duktape

Commit Graph

Author	SHA1	Message	Date
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
Sami Vaarala	77e9d8dcd2	Releases: assign bytecode improvements	9 years ago
Sami Vaarala	fb6b39a4ba	Suppress "unused xxx" compilation warnings	9 years ago
Sami Vaarala	465236ed6f	Fix dump/load API testcase expect Changed again because assignment handling code was changed.	9 years ago
Sami Vaarala	e5f122f5b4	Improve compiler handling of assignment expression - Allow plain value, const, temp reg as the expression value without forcing into a const/temp which improves some bytecode. - Also allow a non-temp reg (reg-bound identifier) for "top level assignment" statements like "x = 123;". This kind of assignment can safely use a reg- bound identifier as the expression value because the register in question won't be used/mutated by the surrounding expression. - Also allow an unresolved operation (e.g. addition) as an initial expression value (later coerced to a side effect free plain value). This avoids an unnecessary temporary register for assigments like "x = y + z;". For a top level assignment the addition is resolved directly into "x" and the assignment expression value becomes the register of "x". This is safe for top level assignments. For other assignments a temp register is used. - Similar improvements for assign-with-op (e.g. "x += y") so that top level assign-with-op expressions produce better bytecode. The top level assignment case is detected using nud/led counts. The top level assignment optimization should be safe because: - Assignment is parsed in a right associative manner, so that the RHS can't contain any more assignment level expressions. - The only relevant lower binding power construct is comma expression. It's possible that the assignment happens inside a comma expression, e.g. "x = y, z, y", but that's safe because the expression value of "x = y" is ignored by the comma expression so it doesn't matter if it mutates later on in the comma expression.	9 years ago
Sami Vaarala	99ccab941b	Perf tests: const load, continue fast/slow, mandel	9 years ago
Sami Vaarala	09abe09569	Testcases for assignment handling	9 years ago
Sami Vaarala	0bfb61ce94	Merge pull request #381 from svaarala/fix-assign-regbound-rhs Fix assignment expression reg-bound RHS handling	9 years ago
Sami Vaarala	de19a9e6b0	Releases: register-bound rhs in assignment	9 years ago
Sami Vaarala	e19039c826	Fix dump/load API testcase expect The expected output changed after the assign RHS fix.	9 years ago
Sami Vaarala	a8c03987bf	Fix assignment RHS handling to avoid GH-381	9 years ago
Sami Vaarala	5ae7ab8743	Improve regconst ispec coercion handling Allow simultaneous "allow const" and "require temporary register" meaning that the only regconst rejected is a non-temp register (register-bound identifier).	9 years ago
Sami Vaarala	8966c67c93	Bug testcase for assign + mutation	9 years ago
Sami Vaarala	1cb5960f00	Merge pull request #377 from svaarala/perf-test-consistency-fix Trivial fix for perf test loop count	9 years ago
Sami Vaarala	bd70eb7ab8	Merge pull request #378 from svaarala/debugger-doc-pause-on-error Remove debugger pause-on-error future work	9 years ago
Sami Vaarala	45cc8dd819	Remove debugger pause-on-error future work	9 years ago
Sami Vaarala	fa62caa363	Loop count same for native and Ecma call tests	9 years ago
Sami Vaarala	c8637f7f5b	Merge pull request #371 from svaarala/perf-rework-debugger-skip-interrupt Fix debugger "lockout" in interrupt handling for better performance	9 years ago
Sami Vaarala	684a155e30	Assert for dbg_processing state on halt	9 years ago
Sami Vaarala	ad67b32914	Avoid recursive error notify/uncaught pause	9 years ago
Sami Vaarala	a7cadfedbe	Refactor debugger "lockout" for better performance	9 years ago
Sami Vaarala	a661ff6608	Config option for interrupt fixup	9 years ago
Sami Vaarala	70315f9e41	Merge pull request #348 from svaarala/perf-rework-breakcontinue-handling Rework BREAK and CONTINUE opcode handling to avoid longjmp entirely	9 years ago
