Re: [hatari-devel] Hatari data cache tests

[Nicolas Pomarède <npomarede@xxxxxxxxxxxx>]

Le 19/06/2015 10:03, Nicolas Pomarède a écrit :
> Le 19/06/2015 01:22, Douglas Little a écrit :
> > Hi,
> >
> > Here are some results from my Falcon using -f030 -videl config switches.
> > It was captured on an RGB monitor but it shouldn't matter since videl
> > isn't fetching.
> >
> > My Falcon has failed NVRAM battery so it boots into random 50/60hz. It's
> > not worth trying to capture metrics with Videl enabled without knowing
> > which mode its in since the results will differ, and my VGA is not handy
> > to connect right now. So somebody else should do that bit if possible :)
> > Might be better to wait until the tests are more exhaustive anyway.
> >
> > The tests themselves are all 'fragment' tests i.e. its set up to test
> > groups of instructions executed together - but many of them are simple
> > single instruction fragments (repeated adjacently many times of course).
> > There are a few multi-instruction fragments (like movem
> > src,[regpattern], movem regpattern,[dst]), each of which would need
> > detailing to make sensible comparisons but there's probably enough to
> > chew on for now with the single moves and addressing modes alone :)
> >
> > D.
>
> Hi
>
> That's great, thanks !
> I don't think I will have time to take this into account for next Hatari
> 1.9, but it will definitely be useful after that.
>
> Nicolas

Hi

I tried your nimbench program, and I get less results than what you 
posted (even if it's the same built from 18/6/2015).

In Hatari CE mode :

Atari NIMBench 1.0 (nimble benchmark)
 built on Jun 18 2015 @ 20:42:16
 - d.m.l -

tasks:                       HW interactions:
                             I|I|I|D|D|D|D|F|B|D
                             C|C|C|C|R|W|W|p|l|s
                             e|i|m|e|i|i|a|u|t|p

addl Dn,Dn                   - - - - - - - - - - 8.03 MIPS (1.00 cycles)
moveb (An)+,Dn               - - - - - - - - - - 2.01 MIPS (4.00 cycles)
movel (An)+,(Ay) x10         - - - - - - - - - - 1.33 MIPS (6.00 cycles)
movel (An)+,Dn               - - - - - - - - - - 2.01 MIPS (4.00 cycles)
moveml (An)+:(Ay) x10        - - - - - - - - - - 1.75 MIPS (4.58 cycles)
movew #imm,(An)              - - - - - - - - - - 2.00 MIPS (4.00 cycles)
movew (An)+,Dn               - - - - - - - - - - 2.01 MIPS (4.00 cycles)
movew (An),(An)              - - - - - - - - - - 2.01 MIPS (4.00 cycles)
movew (An),Dn                - - - - - - - - - - 4.00 MIPS (2.00 cycles)
movew 16(An),Dn              - - - - - - - - - - 4.00 MIPS (2.00 cycles)
movew 8(An,Dn),Dn            - - - - - - - - - - 2.00 MIPS (4.00 cycles)
movew Dn,(An)                - - - - - - - - - - 4.01 MIPS (2.00 cycles)
movew Dn,16(An)              - - - - - - - - - - 2.00 MIPS (4.00 cycles)
nop                          - - - - - - - - - - 8.03 MIPS (1.00 cycles)

tests complete!

There's no DSP test, an no cache on / cache off tests as you posted in 
your results. Should I add more parameters when starting nb.ttp ?

Also, I don't get the mips / cycles conversion : If the falcon runs at 
16 MHZ, 8.03 mips for NOP should give ~2 cycles as in your results :

nop                          x - - - - - - - - - 8.03 MIPS (2.00 cycles)

Why does it give 1.00 cycles when ran under Hatari ?

Don't you use 16 MHz as a fixed constant to convert MIPS to cycles, or 
do you try to measure cpu freq with a timer ?

Nicolas