rs6000.c (struct processor_costs): Add new fields for simple floating point operation...
authorRoger Sayle <roger@eyesopen.com>
Wed, 7 Jul 2004 20:19:43 +0000 (20:19 +0000)
committerRoger Sayle <sayle@gcc.gnu.org>
Wed, 7 Jul 2004 20:19:43 +0000 (20:19 +0000)
commitf0517163a9b573e1e5858490d95787a48c8a4a10
tree695c690ccd4077b3e039f00d90c5f77e390acb45
parent63b8825280dc6eac8ff82c577c15831a150c1c40
rs6000.c (struct processor_costs): Add new fields for simple floating point operations "fp"...

* config/rs6000/rs6000.c (struct processor_costs): Add new fields
for simple floating point operations "fp", double precision
multiplication "dmul", and single and double precision division
"sdiv" and "ddiv".  Update all CPU variant tables as appropriate.
(ppc630_cost): New table split from ppc620_cost, to distinguish
differences in floating point latencies.
(rs6000_override_options): Use ppc630_cost for PROCESSOR_PPC630.
(rs6000_rtx_costs): Add support for single and double precision
floating point addition, subtraction, multiplication, division,
fused-multiply-add, fused-multiply-sub, negation, absolute value
and negative absolute value.  Tweak MEM case to use prefered idiom.

From-SVN: r84226
gcc/ChangeLog
gcc/config/rs6000/rs6000.c