Pages:
Author

Topic: BitCrack - A tool for brute-forcing private keys - page 60. (Read 74450 times)

jr. member
Activity: 56
Merit: 8
16jY7qLJnN4eqD1eArr9vRVABixvcbLyXY   KwDiBf89QgGbjEhKnhXJuH7LrciVrZi3qYxZTMH89fVdXM613vVD   
may be this will help someone to solve P 64

how to find those address ?? .....
16jY7qLJ*********************
all in range 8000000000000000:10000000000000000
I think you solved it  #64

Code:
16jY7qLJnN4eqD1eArr9vRVABixvcbLyXY  09A97E3E4A6A15391
16jY7qLJNSJnur5KVhoaKC25ZVyP2LpGuj  09CA0EFE4DC9C4B72
16jY7qLJuZ4WrwBbbAoHSsd9tf74x55oCG  0A199E4EEF9FABBB7
16jY7qLJpeQTfzX1SYgoNWDmqwsq9kCeJ1  0A3ABFF4956D1E3CF
16jY7qLJzxMfbfrxVjdzKVcMYuABMT3JUg  0A447E8DC181F71AF
16jY7qLJFMG4uGLQ6p2NK71KSMn1ZsHnR1  0A792E0E2AB63BC53
16jY7qLJ1itzy2eqPDW3EnDrFAMi94pkRw  0A9F727369AB8A72C
16jY7qLJ48zRMcRFt25ZgMwaLGUS4xnGp9  0B1E3FF83D3C2CADA
16jY7qLJ2aBYPJHvYZCexMtX9As7YXKTXa  0B22925E55D30526E
16jY7qLJf5WRNQsr4M7PHw3CcdUkjjiTcD  0B24147E4FA620801
16jY7qLJyE4Cspx2xJZimnKFDu1mA5v8q4  0BC9264E45AB665AF
16jY7qLJfn6u2bhsqryEBfJrUqmyYtwvMw  0BCADEFE466EB45B3
16jY7qLJD6fvZebckbnmBCqyu2uDwRqii6  0BE0F6652B041E9C5
16jY7qLJwdUDparg4Y3EgFpAhZNHTSkPFt  0C06D61514D867204
16jY7QLjn6MEodyvdGc1w1oiKSBoC4beMa  0C16D9AFCE49A235B
16jY7qLJSYVeyhvHbEC7LwgFQnVgPai56w  0C447A5E4210BCE22
16jY7qLJ8BQQDJJjRUop7FEiEyJJCR2ZE9  0C44875E4EF637660
16jY7qLJn19vvyctGxbtBuJrrpRCt5er9H  0C44B91E42375905B
16jY7qLJpG8FVrMLbk6NBtWMiejrKgEd5q  0C653BCE4B4967446
16jY7qLJxw7aXRjDTLutu7BC5Wp1LqKBPc  0C87506E4F8BA10EC
16jY7qLJzqL7qGrCz9joUf3qkTqUwygnez  0C8E6ACD0039891A4
16jY7qLJiGvJJajVN5anM8wbkW2x6JDsbx  0C9C883940AF8DF68
16jY7qLJUV3uavsiC9F69PtoCvXzcnrQpw  0CC910FE39C32DC0B
16jY7qLJSabEVThGZ58M6KmuKvaYKbkJqJ  0CC931AE3E13C6F30
16jY7qLJavMWDu8C9AGuH9LXHFyJVZHSMy  0CC9525E3226024A4
