This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: K6 scheduling definitions for Nnew_ia32_branch
- To: Jan Hubicka <hubicka at atrey dot karlin dot mff dot cuni dot cz>
- Subject: Re: K6 scheduling definitions for Nnew_ia32_branch
- From: Richard Henderson <rth at cygnus dot com>
- Date: Mon, 19 Jul 1999 22:29:10 -0700
- Cc: egcs-patches at egcs dot cygnus dot com
- References: <19990712032646.29172@atrey.karlin.mff.cuni.cz>
On Mon, Jul 12, 1999 at 03:26:46AM +0200, Jan Hubicka wrote:
> Sun Jul 11 18:06:37 EDT 1999 Jan Hubicka <hubicka@freesoft.cz>
> * i386.c (long_memory_operand): New function.
> (ix86_attr_default): Handle NEGNOT type.
> (ix86_adjust_cost): Handle PROCESSOR_K6
> * i386.h (TARGET_SWITCHES): Accept -mk6.
> (CC1_CPU_SPECS): Define -mk6 and -mno-k6.
> (long_memory_operand): Declare.
> * i386.md (type attribute): New type NEGNOT.
> (scheduling definitions): Handle NEGNOT attribute.
> (K6 scheduling parameters): New.
> (mov?i patterns): Use mov $0, reg instead of xor reg,reg.
> (extend?i?i patterns): Do not use vector decoded c?t? instructions
> for K6.
> (mulsi3): Fix length_opcode attribute.
> (xorsi3): Do not use prefixed xorw instruction.
> (neg and not patterns): Use NEGNOT attribute, do not set pent_pair
> explicitly.
> (sh?d patterns): Fix length_opcode attribute.
> (bsf pattern): Fix length_opcode attribute.
> (not to xor peep2): Do conversion for K6 as well.
Applied, minus the -mk6 thing, and plus TARGET_USE_MOV0
predicating the mov?i change.
r~