Messages in this thread | | | From | Paul Mackerras <> | Subject | Re: PPC should not use -fno-builtin (was Re: 2.3.47 imac build?) | Date | Mon, 28 Feb 2000 11:20:25 +1100 |
| |
On Mon, 28 Feb 2000, Jeff Garzik wrote:
> So... you are choosing to lose extremely valuable optimizations because > of a few prototype warnings? This logic seems bogus to me... it also
OK, the problem is that gcc expects size_t to be an unsigned int. We use unsigned long for size_t in the kernel because we want it to be 64 bits on 64-bit PPCs (like it is on the other 64-bit platforms). Gcc doesn't like a prototype for memcpy which has an unsigned long for the 3rd argument (unless you use -fno-builtin).
The only optimization we lose with -fno-builtin (that I can see) is that the compiler won't turn explicit calls to memcpy with a small constant 3rd argument (<= 32) into a sequence of inline loads and stores. (It still does small struct copies with inline loads and stores even with -fno-builtin.)
-- Paul Mackerras, Senior Open Source Researcher, Linuxcare, Inc. +61 2 6262 8990 tel, +61 2 6262 8991 fax paulus@linuxcare.com.au, http://www.linuxcare.com.au/ Linuxcare. Support for the revolution.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |