larserio@IFI.UIO.NO (LarsErikOsterud) (06/16/90)
The version of DHRY.TTP posted on comp.binaries must be insane... It give 630 on a standard ST and 1300 on a 16 Mhz ST, but all tests made by ST freaks here shows over 1000 on a standard ST ??? Lars-Erik / ABK-BBS +47 2132659 / ____ ______ ________________________ Osterud / larserio@ifi.uio.no / /___ / The norwegian ST __________/ ______________________/ ____/ / Klubben, user association
gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) (06/16/90)
In article <CMM.0.88.645477249.larserio@kvart.uio.no> larserio@IFI.UIO.NO (LarsErikOsterud) writes: >The version of DHRY.TTP posted on comp.binaries must be insane... >It give 630 on a standard ST and 1300 on a 16 Mhz ST, but all tests >made by ST freaks here shows over 1000 on a standard ST ??? Different compilers give different Dhrystone ratings depending on how agressive their optimization is. Dhrystone does some very strange things, including copying constant strings into word-aligned buffers. Compilers can "cheat" on this by turning strcpy() into a simple loop. As a result, you can get results as high as 1500 on a standard ST. There are some very interesting and real differences between compilers, but it's hard to use Dhrystone to find which one is the best. Dhrystone is useful to compare the same processor at different clock speeds, but use the same compiler. For example, if you want to compare the speeds of various ST's and Amigas, you could use gcc to compile Dhrystone, and then you would be looking at only the hardware difference, not compilers that may be using optimizations great for Dhrystone but useless for real code. -- "Perhaps I'm commenting a bit cynically, but I think I'm qualified to." - Dan Bernstein
rehrauer@apollo.HP.COM (Steve Rehrauer) (06/16/90)
In article <1990Jun15.194338.8548@murdoch.acc.Virginia.EDU> gl8f@astsun.astro.Virginia.EDU (Greg Lindahl) writes: >In article <CMM.0.88.645477249.larserio@kvart.uio.no> larserio@IFI.UIO.NO (LarsErikOsterud) writes: > >>The version of DHRY.TTP posted on comp.binaries must be insane... >>It give 630 on a standard ST and 1300 on a 16 Mhz ST, but all tests >>made by ST freaks here shows over 1000 on a standard ST ??? > >Different compilers give different Dhrystone ratings depending on how >agressive their optimization is. And what version of Dhrystones you're compiling. "Don't believe Dhrystones. Don't believe Dhrystones. Don't believe..." > Dhrystone does some very strange >things, including copying constant strings into word-aligned buffers. >Compilers can "cheat" on this by turning strcpy() into a simple loop. Pshaw -- true cheaters do it as a mess of longword moves, and forego loop overhead altogether. Veritable cheat-masters take v1.1 of Dhry and do the strcpys on a 680x0 as a pair of MOVEMs that burn all 8 D-regs (taking advantage of the fact that the strings are word-aligned, are 31 bytes long, and that the allocated size of source and destination can be padded to 32 bytes (to keep everything word-aligned)). >As a result, you can get results as high as 1500 on a standard ST. As Greg is suggesting: "Don't believe Dhrystones. Don't believe Dhrystones. Don't believe..." -- >>"Aaiiyeeee! Death from above!"<< | (Steve) rehrauer@apollo.hp.com "Spontaneous human combustion - what luck!"| Apollo Computer (Hewlett-Packard)
hyc@math.lsa.umich.edu (Howard Chu) (06/16/90)
In article <CMM.0.88.645477249.larserio@kvart.uio.no> larserio@IFI.UIO.NO (LarsErikOsterud) writes: >The version of DHRY.TTP posted on comp.binaries must be insane... >It give 630 on a standard ST and 1300 on a 16 Mhz ST, but all tests >made by ST freaks here shows over 1000 on a standard ST ??? I was going to make the same comment... But in fact, it's even worse than that - on my machine with Turbo-16, the figure alternately comes up as 1299.9 or 838.8. (Now, granted, I *might* have done something wrong when I installed my cache-disable switch, but....) I think there's something decidedly wrong with the comp.binaries version. -- -- Howard Chu @ University of Michigan ... the glass is always greener on the side ...