cvs commit: src/sys/i386/i386 apic_vector.s local_apic.c mp_machdep.c src/sys/i386/include apicvar.h smp.h src/sys/i386/isa clock.c

[Alan Cox]

On Mon, Feb 14, 2005 at 04:30:56PM -0500, John Baldwin wrote:
> On Monday 14 February 2005 01:42 pm, Coleman Kane wrote:
> > I've been testing these since they were posted. All is well thus far!
> 
> I need perf numbers though.  They should work fine, but I really don't know if 
> they will help or hurt performance.  Also, the perf characteristics are 
> likely different on UP vs SMP.
> 

For a 2-processor SMP there is a small but consistent improvement:

Before

while? time make -j4 buildworld > & /dev/null 
while? end
1624.417u 568.837s 20:21.84 179.5%      4198+3213k 21574+3973io 1579pf+0w
1623.516u 593.309s 19:30.11 189.4%      4185+3186k 1702+4248io 364pf+0w
1618.944u 594.313s 19:31.65 188.9%      4173+3190k 1687+5088io 364pf+0w
1622.615u 585.424s 19:32.84 188.2%      4187+3197k 1652+7268io 364pf+0w
1618.329u 597.495s 19:30.06 189.3%      4183+3197k 1595+3850io 364pf+0w
1623.015u 595.645s 19:30.30 189.5%      4177+3187k 1584+3844io 364pf+0w
1621.247u 599.443s 19:31.61 189.5%      4173+3191k 1582+3861io 364pf+0w
1624.092u 593.148s 19:31.17 189.3%      4176+3186k 1603+3860io 364pf+0w
1619.299u 594.466s 19:28.27 189.4%      4185+3190k 1593+3852io 364pf+0w
1622.323u 595.654s 19:33.75 188.9%      4175+3185k 1587+3858io 364pf+0w

After

while? time make -j4 buildworld > & /dev/null 
while? end
1626.759u 560.376s 20:31.59 177.5%      4193+3207k 22718+5042io 1576pf+0w
1618.035u 583.427s 19:25.52 188.8%      4186+3200k 1656+7129io 364pf+0w
1617.674u 589.693s 19:29.04 188.8%      4173+3190k 1694+3868io 364pf+0w
1620.134u 586.273s 19:26.56 189.1%      4182+3195k 1590+3853io 364pf+0w
1618.173u 586.320s 19:28.64 188.6%      4178+3195k 1584+3857io 364pf+0w
1620.728u 584.587s 19:26.34 189.0%      4182+3197k 1582+3857io 364pf+0w
1617.428u 585.268s 19:22.33 189.5%      4186+3195k 1638+3866io 364pf+0w
1618.586u 591.811s 19:27.68 189.2%      4175+3189k 1611+3842io 364pf+0w