summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
...
* Remove legacy GC encodings (#5874)Thomas Lively2024-03-122-151/+31
| | | | | It was previously possible to opt in to using the legacy GC opcodes with a build time flag. Now that WasmGC has shipped and users have migrated to the standard opcodes, remove the option to use the legacy encodings.
* Fix Emscripten build with -Wno-unused-command-line-argument (#6392)Thomas Lively2024-03-121-1/+5
| | | | Emscripten had started complaining about the repeated NODERAWFS arguments in the link command, but they would be nontrivial to deduplicate.
* Fuzzer: Fix up null outputs in wasm2js optimized builds (#6374)Alon Zakai2024-03-081-0/+15
| | | | | | | | This is fallout from #6310 where we moved to use fuzz_shell.js for all fuzzing purposes. That script doesn't know wasm types, all it has on the JS side is the number of arguments to a function, and it passes in null for them all regardless of their type. That normally works fine - null is cast to the right type upon use - but in wasm2js optimized builds we can remove casts, which can make that noticeable.
* Check for unreachable in `Select::finalize(Type)` (#6389)Thomas Lively2024-03-081-1/+9
| | | | Previously selects finalized with explicit types would never be marked unreachable, even when they should have been.
* [NFC] Clean up the unreachable replacement code in Print.cpp (#6388)Thomas Lively2024-03-081-108/+56
| | | | | | | When instructions cannot be printed because the children from which they are supposed to get their type immediates are unreachable or null, we print blocks of their dropped children followed by unreachables. But the logic for making this happen was more complicated than necessary and in fact included dead code. Clean it up.
* Fix printing of bulk array ops (#6387)Thomas Lively2024-03-084-66/+212
| | | | | | | | | When the bulk array ops had unreachable or null array types, they were replaced with blocks, but not using the correct code that also prints all their children as dropped followed by an unreachable. This meant that the text output in those cases did not parse as a valid module. Fix the bug. A follow-up PR will simplify the code to prevent similar bugs from occurring in the future.
* Regenerate test output (#6385)Thomas Lively2024-03-073-26/+30
| | | | The checked in test outputs were out of sync with what the auto update script produces.
* [IRBuilder] Validate tuple arities (#6384)Thomas Lively2024-03-071-0/+12
| | | | Throw errors if tuple arity immediates are less than 2 or if tuple index immediates are out of bounds.
* Expose features option in C API binary reading (#6380)Surma2024-03-074-4/+41
| | | | This allows reading a module that requires a particular feature set. The old API assumed only MVP features.
* Do not write assertions to split.wast for spec tests (#6383)Thomas Lively2024-03-072-3/+3
| | | | | | | | | | As part of our running of spec tests, we split out each module in a test script into a separate text file for processing with wasm-opt. We previously included the test assertions corresponding to the module into that text file, where they were ignored by the legacy text parser. The new parser errors out due to the extra tokens after the module, though, so to avoid problems once we switch to the new parser, stop including the assertions in those text files. Also remove a nearby unused argument as a drive-by cleanup.
* Handle extended const segment offsets in the fuzzer (#6382)Thomas Lively2024-03-072-13/+14
| | | | | | The fuzzer already had logic to remove all references to non-imported globals from global initializers and data segment offsets, but it was missing for element segment offsets. Add it, and also add a missing check line for the new test that uncovered this bug as initial fuzzer input.
* Fix EH fuzz bugs (#6381)Thomas Lively2024-03-072-2/+2
| | | | | Due to a typo, the fuzzer was making externrefs when it should have been making exnrefs. Fix that and also let eh-utils.cpp know that TryTable exists to avoid an assertion failure.
* Print '(offset ...)` in data and element segments (#6379)Thomas Lively2024-03-062-2/+23
| | | | | | | Previously we just printed the offset instruction(s) directly, which is a valid shorthand only when there is a single instruction. In the case of extended constant instructions, there can potentially be multiple instructions, in which case the explicit `offset` clause is required. Print the full clause when necessary.
* Add sourcemap support to wasm-metadce and wasm-merge (#6372)Jérôme Vouillon2024-03-0612-12/+271
|
* [Parser] Improve parsed IR for multivalue returns (#6378)Thomas Lively2024-03-052-11/+13
| | | | | Rather than reassembling a tuple from multiple pops, let the pop implementation assemble the tuple. This produces less code in cases where there is already a tuple of the proper size on top of the stack. It also simplifies the code.
* Fuzzer: Standardize notation for exception prefixes (#6369)Alon Zakai2024-03-052-4/+10
| | | | | | | | | We had exception: in one and exception thrown: in another. Making those consistent allows fuzz_shell.js to print the exception after that prefix, which makes debugging easier sometimes. Also canonicalize tag names. Like funcref names, JS VMs print out the internal name, which can change after opts, so canonicalize it.
