Andrew Cooke | Contents | Latest | RSS | Previous | Next

C[omp]ute

Welcome to my blog, which was once a mailing list of the same name and is still generated by mail. Please reply via the "comment" links.

Always interested in offers/projects/new ideas. Eclectic experience in fields like: numerical computing; Python web; Java enterprise; functional languages; GPGPU; SQL databases; etc. Based in Santiago, Chile; telecommute worldwide. CV; email.

Personal Projects

Choochoo Training Diary

Last 100 entries

[Programming] React Leaflet; AliExpress Independent Sellers; Applebaum - Twilight of Democracy; [Politics] Back + US Elections; [Programming,Exercise] Simple Timer Script; [News] 2019: The year revolt went global; [Politics] The world's most-surveilled cities; [Bike] Hope Freehub; [Restaurant] Mama Chau's (Chinese, Providencia); [Politics] Brexit Podcast; [Diary] Pneumonia; [Politics] Britain's Reichstag Fire moment; install cairo; [Programming] GCC Sanitizer Flags; [GPU, Programming] Per-Thread Program Counters; My Bike Accident - Looking Back One Year; [Python] Geographic heights are incredibly easy!; [Cooking] Cookie Recipe; Efficient, Simple, Directed Maximisation of Noisy Function; And for argparse; Bash Completion in Python; [Computing] Configuring Github Jekyll Locally; [Maths, Link] The Napkin Project; You can Masquerade in Firewalld; [Bike] Servicing Budget (Spring) Forks; [Crypto] CIA Internet Comms Failure; [Python] Cute Rate Limiting API; [Causality] Judea Pearl Lecture; [Security, Computing] Chinese Hardware Hack Of Supermicro Boards; SQLAlchemy Joined Table Inheritance and Delete Cascade; [Translation] The Club; [Computing] Super Potato Bruh; [Computing] Extending Jupyter; Further HRM Details; [Computing, Bike] Activities in ch2; [Books, Link] Modern Japanese Lit; What ended up there; [Link, Book] Logic Book; Update - Garmin Express / Connect; Garmin Forerunner 35 v 230; [Link, Politics, Internet] Government Trolls; [Link, Politics] Why identity politics benefits the right more than the left; SSH Forwarding; A Specification For Repeating Events; A Fight for the Soul of Science; [Science, Book, Link] Lost In Math; OpenSuse Leap 15 Network Fixes; Update; [Book] Galileo's Middle Finger; [Bike] Chinese Carbon Rims; [Bike] Servicing Shimano XT Front Hub HB-M8010; [Bike] Aliexpress Cycling Tops; [Computing] Change to ssh handling of multiple identities?; [Bike] Endura Hummvee Lite II; [Computing] Marble Based Logic; [Link, Politics] Sanity Check For Nuclear Launch; [Link, Science] Entropy and Life; [Link, Bike] Cheap Cycling Jerseys; [Link, Music] Music To Steal 2017; [Link, Future] Simulated Brain Drives Robot; [Link, Computing] Learned Index Structures; Solo Air Equalization; Update: Higher Pressures; Psychology; [Bike] Exercise And Fuel; Continental Race King 2.2; Removing Lowers; Mnesiacs; [Maths, Link] Dividing By Zero; [Book, Review] Ray Monk - Ludwig Wittgenstein: The Duty Of Genius; [Link, Bike, Computing] Evolving Lacing Patterns; [Jam] Strawberry and Orange Jam; [Chile, Privacy] Biometric Check During Mail Delivery; [Link, Chile, Spanish] Article on the Chilean Drought; [Bike] Extended Gear Ratios, Shimano XT M8000 (24/36 Chainring); [Link, Politics, USA] The Future Of American Democracy; Mass Hysteria; [Review, Books, Links] Kazuo Ishiguro - Never Let Me Go; [Link, Books] David Mitchell's Favourite Japanese Fiction; [Link, Bike] Rear Suspension Geometry; [Link, Cycling, Art] Strava Artwork; [Link, Computing] Useful gcc flags; [Link] Voynich Manuscript Decoded; [Bike] Notes on Servicing Suspension Forks; [Links, Computing] Snap, Flatpack, Appimage; [Link, Computing] Oracle is leaving Java (to die); [Link, Politics] Cubans + Ultrasonics; [Book, Link] Laurent Binet; VirtualBox; [Book, Link] No One's Ways; [Link] The Biggest Problem For Cyclists Is Bad Driving; [Computing] Doxygen, Sphinx, Breathe; [Admin] Brokw Recent Permalinks; [Bike, Chile] Buying Bearings in Santiago; [Computing, Opensuse] Upgrading to 42.3; [Link, Physics] First Support for a Physics Theory of Life; [Link, Bike] Peruvian Frame Maker; [Link] Awesome Game Theory Tit-For-Tat Thing; [Food, Review] La Fabbrica - Good Italian Food In Santiago; [Link, Programming] MySQL UTF8 Broken; [Link, Books] Latin American Authors

© 2006-2017 Andrew Cooke (site) / post authors (content).

Another GPU Success - PyCUDA, Cross-Correlations

From: andrew cooke <andrew@...>

Date: Thu, 6 Dec 2012 19:58:38 -0300

I've spent the last week investigating using GPUs to calculate
cross-correlaions - something that's important for a couple of clients (it's a
very common operation in seismology because it tells you the relative delay
between two seismic signals recorded in different locations).

The fastest way to calculate correlation coefficients (that I know of, at
least), is to work with FFTs, because you can correlate two signals by
multiplying their Fourier transforms.

So I started with some simple Python numpy code that does exactly that for
synthetic data - it takes about 12s to cross-correlate all the different pairs
when you have 100 spectra, each with 10,000 bins.

Replacing that with code on the GPU, using PyCUDA and PyFFT, reduces the time
to 1.5s.  Taking a little more care, with streams, so that some data is being
loaded/unloaded to/from the GPU while other data is being calculated, lowers
that to 1.2s.  And using FFTs tuned for real (ie zero imaginary part) data
(scikits wrapper for CUBLAS) gets the time down to 0.6s.  So the final speedup
is a healthy factor of 20.

Obviously exact timings depend on the amount of data and the hardware used -
but the above is arguably fair, in that both the CPU and GPU are "once good,
now middle of the road".

And that's just for simple correlations.  I could imagine adding extra
processing to (1) reduce time further and (2) reduce the amount of data that
needs to be transferred back to the CPU.  For example: detecting and returning
the maximum value.

PyCUDA was great.  It took some learning, so initial development was slower,
but once I had things running, it was so much easier to play around, swapping
different options, than with C.  If necessary, I could now translate the
system to C, but my hope is that it will work as a standalone "correlation
service", if we can find a suitable interface (eg reading / writing files,
perhaps to an SSD).

One odd wrinkle, that perhaps is me simply not understanding something, is
that the scikits interface seems to be defined in terms of cudaarray
instances, which seem to me to be sugar in PyCUDA - the "fundamental"
interface being an opaque pointer to memory.  This complicates things when you
want low-level control in other parts of the code.

Andrew

Comment on this post