[mono][jit] Use an mrgctx for all gshared methods on WASM. #82981
Conversation
vargaz
added
NO-REVIEW
|
/azp run runtime-wasm |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-wasm |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-wasm |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-extra-platforms |
|
Azure Pipelines successfully started running 1 pipeline(s). |
vargaz
force-pushed
the
gshared-new
branch
2 times, most recently
from
4b26b61
to
fe193bd
Compare
|
/azp run runtime-wasm |
|
/azp run runtime-extra-platforms |
|
Azure Pipelines successfully started running 1 pipeline(s). |
1 similar comment
|
Azure Pipelines successfully started running 1 pipeline(s). |
vargaz
force-pushed
the
gshared-new
branch
2 times, most recently
from
c1f47c8
to
8203760
Compare
|
/azp run runtime-wasm |
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
/azp run runtime-wasm |
|
Azure Pipelines successfully started running 1 pipeline(s). |
vargaz
changed the title
vargaz
force-pushed
the
gshared-new
branch
2 times, most recently
from
40f1f7e
to
13b04b0
Compare
|
/azp run runtime-ioslike |
|
No pipelines are associated with this pull request. |
| @@ -142,6 +142,12 @@ DEFINE_INT(jiterpreter_interp_entry_queue_flush_threshold, "jiterpreter-interp-e | |||
| DEFINE_INT(jiterpreter_wasm_bytes_limit, "jiterpreter-wasm-bytes-limit", 6 * 1024 * 1024, "Disable jiterpreter code generation once this many bytes of WASM have been generated") | |||
| #endif // HOST_BROWSER | |||
|
|
|||
| #ifdef HOST_WASM | |||
There was a problem hiding this comment.
Should it include HOST_IOS as well?
There was a problem hiding this comment.
Its better to start with one platform to see how it goes.
|
Azure Pipelines successfully started running 1 pipeline(s). |
|
As it might have performance disadvantages, is there a way to test it locally before merging? |
|
The benchmarking suite can be run on wasm locally, not sure about ios. |
|
Also, it could be merged and later reverted if the perf impact is significant. |
Enable it by default on WASM. In this mode, all gshared methods get an mrgctx, which means they can access their data using a simple load from the mrgctx instead of having to call a rgctx fetch trampoline. Upsides: - much simpler. - faster access to gshared data - smaller code and data size in the AOT case - if enabled by default on all platforms, large amount of gshared code can be removed Downsides: - the methods have to initialize their mrgctx in their prolog - on non-wasm platforms, indirect calls to gshared methods (like virtual calls) will need to use rgctx trampolines more often to pass the mrgctx.
|
@vargaz is it ready to be merged after the Preview 4 snap? Is there anything else missing? |
|
It could be merged. |
Good. I will approve it after the Preview 4 snap. |
There was a problem hiding this comment.
LGTM, thank you! Hopefully, we can reduce the size with reasonable speed tradeoff.
P.S. I would leave this to @lambdageek for final approval.
/cc: @SamMonoRT
|
Will see what changes this will do to benchmarks. |
ghost
locked as resolved and limited conversation to collaborators
No description provided.