -07:00
Andreas Gal
1343901446
Add ExitFilter, which builds side exit typemaps for us. This had to be moved into a filter, because this has to happen after we eliminate redundant i2f-f2i chains as we can sink those into the side exit by simply storing the unpromoted/undemoted value and just flip the type in the exit map.
2008-07-06 11:48:41 -07:00
Andreas Gal
84ec537a0c
Merge.
2008-07-06 11:19:06 -07:00
Andreas Gal
9501bcd749
Added doubleToUint32 builtin and make tracker a template.
2008-07-06 10:59:55 -07:00
shaver@mozilla.org
d881a494aa
fix builtin_UnboxInt32 signature and name shape_ins for debugging
2008-07-06 13:58:59 -04:00
shaver@mozilla.org
45b3826103
Trace JSOP_NAME, and refactor out jsval unboxing.
2008-07-06 13:38:55 -04:00
shaver@mozilla.org
06a27ca367
begin on JSOP_NAME, refactor stobj_get_slot to chain better
2008-07-06 12:15:55 -04:00
shaver@mozilla.org
f463dc6891
actually record at HOTLOOP1 (fencepost)
2008-07-06 09:56:25 -04:00
shaver@mozilla.org
7f5f2ae1c8
some remaining low-hanging ops
2008-07-06 09:52:50 -04:00
shaver@mozilla.org
38235e2eec
move JSOP_DOUBLE impl to, er, JSOP_DOUBLE (wtf?)
2008-07-06 09:33:13 -04:00
shaver@mozilla.org
2d307b4b59
Implement JSOP_DOUBLE.
2008-07-06 07:28:36 -04:00
Andreas Gal
100f4ea5e0
Added support for semi-stable loop variables. Compiling for(...) q += 2.5; is ridiculously difficult because it flip-flops between int and double. Add support to promote integer values to doubles at the loop tail if at loop entry we expect a double. Since this isn't possible the other way around, we have to get luck that we catch a path into the loop where q is already double. For this we add 3 trigger points (10, 13, 37). We will try three times to record a trace at those iteration counts of a loop. If none succeed the loop is blacklisted. This probably needs more tuning down the road.
2008-07-05 23:21:53 -07:00
Andreas Gal
41efb6c488
Merge and rename jsIf to ifop.
2008-07-05 22:23:34 -07:00
Andreas Gal
a48c36fae3
Added FuncFilter from tamarin core (not part of nanojit yet since its slightly VM dependant).
2008-07-05 22:20:35 -07:00
Andreas Gal
0b28897ad3
Add helpers for i->f and f->i conversion. The f->i path goes via builtin calls.
2008-07-05 22:04:58 -07:00
Andreas Gal
68359155b9
Add type primitives for numbers (isNumber and asNumber).
2008-07-05 21:51:26 -07:00
shaver@mozilla.org
a268883f3c
Implement JSOP_IFEQ, JSOP_IFNE, JSOP_DUP, JSOP_DUP2.
2008-07-05 23:47:24 -04:00
Andreas Gal
075e3f00a6
Add casting for objects and cleanup casting code.
2008-07-05 20:30:11 -07:00
shaver@mozilla.org
501239ca11
let JSOP_GETELEM handle boolean values too
2008-07-05 23:04:42 -04:00
shaver@mozilla.org
26cc9c9e54
Handle boolean lval in SETELEM; now runs access-nsieve unmodified.
2008-07-05 22:39:34 -04:00
Andreas Gal
0245a6ba7c
Believe it or not NEG can actually overflow the int32 range, so add an overflow bailout. This will only trap for -0x80000000.
2008-07-05 19:25:38 -07:00
Andreas Gal
54a349ce1f
Merge.
2008-07-05 19:18:14 -07:00
Andreas Gal
48e459ba73
Added blacklisting of recording points where we failed to complete a trace (overly aggressive at this point, needs tuning, we want to try several times for each point.)
2008-07-05 19:15:00 -07:00
shaver@mozilla.org
1c4f82c592
Allow hole-filling JSOP_SETELEM to remain on trace.
2008-07-05 21:50:54 -04:00
Andreas Gal
04ae2a44d4
Fixed boolean boxing.
2008-07-05 18:36:40 -07:00
Andreas Gal
cf8e2a555b
Added boolean boxing/unboxing code.
2008-07-05 18:34:41 -07:00
Andreas Gal
8ac9e3ca57
Use default parameter in LSH/RSH/URSH to indicate we don't care for the overflow.
2008-07-05 18:18:06 -07:00
Andreas Gal
888d2f9aee
Add support for some misc opcodes including binary and arithmetic and/or/not.
2008-07-05 17:26:00 -07:00
Andreas Gal
810d54beac
Track trace entry/exit in debug mode and count cycles.
2008-07-05 17:10:58 -07:00
Andreas Gal
19152ffb1f
Signal error from the boxing/unboxing using magic cookies since gcc seems to very seriously object to the use of uint64 return values during a fastcall (horribly inefficient code).
2008-07-05 16:28:03 -07:00
Andreas Gal
05c4c4bb27
Help the branch predictor in the builtins.
2008-07-05 15:46:58 -07:00
shaver@mozilla.org
00b1baa734
use FASTCALL for builtins
2008-07-05 17:37:15 -04:00
Andreas Gal
16b463a6cb
Style nitpicking. Fix overlong lines.
2008-07-05 14:05:25 -07:00
Andreas Gal
d765c3ead2
Finish SETELEM for int and double values.
2008-07-05 14:00:32 -07:00
Andreas Gal
1e9c9451e0
Merge unboxing code into GETELEM code.
2008-07-05 13:44:48 -07:00
Andreas Gal
503bf13dac
Merge.
2008-07-05 11:46:08 -07:00
Andreas Gal
f9981c8870
Added box_int_jsval for the store path. Use only there.
2008-07-05 11:45:53 -07:00
Andreas Gal
ba7e66df92
Code generators to access object slots and native code callouts (builtins) for boxing doubles and ints. Ints have to be boxed through a native code helper on read and write (BoxInt32 and UnboxInt32), because we sometimes have to cast internally to double to store 32-bit values. We don't want a separate trace in this case, so we have to do this inline in a helper. Also a couple of modifications to shaver's code. Always make sure to check types (JSOP_NEG).
2008-07-05 11:35:55 -07:00
Brendan Eich
2e457ec7ab
Fix dumb bugs I just committed, use JS_NOT_REACHED.
2008-07-05 11:29:28 -07:00
Brendan Eich
ab0e34e2ae
Try to keep 64-bit portability via size_t instead of unsigned, jsuword for uintptr instead of long, etc.
2008-07-05 11:18:26 -07:00
Brendan Eich
e544b391b2
Spacing and comment nits picked while reading.
2008-07-05 10:41:35 -07:00
shaver@mozilla.org
ef07fe09b8
shift ops
2008-07-05 09:59:32 -04:00
shaver@mozilla.org
de59f88d9c
Beginning of SETELEM/GETELEM tracing for dense arrays. Needs computed-offset
...
insLoad before they'll work.
2008-07-05 09:47:59 -04:00
shaver@mozilla.org
f298721bf2
update to isInt
2008-07-05 09:43:00 -04:00
shaver@mozilla.org
13b4f67ef5
merge
2008-07-05 09:34:19 -04:00
shaver@mozilla.org
92e9aa44c0
[mq]: simple-ops
2008-07-05 09:15:33 -04:00
Andreas Gal
2df2214a9e
Introduce asInt and asDouble to check for the type of values based on the actual value since some 32-bit integers hide out in doubles.
2008-07-04 23:53:29 -07:00
Andreas Gal
cef5e9becc
Make trace-code 32-bit clean and extend interpreter state to carry the current context (cx). The recorder still has to record a 31-bit int path through the loop, but the emitted code is able to stay in the tree even if values bump over to 32-bit ints.
