aboutsummaryrefslogtreecommitdiff
AgeCommit message (Collapse)Author
2010-10-15recurse blocks in loop optimizer; 1% speedupalon@honor
2010-10-15BNOPP in some simple emulated blocks; 1% speedupalon@honor
2010-10-14sanity checks for benchmarksalon@honor
2010-10-14remove unnecessary boolean ==> int conversions; fix uncovered bug; 20% speedupalon@honor
2010-10-13use tracemonkey in settings.pyalon@honor
2010-10-13aggressively nativize variables; stop doing optimizations that are no longer ↵alon@honor
needed; fix some uncovered bugs; 49% speedup
2010-10-12use closure compiler in benchmarks; 13% speedupalon@honor
2010-10-12commentalon@honor
2010-10-11clean up relooper label storagealon@honor
2010-10-11increment stack all at once; 2% speedupalon@honor
2010-10-10typed arrays support; function __index__ing ; 36% speedupalon@honor
2010-10-10clean up flattener code and remove unneeded flattener definitions; 3% speedupalon@honor
2010-10-10spidermonkey optsalon@honor
2010-10-09pass types to makeGetValue/makeSetValuealon@honor
2010-10-09cleanup: remove need for .text in item.pointeralon@honor
2010-10-09cleanup: remove need for .text in item.typealon@honor
2010-10-09optimize away 1's and 0's in getelementptr; 1% speedupalon@honor
2010-10-09do not GUARD_MEMORY in benchmark; 7% speedupalon@honor
2010-10-09benchmarking setupalon@honor
2010-10-09allow test runner to not check timeouts when running codealon@honor
2010-10-09fix memory alignment/padding of structures | TESTS FIXEDalon@honor
2010-10-09make llvm 2.7/clang work againalon@honor
2010-10-08add entry labels only when necessary | all tests pass but fasta on clangalon@honor
2010-10-08fix bug with last commit, and better GUARD_MEMORY optionalon@honor
2010-10-08initial support for |extractvalue|alon@honor
2010-10-08beginnings of fixes for clang 2.8alon@honor
2010-10-07all llvm-gcc tests passalon@honor
2010-10-07handle |linkonce| constants | all llvm_gcc tests pass but saueralon@honor
2010-10-07begin work to adapt to llvm 2.8 | TESTS BROKENalon@honor
2010-10-07separate running stage out in test runneralon@honor
2010-10-07separate build stage out in test runneralon@honor
2010-10-06optimize raytrace demoalon@honor
2010-10-06inline stack allocationsalon@honor
2010-10-06Runtime system, and inlining of stackEnter/Exitalon@honor
2010-10-05fix relooper bug with not noticing other entries as invalidating; now loops ↵alon@honor
are only created when truly needed
2010-10-05version 0.3alon@honor
2010-10-04higher memory defaults, for raytracing demoalon@honor
2010-10-04temporary hackish removal of unneeded breaks outside of complex emulated blocksalon@honor
2010-10-04GUARD_LABELS option + further cleanupalon@honor
2010-10-04clean up relooperalon@honor
2010-10-04initial working version of 'multiple' block in relooperalon@honor
2010-10-03cleanup fallback when no RELOOPalon@honor
2010-10-03proper indentation of generated codealon@honor
2010-10-03reloop blocks with multiple entriesalon@honor
2010-10-03minor relooper fixesalon@honor
2010-10-03clean up relooperalon@honor
2010-10-03cleanup + test both optimized and non-optimizedalon@honor
2010-10-02add missing branch detection of assign&invoke. all tests pass with reloopingalon@honor
2010-10-02add missing branch detection of default in switch. all reloop but sauer, ↵alon@honor
dlmalloc/clang
2010-10-02test both reloop and non-reloopalon@honor