[Rtk-users] [EXTERNAL] Help with CudaForwardProjectionImageFilter

[Rahman, Obaid]

Thanks.

> On Sep 24, 2024, at 1:03 AM, Simon Rit <simon.rit at creatis.insa-lyon.fr> wrote:
> 
> Hi,
> I think there is still a memory bug that has not been fixed but not released yet see here <https://urldefense.us/v2/url?u=https-3A__github.com_RTKConsortium_RTK_issues_612&d=DwMFaQ&c=v4IIwRuZAmwupIjowmMWUmLasxPEgYsgNI-O7C4ViYc&r=J7uT21mkGp7aMwIrHQkTLGwy72wKx_bOB0IkoGp__bQ&m=ICcGM5WJ6bKBe9ILgWociSYXBip6q9mcBsslpccWs-f_tamP8sx2yeDsuRYdL8wG&s=pEgV6drYMSYi1HmTbLLsb7Lq8se32VmWcnnoJGVRFzo&e=>.
> Iterative reconstruction techniques currently load the whole volume and projections in memory so the limit is how much you can fit in GPU memory + intermediate volumes.
> Simon
> 
> On Mon, Sep 23, 2024 at 9:53 PM Rahman, Obaid <rahmano at ornl.gov <mailto:rahmano at ornl.gov>> wrote:
>> Hi Simon,
>> 
>> I have installed the latest openRTK, and the problem has been resolved.
>> I can reconstruct volumes of 1200X1200X1200 which is more than 2^31-1.
>> 
>> Do you know what is the max recon size I can get?
>> Thanks.
>> 
>> Best,
>> Obaid
>> 
>>> On Apr 29, 2024, at 1:50 PM, Rahman, Obaid <rahmano at ornl.gov <mailto:rahmano at ornl.gov>> wrote:
>>> 
>>> Hi Simon,
>>> 
>>> Thank you again for the prompt reply and for catching the issue.
>>> I have opened a new issue <https://urldefense.us/v2/url?u=https-3A__github.com_RTKConsortium_RTK_issues_594&d=DwMFaQ&c=v4IIwRuZAmwupIjowmMWUmLasxPEgYsgNI-O7C4ViYc&r=J7uT21mkGp7aMwIrHQkTLGwy72wKx_bOB0IkoGp__bQ&m=ICcGM5WJ6bKBe9ILgWociSYXBip6q9mcBsslpccWs-f_tamP8sx2yeDsuRYdL8wG&s=VtfdNavyCjBJDPt8GNpEUW-uu6blyoaFNyGi8zlBMxA&e=> on Github.
>>> Thanks!
>>> 
>>> Best,
>>> Obaid
>>> 
>>>> On Apr 29, 2024, at 11:40 AM, Simon Rit <simon.rit at creatis.insa-lyon.fr <mailto:simon.rit at creatis.insa-lyon.fr>> wrote:
>>>> 
>>>> Thanks. I think I understood the issue. Your code indicates this line for the initial error:
>>>> https://github.com/SimonRit/RTK/blob/86d1e83d5341a6a7a85400d550d4759b46653522/src/rtkCudaUtilities.cu#L115 <https://urldefense.us/v2/url?u=https-3A__github.com_SimonRit_RTK_blob_86d1e83d5341a6a7a85400d550d4759b46653522_src_rtkCudaUtilities.cu-23L115&d=DwMFaQ&c=v4IIwRuZAmwupIjowmMWUmLasxPEgYsgNI-O7C4ViYc&r=J7uT21mkGp7aMwIrHQkTLGwy72wKx_bOB0IkoGp__bQ&m=dX-bgc0aI76oeRMh3MT7_0hyVzTmxOdz7j3ilY3C85ebAPukM8PNxaaKOMC36gHw&s=lQAUBe9IGLcHtcPqqiWGjq1BYkPWmQBQsyqrbZGZssA&e=>
>>>> The numel variable two lines above is computed as an int, which max value (2^31-1=2147483647) is below 1264*1356*1356=2324162304. Larger GPU memory comes with new bugs! Can you report your issue on github and I'll fix this quickly?
>>>> Thanks in advance,
>>>> Simon
>>> 
>> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.creatis.insa-lyon.fr/pipermail/rtk-users/attachments/20240924/a8ea4d5d/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 1611 bytes
Desc: not available
URL: <http://www.creatis.insa-lyon.fr/pipermail/rtk-users/attachments/20240924/a8ea4d5d/attachment.p7s>