GPGPU using Opencl

dvlajkovic · Monday at 08:32

When I run OCLBiss_021 it shows this error:
-ERROR-clEnqueueReadBuffer for counter failed

Code:

C:\Apps\home\OCLBiss_021_test>OCLBiss.exe
Reading from OCLBiss.cfg...Done...

Today is Mon Sep 18 09:23:31 2023
Connected Platform:                                     NVIDIA Corporation
Connected Device:                                       NVIDIA GeForce RTX 4090

PROGRAM_FILE selected:                                  csa_decrypt_1block_004.cl

Device Kernel properties:
        number of Compute Units (CU):                   128
        recommended work group size (local threads):    32
        max work group size:                            256
        Global Memory size:                             25756696576

Number Loops per thread:                                8192
Number of keys per thread:                              8192
Local threads:                                          256
Global threads:                                         1536
Keys per kernel:                                        12582912(C00000)

                *****Important: To stop program and save current status*******
                                PRESS a LONG [ESC] Key and then say 'Y'

                Scroll Lock LED in KeyBoard will be used to Signal Key Found event.
                        Please ensure it is off before continuing

                File Log from Last Searched present:                            (11C16EF51790101)
                Do you what to continue from here?
                If you type 'N' or 'n' will start from config.ini value:        (0000385F624F0000)
                Press Y/N + [ENTER] to Continue (Y)y

 base = (385F750F0000)
 Looking for (0x385F62F94FCABCD5)

BruteForcing for:
SB01:           3C EB DC 17 3C 2B D6 4F 65 16 88 F2 58 D5 97 05

Range: 0000000000C00000

Loop             From             To                kernel Time  Keys per seconds        #Keys Found
00000001 0000385F750F0000 0000385F75CEFFFF [...] 09:23:37

-ERROR-clEnqueueReadBuffer for counter failed
C:\Apps\home\OCLBiss_021_test>

moonbase · Monday at 10:00

cayoenrique said:
....I know moonbase is going to get upset with me.....

@cayoenrique

Not upset at all, I think that you are doing some great work on this subject and I wish you success.
If it would help you to have a faster more modern GPU then we could put together a small crowd funding page and perhaps see if it was practical to post a GPU to you or for you to source one locally.

To the best of my knowledge, your topics recently have been the first for possibly over 10 years to release a new biss BF tool into the public domain.
Nowadays with the correct home/enthusiast PC equipment and CudaBISS, it is practical and possible to BF a biss CW in time to make it useful for a feed, ie, before the feed has ended.
If your OpenCL project progresses to the same level or beyond then it could provide a very useful tool for BF.
The added advantage of your OpenCL tool if I understand correctly is that it can be used for both AMD and NVIDIA GPU, this is a big bonus point.

One question, is it possible to run multiple instances of your OpenCL tool on the same PC at the same time and to nominate which GPU in the PC is used per instance?
With CudaBISS it is possible to use commmand lines to instruct CudaBISS which GPU to use in a multi GPU setup.

azboxgo · Monday at 11:45

dvlajkovic said:
When I run OCLBiss_021 it shows this error:
-ERROR-clEnqueueReadBuffer for counter failed

I got this error too, it goes away if you lower the Global threads value.
It works with 16128 , with 16384 it gives that error, I did not try values in between.

Device Kernel properties:
number of Compute Units (CU): 128
recommended work group size (local threads): 32
max work group size: 256
Global Memory size: 25390809088

Number Loops per thread: 8000
Number of keys per thread: 8000
Local threads: 256
Global threads: 16128
Keys per kernel: 129024000(7B0C000)

@moonbase
See OCLBiss.cfg to select the device, even though I'm not sure what P(latform) encodes... but I have only one 4090 card, maybe it is possible to use also the integrated Intel graphics...

#DEFINITIONS
#DETECTDEVICEENABLE:1 #0=disable 1=enable
#SELECTEDDEVICE: a two digit number "PD" where P is selected Platform and D is selected device. Please note that D nor P can be bigger that 9.
# Please enable for once DETECTDEVICE:1 to detect your platform and devices. And make a note of the desire device.

t51w · Monday at 14:03

moonbase said:
to BF a biss CW in time to make it useful for a feed, ie, before the feed has ended.

you could always record encrypted TS and decrypt later

cayoenrique · Monday at 14:29

@azboxgo
He is already doing all you said:

Code:

Connected Device:                                       NVIDIA GeForce RTX 4090
***
PROGRAM_FILE selected:                                  csa_decrypt_1block_004.cl
***
Number Loops per thread:                                8192
***
Local threads:                                          256
Global threads:                                         1536

Yes you can use Intel internal GPU if available. Never done it but I guess If you have a AMD APU it should show also as an alternative.

