summaryrefslogtreecommitdiff
path: root/src/passes
Commit message (Collapse)AuthorAgeFilesLines
* Fuzz fix for MemoryPacking on trampled data (#3222)Alon Zakai2020-10-151-3/+66
| | | | | | | | | | | | | I believe originally wasm did not allow overlapping segments, that is, where one memory segment tramples the data from a previous one. But then the spec changed its mind and we allowed it. Binaryen seems to have assumed the original case, and not checked for trampling. If there is a chance of trampling, we cannot optimize out zeros - the zero may have an effect if it tramples data from a previous segment. This does not occur in practice in LLVM output, which is why this wasn't a problem so far, I think. An existing testcase hit this issue, so I split it up.
* Log nested pass names in BINARYEN_PASS_DEBUG=2 (#3214)Alon Zakai2020-10-151-9/+15
| | | | We can't validate or print out the wasm in that case, but at least logging the names as they run can help debug some situations.
* [MemoryPacking] Emit the correct segment indices on memory.init (#3239)Thomas Lively2020-10-141-1/+6
| | | | | | | | | | | This PR fixes a bug in which the segment index of a memory.init instruction was incorrect in some circumstances. Specifically, the first segment index used in output memory.init instructions was always the index of the first segment created from splitting up the corresponding input segment. This was incorrect when the input memory.init had an offset that caused it to skip over that first emitted segment so that the first output memory.init should have referred to a subsequent output segment. Fixes #3225.
* Improve partial evaluation (#3236)Max Graey2020-10-141-2/+11
|
* PickLoadSigns fuzz fix: cannot make an atomic operation signed (#3235)Alon Zakai2020-10-131-0/+4
|
* Optimize power of two float divisions (#3018)Max Graey2020-10-131-5/+44
|
* EmscriptenPIC: Remove internalization of GOT entries (#3211)Sam Clegg2020-10-131-82/+3
| | | | | | | wasm-ld now does this better than binaryen and does it by default when linking and executable and optionally with `-Bsymbolic` when linking a shared library. See https://reviews.llvm.org/D89152
* Added Initial Memory64Lowering pass (#3230)Wouter van Oortmerssen2020-10-134-0/+93
| | | | This pass will convert a module with 64-bit loads and stores accessing a 64-bit memory to a regular 32-bit one. Pointers remain 64-bit but are truncated just before use.
* Optimize unsigned divisions when rhs is negative constant (#2991)Max Graey2020-10-131-7/+22
| | | | | | | | `(uint32_t)x / C` --> `x >= C`, where `C > 2^31` `(uint32_t)x / -1` --> `x != -1` and for `shrinkLevel == 0`: `(uint64_t)x / C` --> `uint64_t(x >= C)`, where `C > 2^63` `(uint64_t)x / -1` --> `x != -1`
* Rename Emscripten EHSjLj functions in wasm backend (#3191)Heejin Ahn2020-10-102-19/+2
| | | | | | | | | | | Now that we are renaming invoke wrappers and `emscripten_longjmp_jmpbuf` in the wasm backend, this deletes all related renaming routines and relevant tests. Depends on #3192. Addresses: #3043 and #3081 Companions: https://reviews.llvm.org/D88697 emscripten-core/emscripten#12399
* RemoveUnusedBrs fuzz fix for switches with a single target and with a value ↵Alon Zakai2020-10-091-5/+12
| | | | | | (#3220) We turn a br_table with a single target into a br, but we reverse the order of the condition and the value when doing so, which we forgot to take into account.
* Refactor naming convention for functions handling tuples (#3196)Max Graey2020-10-094-11/+11
| | | When there are two versions of a function, one handling tuples and the other handling non-tuple values, the previous naming convention was to have "Single" in the name of the non-tuple handling function. This PR simplifies the convention and shortens function names by making the names plural for the tuple-handling version and singular for the non-tuple-handling version.
* Fuzz fix for DuplicateFunctionElimination (#3204)Alon Zakai2020-10-081-1/+3
| | | | The replaceFunctions utility replaced exports by name, but did not check the kind, so it could get confused when names happen to overlap.
* Remove RelooperJumpThreading pass, which was just for fastcomp (#3199)Alon Zakai2020-10-084-286/+0
| | | See emscripten-core/emscripten#11860
* Remove old attempted DSL beginnings from OptimizeInstructions (#3200)Alon Zakai2020-10-083-73/+0
| | | | | | Wasm turned out to not be that good for a DSL for such peephole optimizations, so that never made progress. Meanwhile we have the new matcher stuff which works well.
