WasmExecutor takes a cache directory #8057
Conversation
That is useful for executors like wasmtime which produces compiled code and can actually benefit from caching under some circumstances
pepyakin
added
A0-please_review
|
@insipx Potentially useful in |
| .map_err(|err| format!("cannot write the cache config: {:?}", err))?; | ||
|
|
||
| config | ||
| .cache_config_load(cache_config_path) |
There was a problem hiding this comment.
I read in the docs that this needs a config feature enabled on wasmtime, but I can't see it anywhere. Am I missing something?
In 0.19 it is not feature gated, in latest (0.21) it is.
There was a problem hiding this comment.
(I assume you meant the cache feature), it is enabled by default as well.
| @@ -267,6 +283,7 @@ impl<D: NativeExecutionDispatch> NativeExecutor<D> { | |||
| default_heap_pages, | |||
| host_functions, | |||
| max_runtime_instances, | |||
| None, | |||
There was a problem hiding this comment.
Does this mean caching is disabled when running substrate node?
There was a problem hiding this comment.
Yes. I deliberately didn't put it here.
The thing is that this change is dictated by needs of polkadot. I believe it should mitigate the stalls we are seeing there right now. Therefore, it is critical to land it quickly. I figured that we should move the question about the cache out of the way.
|
bot merge |
|
Trying merge. |
ghost
merged commit 
This PR allows to specify a cache directory when creating a
WasmExecutor. If specified and wasmtime executor is used, we configure wasmtime to use it to put its compiled artifacts cache there.This change is motivated by the polkadot validation function which has to execute many different wasm blobs as quickly as possible. Right now, the lack of caching leads to severe stalls.
polkadot companion: paritytech/polkadot#2387