Sami Vaarala	46adf51302	Releases: BREAK and OPCODE rework	9 years ago
Sami Vaarala	3b1bd61b99	Testcase exercising break/continue code paths	9 years ago
Sami Vaarala	e5e172726a	Handle BREAK and CONTINUE opcodes without longjmp Also RETURN and other control flow cleanups. DUK_SETJMP() return value consistency.	9 years ago
Sami Vaarala	91d7830836	Merge pull request #345 from svaarala/perf-rework-return-handling Rework RETURN opcode handling to avoid longjmp() entirely	9 years ago
Sami Vaarala	a2b1644e42	Releases: return rework	9 years ago
Sami Vaarala	fb3165667b	Testcase cleanup and improvement for fast returns	9 years ago
Sami Vaarala	42eb7feee9	Remove fast return references from docs	9 years ago
Sami Vaarala	8d9c09eae6	Update debugger opcode list, no fast return flag	9 years ago
Sami Vaarala	d31ec37fc5	Handle RETURN opcode without longjmp	9 years ago
Sami Vaarala	81cc48152c	Merge pull request #367 from svaarala/fix-website-broken-link Fix broken API doc link	9 years ago
Sami Vaarala	f4152bfaa5	Fix broken API doc link	9 years ago
Sami Vaarala	5b4a77b7d5	Merge pull request #365 from svaarala/move-project-list-to-wiki Migrate project list to Wiki	9 years ago
Sami Vaarala	bd49d12c67	Migrate project list to wiki	9 years ago
Sami Vaarala	6bc74a512a	Merge pull request #364 from svaarala/add-project-soletta Add Soletta project link	9 years ago
Sami Vaarala	1fdc42aec5	Add Soletta project link	9 years ago
Sami Vaarala	e19dec57e3	Merge pull request #355 from svaarala/add-emduk-nodefs Add relative file support to emduk using NODEFS	9 years ago
Sami Vaarala	68de23f247	Runtests workaround for Emduk trailing line Trailing line contains a space and a newline (0x20 0x0a), add a hack runtests option to strip that.	9 years ago
Sami Vaarala	38c53482c9	Add relative path and /tmp support using NODEFS	9 years ago
Sami Vaarala	5b0e715661	Merge pull request #347 from svaarala/debugger-throw-notify-cleanups Debugger throw notify cleanups	9 years ago
Sami Vaarala	ab48b29ba0	Release notes: Throw, uncaught halt, debugger stmt	9 years ago
Sami Vaarala	bbd2346cf1	Remove debugger stmt from future work	9 years ago
Sami Vaarala	7deefa7862	Manual test case for native caught/uncaught error	9 years ago
Sami Vaarala	022352f2d2	Debugger Throw related and other cleanups - Add config guards for debug throw/pause uncaught - Add one missing DUK_USE_DEBUGGER_SUPPORT guard in master - Add config guards for new config options - Use duk_debug_halt_execution() for DEBUGGER stmt, which fixes the off-by-one line number for debugger statement when the executor gets paused - Add internal duk_safe_to_hstring() and combined safe_to_hstring + duk_debug_write_hstring() helpers to debugger code - Debug command capitalization in debug messages - Some debugger assert additions; in particular, assert that when doing a halt or entering the message loop the debugger is attached (if this is not a requirement, it will need special handling)	9 years ago
Sami Vaarala	f82e9aefaf	Fix pc2line information for debugger statement Previous position for bytecode emission caused the line preceding the debugger statement to be attributed for the PC of the DEBUGGER opcode.	9 years ago
Sami Vaarala	ee97ecf15c	Add new config options to matrix compile, Makefile	9 years ago
Sami Vaarala	72c51ada7c	Config metadata for debugger throw, pause uncaught Also add config example for debugger support with pause uncaught enabled.	9 years ago
Sami Vaarala	99a2e78a09	Debug web UI support for Throw notify Duk_debug.js fixes for Throw notify - Ignore unknown Notify messages, this is probably a better default than being strict and dropping the connection. - Add Throw notify command number (earlier GC notify was unused). - Add minimal web UI integration to report "THROW: xxx" or "UNCAUGHT: xxx" from Throw notifys Add 'make rundebug' convenience duk_debug target with logging.	9 years ago

1 2 3 4 5 ...

5596 Commits (8bb0156574c4211cad0e63cbf8d04d9ba49bfd0a) All Branches Search

5596 Commits (8bb0156574c4211cad0e63cbf8d04d9ba49bfd0a)

All Branches