Bug 211029 – [JSC] Add fast path for BigInt32 left-shift

• Component: New Bugs
• Status: RESOLVED FIXED

[Yusuke Suzuki]

[JSC] Add fast path for BigInt32 left-shift

[Yusuke Suzuki]

Created attachment 397581 [details]
Patch

[Saam Barati]

Comment on attachment 397581 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=397581&action=review

> Source/JavaScriptCore/runtime/Operations.h:776
> +            int64_t result64 = static_cast<int64_t>(leftInt32) << rightInt32;

you need to check if this is a left shift, right?

[Saam Barati]

Comment on attachment 397581 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=397581&action=review

Nice. r=me

>> Source/JavaScriptCore/runtime/Operations.h:776
>> +            int64_t result64 = static_cast<int64_t>(leftInt32) << rightInt32;
> 
> you need to check if this is a left shift, right?

sorry, I missed the above !isLeft branch

[Yusuke Suzuki]

Comment on attachment 397581 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=397581&action=review

> Source/JavaScriptCore/runtime/Operations.h:765
>              rightInt32 = -rightInt32;

BTW, while this is completely separate issue from this, is this right? What happens if rightInt32 is INT32_MIN?

[Yusuke Suzuki]

Comment on attachment 397581 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=397581&action=review

> Source/JavaScriptCore/runtime/Operations.h:-770
> -            return jsBigInt32(leftInt32 >> std::min(rightInt32, 31));

I thought this is right in terms of int32 spec, but it seems that BigInt is not using rounding. So this is wrong. Will just use std::min.

[Yusuke Suzuki]

Created attachment 397582 [details]
Patch for landing

[Yusuke Suzuki]

Created attachment 397583 [details]
Patch for landing

[Yusuke Suzuki]

Created attachment 397584 [details]
Patch for landing

[Yusuke Suzuki]

Committed r260711: <https://trac.webkit.org/changeset/260711>

[Radar WebKit Bug Importer]

<rdar://problem/62377699>