>I have a question. It is OK to turn stack pointer addition into
>pop instructions with a scratch register. But I don't see how you can
>turn stack pointer substraction into push instructions with a
>scratch register since push will change the contents of the stack,
>in addition to stack pointer.
On i386 platforms the contents of the stack aren't changed, because the
memory below where the stack pointer points to aren't part of the stack,
so this optimzation is safe. However, on x86_64 ABI platforms there's
128 byte "red zone" below where the stack pointer points to that is
considered part of the stack and since GCC is using that red zone to
"push' values on the stack, this optimization isn't safe.
Ross Ridge
