kelso@mimsy.umd.edu (Stephen Kelley) (07/23/90)
I compiled and ran dhry 1.1 and diskperf programs (I refuse to use the *B* word) and got some puzzling results (esp w/ diskperf). My config: A3000/25-100, 2MB chip + 2MB 80ns Static Col Fast Manx 5.0(a?) w/ options -DREG=register -c2 -f8 -so (all optimization on) Dhry 1.1: w/ 50000 iterations either 6250 or 5555 dhry/sec w/200000 iterations either 6060 or 5880 dhry/sec Hmmm. I heard rumors of 7700. Anyone care to comment? Also, the -f8 flag didn't do diddley! I compile w/ in on & off and got the same results. PS I ran the same dhry on my Sparc1 @ work and got 18000-20000/sec. Diskperf: now *this* was more disconcerting... I got 26000 bytes/sec read, 19000bytes/sec write @ 512 200000 bytes/sec read @ 8192 580000 bytes/sec read @ 512K What happened here??? Subjectively, my system *seems* fast. When I compiled these programs (500+ lines of code) it took ~10 sec each. (not to mention the general snappiness of commands etc). Deformed minds want to know (;-} -- Real: Stephen Kelley, Welch Library, Johns Hopkins Univ. Internet: stevek@welch.jhu.edu
jesup@cbmvax.commodore.com (Randell Jesup) (07/24/90)
In article <25660@mimsy.umd.edu> kelso@mimsy.umd.edu (Stephen Kelley) writes: >Diskperf: > now *this* was more disconcerting... > I got 26000 bytes/sec read, 19000bytes/sec write @ 512 > 200000 bytes/sec read @ 8192 > 580000 bytes/sec read @ 512K > What happened here??? Subjectively, my system *seems* fast. Diskperf is pretty brain-dead, which is why there are so many variants of it around. DiskSpeed is more reliable as an indicator of true speed, though the Seek/Read test is a bit over-simplistic. That drive you have will probably top 700 or 800 K/s reads on large reads, as opposed to the 580 number from diskperf. Diskperf is also particularily inaccurate at high speeds (one of the reasons DiskSpeed has three different intensity levels). -- Randell Jesup, Keeper of AmigaDos, Commodore Engineering. {uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.cbm.commodore.com BIX: rjesup Common phrase heard at Amiga Devcon '89: "It's in there!"
hclausen@adspdk.CBMNET (Henrik Clausen) (07/27/90)
>In article <25660@mimsy.umd.edu> kelso@mimsy.umd.edu (Stephen Kelley) writes: >My config: > A3000/25-100, 2MB chip + 2MB 80ns Static Col Fast >Diskperf: > now *this* was more disconcerting... > I got 26000 bytes/sec read, 19000bytes/sec write @ 512 > 200000 bytes/sec read @ 8192 > 580000 bytes/sec read @ 512K > What happened here??? Subjectively, my system *seems* fast. Use something other than DiskPerf. DiskSpeed by Michael Sinz seems to be more accurate on fast disks - I got 763685 bytes/sec read @ 256K. Also use the HDToolBox to check the mask value. Mine defaults to 0xfffffe, where 0xfffffffc gives much better speed in fast ram. For instance, the 2.0 Kickstart seems to load in about 1/5 the time it used to. My setup is also A3000 25/100, though only 1M chip/1M fast (yet!). -Henrik -- | Henrik Clausen, Graffiti Data (Fido: 2:230/22.33) | | ...{pyramid|rutgers}!cbmvax!cbmehq!adspdk!hclausen | \_______________AmigaDOS: Amiga's Divine Operating System________________/
jesup@cbmvax.commodore.com (Randell Jesup) (07/31/90)
In article <hclausen.1959@adspdk.CBMNET> hclausen@adspdk.CBMNET (Henrik Clausen) writes: > Also use the HDToolBox to check the mask value. Mine defaults to >0xfffffe, where 0xfffffffc gives much better speed in fast ram. For instance, >the 2.0 Kickstart seems to load in about 1/5 the time it used to. > > My setup is also A3000 25/100, though only 1M chip/1M fast (yet!). If you had a developer machine, make sure you updated your HDToolBox when you upgraded your A3000. The old ones had defaults suited to the A2091, more recent versions are correct for the A3000 (early developer A3000's may have been set up with the old HDToolBox also). HDToolBox is on the A3000 install disk, I think. Henrik is correct that the proper mask for the A3000 is 0xfffffffc, which is what the release A3000 HDToolBox uses. -- Randell Jesup, Keeper of AmigaDos, Commodore Engineering. {uunet|rutgers}!cbmvax!jesup, jesup@cbmvax.cbm.commodore.com BIX: rjesup Common phrase heard at Amiga Devcon '89: "It's in there!"