This is the mail archive of the
mailing list for the GCC project.
Re: mainline performance regression (observered on SPARC)
From: Dan Nicolaescu <dann@godzilla.ICS.UCI.EDU>
Date: Fri, 24 May 2002 14:17:27 -0700
"David S. Miller" <firstname.lastname@example.org> writes:
> From: Dan Nicolaescu <dann@godzilla.ICS.UCI.EDU>
> Date: Fri, 24 May 2002 13:53:33 -0700
> -mcpu=ultrasparc is v8plus not v9
> Check again, -mcpu=ultrasparc is v9.
Hmm, then something is really fishy, look at the assembler command
/opt/contrib/binutils-2.12.1/bin/as --traditional-format -V -Qy -s
-xarch=v8plusa -o /var/tmp//ccAgaiuN.o /var/tmp//ccvfJa4R.s
What Sun refers to as "v9" is what GCC would call "v9 + 64-bit
pointers". What we call v9 in GCC is just the ability to use all the
float registers in v9 plus execute the v9 instructions, Sun calls