Date: Wed, 6 Feb 2013 07:12:34 -0800 (PST) From: deepika dutta <deepikadutta_19@...oo.com> To: "john-dev@...ts.openwall.com" <john-dev@...ts.openwall.com> Subject: Re: mschapv2-bitsliced conversion- openmp support Hi, I came back after a long time to John-dev to see your email. Actually, I worked on coverting the DES_bs_crypt_plain(), which I had added to do single DES encryption, into SSE implementation. I have added this function into the x86-sse.S file. Since the code in this file was in assembly language, I tried some copy-pasting from previous code and added some assembly code myself to get the single DES encryption working. I compiled this program using linux-x86-64 option, everything is working fine and the latest banchmarks for bitsliced implementation are given below. You can test for yourself. The patch is attached. **I have also implemented the single DES encryption code using sse intrinsics taking help from your code. This I did for my personal use :) Bitsliced DES code in mschapv2 banchmark. My machine: 64 bit Ubuntu, Intel Core 2 Duo processor. Generic-Without openmp Benchmarking: MSCHAPv2 C/R MD4 DES [DES_BS_MSCHAPv2]... DONE Many salts: 20068K c/s real, 20068K c/s virtual Only one salt: 3322K c/s real, 3322K c/s virtual Generic-With openmp Benchmarking: MSCHAPv2 C/R MD4 DES [DES_BS_MSCHAPv2]... (2xOMP) DONE Many salts: 37740K c/s real, 19060K c/s virtual Only one salt: 5648K c/s real, 2852K c/s virtual linux-x86-64- Without openmp Benchmarking: MSCHAPv2 C/R MD4 DES [DES_BS_MSCHAPv2-SSE2]... DONE Many salts: 43646K c/s real, 43646K c/s virtual Only one salt: 4112K c/s real, 4112K c/s virtual linux-x86-64- With openmp Benchmarking: MSCHAPv2 C/R MD4 DES [DES_BS_MSCHAPv2-SSE2]... (2xOMP) DONE Many salts: 79691K c/s real, 40452K c/s virtual Only one salt: 6414K c/s real, 3256K c/s virtual Cheers, Deepika ________________________________ From: Solar Designer <solar@...nwall.com> To: john-dev@...ts.openwall.com Sent: Sunday, January 20, 2013 12:50 AM Subject: Re: [john-dev] mschapv2-bitsliced conversion- openmp support magnum - On Sat, Jan 19, 2013 at 05:21:45PM +0100, magnum wrote: > I feel bad about not getting Deepika's work included in Jumbo. It's a very skilled contribution compared to to the trivial OpenSSL-powered formats that we are used to. Leaving it in limbo discourages future contributions - from her and from others. Yes, I feel bad about this too, and I agree that not properly and timely making use of a contribution discourages further contributions. > Besides that, it appears to give 2x on single salt and 10x on multiple salts. > > I understand it's not trivial though. I am not quite sure what needs to be done to actually add it to Jumbo. Apparently it only builds using generic. Is there anything I can do, or anything we could ask Deepika to do, to help this forward? Unfortunately, I'm still of the opinion that I'd need to work on this myself, and it's still not at the top of my list. However, if you see how you can integrate the code in some temporary way, please feel free to go ahead and do it. Thanks, Alexander Content of type "text/html" skipped View attachment "john-1.7.9-jumbo-5-MSCHAPV2-BS-3.diff" of type "text/x-patch" (29777 bytes)
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.