Just start with @azboxgo example he seems to have a 4090 and once it goes over the limit then change to bigger numbers.

Code:

Number Loops per thread: 8000
Local threads: 256
Global threads: 16128

To the best of my knowledge, your topics recently have been the first for possibly over 10 years to release a new biss BF tool into the public domain.

It is funny, for almost same years I have offered to create Opencl thread but no one was interested. To be honest I have a few SU old timers users that kept bulling me here. So for most of the time I kept my self on the background.

The added advantage of your OpenCL tool if I understand correctly is that it can be used for both AMD and NVIDIA GPU, this is a big bonus point.

Yes

One question, is it possible to run multiple instances of your OpenCL tool on the same PC at the same time and to nominate which GPU in the PC is used per instance?

Here is where its get trick. The short answer is Yes. Just copy OCLBiss.exe + OCLBiss.cfg + the *.cl that you use in a folder. Make copies of that folder and change the config to do different start points. Same GPU or different.

Now A program like the 014 where it does not uses memory to share and intermediate value between kernels, my best guess is that it should be find. Problem comes if you do share Global memory like in 021. There was no waranties that thememories are kept reserve by each instance. So one instance of the program may be overriding memories used by the other!!

Even when I have played with OpenCL for years, it does not provide me with enough experience. As I only played with version 1.2 general features. The Official answer was to avoid multiple Instances. There is no warranty that devices do actually separate resources!! Now with OpenCl versions over 2, most likely 3.0 and with new GPU it seems to be OK. As they where build to share.

Now there are users in the net, that claims it is perfectly find!
For example you can use Opencl and the render of your screen does not get affected, if you keep your kernels under certain time. That was what the claim.
You gave the sample of Cudabiss with multiple instances.
Maybe was a problem with AMD GPU! I do not know.

With CudaBISS it is possible to use command lines to instruct CudaBISS ....

I am glad you ask again. I did posted an answer, that again got some how lost. I have a feeling have something to do with my mouse button failing.

I have no experience with cudabis. Never read a tutorial. So I do not know how it works.
My answer was that if you had use Cuddabis in multi-thread faction, meaning different instances, and it works for you, that is fine. In particular that if you find the key always then that proves it does work. Or even if you find the key half of the time, I guess it is a valid used of time and energy. That what I ask you guys to ask the original developer.

Now you are saying it is not multiple instances, but in fact a normal setting. This change my answer, because the developer was on purpose building such features. So the Cook/Generals story do not apply.

... to instruct CudaBISS which GPU to use in a multi GPU setup.

I have not built such features. We could in the future. But different GPU can be set as I explain with copies of the folder, with different settings for each different GPU. So the answer can do it manually.

dvlajkovic · Monday at 22:47

azboxgo said:
I got this error too, it goes away if you lower the Global threads value.
It works with 16128 , with 16384 it gives that error, I did not try values in between.

Device Kernel properties:
number of Compute Units (CU): 128
recommended work group size (local threads): 32
max work group size: 256
Global Memory size: 25390809088

Number Loops per thread: 8000
Number of keys per thread: 8000
Local threads: 256
Global threads: 16128
Keys per kernel: 129024000(7B0C000)

try
LOCALTHREADS:256
GLOBALTHREADS:65280
LOOPSPERTHREAD:8000

#PROGRAM_FILE:"csa_decrypt_1block_004.cl"
PROGRAM_FILE:"csa_decrypt_1block_005.cl"
#PROGRAM_FILE:"csa_decrypt_1block_006.cl"

cayoenrique · Tuesday at 20:52

For those that may still have problems correcting the last issues. Here more or less same file without the errors.
And I saw some weird timenig in windows. So I change the timer function to match Linux. Now the kps are more stable. It has nothing to do with functions but time was unreliable.
Modprograms.zip (86.11 KB)

Code:

https://workupload.com/file/hvdcKk44xGJ

dvlajkovic · Tuesday at 21:30

Got even better results with OCLBiss_021M from Modprograms.

