duktape

Commit Graph

Author	SHA1	Message	Date
Sami Vaarala	5de6133ef1	Direct value stack access in duk_push_this()	9 years ago
Sami Vaarala	16ac453eaf	Merge pull request #398 from svaarala/avoid-rdflib-in-dist Avoid rdflib in non-official dist	9 years ago
Sami Vaarala	de34db7f9c	Makefile: keep compiler.jar, --create-spdx Keeping compiler.jar means that if you once "make dist-src-official" or "make compiler.jar", "make dist" will then closure-minify further builds.	9 years ago
Sami Vaarala	0d2a39cc2f	Add --create-spdx option to dist	9 years ago
Sami Vaarala	e125bae7fd	Merge pull request #397 from svaarala/perf-call-coerce-this Use direct valstack access for "effective this" in call handling	9 years ago
Sami Vaarala	eb1ff24ca5	Releases: effective this binding	9 years ago
Sami Vaarala	4d5abb129b	Direct valstack access for effective 'this' setup Figuring out an effective 'this' value happens for every call. The previous implementation for non-strict functions used stack API calls. The reworked implementation accesses the value stack directly which matters for this performance critical code path.	9 years ago
Sami Vaarala	099c8b0140	Tval refcount macro name typo fixes	9 years ago
Sami Vaarala	67e88f8fdf	Merge pull request #396 from svaarala/perf-unused-marker-tag Add a separate marker tag for 'unused' values	9 years ago
Sami Vaarala	e44c3cf55f	Releases: unused marker tag	9 years ago
Sami Vaarala	5a4a8b0b1a	Internal doc changes for 'unused' marker tag	9 years ago
Sami Vaarala	3fbd785183	Fix isFastint polyfill for new tag numbers The new fastint detection works for both old and new packed boolean tag (0xfff4 in Duktape 1.3.0, 0xfff5 now).	9 years ago
Sami Vaarala	88185d5acf	Test case 'unused' tag fixes The new fastint detection works for both old and new packed boolean tag (0xfff4 in Duktape 1.3.0, 0xfff5 now).	9 years ago
Sami Vaarala	efc52c8e8d	Separate (marker) unused tag Reorder tags to accommodate a separate 'unused' tag so that 'undefined' can become a single tag write (instead of tag + value like booleans). This is good because 'undefined' values are involved in e.g. value stack resizes and are performance relevant. Also reorder tags so that "is heap allocated" check can be a single bit test instead of a comparison when using non-packed duk_tval. This makes every DECREF potentially faster because an "is heap allocated" test appears in every DECREF. Because "unused" is not intended to appear anywhere in actual use (e.g. as a value stack value, as a property value, etc), "unused" values will fall into the default clause of DUK_TAG_xxx switch case statements. Add an assert to every such default clause that the value is not intended to be "unused". Remove duk_push_unused() as it should no longer be used. It was only used by the debugger protocol; refuse an inbound "unused" value in the debugger. This is not breaking compatibility because there was no legitimate usage for the debug client sending requests with "unused" values.	9 years ago
Sami Vaarala	d7ee8df5f0	Merge pull request #389 from svaarala/perf-valstack-init-policy Change value stack init policy (garbage after current top)	9 years ago
Sami Vaarala	36ab813e0a	Merge pull request #368 from svaarala/runtests-windows-tmp Remove /tmp assumption from runtests.js	9 years ago
Sami Vaarala	cd27ca9ad1	Releases: value stack init policy	9 years ago
Sami Vaarala	1a05402fa1	Strip 'leave as garbage' valstack policy 'Leave as undefined' seems to be the best overall value stack initialization policy. While 'leave as garbage' is marginally better in a few cases (mostly when refcounting is disabled) it's probably not worth keeping two policies around.	9 years ago
Sami Vaarala	5603ba84c7	Strip DUK_USE_VALSTACK_GARBAGE option	9 years ago
Sami Vaarala	f85ccf7081	Add valstack init policies: garbage vs. undefined Change the current value stack policy from "unused above top" to either "undefined above top" or "garbage above top", depending on a config option. Change mark-and-sweep and debug print code to only process entries between [0,top[ in either case. Both policies have potential upsides and downsides based on performance measurement. This commit provides both policies; "undefined above top" will probably be the only policy left however, because "garbage above top" is only better in a few cases and mostly without refcounts. Other minor improvements: - Rework index validation to use duk_uidx_t and more shared code. - Add cached thr->valstack_size value	9 years ago
Sami Vaarala	7397c0c320	Prefer size by default, valstack policy config opt - Make DUK_USE_PREFER_SIZE false by default - DUK_USE_VALSTACK_GARBAGE temporary config option (will be removed later)	9 years ago