* [Parser] Propagate debug locations like the old parser (#6377)Thomas Lively2024-03-052-0/+81
| | | | | | | | | Add a pass that propagates debug locations to unannotated child and sibling expressions after parsing. The new parser on its own only attaches debug locations to directly annotated instructions, but this pass, which we run unconditionally, emulates the behavior of the previous parser for compatibility with existing programs. It does unintuitive things to programs using the non-nested format because it runs on nested Binaryen IR, so we may want to rethink this at some point.
* Fuzzer: Ignore fuzz testcases that make VMs run out of stack (#6376)Alon Zakai2024-03-041-8/+19
| | | | | | | | | | | | | | | When the stack runs out is observable and optimizations can change it, so we must ignore such testcases. Also add some logic to help debug stuff like this, as suggested by tlively in the past, to add some metrics on the reasons we ignored a testcase. That emits something like this: (ignored 253 iters, for reasons {'too many errors vs calls': 230, '[host limit ': 20, 'uninitialized non-defaultable local': 3}) As a drive by make the metrics print wasm bytes/iter rather than by second (the former is easy to compute from the latter anyhow, and the latter is more interesting I think).
* [Parser] Support prologue and epilogue sourcemap annotations (#6370)Thomas Lively2024-03-047-33/+91
| | | | | | | and fix a bug with sourcemap annotations on folded `if` conditions. Update IRBuilder to apply prologue and epilogue source locations when beginning and ending a function scope. Add basic support in the parser for explicitly tracking annotations on module fields, although only do anything with them in the case of prologue source location annotations.
* OptimizeAddedConstants: Replace an assert with a proper error (#6375)Alon Zakai2024-03-041-2/+5
| | | See #6373
* Typed continuations: cont.bind instructions (#6365)Frank Emrich2024-03-0430-132/+511
| | | | | | | | | | | | | | | | | | | | | | | | This PR is part of a series that adds basic support for the [typed continuations/wasmfx proposal](https://github.com/wasmfx/specfx). This particular PR adds support for the `cont.bind` instruction for partially applying continuations, documented [here](https://github.com/wasmfx/specfx/blob/main/proposals/continuations/Overview.md#instructions). In short, these instructions are of the form `(cont.bind $ct_before $ct_after)` where `$ct_before` and `$ct_after` are related continuation types. They must only differ in the number of arguments, where `$ct_before` has _n_ additional parameters as compared to `$ct_after`, for some _n_ ≥ 0. The idea is that `(cont.bind $ct_before $ct_after)` then takes a reference to a continuation of type `$ct_before` as well as _n_ operands and returns a (reference to a) continuation of type `$ct_after`. Thus, the folded textual representation looks like `(cont.bind $ct_before $ct_after arg1 ... argn c)`. Support for the instruction is implemented in both the old and the new wat parser. Note that this PR does not implement validation of the new instruction.
* Fuzzer: Mark Roundtrip pass as adding effects (#6366)Alon Zakai2024-02-291-0/+6
|
* [Parser] Support inline data in 64-bit memory declarations (#6364)Thomas Lively2024-02-292-8/+35
| | | | This new form of the abbreviated memory declaration with inline data is introduced in the memory64 proposal.
* [Parser] Do not require a memory for GC string ops (#6363)Thomas Lively2024-02-293-12/+55
| | | | | We previously required a memory to exist while parsing all `StringNew` and `StringEncode` instructions, even though some variants of the instructions use GC arrays instead. Require a memory only for those instructions that use one.
* [NFC] Add the type of the Expression when eliding it (#6362)Alon Zakai2024-02-2830-68/+69
| | | | | | | | | | In some cases we don't print an Expression in full if it is unreachable, so we print something instead as a placeholder. This happens in unreachable code when the children don't provide enough info to print the parent (e.g. a StructGet with an unreachable reference doesn't know what struct type to use). This PR prints out the name of the Expression type of such things, which can help debugging sometimes.
* Fuzzing: Remove some obsolete experimental V8 flags (#6347)Alon Zakai2024-02-281-4/+0
| | | | Those flags were removed in V8 as the features are no longer experimental. This PR removes some warnings from being logged (but V8 does not error on them).
* Fuzz V8 Turboshaft (#6360)Alon Zakai2024-02-281-1/+8
|
* C API: Support adding data segments individually (#6346)Lingming Zhang2024-02-284-0/+36
| | | Fixes #6314.
* [NFC] Add some comments about flow in SubtypingDiscoverer and Unsubtyping ↵Alon Zakai2024-02-282-0/+18
| | | | | | | (#6359) I audited all of SubtypingDiscoverer for flow/non-flow constraints and added some comments to clarify things for our future selves if we ever need to generalize it.