* Add optimization rules for some shift operations (#3099)Max Graey2020-10-071-0/+35
| | | Specifically, truncates constant shift values that are greater than the number of bits available and optimizes out explicit masking of the shift value that is redundant with the implicit masking performed by shift operations.
* Revert some changes for #3193 (#3197)Max Graey2020-10-061-15/+14
| | | `(signed)x % (i32|i64).min_s ==> (x & (i32|i64).max_s)` is not valid unless compared to zero.
* fast-math: Fold `fp * -1` to `-fp` (#3189)Max Graey2020-10-051-2/+5
|
* Generalize transforms for #3153 (#3193)Max Graey2020-10-051-6/+17
| | | | | | | | | | | | | | Implement a more general (additional) version of #3153 which also handles negative constant divisors: `(int32)x % -4 == 0` --> `(x & 3) == 0` `x % -C_pot == 0` --> `(x & (abs(C_pot) - 1)) == 0` and special two-complement values as well: `(int32)x % 0x80000000 == 0` --> `(x & 0x7fffffff) == 0` `(int64)x % 0x8000000000000000 == 0` --> `(x & 0x7fffffffffffffff) == 0` as separete rules: `(int32)x % 0x80000000` --> `x & 0x7fffffff` `(int64)x % 0x8000000000000000` --> `x & 0x7fffffffffffffff` The [previous pr](https://github.com/WebAssembly/binaryen/pull/3153) didn't use these possibilities.
* Ordering correction fix in OptimizeInstructions for #3047 (#3195)Alon Zakai2020-10-051-2/+12
| | | | | | | | | | | | (found by the fuzzer) It is not valid to replace x | (y | x) ==> y | x, if x, y cannot be reordered. It is also not valid to replace x ^ (y ^ x) ==> y, if x, y cannot be reordered, for a more subtle reason: if they cannot be reordered then y can affect the value of x (the opposite is not possible as we checked x for side effects so that we could remove one copy). If so, then the second appearance of x could be different, if e.g. it reads a local y writes to. Whereas, if it's ok to reorder, then it's ok to do x ^ (y ^ x) ==> x ^ (x ^ y) ==> y.
* Let GenerateDynCalls generate dynCalls for invokes (#3192)Heejin Ahn2020-10-021-5/+87
| | | | | | This moves dynCall generating functionaity for invokes from `EmscriptenGlueGenerator` to `GenerateDynCalls` pass. So now `GenerateDynCalls` pass will take care of all cases we need dynCalls: functions in tables and invokes.
* Optimize "clear bit mask" combination to cyclic rotation over preinverted ↵Max Graey2020-10-011-0/+14
| | | | mask (#3184)
* Add comment about signed => unsigned lowering (#3187)Max Graey2020-10-011-0/+5
|
* Clean up support/bits.h (#3177)Thomas Lively2020-09-301-9/+9
| | | | | Use overloads instead of templates where applicable and change function names from PascalCase to camelCase. Also puts the functions in the Bits namespace to avoid naming conflicts.
* Add --fast-math mode (#3155)Alon Zakai2020-09-301-8/+11
| | | | | | | | | | | | Similar to clang and gcc, --fast-math makes us ignore corner cases of floating-point math like NaN changes and (not done yet) lack of associativity and so forth. In the future we may want to have separate fast math flags for each specific thing, like gcc and clang do. This undoes some changes (#2958 and #3096) where we assumed it was ok to not change NaN bits, but @binji corrected us. We can only do such things in fast math mode. This puts those optimizations behind that flag, adds tests for it, and restores the interpreter to the simpler code from before with no special cases.
* Fold i32.eqz(wrap_i64(x)) -> i64.eqz(x) where possible (#3181)Max Graey2020-09-301-0/+12
| | | Specifically, when `x` has at most 32 bits so that wrapping doesn't change its value.
* Simplify signed remainders compared with zero (#3153)Max Graey2020-09-291-9/+29
| | | | | | Specifically when the divisor is a power of two. `eqz((signed)x % C_pot)` -> `eqz(x & (C_pot - 1))` `(signed)x % C_pot != 0` -> `x & (C_pot - 1) != 0`
* GC: Add stubs for the remaining instructions (#3174)Daniel Wirtz2020-09-292-0/+128
| | | NFC, except adding most of the boilerplate for the remaining GC instructions. Each implementation site is marked with a respective `TODO (gc): theInstruction` in between the typical boilerplate code.
