foxhill

46 ms·

421.
by foxhill 13y ago
i don't understand what's horrible about OpenCL for all the vendors? aside from beta platforms, i've had all my codes work across platforms that claim to support the full profile. also, gpuocelot is a terrible idea. it's
422.
by foxhill 13y ago
given that a sensical idea of time couldn't really exist before the universe, i believe it is sound to say that current understandings of the universe already state this. just that "forever" is 13.8 billion years long and cou
423.
by foxhill 13y ago
there are plenty of data parallel problems, but general computer usage isn't really one of them. another issue - say the Mill is released. when the next version comes, with different latencies, belt lengths, etc, what does that mean fo
424.
by foxhill 13y ago
not true. an OoO scheduler has a fixed window size with which to re-order. and as soon as you have data-dependent branches or dependencies, your compiler cannot possibly schedule instructions effectively.
425.
by foxhill 13y ago
i've never came across code that has performed better with manual hints than compiler profiling. at best, you can match the compiler, but.. i find that improbable, unless you have only a handful of branches.
426.
by foxhill 13y ago
until i can see some decent attempt at an implementation (either in an FPGA, or an open-source simulator), where i can see cycle counts for some benchmarks, i will be erring on the side of scepticism, primarily because this CPU design still
427.
by foxhill 13y ago
it's better to let the compiler decide (with gcc, you build a branch profile version, then rebuild the code with profiling results). unless you have something unusual, and have some deep insight into your code (unlikely), you're n
428.
by foxhill 13y ago
as you said, reading it is just undefined. the compiler has to do something, and it's improbable that a read before write, followed by another read will result in two different values. if you could produce a piece of code that exhibits
429.
by foxhill 13y ago
paypal claim to have done $44 billion of trade in Q3 2013(1). so lets assume paypal do $4x44 billion of trade for the year. there will only ever be 21 million BTC in existence* , there are much less than that right now (12 million ish). 4x4
430.
by foxhill 13y ago
as cool as this is, it was dated august. it's still (practically) impossible to do a proper LINPACK benchmark with open source tools on AMD GPUs, although this is a step in the right direction, and more importantly, a big blow to CUDA.
431.
by foxhill 13y ago
actually, aside from a few UI issues, i don't really have any complaints with sublimeclang, however, if the time comes that it no longer works with a build of ST3, i'll be more than happy to try and make a fix.
432.
by foxhill 13y ago
are you sure fglrx gets loaded properly? i'd give it a go and poke around your Xorg log file, and ensure everything is as expected - any radeon from 2006+ should be able to handle unity and sublime..!
433.
by foxhill 13y ago
do you know what GPU driver you're using? many times, i have come across people still using the framebuffer driver..
434.
by foxhill 13y ago
but its open source nature means we can add features/fix bugs ourselves. if they gain enough traction, i'm sure they'll end up in the mainline branch.
435.
by foxhill 13y ago
i think this is quite early work, we're a little bit off having stable binary issues - i think quarnster is only after feedback from people comfortable with coding :)
436.
by foxhill 13y ago
i wish quarnster didn't decide to retire his excellent sublimeclang plugin. it's by far the most useful plugin i've come across for sublime (in fact, i've almost come to rely on it), and i fear that his replacement isn&#
437.
by foxhill 13y ago
is your third source accessible from anywhere that doesn't require an account? i'd like to read it.
438.
by foxhill 13y ago
in his defence, cmake is awful. it is my least favourite of all the build systems.
439.
by foxhill 13y ago
if i never have to use VST GUI again, that wont be long enough. honestly i think steinberg were pushing for using native widgets. i'm confused then, as to why the native UI system looks just.. terrible.
440.
by foxhill 13y ago
it does a lot more than cross-platform GUI. it says what it can be used to do on the front page.
441.
by foxhill 13y ago
as a developer, whose primary role is in GPGPU programming, i can't give you an accurate answer. i can tell you, however, that the specialist cards typically are only in super-computers. consumer GPUs are typically faster , and signif
442.
by foxhill 13y ago
so, you're saying that anyone that wants reasonable 3D performance, 2+ displays, and GPGPU support can go fuck themselves? :/ don't get me wrong, open source is great. we need it now more than ever. but this is exactly the wr
443.
by foxhill 13y ago
OpenCL, CUDA
444.
by foxhill 13y ago
5870 has a theoretical peak of 2.7 tflops, a GTX680 has a theoretical peak of 3.1 tflops. even a 7970 has a peak of 3.8 tflops. you can argue (successfully) that it's much easier to achieve a high proportion of peak on GCN cards, but C
445.
by foxhill 13y ago
> If you know the biases.. indeed. > for instance by asking people directly.. a terrible way of accounting for bias. people are often unaware of the bias they impart on what they say.
446.
by foxhill 13y ago
the plural of anecdote is not data - some games work, some games don't. i've had both successes and failures with older games.
447.
by foxhill 13y ago
but what happens when you view a direct link to an image? imo, the rotation data like that should be associated with the file, and respected all the time. it might break a few things, but i'm fairly certain people could figure out the
448.
by foxhill 13y ago
from what i can see, both chrome and safari respect the rotation data.
449.
by foxhill 13y ago
for all y in Y, Y is the subset of all computer programming languages, y is dangerous if you do not understand Y.
450.
by foxhill 13y ago
i used to think like you do, but this changed my mind; int *a; is saying, when you dereference a, you have an integer. and when you think about it - a pointer is the same length for any data type, even pointers to functions

More ›