Status and incomplete fat Precode pointer.
#2780
Conversation
Incomplete fat pointer implementation using Precode.
| @@ -42,6 +42,11 @@ StackCopyLoop: ; copy the arguments to stack top-down t | |||
|
|
|||
| mov rax, [rbx + CallDescrData__dwRegTypeMap] ; save the reg (arg) type map | |||
|
|
|||
| ; Store stub context | |||
| cmp QWORD PTR [rbx + CallDescrData__pStubContextMD], 0 | |||
| jz NoStubContext | |||
There was a problem hiding this comment.
I wonder if it would be a problem to load the r10 unconditionally here since r10 is volatile.
There was a problem hiding this comment.
How do we know that r10 is volatile in this case? I'm hoping we can simply avoid stub sharing in interpreter scenarios.
There was a problem hiding this comment.
The r10 is volatile in both the windows and unix x64 calling convention. And this function is only called from native runtime code, so there is no case when a special calling convention extension that we use for some stubs could expect r10 to be preserved.
There was a problem hiding this comment.
Oh! You mean volatile in the calling convention sense, not volatile in the memory sense. Yes, I think we can unconditionally play with r10. I added the check just in case it was needed by another scenario that I didn't know about.
Update status document.
Incomplete fat pointer implementation using
Precode.