summaryrefslogtreecommitdiff
path: root/test
Commit message (Collapse)AuthorAgeFilesLines
* optimize loads+shifts into signed loadsAlon Zakai2016-09-077-428/+464
|
* optimize out eqz^2Alon Zakai2016-09-074-588/+304
|
* use eqz as an optimizationAlon Zakai2016-09-0710-4357/+3682
|
* new validation checks for upcoming spec testsAlon Zakai2016-09-075-51/+51
|
* new export syntax in spec repoAlon Zakai2016-09-0740-283/+283
|
* new import syntax in spec repoAlon Zakai2016-09-0743-475/+475
|
* re-enable spec testsAlon Zakai2016-09-071-0/+0
|
* Add a ReFinalize helper, and use that to properly handle asm.js imports ↵Alon Zakai2016-09-075-56/+210
| | | | whose return value is polymorphic
* support HEAP8[x | 0| notation in asm2wasmAlon Zakai2016-09-075-0/+29
|
* import memoryBase and tableBaseAlon Zakai2016-09-0732-16/+72
|
* import tableAlon Zakai2016-09-0732-0/+32
|
* import memory #684Alon Zakai2016-09-0732-32/+32
|
* update empty.fromast test outputAlon Zakai2016-09-071-0/+2
|
* add asm2wasm option to import a mem init fileAlon Zakai2016-09-077-0/+14
| | | | | | | | apply memory segments only if there isn't a memory initializer (which we need for asmjs and asm2wasm modes) use wasm-opt to check recreated wasts for validity, as wasm-shell would try to execute them add testing for combined modes like asmjs,interpret-binary
* do a little more optimization at the end of asm2wasm processingAlon Zakai2016-09-076-136/+68
|
* sink a drop into a single if armAlon Zakai2016-09-074-34/+63
|
* autodrop must be run before we optimize in asm2wasm, as otherwise its input ↵Alon Zakai2016-09-074-76/+102
| | | | | | is not yet valid then after finalizeCalls, we must autodrop again to drop things that finalizeCalls changed
* don't simplify locals out of loops if they contain branching, as it may ↵Alon Zakai2016-09-072-0/+49
| | | | invalidate the branch
* when replacing an if with its condition (when it has no body), we must drop itAlon Zakai2016-09-075-0/+77
|
* add a test for loop finalizationAlon Zakai2016-09-075-0/+108
|
* run vacuum again after autodrop in asm2wasm, if optimizingAlon Zakai2016-09-076-130/+430
|
* handle asm.js globals that are set and the return value usedAlon Zakai2016-09-071-0/+11
|
* fix AutoDrop block handling - the block type might change as we modify its ↵Alon Zakai2016-09-074-168/+128
| | | | contents
* set type of calls to their target, instead of the previous behavior where ↵Alon Zakai2016-09-075-0/+106
| | | | the asm.js context informed us. this lets us add drops where necessary
* use globals in asm2wasmAlon Zakai2016-09-0720-2762/+1450
|
* export kindsAlon Zakai2016-09-0732-35/+35
|
* add a drop for final elements in blocks if they are not usedAlon Zakai2016-09-078-124/+196
|
* update testsAlon Zakai2016-09-071-4/+4
|
* don't depend on order of operations in calls, it varies by compilerAlon Zakai2016-09-0712-1416/+1416
|
* select values must be validAlon Zakai2016-09-071-16/+16
|
* move drop into blocks, dropping all the breaks as well, when possibleAlon Zakai2016-09-071-274/+224
|
* loops no longer have an out label and other upstream loop updatesAlon Zakai2016-09-0738-40507/+41619
|
* call_indirect now has the target at the endAlon Zakai2016-09-0722-302/+303
|
* add drop and tee expressionsAlon Zakai2016-09-0783-11054/+16397
|
* remove lower-if-else, as it's no longer neededAlon Zakai2016-09-072-71/+0
|
* Improvements to build-js.sh and JS API (#679)Rasmus2016-09-022-0/+171
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * Adds command-line argument to build-js.sh for specifying the location of Emscripten. Also spreads out program arguments on separate lines * Exposes WasmBinaryWriter and BufferWithRandomAccess in the JS API to allow writing WASM modules and access the produced bytes via the JS API * Updates bin/binaryen.js and bin/wasm.js from changes to build-js.sh * Adds exclude patterns to .gitignore for some files generated by build-js.sh and .DS_Store for macOS * Changes build-js.sh to use EMSCRIPTEN env var instead of a command-line argument to provide the path to emscripten * Improvements to JS builds - Adds -g flag to build-js.sh that will build unoptimized and unmangled versions of binaryen.js and wasm.js (output has a "-g.js" suffix to allow co-existence with regular optimized builds). - Enables closure compiler for non-debug builds - Adds browser test for s-expression parser + WASM code gen (requires wasm to be enabled in the browser.) - Adds iterator interface to BufferWithRandomAccess (when Symbol.iterator is available) - Adds toArrayBuffer to BufferWithRandomAccess (when TypedArray is available) - Adds compileWast(sourceText :string) :ArrayBuffer to the module, parsing & compiling s-expression code to a WASM module - Changes the way binaryen.js is exported to allow usage in CommonJS, AMD and UMD envionments. * Expose "Binaryen" global in a better way to work with a.js generated by check.py * Fix to binaryen.js to only export a global variable when running the test (a.js) while inside a module (avoids polluting global in e.g. nodejs). Also fixes a spelling mistake. * Better "no WASM detected" message in test/binaryen.js/browser.html * Small change to error message in build-js.sh where $EMSCRIPTEN does not point to a directory * Changes emcc args in build-js.sh after investingating a large number of argument combinations. Also adds a browser benchmark. The result of emcc arguments and the effect on performance is summarized in this doc: https://gist.github.com/rsms/e33c61a25a31c08260161a087be03169 * Enable inferring emscripten path by looking in PATH when EMSCRIPTEN is not set in env
* Update waterfall build to 10372 (#689)Derek Schuff2016-08-31464-38736/+41076
| | | | | | | This brings in LLVM changes up to r28025, which includes a fix for PR29127 and includes disabling the store-result optimization. * remove extraneous wasm-interpreter.h change
* asm.js-style setjmp/longjmp handling for wasm: add handling foraheejin2016-08-302-12/+24
| | | | emscripten_longjmp_jmpbuf name
* Asm.js-style setjmp/longjmp support for wasm (#681)Heejin Ahn2016-08-262-0/+14
| | | | | | | This needs to export realloc as well, in addition to malloc and free handled in #4469. To support asm.js style setjmp/longjmp, wasm needs to export realloc as well, in addition to malloc and free handled in #4469. saveSetjmp() uses realloc within it, and realloc is not implemented in JS glue code.
* Merge pull request #682 from loganchien/fix-empty-deadlockAlon Zakai2016-08-265-0/+20
|\ | | | | Fix asm2wasm dead lock caused by empty module
| * Fix asm2wasm dead lock caused by empty modules.Logan Chien2016-08-265-0/+20
| | | | | | | | | | | | | | | | | | This commit fixes an asm2wasm dead lock when asm2wasm is compiling an empty module, i.e. a module without any functions. Without this commit, worker threads are likely to leave `workerMain()` and decrease `liveWorkers` early. Consequently, `waitUntilAllReady()` will never observe `liveWorkers == numWorkers`.
* | Fix alignment bug in .lcomm (#680)Heejin Ahn2016-08-262-0/+40
|/ | | | | | | | | | | | When parsing .lcomm directives, s2wasm does not parse the alignment number and skip it. This causes alignment bugs in some cases. (In the test case attached, 'buf' should be 4 bytes aligned, but it does not align it properly, so this code was generated: ``` (call $foo (i32.const 13) ) ``` 13 is not 4-bytes aligned. This patch fixes this bug.
* offset support in tableAlon Zakai2016-08-1531-36/+36
|
* support function table initial and max sizes, and new printing formatAlon Zakai2016-08-1231-35/+70
|
* support expressions in segment offsetsAlon Zakai2016-08-1269-237/+169
|
* Implement asm.js style exception handling for Wasm (#664)Heejin Ahn2016-08-115-1/+208
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | * Implement asm.js style exception handling for Wasm Converts invoke wrapper names generated by LLVM backend to real invoke wrapper names that are expected by JavaScript glue code. This is required to support wasm exception handling (asm.js style). LLVM backend lowers invoke @func(arg1, arg2) to label %invoke.cont unwind label %lpad into ... (some code) call @invoke_SIG(func, arg1, arg2) ... (some code) SIG is a mangled string generated based on the LLVM IR-level function signature. In LLVM IR, types are not lowered yet, so this mangling scheme simply takes LLVM's string representtion of parameter types and concatenate them with '_'. For example, the name of an invoke wrapper for function void foo(struct mystruct*, int) will be "__invoke_void_%struct.mystruct*_int". This function converts the names of invoke wrappers based on their lowered argument types and a return type. In the example above, the resulting new wrapper name becomes "invoke_vii". * Address comments Change variable names to camelcase Add a small (semi-)handwritten test case * Export malloc and free from wasm when available * Add a test case for exporting malloc/free feature + cosmetic 'file' name change in text_before_type.s * fixInvokeWrapper -> fixEmExceptionInvoke * Add a TODO
* in DemoteFloat64, if the truncated value is exactly at the limit, return it ↵Alon Zakai2016-08-081-0/+0
| | | | | (#665) and update spec tests
* Fix a parsing bug introduced by #615 (#661)Heejin Ahn2016-08-052-0/+31
| | | | | | | | | | | | | | | Name lhs = getStrToSep(); if (!skipEqual()){ s = strchr(s, '\n'); if (!s) break; continue; } The above code snippet introduced by #615 has a bug when there is only one word (e.g. ".text") in a line. If there is only one word in a line, skipEqual() also skips the newline character at the end of the line, and strchr(s, '\n') moves the cursor to the end of the next line, effectively skipping the whole next line.
* Update waterfall build revision to 9397 (#662)Derek Schuff2016-08-0452-3722/+3604
| | | Includes torture tests and new expected passes
* Create a dummy function to prevent NULL miscomparisons, if necessary (#658)Dominic Chen2016-08-038-28/+49
| | | Resolves WebAssembly/spec#312