* [Outlining] Fixes break reconstruction (#6352)Ashley Nelson2024-02-274-6/+174
| | | Adds new visitBreakWithType and visitSwitchWithType functions to the IRBuilder API. These functions work around an assumption in IRBuilder that the module is being traversed in the fully nested format, i.e., that the destination scope of a break or switch has been visited before visiting the break or switch. Instead, the type of the destination scope is passed to IRBuilder.
* Release 117 (#6358)Alon Zakai2024-02-272-1/+4
|
* Fuzzer: Separate arguments used to make the fuzz wasm from the opts we run ↵Alon Zakai2024-02-271-8/+16
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | on it (#6357) Before FUZZ_OPTS was used both when doing --translate-to-fuzz/-ttf to generate the wasm from the random bytes and also when later running optimizations to generate a second wasm file for comparison. That is, we ended up doing this, if the opts were -O3: wasm-opt random.input -ttf -o a.wasm -O3 wasm-opt a.wasm -O3 -o b.wasm Now we have a pair a.wasm,b.wasm which we can test. However, we have run -O3 on both which is a little silly - the second -O3 might not actually have anything left to do, which would mean we compare the same wasm to itself. Worse, this is incorrect, as there are things we need to do only during the generation phase, like --denan. We need that in order to generate a valid wasm to test on, but it is "destructive" in itself: when removing NaNs (to avoid nondeterminism) if replaces them with 0, which is different. As a result, running --denan when generating the second wasm from the first could lead to different execution in them. This was always a problem, but became more noticable recently now that DeNaN modifies SIMD operations, as one optimization we do is to replace a memory.copy with v128.load + v128.store, and --denan will make sure the loaded value has no NaNs... To fix this, separate the generation and optimization phase. Instead of wasm-opt random.input -ttf -o a.wasm --denan -O3 wasm-opt a.wasm --denan -O3 -o b.wasm (note how --denan -O3 appears twice), do this: wasm-opt random.input -ttf -o a.wasm --denan wasm-opt a.wasm -O3 -o b.wasm (note how --denan appears in generation, and -O3 in optimization).
* SubtypingDiscoverer: Differentiate non-flow subtyping constraints (#6344)Alon Zakai2024-02-274-2/+141
| | | | | | | | | | | | | | | | | | When we do a local.set of a value into a local then we have both a subtyping constraint - for the value to be valid to put in that local - and also a flow of a value, which can then reach more places. Such flow then interacts with casts in Unsubtyping, since it needs to know what can flow where in order to know how casts force us to keep subtyping relations. That regressed in the not-actually-NFC #6323 in which I added the innocuous lines to add subtyping constraints in ref.eq. It seems fine to require that the arms of a RefEq must be of type eqref, but Unsubtyping then assuming those arms flowed into a location of type eqref... which means casts might force us to not optimize some things. To fix this, differentiate the rare case of non-flowing subtyping constraints, which is basically only RefEq. There are perhaps a few more cases (like i31 operations) but they do not matter in practice for Unsubtyping anyhow; I suggest we land this first to undo the regression and then at our leisure investigate the other instructions.
* Update changlog before release [NFC] (#6356)Alon Zakai2024-02-271-3/+18
|
* [NFC] Use ifdef-else in threads.cpp (#6355)Alon Zakai2024-02-271-2/+2
|
* Add a WebAssembly build to release (#6351)Alon Zakai2024-02-272-1/+98
| | | | | | | | | | | | | | | | | | | Simply build wasm-opt with Emscripten and bundle that up. Example build: https://github.com/kripken/binaryen/releases/tag/wasm-build-1 Specifically binaryen-wasm-build-1-wasm.tar.gz Only 1.72 MB, as it's just wasm-opt and not any other tool, so it is much smaller than our other targets. Perhaps we will add more of the tools later as needed (wasm-metadce, wasm-split, etc.). Also update the readme regarding which toolchains use us as a library, that I noticed while editing it to add the release platforms.
* [StringLowering] Lower `stringview_wtf16.get_codeunit` to `charCodeAt` (#6353)Thomas Lively2024-02-263-8/+8
| | | | Previously we lowered this to `getCodePointAt`, which has different semantics around surrogate pairs.
* [Parser] Parse annotations, including source map comments (#6345)Thomas Lively2024-02-269-1188/+2400
| | | | | | | | | | Parse annotations using the standards-track `(@annotation ...)` format as well as the `;;@ source-map:0:1` format. Have the lexer implicitly collect annotations while it skips whitespace and add lexer APIs to access the annotations since the last token was parsed. Collect annotations before parsing each instruction and pass the annotations explicitly to the parser and parser context functions for instructions. Add an API to `IRBuilder` to set a debug location to be attached to the next visited or created instruction and use it from the parser.