Me2019H · Tuesday at 23:32

Reading from OCLBiss.cfg...Done...
Today is Tue Sep 19 23:29:22 2023
OS TYPE:t Windows
OS: Windows_NT
Timer 1 seconds: 1.014002
Connected Platform: Intel(R) Corporation
Connected Device: Intel(R) HD Graphics 4400

OCLBISSVERSION: 021M
PROGRAM_FILE selected: csa_decrypt_1block_005.cl

Device Kernel properties:
number of Compute Units (CU): 20
recommended work group size (local threads): 8
max work group size: 512
Global Memory size: 1360632218

Number Loops per thread: 1024
Number of keys per thread: 1024
Local threads: 64
Global threads: 1280
Keys per kernel: 1310720(140000)

*****Important: To stop program and save current status*******
PRESS a LONG [ESC] Key and then say 'Y'

Scroll Lock LED in KeyBoard will be used to Signal Key Found event.
Please ensure it is off before continuing

File Log from Last Searched present: (11C16EF5E380101)
Do you what to continue from here?
If you type 'N' or 'n' will start from config.ini value: (0000385F624F0000)
Press Y/N + [ENTER] to Continue (Y)y

base = (385F75E70000)
Looking for (0x385F62F94FCABCD5)

BruteForcing for:
SB01: 3C EB DC 17 3C 2B D6 4F 65 16 88 F2 58 D5 97 05

Range: 0000000000140000

Loop From To Time Kernel Total Keys Found
00000001 0000385F75E70000 0000385F75FAFFFF [...] 23:29:31 kps:8.57e+005 8.49e+005 00000000 Total 0:
00000002 0000385F75FB0000 0000385F760EFFFF [...] 23:29:33 kps:7.64e+005 7.57e+005 00000000 Total 0:
00000003 0000385F760F0000 0000385F7622FFFF [...] 23:29:34 kps:8.49e+005 8.40e+005 00000000 Total 0:
00000004 0000385F76230000 0000385F7636FFFF [...] 23:29:36 kps:8.49e+005 8.40e+005 00000001 Total 1: Key 001:38 5F 76 0D 23 BB 58 36
00000005 0000385F76370000 0000385F764AFFFF [...] 23:29:37 kps:8.49e+005 8.32e+005 00000000 Total 1:
00000006 0000385F764B0000 0000385F765EFFFF [...] 23:29:39 kps:7.64e+005 7.57e+005 00000000 Total 1:
00000007 0000385F765F0000 0000385F7672FFFF [...] 23:29:41 kps:7.64e+005 7.64e+005 00000000 Total 1:
00000008 0000385F76730000 0000385F7686FFFF [...] 23:29:42 kps:8.49e+005 8.40e+005 00000000 Total 1:
00000009 0000385F76870000 0000385F769AFFFF [...] 23:29:44 kps:7.64e+005 7.57e+005 00000000 Total 1:
0000000A 0000385F769B0000 0000385F76AEFFFF [...] 23:29:46 kps:7.71e+005 7.64e+005 00000000 Total 1:
0000000B 0000385F76AF0000 0000385F76C2FFFF [...] 23:29:48 kps:7.64e+005 7.57e+005 00000000 Total 1:
0000000C 0000385F76C30000 0000385F76D6FFFF [...] 23:29:49 kps:8.48e+005 8.39e+005 00000001 Total 2: Key 001:38 5F 76 0D CA C6 CA 5A
0000000D 0000385F76D70000 0000385F76EAFFFF [...] 23:29:51 kps:8.49e+005 8.40e+005 00000000 Total 2:
0000000E 0000385F76EB0000 0000385F76FEFFFF [...] 23:29:52 kps:8.40e+005 8.32e+005 00000000 Total 2:
0000000F 0000385F76FF0000 0000385F7712FFFF [...] 23:29:54 kps:8.49e+005 8.40e+005 00000000 Total 2:
00000010 0000385F77130000 0000385F7726FFFF [...] 23:29:56 kps:7.70e+005 7.63e+005 00000000 Total 2:
00000011 0000385F77270000 0000385F773AFFFF [...] 23:29:57 kps:7.62e+005 7.56e+005 00000000 Total 2:
00000012 0000385F773B0000 0000385F774EFFFF [...] 23:29:59 kps:8.40e+005 8.40e+005 00000000 Total 2:
00000013 0000385F774F0000 0000385F7762FFFF [...] 23:30:00 kps:8.40e+005 8.40e+005 00000001 Total 3: Key 001:38 5F 77 0E 5C E4 31 71
00000014 0000385F77630000 0000385F7776FFFF [...] 23:30:02 kps:7.64e+005 7.57e+005 00000001 Total 4: Key 001:38 5F 77 0E 76 5F 8F 64
00000015 0000385F77770000 0000385F778AFFFF [...] 23:30:04 kps:8.49e+005 8.40e+005 00000000 Total 4:
00000016 0000385F778B0000 0000385F779EFFFF [...] 23:30:05 kps:8.49e+005 8.40e+005 00000000 Total 4:
00000017 0000385F779F0000 0000385F77B2FFFF [...] 23:30:07 kps:7.64e+005 7.57e+005 00000000 Total 4:
00000018 0000385F77B30000 0000385F77C6FFFF [...] 23:30:09 kps:7.65e+005 7.59e+005 00000000 Total 4:
00000019 0000385F77C70000 0000385F77DAFFFF [...] 23:30:10 kps:7.68e+005 7.62e+005 00000000 Total 4:

why when i choose
10 NVIDIA Corporation GeForce 820M GPU

DETECTDEVICEENABLE:1
SELECTEDDEVICE:10

Reading from OCLBiss.cfg...Done...

Inspecting System for OpenCL devices...Platforms found: 2
00 Intel(R) Corporation Intel(R) HD Graphics 4400 GPU
01 Intel(R) Corporation Intel(R) Core(TM) i5-4210U CPU @ 1.70GHz CPU
10 NVIDIA Corporation GeForce 820M GPU

Today is Tue Sep 19 23:40:19 2023
OS TYPE:t Windows
OS: Windows_NT
Timer 1 seconds: 1.014002
Connected Platform: NVIDIA Corporation
Connected Device: GeForce 820M

OCLBISSVERSION: 021M
PROGRAM_FILE selected: csa_decrypt_1block_005.cl

Device Kernel properties:
number of Compute Units (CU): 2
recommended work group size (local threads): 32
max work group size: 512
Global Memory size: 2147483648

Number Loops per thread: 1024
Number of keys per thread: 1024
Local threads: 32
Global threads: 256
Keys per kernel: 262144(40000)

*****Important: To stop program and save current status*******
PRESS a LONG [ESC] Key and then say 'Y'

Scroll Lock LED in KeyBoard will be used to Signal Key Found event.
Please ensure it is off before continuing

File Log from Last Searched present: (11C16EFC4080101)
Do you what to continue from here?
If you type 'N' or 'n' will start from config.ini value: (0000385F624F0000)
Press Y/N + [ENTER] to Continue (Y)

The program is not working(No connection to NVIDIA)

Press Y/N + [ENTER] to Continue (Y)y

base = (385F7C410000)
Looking for (0x385F62F94FCABCD5)

BruteForcing for:
SB01: 3C EB DC 17 3C 2B D6 4F 65 16 88 F2 58 D5 97 05

Range: 0000000000040000

Loop From To Time Kernel Total Keys Found
00000001 0000385F7C410000 0000385F7C44FFFF [...] 23:44:57

-ERROR-clEnqueueReadBuffer for counter failed
Process returned 1 (0x1) execution time : 278.374 s
Press any key to continue.

cayoenrique · Wednesday at 02:02

But for you it does work 10 NVIDIA Corporation GeForce 820M GPU for OCLBiss.exe 014M Right?

cayoenrique · Wednesday at 09:00

There are limits at what we can do on old GPU using Opencl 1.0.
ModprogramsOCLV100.zip (63.49 KB)

Code:

https://workupload.com/file/wJDcPYLuAKf

Here 012 & 0014 build for Opencl 1.0,
Now I could not make work
csa_decrypt_1block_002.cl
csa_decrypt_1block_003.cl

In my very old GPU, The cl compilation just hang up to never end. There is no error that I can try to solve. Sorry.

dvlajkovic · Wednesday at 10:37

Don't quite understand if it's meant to work, but on my GPU
OCLBiss_014M_modOCLV100
runs without errors.

