home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.sys.sun.hardware:3523 alt.sys.sun:2996
- Newsgroups: comp.sys.sun.hardware,alt.sys.sun
- Path: sparky!uunet!utcsri!helios.physics.utoronto.ca!sysmark
- From: sysmark@helios.physics.utoronto.ca (Mark Bartelt)
- Subject: Sparc-10 performance oddity
- Message-ID: <Brw9Dw.Bn@helios.physics.utoronto.ca>
- Sender: news@helios.physics.utoronto.ca (News Administrator)
- Reply-To: mark@cita.toronto.edu
- Organization: University of Toronto Physics/Astronomy/CITA
- Date: Fri, 24 Jul 1992 12:58:43 GMT
- Lines: 54
-
- I've noticed a peculiarity (described below) with the Sparc-10/41. Does
- anyone have a clue as to what's going on? To compare the Sparc-10 with
- older Sparcs, I ran an admittedly crude zeroth-order test of non-FP CPU
- horsepower, namely how much CPU time does it take for "dc" to calculate
- sqrt(2) N times to D decimal places, for various combinations of N and D.
- The test was run three times on a Sparc-1+, a Sparc-2, and a Sparc-10/41.
- (At least I'm *told* it was a 10/41; it was in the booth at a trade show
- earlier this summer.) Here are the performance ratios:
-
- RelativePerf
- Prec Loops 2:1+ 41:2
- ---- ----- ---- ----
- 5 50000 1.72 1.64
- 20 7000 1.73 1.35
- 50 1400 1.72 1.24
- 99 400 1.74 1.20
-
- Note that the performance of the Sparc-2 relative to the Sparc-1+ is
- fairly constant, independent of which test was run. But for the 10/41,
- the performance (relative to the Sparc-2) differed, depending on which
- test was run. (Also, running the *same* test repeatedly gave a wider
- variance of CPU times than one sees on the older Sparcs.)
-
- So, a couple questions: First, even in the best case above, the 10/41
- doesn't perform anywhere near as well as one might hope, based on the
- published MIPS/SPECint figures. Is this expected?
-
- Secondly, why should the performance ratios depend on the precision that
- dc is using? I suppose it could be related to cache hit rates or some
- such, but I can't believe that dc, even at 99 decimal places of precision,
- needs so much memory that the cache hit ratio would drop dramatically.
-
- The script used to get the times is appended below. Does anyone have any
- insights?
-
- Mark Bartelt 416/978-5619
- Canadian Institute for mark@cita.toronto.edu
- Theoretical Astrophysics mark@cita.utoronto.ca
-
- "Clothes not busy being worn are busy drying." - Dylan, on laundry day
-
- ---------------
-
- #!/bin/sh
-
- x1=2vc2vc2vc2vc2vc2vc2vc2vc2vc2vc # 10
- x2=$x1$x1$x1$x1$x1$x1$x1$x1$x1$x1 # 100
- x3=$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2 # 1000
- x4=$x3$x3$x3$x3$x3$x3$x3$x3$x3$x3 # 10000
-
- echo 5k$x4$x4$x4$x4$x4 | time dc
- echo 20k$x3$x3$x3$x3$x3$x3$x3 | time dc
- echo 50k$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2$x2 | time dc
- echo 99k$x2$x2$x2$x2 | time dc
-