Forums

LPC2214 MIPs?

Started by Paulo de Almeida December 19, 2004


Hi,

How fast can the LPC2214 go in term of real MIPs (not
Drystone MIPs) with a 60Mhz clock ?

Thanks

Paulo A.




An Engineer's Guide to the LPC2100 Series


"Real MIPS" is a completely meaningless measure.

With perfect pipelining and no branching or stalling, the ARM core
can 1 instruction per clock (ie. 60 million instructions/sec). That
would, however, be pretty useless code.

Real code will not have perfect pipelining, will branch and will
stall the core so you will get less than that.

Sorry for a "non-answer".

-- CHarles

do --- In , "Paulo de Almeida" <listas@a...>
wrote:
>
>
> Hi,
>
> How fast can the LPC2214 go in term of real MIPs (not
> Drystone MIPs) with a 60Mhz clock ?
>
> Thanks
>
> Paulo A.




??? REAL MIPS ????

We probably all agree (at tleast those who have been dealing with
benchmarks ans MIPS numbers) that MIPS is probably the most useless,
yet most use measure for performance. However, "real MIPS" is
something new to me. How would you propose to measure them?

Somewhat useful benchmarks are those from EEMBC (www.eembc.org) but
these benchmarks are also meaningless for lower end architectures like
the ARM7 because they obviously focus on latest and greatest chips or
simulations, nothing that you or I might actually use today.

Cheers, Bob

--- In , "Paulo de Almeida" <listas@a...> wrote:
>
>
> Hi,
>
> How fast can the LPC2214 go in term of real MIPs (not
> Drystone MIPs) with a 60Mhz clock ?
>
> Thanks
>
> Paulo A.




Real MIPS as opposed to virtual MIPS perhaps? I have a different measure
of processor capability, being biased towards low power consumption.
Thats is given task A how long does it take to complete the task, and
what is the current consumption during execution. This gives you time at
a known current so divide seconds by time, and multiply the mean current
by this and you get something like:-

A 256 tap FIR with precision of n bits requires XMa/second.

The FIR could be replaced with an FFT, or any function that the current
design needs, that is critical

Far more meaningful to me than MIPS, MHz clock, MFLOPS etc. Knowing the
FIR code allows me to rapidly assess the capabilities of the system, and
its suitability in any given design, since I also know the absolute
execution time of the function being tested, and I can then customise
the test for any app.

Al

lpc2100_fan wrote:

>
> ??? REAL MIPS ????
>
> We probably all agree (at tleast those who have been dealing with
> benchmarks ans MIPS numbers) that MIPS is probably the most useless,
> yet most use measure for performance. However, "real MIPS" is
> something new to me. How would you propose to measure them?
>
> Somewhat useful benchmarks are those from EEMBC (www.eembc.org) but
> these benchmarks are also meaningless for lower end architectures like
> the ARM7 because they obviously focus on latest and greatest chips or
> simulations, nothing that you or I might actually use today.
>
> Cheers, Bob
>
> --- In , "Paulo de Almeida" <listas@a...> wrote:
> >
> >
> > Hi,
> >
> > How fast can the LPC2214 go in term of real MIPs (not
> > Drystone MIPs) with a 60Mhz clock ?
> >
> > Thanks
> >
> > Paulo A. > <http://www.netflix.com/Default?mqso`188913" target="_blank" rel="nofollow">http://us.ard.yahoo.com/SIG9qd63gd/M)8184.5639630.6699735.3001176/D=groups/S06554205:HM/EXP03594910/A$95202/R=0/SIGevjk50a/*http://www.netflix.com/Default?mqso`188913 >
> >.





There is only one true test:
Get an eval board and run your code.

The Philips parts have a MAM and wide flash bus which makes them fast
compared to other ARM parts like Atmel.

However, the Philips parts have slow GPIO compared with the Atmel parts.

It all comes down to trying to match your needs. On Monday 20 December 2004 20:40, onestone wrote:
> Real MIPS as opposed to virtual MIPS perhaps? I have a different measure
> of processor capability, being biased towards low power consumption.
> Thats is given task A how long does it take to complete the task, and
> what is the current consumption during execution. This gives you time at
> a known current so divide seconds by time, and multiply the mean current
> by this and you get something like:-
>
> A 256 tap FIR with precision of n bits requires XMa/second.
>
> The FIR could be replaced with an FFT, or any function that the current
> design needs, that is critical
>
> Far more meaningful to me than MIPS, MHz clock, MFLOPS etc. Knowing the
> FIR code allows me to rapidly assess the capabilities of the system, and
> its suitability in any given design, since I also know the absolute
> execution time of the function being tested, and I can then customise
> the test for any app.
>
> Al
>
> lpc2100_fan wrote:
> > ??? REAL MIPS ????
> >
> > We probably all agree (at tleast those who have been dealing with
> > benchmarks ans MIPS numbers) that MIPS is probably the most useless,
> > yet most use measure for performance. However, "real MIPS" is
> > something new to me. How would you propose to measure them?
> >
> > Somewhat useful benchmarks are those from EEMBC (www.eembc.org) but
> > these benchmarks are also meaningless for lower end architectures like
> > the ARM7 because they obviously focus on latest and greatest chips or
> > simulations, nothing that you or I might actually use today.
> >
> > Cheers, Bob
> >
> > --- In , "Paulo de Almeida" <listas@a...> wrote:
> > > Hi,
> > >
> > > How fast can the LPC2214 go in term of real MIPs (not
> > > Drystone MIPs) with a 60Mhz clock ?
> > >
> > > Thanks
> > >
> > > Paulo A.
> >
> >
> > <http://us.ard.yahoo.com/SIG9qd63gd/M)8184.5639630.6699735.3001176/D
> >=groups/S06554205:HM/EXP03594910/A$95202/R=0/SIGevjk50a/*http:/
> >/www.netflix.com/Default?mqso`188913>
> >
> >
> >
> > >. >
> Yahoo! Groups Links >