tmurray

10 ms·

31.
by tmurray 14y ago
Looks like too little too late. If Steamroller is still a year or more out, that will put it in competition with Haswell (DDR4 in the server part, AVX2 with FMA ops, etc.). Considering how much Ivy Bridge is already dominating and those kin
32.
by tmurray 14y ago
I'm surprised nobody's linked the gdb reference card. It's fairly old now, but it's still handy if you don't use gdb that often: http://users.ece.utexas.edu/~adnan/gdb-refcard.pdf
33.
by tmurray 14y ago
quick changes that don't require a full recompilation generally mean intermediate files are lying around, which means you only need to recompile changed files and relink. this is usually pretty fast until you start changing headers or somet
34.
by tmurray 14y ago
I don't think most people would dispute that games in general are worthwhile, but I think there are also arguments to be made about the quality of what Zynga does. Obvious example: http://www.bogost.com/blog/cow_clicker_1.shtml
35.
by tmurray 14y ago
it's not just that you have to recompile for ARM; that would be too easy. Third-party apps on Win8 ARM are Metro-only (or whatever Microsoft is calling Metro now). So okay, port to Metro, you think! Well, you can't use the Win32 API at all
36.
by tmurray 14y ago
any chance you can comment on the fault tolerance proposal in MPI 3? also, do you have any examples of custom network engines on top of IP?
37.
by tmurray 14y ago
do you think we'll see a lot of HPC apps using C++11 for parallelism within a node, or will they stick to MPI for that? most of the apps I've seen are MPI only or OpenMP + MPI--I'm not convinced that C++11 will be very relevant to them beca
38.
by tmurray 14y ago
Foreword: I'm biased, as I worked on CUDA for several years. The conclusions offered by this deck are mostly FUD. First of all, Haswell, the architecture where those transactional memory primitives are available, isn't out for another year.
39.
by tmurray 14y ago
It's good to be exposed to this sort of thing as an engineer. What it teaches you to do is how to make an impact--you can figure out that feature X, despite seeming boring and uncool, will dramatically change your business. That's a great s
40.
by tmurray 14y ago
You don't need preemptive multitasking for a known set of games, though; context switching can occur at frame boundaries (or even finer-grained than that, the GPU just has to go idle in some sense). It's necessary for GPGPU because you may
41.
by tmurray 14y ago
already posted: http://news.ycombinator.com/item?id=4373461
42.
by tmurray 14y ago
First, continuous deployment works fine for web projects, but it's not an option if you've got a packaged product that you have to ship at some point. Agile is not a panacea for these environments. It's important to differentiate between th
43.
by tmurray 14y ago
Gary Frost, the AMD person involved, was also the lead on aparapi, which seems to be in the same vein: http://code.google.com/p/aparapi/
44.
by tmurray 14y ago
Citation needed on the 1/5 power consumption thing. GDDR5 isn't that power hungry, most modern GPUs can basically turn themselves off, and PCIe isn't that power hungry either. Anyway, Intel's in better shape than you think and in a signific
45.
by tmurray 14y ago
it's a horrible article; besides some PR fluff, the description of the laptop is "it's a completely standard Ivy Bridge ultrabook except with Linux preinstalled." no examination of the bizarre claims about designing for Linux versus Windows
46.
by tmurray 14y ago
this sort of thing is why NVIDIA is supporting LLVM: http://nvidianews.nvidia.com/Releases/NVIDIA-Contributes-CUD... in other words, with that and the right frontend, you can take Language X, compile to LLVM IR, and run it through the PTX
47.
by tmurray 14y ago
Buh, I think this article has totally missed the point. First of all, I think the reason why the iPad and iPhone have been so good hardware-wise while most Android devices have been lackluster in comparison is that the vertical integration
48.
by tmurray 14y ago
GPUs do a lot more in mobile today than probably occurs to you off the top of your head if you're not extremely familiar with them. First of all, you have to worry about the actual act of running OpenGL ES applications. This is a non-trivia
49.
by tmurray 14y ago
Microsoft doesn't have its own team to write its own GPU drivers. They have conformance tests for third parties and meaningful penalties for drivers that don't conform (no automatic installing, etc). That's the model you should be thinking
50.
by tmurray 14y ago
I wonder how much of this is hardware versus software. Keep in mind that end-to-end latency has a number of possible sources: - touch digitizer - CPU to do something with the input - GPU to start rendering the result - display to show the c
51.
by tmurray 14y ago
Speaking as an ex-GPU driver guy, you're off by an order of magnitude about the number of engineers it takes to make a solid graphics driver. For a good GPU driver that makes smart choices about display, power management, OGLES, etc, you're
52.
by tmurray 14y ago
reasonable SIMD support (let's limit it to SSE2 and above for "reasonable") has been in every Intel processor since what, Pentium 4? it's not much of a bolt-on anymore. one of the causes of the lack of good programming models for SIMD is t
53.
by tmurray 14y ago
Using a decent burr grinder + AeroPress has completely changed my coffee drinking habits in the past six months in that I almost exclusively make coffee at home versus buying a latte somewhere. I also switched from the normal AeroPress brew
54.
by tmurray 14y ago
whoa: recursion has been supported since Fermi support was released in CUDA 3.0 (there's a stack pointer and a stack frame and everything). what's not supported until GK110 (Kepler 2) is GPU kernels launching/waiting on GPU kernels. general
55.
by tmurray 14y ago
Why couldn't they fork Android in exactly the same way Amazon has done with the Kindle Fire?
56.
by tmurray 14y ago
"Trust but verify" is not micromanagement, it's simply good practice (in management or in engineering).
57.
by tmurray 14y ago
"API is forever. If you expose too much up-front then it increases the surface area of getting things wrong, which developers will be stuck with for a long time or need to deal with eventual deprecations which is fun for no one. Also, more
58.
by tmurray 14y ago
(insert standard disclaimer about being responsible for CUDA here) I'm definitely happy to see more languages with GPU support, but schedulers to distribute work between CPUs and GPUs are a particular interest of mine. The most full-feature
59.
by tmurray 14y ago
I read your post and this comment as arguing for resumes to focus on achievements ("this awesome thing is my fault") rather than tasks ("I did thing X as a means to an end while in role Y"). Is that a fair summary? If so, I think you're off
60.
by tmurray 14y ago
I understand the notion that corporations are amoral entities that exist solely to profit and as such are immune to moral judgments, but I don't think an employee should ignore the final product of the company where he or she is employed an

More ›