16jY7qLJpBTbteT7BCKPPrdtgwRbd2g6xe  0CC9633E407F8086A
16jY7qLJsaainy7EAxWCmsDGg934UdQujU  0CC98E9E362CC4542
16jY7qLJbYbK2DTYqDih9d1fhui4eELaBr  0CD51EFE3E898CE22
16jY7qLJjWKajipMwHwgFd7R5gYWDRVxvX  0CEAEFDE46E720BA0
16jY7qLJHmJ1N3frmsNRJhXYwXk4pz9dBP  0D64F13C3A8A5A56C
16jY7qLJrGLxGSzddR5KcJHe8wMXPKvoZ5  0D6DF549B417D5F5F
16jY7qLJeBU1dWv6Gw7cNFdpEfssB6nM6p  0D7C3610CACD10E2F
16jY7qLJ3ueYvr33g8f3WMxMUKubDfyTYt  0DBB6EEE24AF250E0
16jY7qLJySCNV747Z89CXnYnhC8sBETpAq  0DC7A6954459982B4
16jY7qLJtSouCs3cVpeZFMynfCZcUYhBJL  0E2C7BCE4D98A1CEF
16jY7qLJg8UVMsNVgmDkU5LYfnUvcHFgHE  0E7734AEB1A2C68A4
16jY7qLJbpczaisuZfaGAwGtmyYgpyikCa  0E859E5E2400E0E2D
16jY7qLJZRNXy8sYPcy8icoaHTszZzrmxY  0E9ACEEE0A95849D1
16jY7qLJceFLCRJ3W5VDvFfBagQVpde5Pj  0ECA0EFE42AB81E7D
16jY7qLJceFLCRJ3W5VDvFfBagQVpde5Pj  0ECA0EFE42AB81E7D
16jY7qLJTeYW5FumLwXdjQVG9siirnups8  0ECA7EFE46C1094D9
16jY7qLJ6AkxWNMXHHiWL6nQS8sAfijdjm  0F04C4D5648DC1F94
16jY7qLJmHoEkEbsaw8rMXUKT6VxJ6tCV1  0F07A6E5B3C4671AA
16jY7qLJZTyb8ammZufMPy1gTktDm6tKSx  0F257E7D91DC58530
16jY7qLJ7xMpwdjzoW3zsaESTwy9tekRxA  0F2624F96E4B0F443
16jY7qLJnNtLkE51fgNsMqYAZ5i6LqURJm  0F262E8D54F236F43
16jY7qLJBSqbwDwVa11q4HFr4237sEypfZ  0F2CDEF415A50F2BC
16jY7qLJJa2D8Q7aNSt818G47epM2rwsou  0F618E4E0BEFB65C1
16jY7qLJevsjxqTvRGUfjEhSKjnceRew5h  0FB72665B09D2E273
16jY7qLJWQnhf9uUKSyAW1WbgGVYvBmLgy  0FE9264E43F208CCF
16jY7qLJeSKNp2iZDStRSn4Vq9Qp9N76cR  0FF854CE413430A01
16jY7qLJxeuWjZKR25E4v8ZBnEPvxwg3hu  0FFBD79E6D341D584
16jY7qLJAi1LdLnfSX6BPkntA4NypEXnZr  0F02DB8E20CFB42B0
legendary
Activity: 1568
Merit: 6660
bitcoincleanup.com / bitmixlist.org
Not sure at what rate you're leasing a Tesla A. But isn't it cheaper to buy, for example, a 3060?
Ofc I'm not familiar with shipping and stocking in your country Smiley

