Re: [amibroker] Re: Cuda prototype using 64 cores on graphics card, AmiBroker Email List Archive

That's what I gathered from perusing the info previously i.e. that it's really meant for lower level processing i.e. something that AB could possibly take advantage of internally but not that we could take advantage of at the higher level I would be interested in ...

----- Original Message -----
From: dloyer123
Date: Thursday, July 24, 2008 4:11 pm
Subject: [amibroker] Re: Cuda prototype using 64 cores on graphics card
To: amibroker@xxxxxxxxxxxxxxx

> Not a chance. C code can be run on the cores, but they do not
> support the win32 api. They are optimized for performing
> repeated
> calculations over and over on large chunks of data.
>
> Some products such as photoshop and matlab have plug ins that
> allow
> Cuda to be used as a co-processor. It would be possible to do
> the
> same with Ami broker, but I doubt that the market is large
> enough to
> support the effort. I would love to be proven wrong.
>
> A good description can be found in the programing guide here:
> http://www.nvidia.com/object/cuda_develop.html
>
> The general idea is that they have a similar transister count as
> the
> CPU, but spend the transister budget on execution units rather
> than
> cache, then use a fat pipe to memory.
>
> The current high end cards have 240 cores and new models come
> out
> every 6 months. The card I am using is very modest.
>
> I would not be surpised if it is possible to perform a backtest
> over
> a few years of 5 minute bars at video rates (say >20/sec)
>
> --- In amibroker@xxxxxxxxxxxxxxx, ftonetti@xxx wrote:
> >
> > This is very interesting ... AB Dll's are one thing ... Do you
> think it's possible to run individual instances of AB itself
> with
> CUDA ?
> >
> > ----- Original Message -----
> > From: dloyer123
> > Date: Thursday, July 24, 2008 3:13 pm
> > Subject: [amibroker] Cuda prototype using 64 cores on graphics card
> > To: amibroker@xxxxxxxxxxxxxxx
> >
> > > I was able to get a AmiBroker dll to work with Nvidia CUDA
> drivers.
> > >
> > > These drivers allow C code to run on the graphics shares of
> a
> > > modern
> > > video card. These are the same processors that allow high
> speed
> > > 3d
> > > graphics. Several math intensive applications report a 50-
> 100
> > > fold
> > > performance improvement over running on the host cpu.
> > >
> > > The mid range card that came on my system has 64 cores, each
> > > able to
> > > perform one floating point operation per clock.
> > >
> > > As a simple test, I wrote a AmiBroker plug in, called by
> AFL.
> > >
> > > It calculated the average price (H+L+C)/3 for 60464 bars in 21us.
> > >
> > > This works out to about 8.5GF (billion floating point
> operations
> > > per
> > > second) and 46GB/s memory transfer speed. (read 3 floats and
> > > write
> > > one per bar), (2 floating point adds and 1 multiply per bar)
> > >
> > > The 46GB/s transfer rate is not far from the available
> memory
> > > bandwidth on the card, but the simple test calculation is
> not
> > > very "dense" so, I should be able to get a much higher
> > > calculation
> > > rate once I move more of my code to the graphics cores.
> Several
> > > of
> > > the CUDA demos report > 150GF/s. Memory is the bottleneck of
> > > this
> > > simple test. I used one thread per bar.
> > >
> > > High end graphics cards are available now that would improve
> > > performance by another factor of 2 to 4.
> > >
> > > A few problems:
> > > * The above numbers do not include the time needed to copy
> the
> > > data
> > > from ami to the graphics card or copy the results back. This
> > > time is
> > > much greater than the calculation time in this simple test.
> > > * This is not a general AFL accelerator.
> > >
> > > My goal is to reduce my current 25s backtest time down to <
> 1s
> > > per
> > > pass. To do this, I will need to move the data set for all
> > > symbols
> > > to the graphics card once and make many passes over the data
> > > with
> > > different optimization values. Each CUDA thread will work on
> > > one
> > > symbol, rather than a thread per bar as in my first test.
> > >
> > > There is not much point in writing a CUDA routine to just
> > > execute
> > > directly from AFL code. There is too much overhead. In my
> > > application, the AFL code is a very small part of the total
> time
> > > for
> > > each backtest. Even if I reduced the time to zero, it would
> not
> > > reduce the time per pass very much. Also, the time needed to
> > > copy
> > > the price data on each pass would greatly reduce the
> benefit.
> > > As far
> > > as I can tell, the current Ami API does not allow injecting
> a
> > > externally generated trade list into the backtest, so I will
> > > need to
> > > perform the full backtest and fitness function calculation
> > > externally.
> > >
> > > I had no compatibility problems getting the CUDA api to run
> as a
> > > Ami
> > > plug in.
> > >
> > > Why go to the trouble? Using Fred's IO program would get
> much
> > > of the
> > > same benefit for less trouble, or I could wait until Ami
> finally
> > > supports multi cores, or finds other clever ways to reduce
> the
> > > per
> > > pass overhead. The real answer is that I just had to try it....
> > >
> > >
> > >
> > >
> > >
> > >
> >
>
>
>