my setup is
LOCALTHREADS:256
GLOBALTHREADS:65536
LOOPSPERTHREAD:8000

I've tried:
PROGRAM_FILE:"csa_decrypt_1block_000.cl"
PROGRAM_FILE:"csa_decrypt_1block_001.cl"
#PROGRAM_FILE:"csa_decrypt_1block_002.cl"
#PROGRAM_FILE:"csa_decrypt_1block_003.cl"

and then:
#PROGRAM_FILE:"csa_decrypt_1block_000.cl"
#PROGRAM_FILE:"csa_decrypt_1block_001.cl"
PROGRAM_FILE:"csa_decrypt_1block_002.cl"
PROGRAM_FILE:"csa_decrypt_1block_003.cl"

Both give nice kps.

cayoenrique · Wednesday at 14:41

Explanation of _modOCLV100
OCLBiss_014M_modOCLV100 was build mainly for @Me2019H. He was the one that ask me to return the learning tutorials.
Sadly he has no Discrete GPU. Only old Nvidia 96 and Intel 160 cores Gpu. Been old means that It can only handle OpenCl API version 1.0. ( 100).

He has been behind because has been some trouble to setup his GPU.

Now that they work, he is finding that OLD API 1.0 has its limits. So He can ONLY execute "csa_decrypt_1block_000.cl" & "csa_decrypt_1block_001.cl" NONE of the csadvb lib cl kernels.

=========================================
I bet you find out that 021 split kernels where the fastest. But they consume too much memory. That is the reason why big kernels are rejected, the memory requirement is over what you have.

To your disappointment, my next step will be NOT using Split Kernels. But I will be using GLOBAL memory. Instead of bringing the back the results of all those FAKE keys. I plan to save them in Global memory.
@C0der brought us the fact that we should get 1 FAKE key every 2^24 keys tested, more or less ( 1/ 16777216). So we save those Fake keys, when we have enough FAKE keys then we launch a Second Kernel with a different SB ( second TS ) to reduce the number or fake to almost 0. Then we read the last remaining surviving FAKE keys. CUDABISS seems to suggest it does this 3 times to get the final GOO Key.

Well the problem at the moment with _modOCLV100 is that they can not setup buffers with CL_MEM_HOST_NO_ACCESS. So lets see what we can do for @Me2019H

cayoenrique · Thursday at 02:16

It was hard but finally found a way for @Me2019H to be able to execute 021.
As simple that can be just removing "CL_MEM_HOST_NO_ACCESS" from the code is enough to allow device use of internal memory.
And the hung-up can be override by disabling compiler optimization. cl-opt-disable

OCLBiss_021_modOCLV100_000.zip (38.12 KB)

Code:

https://workupload.com/file/sVNSALuEVGa

same pass www.sat-universe.com
Listen I have a filling that my compilation failed because my GPU has o little memory, Just try as is. the go to line OCLBiss.c line 283 and excahnge those lines.

Code:

//    err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
    err = clBuildProgram(program, 0, NULL, "-cl-opt-disable", NULL, NULL);

By uncommenting 1rst and commenting second. You will enable optimization. Hope it works for you.

Me2019H · Thursday at 23:23

Thank YOU
I'm grateful for all that you do and I appreciate your time.

it works

cayoenrique · Friday at 13:37

Your welcome.

I did changes so you do not fall behind. But in my laptop that "clBuildProgram(program, 0, NULL, "-cl-opt-disable", NULL, NULL);" speed is very bad.

Now that you can execute more or less all I posted, you should recalled that I said:
Intel HD Graphics 4400

Render Config
Shading Units 160
TMUs 20
ROPs 2
Execution Units 20
Theoretical Performance
Pixel Rate 2.300 GPixel/s
Texture Rate 23.00 GTexel/s
FP32 (float) 368.0 GFLOPS
FP64 (double) 92.00 GFLOPS (1:4)

Dell GeForce 820M 2 GB

Render Config
Shading Units 96
TMUs 16
ROPs 8
SM Count 2
L1 Cache 64 KB (per SM)
L2 Cache 128 KB
Theoretical Performance
Pixel Rate 2.500 GPixel/s
Texture Rate 10.00 GTexel/s
FP32 (float) 240.0 GFLOPS
FP64 (double) 20.00 GFLOPS (1:12)

This Shading Units used in screen graphics are in fact the cores. You have more cores that the discrete GPU. I suspect that if you Google your Laptop by model, your original CPU was not an i5. Bu a less powerful CPU.

I hope you install your newest GPU Driver from DELL. Once install there should be a menu by Nvidia. Usually in the lower left side of toolbar. In some Win7 PC you can also right click on desktop and you can get Nvidia preference.
Well in that Nvidia Graphics Setup you should find an option to setup witch GPU should be your default. You need to select Nvidia GPU.