I'm overseas ATM so I'm in the process of leasing a custom server from walkerservers.com. They're the only guys who accept crypto and offer GPUs at the same time.

They also only have Teslas, so no GeForce or Quadro for me.  Cry (you gotta request a quote from their sales dept for one, they aren't available in the product listings.)

PPS. do not get a 3060 until someone confirms whether they throttle on these programs, that card in particular throttles itself for ethereum mining, and I have no idea if it classifies private key bruteforcing as ethereum mining, as they both use ECC, and'll throttle on these programs too.

I see, I could free up a windows desktop with a 3080 or 3070 if thats something you're willing to work with?
Would need to request and need a time-window (1 week, 2 weeks?) for it, but could work it out for free.

Offer noted if I need it in the future, but someone else here contacted me first and got me a GPU to work with. Thanks though.

Why waste so much electricity with the brute force attack when you just need to solve a simple mathematical problem?  Grin

https://decrypt.co/59340/ancient-bitcoin-whale-just-moved-5-million

This explains nothing. He probably owned the private keys to that wallet already.

EDIT: you should back that claim up with more than a decrypt.co news story.
newbie
Activity: 7
Merit: 0
Why waste so much electricity with the brute force attack when you just need to solve a simple mathematical problem?  Grin

https://decrypt.co/59340/ancient-bitcoin-whale-just-moved-5-million
jr. member
Activity: 36
Merit: 3
Not sure at what rate you're leasing a Tesla A. But isn't it cheaper to buy, for example, a 3060?
Ofc I'm not familiar with shipping and stocking in your country Smiley

I'm overseas ATM so I'm in the process of leasing a custom server from walkerservers.com. They're the only guys who accept crypto and offer GPUs at the same time.

They also only have Teslas, so no GeForce or Quadro for me.  Cry (you gotta request a quote from their sales dept for one, they aren't available in the product listings.)

PPS. do not get a 3060 until someone confirms whether they throttle on these programs, that card in particular throttles itself for ethereum mining, and I have no idea if it classifies private key bruteforcing as ethereum mining, as they both use ECC, and'll throttle on these programs too.

I see, I could free up a windows desktop with a 3080 or 3070 if thats something you're willing to work with?
Would need to request and need a time-window (1 week, 2 weeks?) for it, but could work it out for free.
legendary
Activity: 1568
Merit: 6660
bitcoincleanup.com / bitmixlist.org
Not sure at what rate you're leasing a Tesla A. But isn't it cheaper to buy, for example, a 3060?
Ofc I'm not familiar with shipping and stocking in your country Smiley

I'm overseas ATM so I'm in the process of leasing a custom server from walkerservers.com. They're the only guys who accept crypto and offer GPUs at the same time.

They also only have Teslas, so no GeForce or Quadro for me.  Cry (you gotta request a quote from their sales dept for one, they aren't available in the product listings.)

PPS. do not get a 3060 until someone confirms whether they throttle on these programs, that card in particular throttles itself for ethereum mining, and I have no idea if it classifies private key bruteforcing as ethereum mining, as they both use ECC, and'll throttle on these programs too.
jr. member
Activity: 36
Merit: 3

10xx series cards work with cubitcrack without issue, regardless of drivers.
20xx series cards work with cubitcrack, if you roll back drivers to 452.06 or 456.02 (can't remember which)
30xx series cards do not work with cubitcrack no matter the driver

Got it. Better avoid getting a Pascal or Maxwell then.

Not sure at what rate you're leasing a Tesla A. But isn't it cheaper to buy, for example, a 3060?
Ofc I'm not familiar with shipping and stocking in your country Smiley
legendary
Activity: 1568
Merit: 6660
bitcoincleanup.com / bitmixlist.org
The Tesla I'm leasing for Kangaroo will also be used to patch the long standing misaligned address CUDA error, so I need to know if anyone here has experienced this error on a Pascal or Maxwell GPU so I don't lease one of those cards unknowingly and kick myself later, I have a limited budget allocated for this and these things 'ain't cheap.
10xx series cards work with cubitcrack without issue, regardless of drivers.
20xx series cards work with cubitcrack, if you roll back drivers to 452.06 or 456.02 (can't remember which)
30xx series cards do not work with cubitcrack no matter the driver

Got it. Better avoid getting a Pascal or Maxwell then.

PS. It was 452.06 you roll back to.

