This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to improve x86 FP -> integer conversions
- To: John Wehle <john at feith dot com>, egcs-patches at egcs dot cygnus dot com
- Subject: Re: Patch to improve x86 FP -> integer conversions
- From: Richard Henderson <rth at cygnus dot com>
- Date: Thu, 15 Apr 1999 18:42:43 -0700
- References: <199904150502.BAA16745@jwlab.FEITH.COM>
On Thu, Apr 15, 1999 at 01:02:03AM -0400, John Wehle wrote:
> Old uops New uops
> fnstcw -4(%ebp) 3 fnstcw -4(%ebp) 3
> movl -4(%ebp),%ecx 1 movb $12,-3(%ebp) 1
Both the old and your proposed code loses in that it writes and
reads from the same memory in different widths. This causes
store buffers to be flushed. Since L1 is write-through, we are
now (potentially) forced out to L2 cache to make this happen.
Using an extra HImode register is definitely the best option.
r~