* [Emscripten port] Fix core count logic for Emscripten+pthreads (#6350)Alon Zakai2024-02-261-3/+5
| | | | Before this all Emscripten builds would use 1 core, but it is important to allow pthreads builds there to use more.
* Fuzzer: Fix BigInt printing (#6348)Alon Zakai2024-02-261-0/+4
| | | | Before this we only printed the type of a BigInt and not the value.
* [Emscripten port] Improve emcc flags (#6349)Alon Zakai2024-02-261-1/+18
| | | | | | | | | | | | | | | No changes here to binaryen.js/wasm builds. 1. Add a flag to enable pthreads. 2. Use SINGLE_FILE on binaryen.js/.wasm as before, which is nice for library users as they want just a single file to distribute for Binaryen support. For other builds like wasm-opt.js etc. no longer set SINGLE_FILE, as that type of build wants to be a replacement for a normal wasm-opt build as much as possible, so avoid the overhead of SINGLE_FILE. (Previously we disabled SINGLE_FILE also in the case of BUILD_FOR_BROWSER but I don't think we need to special-case that any more.)
* Build and release on aarch64 (#6334)Darren Worrall2024-02-231-3/+12
| | | Fixes #6311
* Implement dropping of active Element Segments (#6343)Alon Zakai2024-02-232-10/+121
| | | | Also rename the existing droppedSegments to droppedDataSegments for clarity.
* Fuzzer: Handle negative i31s (#6341)Alon Zakai2024-02-231-1/+1
|
* [Parser] Condense redundant pop values (#6339)Ashley Nelson2024-02-221-13/+1
| | | A bit of clean-up, changes getBranchValue to use pop().
* Fuzzer: Ignore V8 errors on uninitialized non-defaultable locals (#6337)Alon Zakai2024-02-221-0/+9
| | | | | | | | | | See #5665 #5599, this is an existing issue and we have a workaround for it using --dce, but it does not always work. I seem to be seeing this in higher frequency since landing recent fuzzer improvements, so ignore it. There is some risk of us missing real bugs here (that we validate and V8 does not), but this is a validation error which is not as serious as a difference in behavior. And this is a long-standing issue that hasn't bitten us yet.
* Typed continuations: cont.new instructions (#6308)Frank Emrich2024-02-2230-38/+280
| | | | | | | | | | | | | | | | | This PR is part of a series that adds basic support for the [typed continuations/wasmfx proposal](https://github.com/wasmfx/specfx). This particular PR adds support for the `cont.new` instruction for creating continuations, documented [here(https://github.com/wasmfx/specfx/blob/main/proposals/continuations/Overview.md#instructions). In short, these instructions are of the form `(cont.new $ct)` where `$ct` must be a continuation type. The instruction takes a single (nullable) function reference as its argument, which means that the folded representation of the instruction is of the form `(cont.new $ct (foo ...))`. Support for the instruction is implemented in both the old and the new wat parser. Note that this PR does not implement validation of the new instruction.
* Fuzzer: Adjust feature fuzzing frequency (#6305)Alon Zakai2024-02-221-14/+21
| | | | | | | | | We used to fuzz MVP 1/3, all 1/3, and a mixture 1/3, but that gives far too much priority to the MVP which is increasingly less important. It is also a good idea to give "all" more priority as that enables more initial content to run (the fuzzer will discard initial content if it doesn't validate with the features chosen in the current iteration). Also (NFC) rename POSSIBLE_FEATURE_OPTS to make the code easier to follow.
* Fuzzer: Allow using initial content with V8 (#6327)Alon Zakai2024-02-224-5/+17
| | | | | | | | | | | | | | | One problem was that spec testcases had exports with names that are not valid to write as JS exports.name. For example an export with a - in the name would end up as exports.foo-bar etc. Since #6310 that is fixed as we do not emit such JS (we use the generic fuzz_shell.js script which iterates over the keys in exports with exports[name]). Also fix a few trivial fuzzer issues that initial content uncovered: - Ignore a wat file with invalid utf-8. - Print string literals in the same way from JS as from C++. - Enable the stringref flag in V8. - Remove tag imports (the same as we do for global and function and other imports).
* Fuzzer: Match the logging of i31ref between JS and C++ (#6335)Alon Zakai2024-02-223-21/+68
| | | | | | | | | | | | | JS engines print i31ref as just a number, so we need a small regex to standardize the representation (similar to what we do for funcrefs on the code above). On the C++ side, make it actually print the i31ref rather than treat it like a generic reference (for whom we only print "object"). To do that we must unwrap an externalized i31 as necessary, and add a case for i31 in the printing logic. Also move that printing logic to its own function, as it was starting to get quite long.
generated by cgit v1.2.3 (git 2.39.1) at 2025-08-12 00:57:06 +0000