Gc poll without HELPER_METHOD_FRAME #109378
Conversation
…hodDesc fields after the normal fields.
…ss static data. Should result in very similar to native codegen patterns
…tics even when tls jit optimizations are disabled
|
Tagging subscribers to this area: @mangod9 |
src/coreclr/System.Private.CoreLib/src/System/Threading/Thread.CoreCLR.cs
Outdated
Show resolved
Hide resolved
Co-authored-by: Aaron Robinson <[email protected]>
| jnz RarePath | ||
| ret | ||
| RarePath: |
There was a problem hiding this comment.
| jnz RarePath | |
| ret | |
| RarePath: | |
| jnz JIT_PollGCRarePath | |
| ret | |
| JIT_PollGCRarePath: |
| @@ -606,5 +606,6 @@ inline BOOL ClrFlushInstructionCache(LPCVOID pCodeAddr, size_t sizeOfCode, bool | |||
| // | |||
| #define JIT_GetDynamicGCStaticBase JIT_GetDynamicGCStaticBase_SingleAppDomain | |||
| #define JIT_GetDynamicNonGCStaticBase JIT_GetDynamicNonGCStaticBase_SingleAppDomain | |||
| #define JIT_PollGC JIT_PollGC | |||
There was a problem hiding this comment.
Why are we defining the name itself?
| @@ -102,6 +102,8 @@ EXTERN_C void setFPReturn(int fpSize, INT64 retVal); | |||
|
|
|||
| #define FLOAT_REGISTER_SIZE 4 // each register in FloatArgumentRegisters is 4 bytes. | |||
|
|
|||
| #define JIT_PollGC JIT_PollGC | |||
There was a problem hiding this comment.
Same question as above. Why same name?
| cbnz w9, RarePath | ||
| ret | ||
| RarePath |
There was a problem hiding this comment.
| cbnz w9, RarePath | |
| ret | |
| RarePath | |
| cbnz w9, JIT_PollGCRarePath | |
| ret | |
| JIT_PollGCRarePath |
| jnz RarePath | ||
| ret | ||
| RarePath: |
There was a problem hiding this comment.
| jnz RarePath | |
| ret | |
| RarePath: | |
| jnz JIT_PollGCRarePath | |
| ret | |
| JIT_PollGCRarePath: |
| #ifdef JIT_PollGC | ||
| EXTERN_C FCDECL0(void, JIT_PollGC); | ||
| #else | ||
| #define JIT_PollGC NULL | ||
| #endif |
There was a problem hiding this comment.
I assume this has something to do with the defining of JIT_PollGC to the same name. We should avoid this pattern in the final PR and come up with a better mechanism.
| @@ -2608,49 +2608,10 @@ HRESULT EEToProfInterfaceImpl::SetEnterLeaveFunctionHooksForJit(FunctionEnter3 * | |||
|
|
|||
| /*************************************************************/ | |||
| // Slow helper to tailcall from the fast one | |||
| NOINLINE HCIMPL0(void, JIT_PollGC_Framed) | |||
| extern "C" void QCALLTYPE PollGC_Native() | |||
There was a problem hiding this comment.
Is this extra empty function needed? It is not referenced anywhere else. It is labeled with QCALLTYPE but does not appear in the qcallentrypoints.cpp. And this PR is already adding another empty QCALL function ThreadNative_PollGC in ``comsynchronizable.cppthat _is_ referenced inqcallentrypoints.cpp`.
| #else | ||
| HCIMPL0(void, JIT_RareDisableHelper) | ||
| void JIT_RareDisableHelper() | ||
| #endif | ||
| { | ||
| // We do this here (before we set up a frame), because the following scenario |
There was a problem hiding this comment.
Does this comment need to be updated?
Rewrite gc polling logic as a managed helper and remove the stress helper
NOTE: I intend to build a set of assembly stubs to cover the fast path case where we just check the g_TrapReturningThreads flag, but I'd like to see how this works without that change.