2008-07-04 16:24:42 -07:00
Andreas Gal
e35d351b28
Added support for SET_VAR again. bitwise.js working now.
2008-07-04 15:21:56 -07:00
Andreas Gal
ab1dc289ba
Fixed some stack handling and trace activation issues. We can run trace.js again.
2008-07-04 15:07:05 -07:00
Andreas Gal
5e9ca6ef0e
Cleanup of stack handling.
2008-07-04 13:23:42 -07:00
Andreas Gal
9b8f5f75a8
Added back support for the instructions required to compile trace.js.
2008-07-04 03:06:18 -07:00
Andreas Gal
1503932551
Properly switch tracer on and off depending on loopEdge and abort signaling from trace recorder recording functions.
2008-07-04 01:10:57 -07:00
Andreas Gal
efb72df02a
Re-integrate trace recording and trace activation infrastructure. Model more closely after Tamarin.
2008-07-04 00:51:30 -07:00
Andreas Gal
731614bfe9
Create a stub for each opcode in TraceRecorder and invoke them from the stubs that are pointed to by the recorder dispatch-table.
2008-07-03 23:57:57 -07:00
Andreas Gal
62d0c7bbfb
Switch back to mozilla-central jsinter.cpp. Lets try a different approach to attach the tracer to the interpreter.
2008-07-03 22:08:13 -07:00
Andreas Gal
8f4707fd29
Added dmod builtin and flag broken builtins for removal (shaver).
2008-07-02 22:05:30 -07:00
Andreas Gal
6eb6314e75
Support recording of ValueToECMAInt32, ValueToECMAUint32 and ValueToBoolean.
2008-07-02 21:48:42 -07:00
Andreas Gal
2cd7b6993b
VOIDs hang out in the BOOLEAN value space, so we have to treat them like a boolean on the trace. For ValuetoECMAInt32 all special boolean values except TRUE produce 0, and TRUE produces 1. Use a conditional move to implement this.
2008-07-02 19:11:07 -07:00
Andreas Gal
b137f2da3b
Introduce a VMSideExitInfo structure (holds the exit typemap only for now).
2008-07-02 18:06:24 -07:00
Andreas Gal
523ba92b3d
Use fragmento to track fragments and add proper trace activiation code.
2008-07-02 17:19:07 -07:00
Andreas Gal
47c6075eb2
Fix unstable trace rejection for optimized build.
2008-07-02 15:31:40 -07:00
Andreas Gal
7906238314
Fix type stability check.
2008-07-02 15:17:53 -07:00
Andreas Gal
6f3af33edf
Add type stability check for loop variables. Rename readstack to import.
2008-07-02 15:14:43 -07:00
Andreas Gal
e70272bfab
Removed attempt to imply ints into doubles on the fly. This can't work. Use proper float loads where needed and add some initial code for builtin functions.
2008-07-02 14:38:12 -07:00
Andreas Gal
8907cd354b
Flag ints as type INT in the incoming context if its merely a double holding an int that was too large to fit into jsval directly. Along the side exits make sure we can properly box oversized ints (by casting them to doubles). The fetch_int/store_int paths still need fixing to ensure that values are merely passed through (prim_copy) instead of explicit casting or calling to ValueToECMAInt32 etc.
2008-07-02 02:53:24 -07:00
Andreas Gal
08bd8d4245
Generate proper overflow detection code. Requires a trivial fix in nanojit (included, pending review to be pushed upstream).
2008-07-02 00:13:05 -07:00
Andreas Gal
7eba55b32b
Merge.
2008-07-01 19:44:24 -07:00
Andreas Gal
f3fba20c9c
Create type maps during trace entry and in each side exit and store them in the LIR using LIR_skip. Use these type maps during trace entry and exit.
