Hi All,
While reading the code of usr/src/uts/intel/ia32/sys/stack.h, I found
the STACK_ALIGN on amd64 is defined as 16.
In the comments of the same file, it says,
* The end of the input argument area must be aligned on a 16-byte
* boundary; i.e. (%rsp - 8) % 16 == 0 at function entry.
My question is what's the meaning or purpose for end of the input
argument alignment?
The ASSERT_STACK_ALIGNED will use this definition,
#define ASSERT_STACK_ALIGNED() \
{ \
long double __tmp; \
ASSERT((((uintptr_t)&__tmp) & (STACK_ALIGN - 1)) == 0); \
}
#endif
Since ASSERT_STACK_ALIGNED is used in many critical code path in the
kernel, I think the 16bytes alignments must be verify important for
performance on AMD64 platform.
But on AMD64 manual, I just found the 16bytes stack alignment for
16-byte XMM registers load/store optimizations, and I don't think it's
the related to ASSERT_STACK_ALIGNED.
_______________________________________________
opensolaris-code mailing list
opensolaris-code@opensolaris.org
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code
