================
@@ -868,6 +868,41 @@ RegisterContextUnwind::GetFullUnwindPlanForFrame() {
return arch_default_unwind_plan_sp;
}
+ // Function outlining is a clang feature where common blocks of instructions
+ // from separate functions can be put in a separate utility function, and
+ // the original functions call into the utility function to execute them,
+ // resulting in fewer bytes used for the code section overall.
+ //
+ // The call to the OUTLINED_FUNCTION may not be a normal ABI call (e.g.
+ // on RISCV it might be called `jal t0, OUTLINED_FUNCTION_<nn>` putting the
+ // return address in a temporary register instead of $ra). Tthe unwind
+ // instructions in eh_frame/debug_frame are not correct today for an
+ // OUTLINED_FUNCTION, even when a normal ABI call is made.
+ // Clang does not insert CFI directives in the OUTLINED_FUNCTION so it
----------------
jasonmolenda wrote:
The OUTLINED_FUNCTIONs can make function calls - they can set up a stack frame
in some cases. I don't believe this is solvable in clang today because the
outlining is done at the machine instruction layer, my naive understanding is
that there is a pass looking at the instructions for multiple functions and
finding blocks of identical insns with certain constraints (like not using
pc-relative data or something I'd guess), moving those into an outlined
function. But it doesn't know that those instructions are doing
prologue/epilogue things. It could track that e.g. a non-standard return
address register is being used (it had to insert the `ret t0` at the end of the
instruction block, as well as the `jal t0, OUTLINED_FUNCTION_n` call to it) but
there are definitely cases where prologue/epilogue instructions are outlined
that are very difficult to handle correctly at this layer.
https://github.com/llvm/llvm-project/pull/204500
_______________________________________________
lldb-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