* Add also non-equal with zero simplification for boolean context (#3178)Max Graey2020-09-291-2/+3
|
* Lower signed binops to unsigned binops when possible (#2988)Max Graey2020-09-281-10/+59
| | | This can unlock further instruction optimizations that do not apply to signed operations.
* GC: Add i31 instructions (#3154)Daniel Wirtz2020-09-242-0/+22
| | | Adds the `i31.new` and `i31.get_s/u` instructions for creating and working with `i31ref` typed values. Does not include fuzzer integration just yet because the fuzzer expects that trivial values it creates are suitable in global initializers, which is not the case for trivial `i31ref` expressions.
* Remove unused stack-pointer and emscripten-sbrk-ptr pass args from ↵Sam Clegg2020-09-221-52/+0
| | | | | PostEmscripten (#3161) These were removed completely from the emscripten side in #12057
* GC: Add ref.eq instruction (#3145)Daniel Wirtz2020-09-212-0/+19
| | | With `eqref` now integrated, the `ref.eq` instruction can be implemented. The only valid LHS and RHS value is `(ref.null eq)` for now, but implementation and fuzzer integration is otherwise complete.
* GC: Integrate eqref and i31ref types (#3141)Daniel Wirtz2020-09-193-3/+33
| | | Adds the `eqref` and `i31ref` types to their respective code locations. Implements what can be implemented trivially and otherwise traps with a TODO for now. Integration of `eqref` is mostly complete due to it being nullable, just like `anyref`, but `i31ref` needs to remain disabled in the fuzzer because we are lacking the functionality to create trivial `i31ref` values, i.e. `(i31.new (i32.const 0))`, which is left for follow-ups to implement.
* Expression matching API (#3134)Thomas Lively2020-09-181-352/+368
| | | | | | | | | | | Provides an easily extensible layered API for matching expression patterns and extracting their components. The low-level API provides modular building blocks for creating matchers for any data type and the high-level API provides a succinct and flexible interface for matching expressions and extracting useful information from them. Matchers are currently provided for Const, Unary, Binary, and Select instructions. Adding a matcher for a new type of expression is straightforward enough that I expect to add them as they become useful as part of other changes.
* Initial implementation of "Memory64" proposal (#3130)Wouter van Oortmerssen2020-09-188-98/+129
| | | Also includes a lot of new spec tests that eventually need to go into the spec repo
* Optimize binary operations with 1-bit on lhs and 1 const on rhs (#2948)Max Graey2020-09-171-10/+45
| | | | | | | `expr | 1` --> `1` `expr & 1` --> `expr` `expr == 1` --> `expr` `expr != 1` --> `!expr` where `maxBits(expr) == 1` i.e `expr` is boolean
* Improve testing on Windows (#3142)Wouter van Oortmerssen2020-09-172-12/+9
| | | | | | This PR contains: - Changes that enable/disable tests on Windows to allow for better local testing. - Also changes many abort() into Fatal() when it is really just exiting on error. This is because abort() generates a dialog window on Windows which is not great in automated scripts. - Improvements to CMake to better work with the project in IDEs (VS).
* Unary and binary duplicate expression elimination (#3047)Max Graey2020-09-171-15/+139
| | | | | | | | | | | | | | | | | | Simplifies patterns in which an expression is applied twice to its operands. `abs(abs(x))` -> `abs(x)` `ceil(ceil(x))` -> `ceil(x)` `floor(floor(x))` -> `floor(x)` `trunc(trunc(x))` -> `trunc(x)` `nearest(nearest(x))` -> `nearest(x)` `eqz(eqz(bool(x)))` -> `bool(x)` `sext(sext(x))` -> `sext(x)` `neg(neg(x))` -> `x` `y - (y - x)` -> `x` `(x ^ y) ^ y` -> `x` `(x | y) | y` -> `x | y` `(x & y) & y` -> `x & y` `(x % y) % y` -> `x % y`
* Refactor Host expression to MemorySize and MemoryGrow (#3137)Daniel Wirtz2020-09-175-35/+29
| | | Aligns the internal representations of `memory.size` and `memory.grow` with other more recent memory instructions by removing the legacy `Host` expression class and adding separate expression classes for `MemorySize` and `MemoryGrow`. Simplifies related APIs, but is also a breaking API change.