At this point you should have find witch is you faster OCLBISS program and witch what GPU is the fastest. Intell having 64 more cores should be the winner I suspect.

What we trying to do is to assigned Nvidia to be the Win7 default Graphic Render. Why? so that Intel HD4400 gets free from rendering. So it is not only the fastest but also Is free for you to use. No more window black screen if you assign too much work. And you GPU gets to be even faster in Opencl. See the point. Then after the change. Check the speed it should be quicker. Another advantage is that I belive HD4400 is Opencl API V1.1. I suspect you will not need for -cl-opt-disable.
So once you change your Win7 GPU also try to run same OCLBiss_021_modOCLV100_000. Then go to OCLBiss.c line 283 and reverse the commets.

err = clBuildProgram(program, 0, NULL, NULL, NULL, NULL);
// err = clBuildProgram(program, 0, NULL, "-cl-opt-disable", NULL, NULL);

And this time you should find out you no longer need cl-opt-disable and the program is at least 10X faster.

Finally. I never done this but it should be tru. Your HD4400 Global memory is in fact you PC RAM. This is interesting as your opencl data buffers do not have to pass thru PCI Express read/write! You should be able write INITs or Read results much faster changing operations to MAP. In MAP the same variable that exist inside GPU also can exist in PC C main program, ( Shared Memory). So no wait time in PCI Express operations. You program will get a littel faster. Once you have your Nvidia drive as default latter I can show you. 1rst do Win7 default change, and tell us the result.

cayoenrique · Friday at 13:44

@all
This week I been busy with personal issues. I hope to have time today and the following days. As I said I intent to launch multiple kernels so that there is always work in the queue. This technique I suspect will render useless the need for launching different programs at same time. Just as yo do in CUDABISS.

cayoenrique · 2023-09-23T19:06:24+0100

Update. Be sure I been working on the program. Just that you understand. I am no professional C programmer. In fact I code rather slow as I even my typing speed is slow and I make a lot of type errors.

Now introducing multithreads is confuse topic itself. After giving some thought I realize that the future program may get confuse for a newbie. So i am changing the general structure of our program. I know I told you guys that we can write our code between the lines I posted in the program. Instead I am placing the OpenCl, and thread outside in their own C file. The idea is to let in Main ONLY the logic for the program it self. At the end a newbie should be able to follow what the program does more easily. This change had taken ma day and I guess I need at least one more.

If you are trying to understand the code and want to learn PLEASE post you questions so that I know that there are more people trying to learn to code. At the moment I know @dvlajkovic, @Me2019H & @azboxgo are trying to understand the C code logic.

Me2019H · 2023-09-23T23:16:26+0100

I update GeForce 820M

Platform #0
Name: Intel(R) OpenCL
Version: OpenCL 1.2

Total num_devices 2
Device #0
Name: Intel(R) HD Graphics 4400
Type: GPU
Version: OpenCL 1.2
Global memory size: 1 GB 273 MB 614 kB 410 B
Local memory size: 64 kB
Max work group size: 512
Max work item sizes: (512, 512, 512)

Device #1
Name: Intel(R) Core(TM) i5-4210U CPU @ 1.70GHz
Type: CPU
Version: OpenCL 1.2 (Build 148)
Global memory size: 7 GB 927 MB 844 kB
Local memory size: 32 kB
Max work group size: 8192
Max work item sizes: (8192, 8192, 8192)

Platform #1
Name: NVIDIA CUDA
Version: OpenCL 1.2 CUDA 9.1.84

Total num_devices 1
Device #0
Name: GeForce 820M
Type: GPU
Version: OpenCL 1.1 CUDA
Global memory size: 2 GB
Local memory size: 48 kB
Max work group size: 1024
Max work item sizes: (1024, 1024, 64)

Process returned 0 (0x0) execution time : 0.889 s
Press any key to continue.

