Add proposal for skipping data segment initialization #65
Conversation
| initializer expression for its offset, and its raw data. | ||
|
|
||
| Since WebAssembly currently does not allow for multiple memories, the memory | ||
| index must be zero. We can repurpose this field as a flags field and use it to |
There was a problem hiding this comment.
Repurpose as in make it impossible to specify multiple memories? Or something else?
There was a problem hiding this comment.
If it means not allowing multiple memories "forever", that doesn't seem wise. IMHO solution 1 looks to be a better approach.
There was a problem hiding this comment.
Sorry, I should make this more clear. The intention is that it becomes a flags field, but one of the bits can be used to specify that there exists a non-zero memory index, which would be a separate field. I'll add that to the proposal.
| @@ -0,0 +1,135 @@ | |||
| # Data segment initialization flag proposal | |||
|
|
|||
| This page describes a proposal for providing a flag to skip data segment | |||
There was a problem hiding this comment.
Although it's not technically an issue yet b/c tables can't be shared, it would be good to keep elem and data segments symmetric and make the same change to both.
| }); | ||
|
|
||
| // worker.js | ||
| let moduleBytes = ...; |
There was a problem hiding this comment.
Mostly a nit, but if the example becomes long-lived, it'd be nice if main.js would postMessage the module to the worker which the worker would receive via onmessage. This also drives home the point of "one module"
* Apply changes to element segments too * Pass module to worker in example * Clarify that resuing the memory/table index as flag field doesn't prevent multiple memories/tables
|
Addressed comments, PTAL |
|
Beautiful, thanks! lgtm as a proposal |
|
Yeah as a proposal this is good. I like 2 better. |
| (memory (export "memory") 1) | ||
|
|
||
| ;; Some value used as a counter. | ||
| (data (i32.const 1) "\0") |
There was a problem hiding this comment.
Is there should be i32.const 0? IIUC, addOne expects counter to be at 0.
This adds a new proposal doc, with two potential solutions as discussed in #62.