LTC miner optimizations for PowerPC (Power Mac) and Cell/BE (PlayStation 3)

[ZedZedNova]

After messing around with Ubuntu 11.10 that didn't seem to like the nVidia graphics card, even when not using KMS (booting with "nouveau.modestate=0"), I finally just installed a CLI version, booted with "nouveau.modestate=0" and all is well. I installed the "missing" libraries and gcc, and built cpuminer.

I'm seeing ~3.10 - 3.25 kH/s now. This is about equivalent to one thread on my 2.3 GHz Core i5 powered Mac Mini.

Edit: Second machine installed and churning out the same as the first.

- Zed

[1714945667]

1714945667

[1714945667]

1714945667

[1714945667]

1714945667

[ssvb]

What kind of hash rate would we get on the xbox 360?

Assuming that it's possible to get linux running on it, one may expect something like 3 * 3.6 khash/s = ~10.8 khash/s hashing power from xbox 360. At least I'm getting ~3.6 khash/s from two threads of one Cell PPE core. And the 3 cores in Xenon processor are supposedly slightly modified versions of the PPE in the Cell processor according to wikipedia.

The current code is using gcc vector extensions for quick prototyping. This makes it possible to reuse significant parts of the same code on Cell PPE (PowerPC Altivec), Cell SPE (has it's own SIMD architecture) and even x86 (SSE2) architectures. But relying on the compiler to do registers allocation and instructions scheduling sacrifices some of the performance. And this can be easily seen if benchmarking against pooler's hand written SSE2 code on x86

It should be possible to improve Altivec performance by adding hand written assembly optimizations. But realistically, from the whole PowerPC camp, only PlayStation3 has chances to be really competitive in terms of absolute performance per machine or performance per watt.

[ssvb]

The first feedback and actually a patch for the Cell miner sent to me as a private message: https://github.com/shakt1/cpuminer
But one problem is that shakti is a newly registered user and has no permissions to post anything here yet.

edit: here is his thread in the Newbies section: https://bitcointalk.org/index.php?topic=64222.msg753270#msg753270
If there are more people "on probation" (maybe newcomers from the playstation3 linux community), then it could be a good place for them to post some comments.

[shakti]

The first feedback and actually a patch for the Cell miner sent to me as a private message: https://github.com/shakt1/cpuminer
But one problem is that shakti is a newly registered user and has no permissions to post anything here yet.

edit: here is his thread in the Newbies section: https://bitcointalk.org/index.php?topic=64222.msg753270#msg753270
If there are more people "on probation" (maybe newcomers from the playstation3 linux community), then it could be a good place for them to post some comments.

Finally i can post here Nice work ssvb to make miner for PS3, i hope we can tweak it together to make it even better

[ssvb]

Finally i can post here Nice work ssvb to make miner for PS3, i hope we can tweak it together to make it even better

Sure, more improvements for PS3 miner are definitely possible. I have pushed some of my old unfinished code which implements parts of scrypt in SPU assembly (~5.4 khash/s -> 5.9 khash/s improvement per SPE core) to github. This is only an optimization for the first big loop, handling the second big loop in a similar way is expected to provide about the same speedup. This is just better instructions scheduling and keeping all data in registers avoiding unnecessary spills. Your improvements are focusing on a different aspect - better data layout for less scattered writes and more unrolling to completely eliminate any stray stalls waiting for DMA completion (up to ~6.1 khash/s per SPE core). Combining both optimizations should provide quite good results, maybe even changing to handle 10 hashes at once (5+5) would be beneficial for the second loop. The theoretical peak performance per SPU core, based on counting 128-bit vector ADD/ROL/XOR operations is ~7.35 khash/s (an optimistic estimate, not even taking negligible SHA256 part and other overhead into account). Which means that we are already at >80% of the theoretical peak performance.

Just GPU miners are a bit more hot topic at the moment. During the last weekend I was busy installing a new graphics card and then doing some OpenCL coding with it. But I'm going to revisit Cell/BE code after I get my GPU miner up and running

[tacotime]

Finally i can post here Nice work ssvb to make miner for PS3, i hope we can tweak it together to make it even better

Just GPU miners are a bit more hot topic at the moment. During the last weekend I was busy installing a new graphics card and then doing some OpenCL coding with it. But I'm going to revisit Cell/BE code after I get my GPU miner up and running

I look forward to the performance numbers!  The previously much talked about miner (if it even exists) showed only marginal per-watt performance over CPUs.

[ssvb]

I look forward to the performance numbers!  The previously much talked about miner (if it even exists) showed only marginal per-watt performance over CPUs.

There are better forum threads for speculating about the alleged performance numbers. The GPU miner is going to be released with full sources included when/if I get something ready. No other comments for now.

[Bananahands]

Sorry to revive such an old post but where can I find the Cell/PPC optimized miner?  Will it run on OS X 10.5.8?

Thanks,
Mike

[meti]

It's still there... haven't you read the first post?
https://github.com/ssvb/cpuminer

Code:

$ git clone git://github.com/ssvb/cpuminer.git
$ cd cpuminer
$ export CFLAGS="-O3 -mcpu=G4 -maltivec -fstrict-aliasing"
$ ./autogen.sh && ./configure && make

[Bananahands]

Thanks, I'm new to compiling so I overlooked that the first time.

Anyway, I've tried compiling it on both Mac OS X 10.5.8 and Debian Linux (whatever one is for my PPC G5) and I get stuck with

./configure: line 5010: syntax error near unexpected token 'LIBCURL_CHECK_CONFIG'
./configure: line 5010: 'LIBCURL_CHECK_CONFIG(, 7.10.1, ,'

Any idea how to fix this?

Mike

[ssvb]

In debian you can try "apt-get install build-essential libtool libltdl-dev automake" (I hope I did not forget something in this list). The autogen.sh script needs autotools installed to generate the configure script. This stuff is often already installed on the computers used for software development, but may be indeed missing on some systems.

edit: hmm, actually based on the error message, the autotools might be already fine. Try "apt-get install libcurl-dev" to see if this resolves the curl related issues.

[Bananahands]

Thank you for the reply, I tried both of your suggestions and they worked but I am still getting the same error....

Edit:I guess I was in a hurry, it was asking me to select which libcurl I wanted to install.  Now I just need to figure out how to use it....

Mike

[Bananahands]

It works!  Thank you meti, thank you ssvb.  One final question though.  I don't see any improvement over Pooler's CPUMiner, I'm still getting ~1.45 khash/sec on each processor.  I was thinking I would get a lot more than that seeing how they are 2.0GHz G5 processors.  Does the fact that I did this over many days with the computer being either off or booted in OS X make any difference?  Would deleting everything in the cpuminer folder and starting from scratch do anything?

Mike

Edit: I fixed it.  I guess i first tried using -mcpuG5 because they are G5s but -mcpuG4 is giving me ~3.80 khash/sec!

[tuaris]

Does anyone know if these optimization were ever backported into pooler's CPU miner?
I have a few G5's sitting around doing nothing.

[Eudardoq]

Hello .. please help to optimize ps3 ... I already managed to mine ... and mine to 1.46 Khas / s but by there I read that they managed to optimize it to an average of 30 Khas / s someone could help me please ..