Update Binaryen and other dependencies #1224
Conversation
| global.set $abi/condition | ||
| i32.const 0 | ||
| global.set $abi/y | ||
| ) |
There was a problem hiding this comment.
it's ok. I told you earlier that binaryen improved some passes)
| i32.lt_s | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 | ||
| i32.gt_s | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 | ||
| i32.le_s | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 | ||
| i32.ge_s | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 | ||
| i32.eq | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 | ||
| i32.eq | ||
| global.set $binary/b | ||
| global.get $binary/i | ||
| i32.const 1 |
| (global $constructor/emptyCtor (mut i32) (i32.const 0)) | ||
| (global $constructor/emptyCtorWithFieldInit (mut i32) (i32.const 0)) | ||
| (global $constructor/emptyCtorWithFieldNoInit (mut i32) (i32.const 0)) | ||
| (global $constructor/none (mut i32) (i32.const 0)) | ||
| (global $constructor/justFieldInit (mut i32) (i32.const 0)) | ||
| (global $constructor/justFieldNoInit (mut i32) (i32.const 0)) | ||
| (global $constructor/ctorReturns (mut i32) (i32.const 0)) | ||
| (global $constructor/ctorConditionallyReturns (mut i32) (i32.const 0)) | ||
| (global $constructor/ctorAllocates (mut i32) (i32.const 0)) | ||
| (global $constructor/ctorConditionallyAllocates (mut i32) (i32.const 0)) |
| global.set $~argumentsLength | ||
| i32.const 2 | ||
| global.set $~argumentsLength | ||
| ) |
There was a problem hiding this comment.
This fixture apparently evaporated entirely
| i32.const 2 | ||
| i32.const 3 | ||
| i32.const 1 | ||
| call $function-types/doAddWithFn<i32> | ||
| call_indirect (type $i32_i32_=>_i32) |
There was a problem hiding this comment.
This is unfortunate, perhaps a regression?
There was a problem hiding this comment.
Hmm, perhaps need some extra tune of passes pipeline. Will try it after this pr
There was a problem hiding this comment.
I get the point!
$function-types/doAddWithFn<i32> was:
(func $function-types/doAddWithFn<i32> (; 5 ;) (param $0 i32) (param $1 i32) (param $2 i32) (result i32)
i32.const 2
global.set $~argumentsLength
local.get $0
local.get $1
local.get $2
call_indirect (type $i32_i32_=>_i32)
)after current optimizations it became:
(func $function-types/doAddWithFn<i32> (; 5 ;) (param $0 i32) (param $1 i32) (param $2 i32) (result i32)
call_indirect (type $i32_i32_=>_i32)
)and inlined so everything alright!
PS Perhaps make sense add one extra directize pass for further optimization. Will investigate later.
| (func $~start (; 1 ;) | ||
| i32.const 0 | ||
| global.set $~argumentsLength | ||
| (func $~setArgumentsLength (param $0 i32) |
There was a problem hiding this comment.
Isn't this a bug? I suppose if the binary never uses the global that's fine, but it seems like anyone inspecting the global would expect this function to modify it.
Just curious what this means.
There was a problem hiding this comment.
Looks like the global is never read, so Binaryen figured it can remove it entirely. If there'd be inspection code, it would be read (somewhere, somewhen), but otherwise it's a blackbox anyway and should be fine to remove.
As usual, this updates Binaryen and other dependencies to the latest and greatest.
Includes
v128.abs<T>for 8/16/32 bit integers incl.i8x16.abs,i16x8.abs,i32x4.absv128.bitmask<T>includingi8x16.bitmask,i16x8.bitmask,i32x4.bitmasktuple.create/tuple.extractpseudo instructions for multi value typesBinaryenTypenow beingusize, unfortunately triggering lots of 32/64 conversion warnings in switches when compiling to Wasm