All programs worked
Modprograms(csa_decrypt_1block_006.cl)

Today is Sat Sep 23 23:01:23 2023

OS TYPE: Windows
OS: Windows_NT
Timer 1 seconds: 1.000001
Connected Platform: NVIDIA Corporation
Connected Device: GeForce 820M
OCLBISSVERSION: 021M
PROGRAM_FILE selected: csa_decrypt_1block_006.cl
Device Kernel properties:
number of Compute Units (CU): 2
recommended work group size (local threads): 32
max work group size: 512
Global Memory size: 2147483648
Number Loops per thread: 574
Number of keys per thread: 574
Local threads: 128
Global threads: 128
Keys per kernel: 73472(11F00)

OS TYPE: Windows
OS: Windows_NT
File Log from Last Searched present: (11C16F2261F2C01)
Start from config.ini value: (0000385F624F0000)
Base Selected: (385F9263CB00)
Looking for: (0x385F62F94FCABCD5)

BruteForcing for:
SB01: 3C EB DC 17 3C 2B D6 4F 65 16 88 F2 58 D5 97 05

Range: 0000000000011F00

Loop From To Time Kernel Total Keys Found

00000001 00385F9263CB00 00385F9264E9FF [...] 23:01:32 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000002 00385F9264EA00 00385F926608FF [...] 23:01:33 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000003 00385F92660900 00385F926727FF [...] 23:01:33 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000004 00385F92672800 00385F926846FF [...] 23:01:33 kps:4.32e+005 4.20e+005 00000000 Total 0:
00000005 00385F92684700 00385F926965FF [...] 23:01:33 kps:4.32e+005 4.20e+005 00000000 Total 0:
00000006 00385F92696600 00385F926A84FF [...] 23:01:33 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000007 00385F926A8500 00385F926BA3FF [...] 23:01:33 kps:4.45e+005 4.20e+005 00000000 Total 0:
00000008 00385F926BA400 00385F926CC2FF [...] 23:01:34 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000009 00385F926CC300 00385F926DE1FF [...] 23:01:34 kps:4.32e+005 4.08e+005 00000000 Total 0:
0000000A 00385F926DE200 00385F926F00FF [...] 23:01:34 kps:4.32e+005 4.20e+005 00000000 Total 0:
0000000B 00385F926F0100 00385F92701FFF [...] 23:01:34 kps:4.32e+005 4.08e+005 00000000 Total 0:
0000000C 00385F92702000 00385F92713EFF [...] 23:01:34 kps:4.45e+005 4.20e+005 00000000 Total 0:
0000000D 00385F92713F00 00385F92725DFF [...] 23:01:35 kps:4.32e+005 4.08e+005 00000000 Total 0:
0000000E 00385F92725E00 00385F92737CFF [...] 23:01:35 kps:4.32e+005 3.87e+005 00000000 Total 0:
0000000F 00385F92737D00 00385F92749BFF [...] 23:01:35 kps:4.32e+005 4.20e+005 00000000 Total 0:
00000010 00385F92749C00 00385F9275BAFF [...] 23:01:35 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000011 00385F9275BB00 00385F9276D9FF [...] 23:01:35 kps:4.45e+005 4.20e+005 00000000 Total 0:
00000012 00385F9276DA00 00385F9277F8FF [...] 23:01:35 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000013 00385F9277F900 00385F927917FF [...] 23:01:36 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000014 00385F92791800 00385F927A36FF [...] 23:01:36 kps:4.45e+005 4.20e+005 00000000 Total 0:
00000033 00385F929BD900 00385F929CF7FF [...] 23:01:41 kps:4.32e+005 4.08e+005 00000000 Total 0:
00000065 00385F92D3E700 00385F92D505FF [...] 23:01:50 kps:4.45e+005 4.20e+005 00000000 Total 0:
0000008F 00385F9302FD00 00385F93041BFF [...] 23:01:58 kps:4.32e+005 4.08e+005 00000001 Total 1: Key 001:38 5F 93 2A 03 4B EA 38
00000097 00385F930BF500 00385F930D13FF [...] 23:01:59 kps:4.45e+005 4.20e+005 00000000 Total 1:
Finish

C0der Add your touch to the code

C0der · 2023-09-24T10:23:44+0100

Which of the 7 csa_decrypt_1block_*.cl gives the best kps?

GPGPU using Opencl

dvlajkovic

Senior Member

moonbase

azboxgo

t51w

cayoenrique

Senior Member

dvlajkovic

Senior Member

cayoenrique

Senior Member

dvlajkovic

Senior Member

Me2019H

cayoenrique

Senior Member

cayoenrique

Senior Member

dvlajkovic

Senior Member

cayoenrique

Senior Member

cayoenrique

Senior Member

Me2019H

cayoenrique

Senior Member

cayoenrique

Senior Member

cayoenrique

Senior Member

Me2019H

C0der

Senior Member