Ok. Is there a discrepancy in the other (previous) list?
I think you said there was none.
We might need to debug this in some other way. Saving and replaying the
data, or replicating this code in the CPU side to see if there's a
difference because of the way the calls to rt_* are performed, etc.
Yup, I completely agree that 'h' index array needs to be updated. This is
definitely something that will be done later. However, if you have noticed,
we write into the pgm file even before copying back the results.
if (BU_LIST_NON_EMPTY(&(finished_segs.l))) {
(void)rt_boolfinal(, ,
On Thu, Jul 14, 2016 at 3:07 PM, Param Hanji
wrote:
>
>
> Try -P1 if you want only 1 worker process/thread.
>>
>> Yup, this did it. Thank you!
>
>> Another interesting thing that has happened is that the GPU pgm files
>> contains an inverted image (and it is possibly
On Mon, Jul 11, 2016 at 4:02 PM, Param Hanji
wrote:
> The other issue is that the number of partitions per pixel is not
>> necessarily the same as the number of segments. IIRC it may be larger. I
>> changed the patch to only copy the first partition per pixel. This
> Hmmm. There seem to be a couple of issues here:
>
> The boolean weaving code expects seg_in.hit_rayp and seg_out.hit_rayp to
> be pointing to something. These values are mostly used for debug printfs in
> bool.c with the exception of bool.c:bool_max_raynum. So I changed this to
> point to a
On Mon, Jul 4, 2016 at 9:33 PM, Param Hanji
wrote:
> On Mon, Jul 4, 2016 at 9:52 PM Vasco Alexandre da Silva Costa <
> vasco.co...@gmail.com> wrote:
>
>> On Mon, Jul 4, 2016 at 12:30 PM, Param Hanji
>> wrote:
>>
>>> Hi Vasco,
>>>
>>> I
On Mon, Jul 4, 2016 at 9:52 PM Vasco Alexandre da Silva Costa <
vasco.co...@gmail.com> wrote:
> On Mon, Jul 4, 2016 at 12:30 PM, Param Hanji
> wrote:
>
>> Hi Vasco,
>>
>> I tried everything and still no luck. I'm pretty sure just copying inhits
>> and outhits is not
On Mon, Jul 4, 2016 at 12:30 PM, Param Hanji
wrote:
> Hi Vasco,
>
> I tried everything and still no luck. I'm pretty sure just copying inhits
> and outhits is not enough. Going by Sean's explanation of segments and
> partitions, it looks like I have no option but to
Hi Vasco,
I tried everything and still no luck. I'm pretty sure just copying inhits
and outhits is not enough. Going by Sean's explanation of segments and
partitions, it looks like I have no option but to implement partitions on
the GPU.
Unless you know what is wrong or have an alternative
On Mon, Jun 27, 2016 at 11:33 AM, Param Hanji
wrote:
>
> Also the GPU side seg_sti doesn't have the same meaning at all. It's a
>> unique solid id. In the soltabs list. So '0' would be the first valid solid
>> in the soltabs list. So your code should work like this:
>
> Also the GPU side seg_sti doesn't have the same meaning at all. It's a
> unique solid id. In the soltabs list. So '0' would be the first valid solid
> in the soltabs list. So your code should work like this:
>
> unsigned int i=0;
> RT_VISIT_ALL_SOLTABS_START(stp, a->a_rt_i) {
> /*
On Thu, Jun 23, 2016 at 1:04 PM, Param Hanji
wrote:
>
>
> On Wed, Jun 22, 2016 at 7:43 PM Vasco Alexandre da Silva Costa <
> vasco.co...@gmail.com> wrote:
>
>> On Wed, Jun 22, 2016 at 2:36 PM, Param Hanji
>> wrote:
>>
>>> - in copy_to_cpu
On Wed, Jun 22, 2016 at 7:43 PM Vasco Alexandre da Silva Costa <
vasco.co...@gmail.com> wrote:
> On Wed, Jun 22, 2016 at 2:36 PM, Param Hanji
> wrote:
>
>> - in copy_to_cpu if gpu_seg->seg_sti is zero the cpu_seg->seg_stp->st_id
>>> will be uninitialized. worse. you
- in copy_to_cpu if gpu_seg->seg_sti is zero the cpu_seg->seg_stp->st_id
> will be uninitialized. worse. you are writing to an uninitialized pointer.
> what I think you want to do is lookup the seg_stp pointer based on the
> seg_sti id and use that to initialize the cpu->seg_stp pointer field.
>
On Thu, Jun 16, 2016 at 12:31 PM, Param Hanji
wrote:
> Hi,
>
> I tried to call boolweave and boolfinal methods using segments generated
> on the GPU. I'm pretty sure the weaving process works fine, but boolfinal()
> not so much. It returns an empty partition for every
Hi,
I tried to call boolweave and boolfinal methods using segments generated on
the GPU. I'm pretty sure the weaving process works fine, but boolfinal()
not so much. It returns an empty partition for every pixel, which I
confirmed using bu_log() statements.
I'll attach some code. It'd be great
On Wed, Jun 8, 2016 at 8:04 AM, Param Hanji
wrote:
> Another thing I realized is that while attempting to call boolweave() and
> boolfinal(), I'm replicating a lot of the code written in librt/shoot.c.
> This is because the structures used by GPU and CPU storage of
On Tue, Jun 7, 2016 at 8:41 PM Christopher Sean Morrison
wrote:
>
> It's called automatically when you invoke through any of the rt* tools and
> is manually in each saveview script as a simple "end;" directive. The
> command table translates the "end" string into a call to
--
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for
In file do.c, cm_end() calls certain functions which ultimately results in
a call to clt_frame() in primitive_util.c. My question is how do I call the
function cm_end() using the command tab.
On Sat, May 28, 2016 at 1:49 AM Vasco Alexandre da Silva Costa <
vasco.co...@gmail.com> wrote:
> On Fri,
On Fri, May 27, 2016 at 2:33 PM, Param Hanji
wrote:
> Thanks a lot! Any leads on how to use the partition list output by
> boolfinal()? I'll need to convert this back into an array of segs to be
> further processed.
>
Can I just copy inhit and outhit directly for
Thanks a lot! Any leads on how to use the partition list output by
boolfinal()? I'll need to convert this back into an array of segs to be
further processed.
Best,
Param Hanji
On Fri, May 27, 2016 at 5:39 PM Vasco Alexandre da Silva Costa <
vasco.co...@gmail.com> wrote:
> On Fri, May 20, 2016
On Fri, May 20, 2016 at 7:05 AM, Param Hanji
wrote:
> Yes I did notice that earlier. I no longer need to be in the same
> directory as .cl file :).
> Anyway I finished up work on ETO. Here's the patch. Let me know if
> everything is fine, I'll submit it on
On Thu, May 19, 2016 at 8:28 PM Vasco Alexandre da Silva Costa <
vasco.co...@gmail.com> wrote:
> On Thu, May 19, 2016 at 1:52 PM, Param Hanji
> wrote:
>
>> Hi,
>>
>> You could basically copy the segments list results from clt_frame() to
>>> the CPU side and then
On Thu, May 19, 2016 at 1:52 PM, Param Hanji
wrote:
> Hi,
>
> You could basically copy the segments list results from clt_frame() to the
>> CPU side and then process the list with ANSI C rt_boolweave and
>> rt_boolfinal. Then send the processed results from the CPU
> The proposal looks fine. The question is how can we do the unit testing. It
> would be nice if you had a simplified ANSI C version (this can be extracted
> from the existing BRL-CAD code) for each step before coding the OpenCL
> prototypes. In this way you could detect any issues early on in
26 matches
Mail list logo