2008-07-01 19:43:10 -07:00
Brendan Eich
9d20e12316
snprintf takes sizeof buffer.
2008-07-01 15:58:47 -07:00
Andreas Gal
39891951f4
Merge.
2008-07-01 13:36:33 -07:00
shaver@mozilla.org
da72d620b9
fix frame offset calculation for args, I think also non-entry frames
2008-07-01 08:11:49 -04:00
shaver@mozilla.org
50d599cb99
label arg/var/stack/sp/state for easier trace-reading
2008-07-01 08:06:02 -04:00
Andreas Gal
d3e7a57a68
Fix deallocation bug in the recorder. We will have to lift more code Tamarin's Interpreter to stabilize the recording.
2008-07-01 02:37:07 -07:00
Andreas Gal
84274f1b33
Save the VM registers into markRegs at opcode entry and restore that state when ending or aborting recording. The same info is pushed into sideexit and used by guards. With this change we can eliminate the hack in MONITOR_BRANCH that had to adjust the stack pointer depending on the opcode type, and it also allows us to trace through the recently added boolean guard opcode fusing. This improves trace code quality, since only the pre-conditional check state is saved. The guard restores into that state, so the trace code doesn't have to emit the value of the conditional evaluation onto the stack, saving a handful ops.
2008-06-30 18:31:29 -07:00
Andreas Gal
87396e1870
Rework memory management, don't leak memory in the filter pipeline.
2008-06-30 18:08:32 -07:00
Andreas Gal
19b4bcdc69
Save the side exit state at entry in to the opcode (since we want to restart ops) and enable store filter.
2008-06-30 17:12:52 -07:00
Andreas Gal
7511e45275
Use #ifdef DEBUG instead of VERBOSE.
2008-06-30 11:44:47 -07:00
Andreas Gal
806d0de678
Must use (&gc) new otherwise memory doesn't get cleared (found by danderson).
2008-06-30 11:39:00 -07:00
Andreas Gal
5d1d997c9d
Housekeeping (add comments, removed some dead code.)
2008-06-30 09:36:10 -07:00
Andreas Gal
d8d25140f9
Mockup of trace execution. Speedup is 10x for a tight loop. Keep in mind that this is a hack and the trace code is not optimized yet.
2008-06-29 23:02:22 -07:00
Andreas Gal
b1e24b8c98
Execute trace code. Boxing back the side exit state is not handled yet.
2008-06-29 20:56:06 -07:00
Andreas Gal
e267cb120c
Fix native frame addressing (reported by vlad.)
2008-06-29 17:17:39 -07:00
Andreas Gal
253b2572b2
Initialize jump table when side-entering the interpreter. Keep track of cx and sp directly in the recorder instead of going through the tracker.
2008-06-29 13:53:40 -07:00
Andreas Gal
ef71223de2
Remember JSContext instead of trying to keep track of frame pointer.
2008-06-28 18:47:12 -07:00
Andreas Gal
628a1f6985
Cleanup trace abort/end code.
2008-06-28 18:19:21 -07:00
Andreas Gal
6df975a5b5
Rewritten frame management. Use cx->fp->regs chain instead of direct passing of those structures.
2008-06-28 17:14:06 -07:00
Andreas Gal
552425ee07
Hook up pc and sp to the tracer in order to generate proper PC/SP adjustment code.
2008-06-28 11:00:18 -07:00
Andreas Gal
1f37ae3e62
Small cleanups and licensing blurb housekeeping.
2008-06-28 09:58:15 -07:00
shaver@mozilla.org
4fc71515e0
Add guards for XML objects, and some tracer bits to accommodate them.
2008-06-27 23:58:06 -04:00
Andreas Gal
40f1a90a78
Move recorder initialization into TraceRecorder and encapsulate its state.
2008-06-27 18:06:50 -07:00
Andreas Gal
2c370239e5
Add a way to calculate the current size of the native frame.
