1413846 - (wasm-bulkmem) [meta] Implement bulk data copy & initialization proposal

Status: RESOLVED FIXED

(Core :: JavaScript: WebAssembly, task, P2)

Description

[Lars T Hansen [:lth]]

There is a pending proposal for a mechanism to control how shared memory is initialized from constant data in a module (namely, not repeatedly).  This proposal is somewhat stalled but it is necessary and we'll need to implement it, when it makes some progress again.

Issue: https://github.com/WebAssembly/threads/issues/62
Prose: https://github.com/WebAssembly/threads/blob/master/proposals/threads/ConditionalSegmentInitialization.md

Comment 1

[Lars T Hansen [:lth]]

According to Luke the CG settled on "Solution 2" from the spec above, so we can now try to implement that.

Comment 2

[Lars T Hansen [:lth]]

This proposal was merged with the bulk data copy proposal and has its own repo: https://github.com/WebAssembly/bulk-memory-operations

Comment 4

[Anne (:annevk)]

It seems this does not need to block bug 1477743. Reach out to me if it should.

Comment 5

[Lars T Hansen [:lth]]

Anne, this bug has been blocking that bug because (a) bulk memory was originally spun out from the shared memory proposal and (b) the shared memory proposal is much less useful without bulk memory. Indeed upcoming google tools are assuming that if shared memory is available, then bulk memory is too, and we've recently landed code to make sure this is also true in firefox.

Don't know if you want to consider that a blocker or not.

Comment 6

[Lars T Hansen [:lth]]

FWIW, there's a discussion about changing the semantics of zero-length fills from dead passive segments / and or change something else so that the semantics match a memory.fill of zero bytes, which is very lenient. https://github.com/WebAssembly/bulk-memory-operations/issues/124

Comment 7

[Lars T Hansen [:lth]]

I belive we're now done with this.