the amd solutions look like the best value/performance to me. R7 260x is 14
cores 110W, and about $100. High double precision performance (but high power
draw) is with the 280 cards, the r9 390 has 40 cores and 8gb ram for $329 US,
and R9 nano has 64 cores (expensive but just 175w)
- Ori
Yes you are correct, only cpu icd is supported for intel xeon.
The gpu icd needs a purchase from intel. But I think money should
be spent on purchasing an amd or nvidia card instead.
Пн, 28 сен 2015, jprogramming написал(а):
> My amd apu has integrated GPU. The cpu part supports f64. the gpu par
though opencl (w f64 cpu) does match 0%0.
- Original Message -
From: 'Pascal Jasmin' via Programming
To: "programm...@jsoftware.com"
Cc:
Sent: Monday, September 28, 2015 1:16 PM
Subject: Re: [Jprogramming] Arrayfire bindings
will update library later, but 0^0 matches. 0%0 is 0 on afcp
will update library later, but 0^0 matches. 0%0 is 0 on afcpu but 2 (???) on
opencl gpu.
is there a standard way of providing infinity to C?
- Original Message -
From: Henry Rich
To: programm...@jsoftware.com
Cc:
Sent: Monday, September 28, 2015 12:56 PM
Subject: Re: [Jprogramming] Ar
Remember the J non-standard math:
0 * _ is 0
0^0 is 1
0%0 is 0
Henry Rich
On 9/28/2015 1:32 AM, 'Pascal Jasmin' via Programming wrote:
these lists should tell you what supports double precision and what doesn't.
For nvidia, it might only be some of the 700 series cards
https://en.wikipedia.
My amd apu has integrated GPU. The cpu part supports f64. the gpu part does
not.
If your i7 has integrated graphics, then its gpu may not support opencl at all
(though intel gpu's above hd4000 should though)
https://en.wikipedia.org/wiki/List_of_Intel_graphics_processing_units
to count open
Does your GPU support fp64? OpenCL should already have support for
fp64, on my intel i7 which has only one device. it shows extensions
with cl_khr_fp64.
Пн, 28 сен 2015, jprogramming написал(а):
> I am cheating on the benchmark,actually. I'm doing double precision in J,
> and float in arrayfi
I am cheating on the benchmark,actually. I'm doing double precision in J, and
float in arrayfire for the 250x speed claim on GPU. It does verify that they
are tolerantly equal to J's results, and gets a 1 on my system.
The functions are matmulsFonly (for floats), and multimatmulsF (for array
This font looks pretty good: http://sourcefoundry.org/hack/ . I've been
using it lately.
On Mon, Sep 28, 2015 at 4:38 AM, 'Jon Hough' via Programming <
programm...@jsoftware.com> wrote:
>
> I was trying to dissect one of the answers to my "Complete Graphs"
> question:
>
> dissect '5 5 #:I.,-.+/\
How does this perform with small arrays?
On Sun, Sep 27, 2015 at 11:32 PM, 'Pascal Jasmin' via Programming <
programm...@jsoftware.com> wrote:
> these lists should tell you what supports double precision and what
> doesn't. For nvidia, it might only be some of the 700 series cards
>
> https://en
I was trying to dissect one of the answers to my "Complete Graphs" question:
dissect '5 5 #:I.,-.+/\=i.5'
In the answer I. looked like |., which threw me off for a brief moment.
However, I just rechecked on a Mac and the I. character has serifs, so it is
very clearly distinguished from |. , an
11 matches
Mail list logo