Roll back your graphics driver to 452.06; only way it will work. Or the only way that I and others have got it to work
full member
Activity: 1050
Merit: 219
Shooters Shoot...
The Tesla I'm leasing for Kangaroo will also be used to patch the long standing misaligned address CUDA error, so I need to know if anyone here has experienced this error on a Pascal or Maxwell GPU so I don't lease one of those cards unknowingly and kick myself later, I have a limited budget allocated for this and these things 'ain't cheap.
10xx series cards work with cubitcrack without issue, regardless of drivers.
20xx series cards work with cubitcrack, if you roll back drivers to 452.06 or 456.02 (can't remember which)
30xx series cards do not work with cubitcrack no matter the driver
legendary
Activity: 1568
Merit: 6660
bitcoincleanup.com / bitmixlist.org
The Tesla I'm leasing for Kangaroo will also be used to patch the long standing misaligned address CUDA error, so I need to know if anyone here has experienced this error on a Pascal or Maxwell GPU so I don't lease one of those cards unknowingly and kick myself later, I have a limited budget allocated for this and these things 'ain't cheap.
full member
Activity: 1050
Merit: 219
Shooters Shoot...

"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.




https://github.com/brichard19/BitCrack/issues/256
Is the fixed I tried before and showed execution of in the above quote.
If I wasn't such a nerd, I would send you the .exe, but I discourage people of doing that. You should really build it yourself for max. security.

The .31 isn't faster. The calculation is wrong, so it looks faster.

So apply that fix to 0.30? And the speed improvements from 0.31 is due to the fact that it's not working properly?
But like they said, the speed on .31 is incorrect...
The crazy thing is, Brichard is a good programmer. He built a kangaroo program using OpenCL that is actually compatible with all my cards and actually works on the 30xx series card. and the speed is accurate. The bad thing, it's not open sourced so I can't make mods to it to suit my needs.
Maybe he doesn't keep working on bitcrack because it has lived its shelf life, meaning brute forcing the keys from 64 on up (the original reason he built the program, was for the puzzle/transaction) is next to impossible without all hands on deck. Who knows...
jr. member
Activity: 36
Merit: 3


So apply that fix to 0.30? And the speed improvements from 0.31 is due to the fact that it's not working properly?

You can just apply it to the master branch. It will work fine.



For now, I am just testing that clbitcrack is working properly by using some of the public test keys against their known keyspace (same list of test keys others are using here). As others have complained, 0.31 has an issue finding private keys and it reaches the end of the keyspace without finding the keys. 0.30 seems to find all keys using the same list, but for me it's half as fast as 0.31 (when I let it run on a private list of addresses where the keys are unknown). With my 3090 using clbitcrack 0.31, I get about 1050MKey/s whereas with 0.30 I get 600MKey/sec. But why run 0.31 if it's never going to find any keys with the known issue.

Another issue people had was clbitcrack would find some of the known keys but not all of them (versus not finding any). Not sure what version has that issue or if it's just with certain GPU's, but it something else I am testing for too.

So I guess my question is, is there an actual fix for clbitcrack 0.31 that doesn't have any of these issues but still has the speed improvements that comes with 0.31? Or, if there's another program that's currently developed and supports 3090's and would work better, please let me know. I am basically just trying to recover private keys from a list of known wallet addresses by incrementally scanning the private keyspace. I know this is exactly what Bitcrack does, but not sure about the others, i.e. VanitySearch, Kangaroo, etc

Fixed code for the RTX would require a full re-write of the existing CUDA codes out there. We investigated && concluded it would be too costly to do, as our devs could be earning way more in that same time.  Wink
There are minor openCL improvements possible, but didn't had the time to check up on them.

We're also in the progress of building a website that publishes all spaces we've ran, so others could do lottery & run other spaces. 'cause yolo.
newbie
Activity: 18
Merit: 0

"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.




https://github.com/brichard19/BitCrack/issues/256
Is the fixed I tried before and showed execution of in the above quote.
If I wasn't such a nerd, I would send you the .exe, but I discourage people of doing that. You should really build it yourself for max. security.

The .31 isn't faster. The calculation is wrong, so it looks faster.

So apply that fix to 0.30? And the speed improvements from 0.31 is due to the fact that it's not working properly?
newbie
Activity: 18
Merit: 0


"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.


What are you searching for? Are you searching for #64 (btc puzzle/transaction 64 bit) or just random keys?

