This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PowerPC register and memory cost model
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Geoff Keating <geoffk at geoffk dot org>, Stan Shebs <shebs at apple dot com>, Dale Johannesen <dalej at apple dot com>
- Cc: Stuart <stuart at apple dot com>, gcc-patches at gcc dot gnu dot org
- Date: Tue, 22 Oct 2002 16:02:19 -0400
- Subject: Re: PowerPC register and memory cost model
- References: <200210221507.LAA24924@makai.watson.ibm.com>
Okay, well this is interesting. I guess this gives us something
more that we need to explore for PowerPC performance, but I do not know
which version to commit in the interim.
I tried the memory_move_cost function I posted yesterday with both
four (4) times the word size (equivalent to GCC default) and six (6) times
the word size (equivalent to current PowerPC port). The interesting
differences are:
COST = 4 COST = 6
ParseVid (INT) 196.0 usecs *166.3 usecs
MPEG (INT) *125.9 usecs 136.7 usecs
PixBlend (INT) 261.4 usecs *254.6 usecs
Rijndael (INT) *120.3 usecs 121.2 usecs
VolInt (FP) *134.3 usecs 140.7 usecs
Comments, recommendations, and suggestions are welcome.
David