Sami Vaarala	84bf7edfc8	Add duk_uidx_t type to config headers Also: suggest DUK_USE_VALSTACK_UNSAFE for low memory. Helps with code footprint.	9 years ago
Sami Vaarala	99078edd13	Add call perf tests with varying nregs/nargs	9 years ago
Sami Vaarala	c43fc1464f	Fix broken assert Triggered when DUK_USE_PREFER_SIZE is not defined.	9 years ago
Sami Vaarala	4a4b587647	Merge pull request #395 from svaarala/fix-duk-ajduk-compile Duk cmdline ajduk compile fix	9 years ago
Sami Vaarala	b082d89558	Duk cmdline ajduk compile fix	9 years ago
Sami Vaarala	dea1b8db62	Merge pull request #393 from svaarala/perf-tval-updref-macro Add DUK_TVAL_SET_xxx_UPDREF() macros	9 years ago
Sami Vaarala	b12e59d9e1	Release checklist: refcount leak regression	9 years ago
Sami Vaarala	0c67d6ef62	Basic (manual) test for detecting refcount leaks	9 years ago
Sami Vaarala	1a09e0b9ac	Add --no-heap-destroy duk option for leak testing When present, run a forced GC and exit without destroying heap.	9 years ago
Sami Vaarala	5f3f958ff6	Remove unnecessary object prop slot cleanup	9 years ago
Sami Vaarala	a1a2f25d74	Use DUK_SET_TVAL_xxx_UPDREF() macro throughout Also cleanups: - Fix a few compile warnings when compiling without refcounting - Fastint support for ENDFIN	9 years ago
Sami Vaarala	3f26826b7e	Add DUK_TVAL_SET_xxx_UPDREF macros These replace the much repeated idiom: duk_tval tv_tmp; DUK_TVAL_SET_TVAL(&tv_tmp, tv_dst); DUK_TVAL_SET_TVAL(tv_dst, tv_src); DUK_TVAL_INCREF(thr, tv_src); DUK_TVAL_DECREF(thr, &tv_tmp); /* side effects */ with (e.g.): DUK_TVAL_SET_TVAL_UPDREF(thr, tv_dst, tv_src); This reduces line count, and also allows the set-and-update-refcount sequence to be optimized in detail.	9 years ago
Sami Vaarala	b8a33a0756	Merge pull request #388 from svaarala/add-luajit-nodejs-perf Add LuaJIT and Node.js (V8) performance comparisons	9 years ago
Sami Vaarala	14a703b0b7	Add node/luajit to makefile perftest target	9 years ago
Sami Vaarala	8e74bd6e6f	Add console.log compatibility to perf tests	9 years ago
Sami Vaarala	616557edef	Merge pull request #370 from svaarala/executor-setjmp-external Move executor setjmp catchpoint to an outer wrapper function	9 years ago
Sami Vaarala	31e727525b	Releases: bytecode setjmp split	9 years ago
Sami Vaarala	ea58277c8b	Double union quiet NaN notes	9 years ago
Sami Vaarala	1973993f60	Code issues notes on setjmp	9 years ago
Sami Vaarala	9bde055868	Config option updates - Add DUK_USE_EXEC_FUN_LOCAL config option - Performance sensitive example updates - Config example typo in option name	9 years ago
Sami Vaarala	1875b54225	Perf test fixes and additions - Add missing 't = 0' initializer to test-assign-addto. Without the initializer 't' becomes NaN and for some reason is over 20x slower on x86 as a result of that. If 't' is initialized performance on x86 is fine. (It's worth investigating why falling out of the fastint fast path is so costly on x86 but not on x64; NaN normalization?) - Add missing test-assign-reg.pl. - Add specific addition tests, where numbers involved are (a) fastints, (b) doubles, (c) NaNs. These tests demonstrate the roughly 10x difference on x86 for NaNs and other IEEE doubles -- with and without packed duk_tval.	9 years ago
Sami Vaarala	6d226a0f6a	Split bytecode executor into outer and inner func Outer function has a setjmp catchpoint and handles all longjmps. Inner function is the plain bytecode executor which is capable of an internal restart but doesn't handle longjmps. Also make 'fun' local from executor config option dependent. x64 is slightly faster without it, x86 is slightly faster with it.	9 years ago
Sami Vaarala	391053e7c2	Merge pull request #382 from svaarala/debugger-bytecode-currpc Minimal support for debugger web UI bytecode "current PC" highlight	9 years ago
Sami Vaarala	f49fe0e074	Rough bytecode curr opcode highlight - Doesn't scroll - Doesn't detect if bytecode doesn't match current function	9 years ago
Sami Vaarala	36443b41bd	Merge pull request #384 from svaarala/perf-test-trycatch Add some try-catch/try-finally perf tests	9 years ago
Sami Vaarala	e2068b50ac	Some try-catch/try-finally perf tests	9 years ago
Sami Vaarala	71adcc7b12	Merge pull request #383 from svaarala/perf-test-loop-unroll Unroll perf test loops to reduce loop overhead	9 years ago
Sami Vaarala	b41371be94	Unroll perf test loops to reduce loop overhead With less loop overhead the micro benchmarks provide clearer results on how the actual operation being tested behaves.	9 years ago
Sami Vaarala	8bb0156574	Merge pull request #380 from svaarala/perf-assign-bytecode-improvement Improve bytecode emitted for assignment expressions with identifier LHS	9 years ago

... 3 4 5 6 7 ...

5849 Commits (13af94da67dac79a8383629b07581ff00de0bb63) All Branches Search

5849 Commits (13af94da67dac79a8383629b07581ff00de0bb63)

All Branches