For now, I am just testing that clbitcrack is working properly by using some of the public test keys against their known keyspace (same list of test keys others are using here). As others have complained, 0.31 has an issue finding private keys and it reaches the end of the keyspace without finding the keys. 0.30 seems to find all keys using the same list, but for me it's half as fast as 0.31 (when I let it run on a private list of addresses where the keys are unknown). With my 3090 using clbitcrack 0.31, I get about 1050MKey/s whereas with 0.30 I get 600MKey/sec. But why run 0.31 if it's never going to find any keys with the known issue.

Another issue people had was clbitcrack would find some of the known keys but not all of them (versus not finding any). Not sure what version has that issue or if it's just with certain GPU's, but it something else I am testing for too.

So I guess my question is, is there an actual fix for clbitcrack 0.31 that doesn't have any of these issues but still has the speed improvements that comes with 0.31? Or, if there's another program that's currently developed and supports 3090's and would work better, please let me know. I am basically just trying to recover private keys from a list of known wallet addresses by incrementally scanning the private keyspace. I know this is exactly what Bitcrack does, but not sure about the others, i.e. VanitySearch, Kangaroo, etc
jr. member
Activity: 36
Merit: 3

"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.




https://github.com/brichard19/BitCrack/issues/256
Is the fixed I tried before and showed execution of in the above quote.
If I wasn't such a nerd, I would send you the .exe, but I discourage people of doing that. You should really build it yourself for max. security.

The .31 isn't faster. The calculation is wrong, so it looks faster.
full member
Activity: 1050
Merit: 219
Shooters Shoot...


"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.


What are you searching for? Are you searching for #64 (btc puzzle/transaction 64 bit) or just random keys?
newbie
Activity: 18
Merit: 0

You just need a fixed CL version. The forum/issues tracker is full of solutions;

Code:
clBitCrack.exe -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv
[2021-02-26.14:35:16] [Info] Compression: compressed
[2021-02-26.14:35:16] [Info] Starting at: 0000000000000000000000000000000000060F4D11574F5DEEE49961D8577307
[2021-02-26.14:35:16] [Info] Ending at:   0000000000000000000000000000000000060F4D11574F5DEEE49961DA609AC5
[2021-02-26.14:35:16] [Info] Counting by: 0000000000000000000000000000000000000000000000000000000000000001
[2021-02-26.14:35:17] [Info] Compiling OpenCL kernels...
[2021-02-26.14:35:17] [Info] Initializing GeForce RTX 3090
[2021-02-26.14:35:17] [Info] Generating 262,144 starting points (10.0MB)
[2021-02-26.14:35:17] [Info] 10.0%
[2021-02-26.14:35:17] [Info] 20.0%
[2021-02-26.14:35:17] [Info] 30.0%
[2021-02-26.14:35:17] [Info] 40.0%
[2021-02-26.14:35:17] [Info] 50.0%
[2021-02-26.14:35:17] [Info] 60.0%
[2021-02-26.14:35:17] [Info] 70.0%
[2021-02-26.14:35:17] [Info] 80.0%
[2021-02-26.14:35:17] [Info] 90.0%
[2021-02-26.14:35:17] [Info] 100.0%
[2021-02-26.14:35:17] [Info] Done
[2021-02-26.14:35:17] [Info] Found key for address '1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv'. Written to 'output.txt'
[2021-02-26.14:35:17] [Info] No targets remaining

"Fixed" clbitcrack as in using an older build like 0.30, or a fixed version of 0.31? I just tried 0.30 and while it finds all keys ok, it's about half as fast as 0.31. I have 3090's so I believe I'm stuck using clbitcrack until the misaligned address issue is fixed for these newer cards that require the newer CUDA.


full member
Activity: 1050
Merit: 219
Shooters Shoot...
0.31 would not find those test keys for me either using clbitcrack. Anyone have a link to 0.30 precompiled so I can test? Can't seem to get 0.30 to compile under Windows.

https://github.com/brichard19/BitCrack/releases/tag/0.30
jr. member
Activity: 36
Merit: 3

reference
https://github.com/brichard19/BitCrack/releases/tag/0.31

I found problem on clBitCrack.exe

Can anyone help to testing if you have BitCrack on you PC/laptop?