* Add GC feature flag (#3135)Daniel Wirtz2020-09-171-1/+1
| | | Adds the `--enable-gc` feature flag, so far enabling the `anyref` type incl. subtyping, and removes the temporary `--enable-anyref` feature flag that it replaces.
* Add float operations for isSymmetric util (#3127)Max Graey2020-09-141-2/+25
| | | Add floating point Eq and Ne operators to Properties::isSymmetric. Also treat additional float ops as symmetric specifically in OptimizeInstructions when their operands are known to be non-NaN.
* Implement module and local names in name section (#3115)Daniel Wirtz2020-09-141-0/+4
| | | | | | | | | | | | | | | Adds support for the module and local subsections of the name section plus the respective C and JS APIs to populate and obtain local names. C API: * BinaryenFunctionGetNumLocals(func) * BinaryenFunctionHasLocalName(func, index) * BinaryenFunctionGetLocalName(func, index) * BinaryenFunctionSetLocalName(func, index, name) JS API: * Function.getNumLocals(func) * Function.hasLocalName(func, index) * Function.getLocalName(func, index) * Function.setLocalName(func, index, name)
* Simplify subtracting zero from float expressions (#3125)Max Graey2020-09-131-0/+21
| | | | | | | `x - 0.0` -> `x` `x + (-0.0)` -> `x` `x - (-0.0)` -> `x + 0.0` where `x` is `f32` or `f64`.
* Fix RefNull issues (#3123)Daniel Wirtz2020-09-131-3/+6
| | | | | | | | | * ExpressionAnalyzer: Fix `ref.null ht` equality check to include `ht`. * Precompute: Fix `ref.null ht` expression reuse to also update `ht`. * Fuzzing: Fix `ref.null func` becoming canonicalized to `ref.func $funcref` when evaluating execution results, by adding a check for `isNull`. * Fuzzing: Print actual and expected execution results when aborting. * Tests: Update `if-arms-subtype` test in `optimize-instructions` to check that identical `if` arms become folded while not identical arms are kept.
* Update Pop text format to handle tuples (#3116)Thomas Lively2020-09-111-2/+5
| | | | | | | | | | | Previously Pops were printed as ({type}.pop), and if the popped type was a tuple, something like ((i32, i64).pop) would get printed. However, the parser didn't support pops of anything besides single basic types. This PR changes the text format to be (pop <type>*) and adds support for parsing pops of tuples of basic types. The text format change is designed to make parsing simpler. This change is necessary for writing Poppy IR tests (see #3059) that contain break or return instructions that consume multiple values, since in Poppy IR that requires tuple-typed pops.
* Stop renaming longjmp in wasm-emscripten-finalize (#3111)Alon Zakai2020-09-111-0/+18
| | | | | | | | | | | Instead of finalize renaming emscripten_longjmp_jmpbuf to emscripten_longjmp, do nothing in finalize. But in the optional --post-emscripten pass, rename it there if both exist, so that we don't end up using two imports (other optimization passes can then remove an unneeded import). Depends on emscripten-core/emscripten#12157 to land first so that emscripten can handle both names, and it is just an optimization to have one or the other. See https://github.com/WebAssembly/binaryen/issues/3043
* Add anyref feature and type (#3109)Daniel Wirtz2020-09-103-9/+26
| | | Adds `anyref` type, which is enabled by a new feature `--enable-anyref`. This type is primarily used for testing that passes correctly handle subtype relationships so that the codebase will continue to be prepared for future subtyping. Since `--enable-anyref` is meaningless without also using `--enable-reference-types`, this PR also makes it a validation error to pass only the former (and similarly makes it a validation error to enable exception handling without enabling reference types).
* Poppy IR wast parsing and validation (#3105)Thomas Lively2020-09-091-1/+2
| | | | | Adds an IR profile to each function so the validator can determine which validation rules to apply and adds a flag to have the wast parser set the profile to Poppy for testing purposes.
* Update reference types (#3084)Daniel Wirtz2020-09-096-26/+10
| | | | | | | Align with the current state of the reference types proposal: * Remove `nullref` * Remove `externref` and `funcref` subtyping * A `Literal` of a nullable reference type can now represent `null` (previously was type `nullref`) * Update the tests and temporarily comment out those tests relying on subtyping
generated by cgit v1.2.3 (git 2.39.1) at 2025-08-25 08:08:56 +0000