2008-06-27 16:24:17 -07:00
Andreas Gal
afd5f3b1e3
Clarified argument names in guard code and explain better why we guard for overflow the way we do.
2008-06-27 07:28:10 -07:00
Andreas Gal
e59076860a
Instead of guarding on null -> boolean conversion just emit a null compare and use the boolean result.
2008-06-27 06:53:10 -07:00
Andreas Gal
a21ca6a788
Prepare LIR_ov to be used as soon Ed adds it and add a few comments that explain how overflow is handled.
2008-06-26 23:34:50 -08:00
Andreas Gal
49fe7a95eb
More cleanup and code move into TraceRecorder and prepare for LIR_ov.
2008-06-26 23:22:16 -08:00
Andreas Gal
f357ff14f0
Move the recorder functionality into TraceRecorder to unclutter trace inlines.
2008-06-26 21:58:57 -08:00
Andreas Gal
0c98250a90
Move set/get into recorder and introduce init.
2008-06-26 21:41:10 -08:00
Andreas Gal
431b7b810a
Moved loading context slots into the recorder (more to follow).
2008-06-26 21:33:33 -08:00
Andreas Gal
2dca610f28
Make sure nativeOffset uses 64-bit slots on the native stack.
2008-06-26 21:25:30 -08:00
Andreas Gal
693ee7f0cc
Turn macros into inline functions and start prepatations for inlining function calls.
2008-06-26 21:19:11 -08:00
Andreas Gal
c7bab6721b
Drop JS prefix from classes related to traceing.
2008-06-26 19:49:04 -08:00
Andreas Gal
ad5ceed9cd
Merge.
2008-06-26 19:46:51 -08:00
Andreas Gal
a1b42735cd
Backed out changeset fa82b7eda72a
2008-06-26 19:44:23 -08:00
Andreas Gal
2fddbfdceb
Removed debug printfs from tracker.
2008-06-26 19:41:58 -08:00
Andreas Gal
5b4f0aaeaa
Allocate recorder dynamically to avoid having a vtable in the struct holding the reference. Emit writebacks for every update to the tracker, not just stack writes.
2008-06-26 19:37:28 -08:00
Andreas Gal
407b9cdf75
Parameterize tracker.
2008-06-26 17:13:54 -08:00
Andreas Gal
4678be95aa
Fix a bug in tracing can_do_fast_inc_dec (wasn't tracking a copy and traced incorrect code). Merge with TT tip.
2008-06-24 15:57:33 -07:00
Andreas Gal
d8e0d5edec
Removed debug code.
2008-06-23 17:17:48 -07:00
Andreas Gal
5e55f5e1f4
Merge.
2008-06-23 16:52:37 -07:00
Andreas Gal
5b30ba137c
Merge.
2008-06-23 16:51:28 -07:00
Andreas Gal
0c1996ce25
Don't setup debugging data structures when not debugging.
2008-06-22 23:59:41 -07:00
Brendan Eich
d5d3c6fb26
Random style nit-picking.
2008-06-22 23:42:54 -07:00
Andreas Gal
17d189ecb9
Eliminate boxing in trace code. To execute such traces all values on the stack must have the same type at execution time as at trace time. Code to detect and enforce these types will be added next.
2008-06-22 21:32:24 -07:00
Andreas Gal
4312eb9b9b
Write back stack/local variable state in the trace code and improved redundand boxing/unboxing elimination.
2008-06-22 19:58:24 -07:00
Andreas Gal
6d192a7054
Added end-of-trace detection and make sure trace loops back to the loop header.
2008-06-22 09:30:04 -07:00
Andreas Gal
2270a14eb0
Fix guard code assembly. This code generates trace code for the first time.
2008-06-21 23:45:16 -07:00
Andreas Gal
78da0668af
Start attaching our tracer to nanojit. This is incomplete and meant for review by David only.
2008-06-21 14:55:32 -07:00
Brendan Eich
77dcedf8cc
Fix slew of gcc warnings, clear pending exception before triggering recorder error.
2008-06-12 12:07:04 -07:00
Andreas Gal
179be35727
Notify the recorder of all primitives as they are recorded. If the recorder doesn't support a primitive we abort the trace. In debugging mode this also stops the VM.
2008-06-11 20:22:00 -07:00
Andreas Gal
448d1ba7e3
Merge with be's last push.
2008-06-11 18:13:43 -07:00
Andreas Gal
bb1b206714
Use an error property in the recorder instead of a return value to indicate errors and make sure DO_OP aborts the trace once we go into the error state in the tracer.
2008-06-11 17:59:28 -07:00
Brendan Eich
a64ed1a5ee
native_pointer_to_jsval should not lose low-order bits (plus random style nits).
2008-06-11 17:21:15 -07:00
Andreas Gal
54dd5103e9
Track data flow through variables in the recorder. Set an error flag if the recorder signals an error.
2008-06-11 17:15:49 -07:00
Brendan Eich
74c80d5597
Kill trailing spaces.
2008-06-11 13:07:24 -07:00
Andreas Gal
fef28e9f56
Allow recorder to abort recording.
2008-06-11 11:29:35 -07:00
Andreas Gal
50bf94c527
Notify recorder when tracing starts/stops. Restore code accidently lost during last merge.
2008-06-11 01:22:39 -07:00
Brendan Eich
45358518de
Minimize JSInterpreterState, fix warning.
2008-06-10 21:40:47 -07:00
Andreas Gal
a1bf0819b2
Removed JSOP_HEADER code. Its too costly. This patch restore essentially the same performance as unmodified SM.
2008-06-10 14:44:51 -07:00
Brendan Eich
04b1ac5592
Merge.
2008-06-04 00:09:57 -07:00
Brendan Eich
894dd03e2a
First stage of loop table work; bitmap free space management and GC hook-up still to come.
2008-06-03 23:52:28 -07:00
Andreas Gal
9684e2070f
Introduced a JavaScript recording script that is loaded dynamically.
2008-06-03 21:01:23 -07:00
Andreas Gal
3094b79c6b
Switch to a uniform style for jstracer.cpp.
2008-06-03 17:15:58 -07:00
Andreas Gal
8b6c4d02c6
Remove InitTacer. Pointless for JSRuntime.
2008-06-03 17:04:42 -07:00
Brendan Eich
9c2b039d6f
JSOP_HEADER takes a byte index of loop header counting from script->loopBase, and related changes.
2008-06-03 14:48:47 -07:00
Brendan Eich
946fa41d8c
Fiddle loop table slot interface and impl in hope of freeing slots over time.
2008-06-02 18:06:33 -07:00
Brendan Eich
01514c4b16
No JS_TRACER ifdefs, keep line len < 80, reparameterize jstracer.cpp functions, js_OpLength helper.
2008-06-02 17:19:41 -07:00
Brendan Eich
5b116ae998
Merge, style nits, no locking for tracing.
2008-05-31 15:29:54 -07:00
Brendan Eich
ef8a767739
Left brace style.
2008-05-31 10:17:40 -07:00
Andreas Gal
e461a72736
The table is now per-thread in a multi-threaded environment, and per-runtime otherwise. During code generation we merely allocate a loop table slot to each loop. Each thread will enlarge the table as needed in JSOP_HEADER.
2008-05-30 18:58:43 -07:00
Andreas Gal
de3107d2ee
Add a per-runtime loop attribute table that associates a jsval attribute with every loop in the code. The jsval is used initially as a counter until a certain threshold is reached, at which point the loop is traced and compiled and the resulting native code object is stored in the jsval to be executed for future encounters of the loop.
2008-05-30 10:11:56 -07:00
Brendan Eich
8bfb8bfef7
Split trace-supporting inlines, add ifdefs/macros for deriving js_TracingInterpret in jstracer.cpp.
2008-05-28 19:07:25 -07:00
