| Andrew Cooke | Contents | Latest | RSS | Twitter | 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

Lepl parser for Python.

Colorless Green.

Photography around Santiago.

SVG experiment.

Professional Portfolio

Calibration of seismometers.

Data access via web services.

Cache rewrite.

Extending OpenSSH.

C-ORM: docs, API.

Last 100 entries

Interesting Argentinian Author - Antonio Di Benedetto; Useful Thread on MetaPhysics; RAND on fighting online anarchy (2001); Now Is Cat Soft LLC's Chance To Save Up To 32% On Mail; NSA Hacked; Call Center Services for Cat Soft LLC; Very Good LRB Article on Brexit; Nussbaum on Anger; Credit Card Processing for Cat Soft LLC; Discover new movies on demand in our online cinema; Tasting; Credit Card Processing for Cat Soft LLC; Apple + Kiwi Jam; Hit Me; Increase Efficiency with GPS Vehicle Tracking for Cat Soft LLC; Sudoku - CSP + Chaos; Recycling Electronics In Santiago; Vector Displays in OpenGL; Call Center Services for Cat Soft LLC; And Anti-Aliased; OpenGL - Render via Intermediate Texture; And Garmin Connect; Using Garmin Forerunner 230 With Linux; Payroll Service Quotes for Cat Soft LLC; (Beating Dead Horse) StackOverflow; Current State of Justice in China; Now Is Cat Soft LLC's Chance To Save Up To 32% On Mail; Axiom of Determinacy; Ewww; Fee Chaos Book; Course on Differential Geometry; Increase Efficiency with GPS Vehicle Tracking for Cat Soft LLC; Okay, but...; Sparse Matrices, Deep Learning; Sounds Bad; Applebaum Rape; Tomato Chutney v4; Have to add...; Culturally Liberal and Nothing More; Weird Finite / Infinite Result; Your diamond is a beaten up mess; Maths Books; Good Bike Route from Providencia / Las Condes to Panul\; Iain Pears (Author of Complex Plots); Plum Jam; Excellent; More Recently; For a moment I forgot StackOverflow sucked; A Few Weeks On...; Chilean Book Recommendations; How To Write Shared Libraries; Jenny Erpenbeck (Author); Dijkstra, Coins, Tables; Python libraries error on OpenSuse; Deserving Trump; And Smugness; McCloskey Economics Trilogy; cmocka - Mocks for C; Concept Creep (Americans); Futhark - OpenCL Language; Moved / Gone; Fan and USB issues; Burgers in Santiago; The Origin of Icosahedral Symmetry in Viruses; autoenum on PyPI; Jars Explains; Tomato Chutney v3; REST; US Elections and Gender: 24 Point Swing; PPPoE on OpenSuse Leap 42.1; SuperMicro X10SDV-TLN4F/F with Opensuse Leap 42.1; Big Data AI Could Be Very Bad Indeed....; Cornering; Postcapitalism (Paul Mason); Black Science Fiction; Git is not a CDN; Mining of Massive Data Sets; Rachel Kaadzi Ghansah; How great republics meet their end; Raspberry, Strawberry and Banana Jam; Interesting Dead Areas of Math; Later Taste; For Sale; Death By Bean; It's Good!; Tomato Chutney v2; Time ATAC MX 2 Pedals - First Impressions; Online Chilean Crafts; Intellectual Variety; Taste + Texture; Time Invariance and Gauge Symmetry; Jodorowsky; Tomato Chutney; Analysis of Support for Trump; Indian SF; TP-Link TL-WR841N DNS TCP Bug; TP-Link TL-WR841N as Wireless Bridge; Sending Email On Time; Maybe run a command; Sterile Neutrinos; Strawberry and Banana Jam

© 2006-2015 Andrew Cooke (site) / post authors (content).

Some Initial Results for Overlapping Tiles with CUDA

From: "andrew cooke" <andrew@...>

Date: Mon, 28 Jul 2008 20:36:24 -0400 (CLT)

I wrote the following code to simulate (perhaps not exactly) the memory
loads that would occur using CUDA if the data were processed using a
tiling that overlaps (so something like the Matrix example, but with
leaking across the boundaries of the box - perhaps for convolving with a
kernel, for example, or, as in my case, calculating "life").

Using the approach shown in the code (large integer types and a single
overlap) is inefficient because (at least for CUDA 1.0 ad 1.1) the reads
cannot be coalesced - either the half weft is the wrong width, or the
shift (which is less than the half-weft to allow for overlapping) is
wrong.

I'm going to see now if using a smaller integer type and overlapping a
whole half-weft makes more sense (sounds crazy, but might work...).

Andrew


(these is just the core block to give some idea of what's happening)

// run through each tile position
int count = 0;
for (int j = 0; j < nY; j++) {
    for (int i = 0; i < nX; i++) {
        // for each tile, run through the half-warps
        for (int k = 0; k < nHalfWarps; k++) {
            for (int l = 0; l < halfWarp; l++) {
                int localOffset = k * halfWarpWidth + l * word;
                int localX = localOffset % windowX;
                int localY = localOffset / windowX;
                int globalX = i * strideX + localX;
                int globalY = j * strideY + localY;
                int globalOffset = globalY * (*paddedX) + globalX;
                int segStart = globalOffset / segment;
                int segEnd =                                        \
                    (globalOffset + halfWarpWidth - word) / segment;

                if (prop.minor < 2) {
                    // 1.0 and 1.1 are really strict about what will
                    // be coalesced.
                    if (segStart == segEnd) {
                        count = count + 1;
                    } else {
                        count = count + halfWarp;
                    }
                } else {
                    // 1.2 is more lenient and simply groups as
                    // necessary
                    count = count + segEnd - segStart + 1;
                }
            }
        }
    }
}


And the output:

Loads for 1234,1234 using  184,  20 stepping  176,  19
8 bytes/word; 128 segments; 16 half-warp
Best count 3180010 for 184, 20 over 1240,1236

See how the stepping here is 8 bytes in 8 because I used 8 byte ints (even
though I only need 1 bit overlap)

The total number of theeads per block would be 184*20/8 = 460.

Better Code + Numbers

From: "andrew cooke" <andrew@...>

Date: Mon, 28 Jul 2008 21:45:23 -0400 (CLT)

There were a fair number of bugs in teh code above.  Not sure I have it
right yet, but I seem to be getting numbers that make more sense.

So, the possible tactics are:

1 - Use a large integer and overlap only as little as possible.
2 - Use a small integer and overlap by a whole segment
3 - Use a large integer and overlap by a whole segment

For a "very large" (ie each dimension significantly larger than the
largest possible tile dimension) data area, searching only over the
largest tiles (ie given X, calculate Y from memory limitations etc) the
relative numbers of memory loads (smaller the better) are:

1 - 10
2 - 2
3 - 1

So it's clearly better to overlap by a whole segment, even though more
memory is "thrown away" (as expected).  The relative speeds for the two
integer sizes just reflects the sizes themselves (4 v 8 bytes).  Since
larger integers load more slowly this may not be significant.

For the original size I was using (1234 x 1234 bytes) things are less
clear because the size of the tile approaches the size of the data in some
configurations, so tweaking tiles shapes becomes significant (in fact [1]
won out because a tile could cover all the data, but [3] was still close).

Andrew

Comment on this post