[Webkit-unassigned] [Bug 221260] [JSC] Enable WasmLLInt on ARMv7

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=221260

--- Comment #8 from Saam Barati <sbarati at apple.com> ---
Comment on attachment 456911
  --> https://bugs.webkit.org/attachment.cgi?id=456911
v4

View in context: https://bugs.webkit.org/attachment.cgi?id=456911&action=review

> Source/JavaScriptCore/interpreter/CalleeBits.h:47
> +        : m_ptr { reinterpret_cast<void*>(static_cast<uintptr_t>(value >> (PayloadOffset * CHAR_BIT))) }
> +        , m_tag { static_cast<uint32_t>(value >> (TagOffset * CHAR_BIT)) }

why not just decode the value and grab payload and tag?

> Source/JavaScriptCore/runtime/JSCJSValue.h:604
> +ALWAYS_INLINE JSValue jsUnboxedFloat(float f)

nit: Maybe "wasmUnboxedFloat"?

> Source/JavaScriptCore/wasm/WasmB3IRGenerator.cpp:928
> +            if (type == B3::Int32 || type == B3::Float)

no need for "type == B3::Float" here since we're a GPR.

> Source/JavaScriptCore/wasm/WasmB3IRGenerator.cpp:932
>              if (type == B3::Int32 || type == B3::Float)

no need for "type == B3::Int32" here since we're a FPR.

> Source/JavaScriptCore/wasm/WasmCalleeGroup.cpp:171
> +#if ENABLE(WEBASSEMBLY_SIGNALING_MEMORY)

Do we really need to compile out this enum value?

Maybe we can achieve the same perf by adding some helpers that take in a mode and check if it's signaling vs bounds checking, and when !WEBASSEMBLY_SIGNALING_MEMORY, we can always return true from "isBoundsChecking"

> Source/JavaScriptCore/wasm/WasmCallingConvention.cpp:60
> +        for (unsigned i = 0; i < numberOfArgumentJSRs; ++i)

let's put braces here, since that's WK style.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20220407/d2032cb8/attachment.htm>