twtw

52 ms·

451.
by twtw 8y ago
> Software was never about making users happy, it was about tricking users, manipulating them into giving you either their money or their time Calm down a bit with your generalities. If you sell the software (or the platform it runs on),
452.
by twtw 8y ago
I'm still upset about the permanent-until-vzeroupper "transition" penalty on nonVEX instructions on skylake after any VEX instruction.
453.
by twtw 8y ago
On a related note, modern GPUs are basically Connection Machines on a chip without the interconnection network.
454.
by twtw 8y ago
Unfortunately I can't think through it now, but I would bet there is a fun connection to be made with 3*5=15. Something about roots more frequently getting repeated, or zeros in the s-plane going to a higher order maybe? I'd be de
455.
by twtw 8y ago
> Most of the people are not using massive complex rendering pipelines that these GPUs have but they are paying in terms of price and wattage. This is not how compute on GPUs works now, or since G80 was released in 2006. The "massiv
456.
by twtw 8y ago
It's on some crazy high resolution display. If you switch to text mode, your resolution drops a ton, and the required pixel pushing power drops accordingly.
457.
by twtw 8y ago
Why? Everything that seems smooth is that way because of the GPU.
458.
by twtw 8y ago
Those useless transforms (are you referring to Laplace and friends?) are pretty useful in EE.
459.
by twtw 8y ago
Except that the Middle Ages in Europe was the same time period as the Islamic Golden Age.
460.
by twtw 8y ago
There are more than 4 companies in Silicon Valley.
461.
by twtw 8y ago
> fremium and ad based internet businesses And the actual silicon businesses. Intel, AMD, Nvidia, Cypress, Broadcom Limited, Marvell, Xilinx, ...
462.
by twtw 8y ago
i dream of cars with an array of speakers around the wheels to cancel the road noise.
463.
by twtw 8y ago
From the paper: "we make no attempt to describe the mechanisms that would be necessary to produce these schedules nor the architectural fea- tures that would be necessary to encode these schedules. We do, however, think that a useful m
464.
by twtw 8y ago
RIP if you are a native speaker too.
465.
by twtw 8y ago
It's a company blog. It exists to highlight applications of the company's products. I don't see why you are bothered, both of those statements are factually true.
466.
by twtw 8y ago
This sort of thing is why I sort of hate energy star everything. Take dishwashers, for example. You can regulate the energy and water usage all you want, but it long ago got to the point where dishwashers that meet regulations basically fai
467.
by twtw 8y ago
Yup, it's awful here, everyone please leave. I hear Seattle is the place to be, go there.
468.
by twtw 8y ago
KPTI hurts (a lot). Syscall latency and architectural impact is dramatically increased with KPTI, in large part due to the TLB flush that is typically required. This is probably why AMD CPUs do not have as big of a drop; KPTI is not (urgent
469.
RenderMan: An Advanced Path Tracing Architecture for Movie Rendering [pdf](graphics.pixar.com)
2 points by twtw 8y ago|0 comments
470.
by twtw 8y ago
Tesla autopilot needs all the short term crutches it can get. Supposed to be fully autonomous since, what, 18 months ago? How about that SF->NY demo?
471.
by twtw 8y ago
Probably current gen GPUs can bridge the gap. The nvidia platform after Drive PX 2 (which is what Tesla ships) uses Volta (with "tensor cores" to speed up fp16) and has a deep learning accelerator ( http://nvdla.org ) fo
472.
by twtw 8y ago
The most interesting question about this is what their motivation was to justify building an in-house ASIC design team. That isn't a super hard thing to do for a company of Tesla's size, but it is nontrivial and seems like a signi
473.
by twtw 8y ago
How did you get one a couple of years ago? Are you talking about the same thing?
474.
by twtw 8y ago
I agree completely, but I would like to point out that FPGAs are fantastic educational tools. Being able to get your hardware design into actual hardware for < $100 is amazing, even though the performance isn't the greatest.
475.
by twtw 8y ago
Not an answer, but I've never heard it called the "Winograd domain." The Winograd transform is an algorithm for computing small discrete Fourier transforms, so its the same Fourier domain.
476.
by twtw 8y ago
Winograd transform probably wouldn't be exposed at the framework layer. I wouldn't be surprised if cuDNN/TensorRT/TPU driver already use it internally.
477.
by twtw 8y ago
https://devblogs.nvidia.com/inside-volta/ Ctrl-F "volta SIMT model"
478.
by twtw 8y ago
In Volta, each "SIMD lane" has its own PC, if that makes it more legit.
479.
by twtw 8y ago
> crucial app like Gmail ignored older browsers; Google started telling me some months ago that my browser is unsupported and random stuff has stopped working every few weeks since. I'm running circa 2015 Safari.
480.
by twtw 8y ago
From https://www.cis.rit.edu/class/simg716/Gauss_History_FFT.pdf > Thus, Gauss' algorithm is as general and powerful as the Cooley-Tukey common factor algorithm and is, in fact, equivalent to a decimation

More ›