I found cuBitCrack.exe can found private key
but, clBitCrack.exe can not found private key

test puzzle #115  1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv

command
cuBitCrack.exe --device 0 -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv
cuBitCrack.exe -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv

clBitCrack.exe --device 0 -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv
clBitCrack.exe -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv

you will be found private key in second
upper both same command options and same keyspace

problem
Result test cuBitCrack.exe can found key
but clBitCrack.exe work fast and never found key

clBitCrack.exe  will be work fast and end with message Reached end of keyspace
but clBitCrack.exe  never found key


cuBitCrack.exe result
GeForce GTX 1050 738 / 4096MB | 1 target 6.11 MKey/s (4,008,181,760 total) [00:10:57] [Info] Found key for address '1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv'. Written to 'output.txt'



You just need a fixed CL version. The forum/issues tracker is full of solutions;

Code:
clBitCrack.exe -o output.txt --keyspace 60F4D11574F5DEEE49961D8577307:60F4D11574F5DEEE49961DA609AC5 1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv
[2021-02-26.14:35:16] [Info] Compression: compressed
[2021-02-26.14:35:16] [Info] Starting at: 0000000000000000000000000000000000060F4D11574F5DEEE49961D8577307
[2021-02-26.14:35:16] [Info] Ending at:   0000000000000000000000000000000000060F4D11574F5DEEE49961DA609AC5
[2021-02-26.14:35:16] [Info] Counting by: 0000000000000000000000000000000000000000000000000000000000000001
[2021-02-26.14:35:17] [Info] Compiling OpenCL kernels...
[2021-02-26.14:35:17] [Info] Initializing GeForce RTX 3090
[2021-02-26.14:35:17] [Info] Generating 262,144 starting points (10.0MB)
[2021-02-26.14:35:17] [Info] 10.0%
[2021-02-26.14:35:17] [Info] 20.0%
[2021-02-26.14:35:17] [Info] 30.0%
[2021-02-26.14:35:17] [Info] 40.0%
[2021-02-26.14:35:17] [Info] 50.0%
[2021-02-26.14:35:17] [Info] 60.0%
[2021-02-26.14:35:17] [Info] 70.0%
[2021-02-26.14:35:17] [Info] 80.0%
[2021-02-26.14:35:17] [Info] 90.0%
[2021-02-26.14:35:17] [Info] 100.0%
[2021-02-26.14:35:17] [Info] Done
[2021-02-26.14:35:17] [Info] Found key for address '1NLbHuJebVwUZ1XqDjsAyfTRUPwDQbemfv'. Written to 'output.txt'
[2021-02-26.14:35:17] [Info] No targets remaining
newbie
Activity: 18
Merit: 0
0.31 would not find those test keys for me either using clbitcrack. Anyone have a link to 0.30 precompiled so I can test? Can't seem to get 0.30 to compile under Windows.
legendary
Activity: 1568
Merit: 6660
bitcoincleanup.com / bitmixlist.org
Can anyone try to test clBitCrack.exe  on you PC/laptop  last version 0.31 (Improve OpenCL performance)

I just looked at those commits leading to 0.31, and while I don't know much about OpenCL, it seems that the only major change he made is to replace the representation of 256-bit numbers from an array of 8 32-bit bits to a struct with an array of 8 32-bit bits (and some loops were unrolled). I don't see how this is supposed to make it run faster because it should've used a larger word size like a 64-bit cl_ulong, which would reduce the number of load and store instructions by half. Nobody uses 32-bit platforms any more so there's no reason to keep a 32-bit version around and therefore constrain all the arrays to use 32-bit ints as well.
Pages:
Jump to: