[QUOTE=Ken_g6;264536]So, does anyone know what might be wrong with using this code in 32-bit mode?
[/QUOTE]
I confirm this problem on 32bit linux.
I have no suggestion.:mellow:

Hi ,
llrCUDA Ver.076 based on llrp 3.8.1.
Only support k*2^b-1(small k,b>26458).
Thank you,
[QUOTE]
llrcuda.0.60$ ./llrCUDA -q"3*2^6090515-1" -d
Starting Lucas Lehmer Riesel prime test of 3*2^6090515-1
Using real irrational base DWT, FFT length = 524288
V1 = 3 ; Computing U0...done.
3*2^6090515-1, iteration : 10000 / 6090515 [0.16%]. Time per iteration : 19.401 ms.
3*2^6090515-1, iteration : 20000 / 6090515 [0.32%]. Time per iteration : 3.197 ms.
3*2^6090515-1, iteration : 30000 / 6090515 [0.49%]. Time per iteration : 3.198 ms.

llrcuda.0.76$ ./llrCUDA -q"3*2^6090515-1" -d
Starting Lucas Lehmer Riesel prime test of 3*2^6090515-1
Using real irrational base DWT, FFT length = 655360
V1 = 3 ; Computing U0...done.
3*2^6090515-1, iteration : 10000 / 6090515 [0.16%]. Time per iteration : 25.600 ms.
3*2^6090515-1, iteration : 20000 / 6090515 [0.32%]. Time per iteration : 8.155 ms.
3*2^6090515-1, iteration : 30000 / 6090515 [0.49%]. Time per iteration : 8.156 ms.
[/QUOTE]

Ver. 0.77 [QUOTE]
llrcuda.0.77$ ./llrCUDA -q"3*2^6090515-1" -d
Starting Lucas Lehmer Riesel prime test of 3*2^6090515-1
Using real irrational base DWT, FFT length = 655360
V1 = 3 ; Computing U0...done.
3*2^6090515-1, iteration : 40000 / 6090515 [0.65%]. Time per iteration : 4.579 ms..
[/QUOTE]

Nice work!
 

Hi Msft,

Many congrats for this nice work, the timing is really impressive!
I hope it works really fine!
Best Regards,
Jean

Hi ,Jean Penné
Your program is a great job.
Thank you,

Ver. 0.80
Only support k*2^b-1(small k,b>26458).
[QUOTE]
llrcuda.0.80$ ./llrCUDA -q"3*2^6090515-1" -d
Starting Lucas Lehmer Riesel prime test of 3*2^6090515-1
Using real irrational base DWT, FFT length = 655360
V1 = 3 ; Computing U0...done.
6090515-1, iteration : 90000 / 6090515 [1.47%]. Time per iteration : 4.236 ms..
[/QUOTE]

Fix 32bit issue.:lol:

Ver 0.83
Only support k*2^b-1(small k,b>26458).

32bitLinux
[code]
tarting Lucas Lehmer Riesel prime test of 3*2^26459-1
Using real irrational base DWT, FFT length = 2304
V1 = 3 ; Computing U0...done.
3*2^26459-1 is prime! Time : 22.069 sec.
Starting Lucas Lehmer Riesel prime test of 3*2^51387-1
Using real irrational base DWT, FFT length = 4608
V1 = 3 ; Computing U0...done.
3*2^51387-1 is prime! Time : 28.759 sec.
[/code]
64bitLinux
[code]
Starting Lucas Lehmer Riesel prime test of 3*2^26459-1
Using real irrational base DWT, FFT length = 2304
V1 = 3 ; Computing U0...done.
3*2^26459-1 is prime! Time : 9.770 sec.
Starting Lucas Lehmer Riesel prime test of 3*2^51387-1
Using real irrational base DWT, FFT length = 4608
V1 = 3 ; Computing U0...done.
3*2^51387-1 is prime! Time : 19.044 sec.
[/code]

[QUOTE=msft;283692]Fix 32bit issue.:lol:[/quote]
Sweet! :big grin:
[QUOTE=msft;283692]
Ver 0.83
Only support k*2^b-1(small k,b>26458).
[/quote]
Why? (Why?! WHY! :down:) Can this fix be applied to 0.60? And/or can 0.8x be made as fast and versatile as 0.60?

[QUOTE=Ken_g6;283847]Sweet! :big grin:

Why? (Why?! WHY! :down:) Can this fix be applied to 0.60? And/or can 0.8x be made as fast and versatile as 0.60?[/QUOTE]

Ken, my integration of version 0.83 in your llrcudaboinc-sources is ready for usage.
All compiler warnings are also solved...
[url=http://primegrid.pytalhost.net/llrcuda/llrcudaboinc.083.ronald.tar.gz]llrcudaboinc.083[/url]

PS: The setting "-DUSE_BOINC" produces the known seg-fault.

[QUOTE=rroonnaalldd;283860]Ken, my integration of version 0.83 in your llrcudaboinc-sources is ready for usage.
...
[url]http://primegrid.pytalhost.net/[/url]
....[/QUOTE]
I noticed that the QR in your avatar leads to the site linked to above.
Initially, I was concerned that it might be an ad.

Hi ,Ken_g6
[QUOTE=Ken_g6;283847]Why? (Why?! WHY! :down:) Can this fix be applied to 0.60? And/or can 0.8x be made as fast and versatile as 0.60?[/QUOTE]
0.60 only support power of 2 fft length.

Ver 0.86
Only support k*2^b+-1(small k,b>26458).
END Development.