Version 0.3.8.1 update for Linux 64-bit

Satoshi Nakamoto August 9, 2010 19:46 UTC Source ·

When we switched to Crypto++ 5.6.0 SHA-256 in version 0.3.6, generation got broken on the Linux 64-bit build.  Version 0.3.8.1 is on SourceForge with the 64-bit binary updated.

Download: http://sourceforge.net/projects/bitcoin/files/Bitcoin/bitcoin-0.3.8/bitcoin-0.3.8.1-linux.tar.gz/download

Future versions after 0.3.8 will probably require SSE2.  Anyone have Pentium 3 or older where this would be a problem?

NewLibertyStandard August 9, 2010 20:26 UTC Source ·

The difficulty has gotten high enough that I wouldn’t worry about dropping block generation support for old computers if it’s hard to maintain compatibility. But the client should definitely provide some feedback explaining why the feature isn’t available. Perhaps you could make a message pop up that explains why generating bitcoins is not supported on old hardware.

Jeff Garzik (jgarzik) August 9, 2010 20:46 UTC Source ·

IMO, SSE should be required for generation, not simply to run the client.

Satoshi Nakamoto August 9, 2010 20:55 UTC Source ·

That’s a good point, I believe you could run with generation off if you don’t have SSE2.

How about add to the top of cryptopp/config.h:

#if !defined(_M_X64) && !defined(x86_64) #define CRYPTOPP_DISABLE_SSE2  1 #endif

that would disable SSE2 for 32-bit builds.  (at least with GCC or MSVC)

dkaparis August 9, 2010 20:57 UTC Source ·

From what I understand, removing the offending define is required only for 64-bit builds and 64-bit architectures are already guaranteed to support SSE2 instructions. 32-bit builds can retain the flag and thus continue to disable SSE2 instructions and remain compatible with older computers.

dkaparis August 9, 2010 21:02 UTC Source ·
Quote from: satoshi on August 09, 2010, 8:55:06 PM UTC

That’s a good point, I believe you could run with generation off if you don’t have SSE2.

How about add to the top of cryptopp/config.h:

#if !defined(_M_X64) && !defined(x86_64) #define CRYPTOPP_DISABLE_SSE2 1 #endif

that would disable SSE2 for 32-bit builds. (at least with GCC or MSVC)

Yes, but better have that specified in the makefile. That way anyone who doesn’t care for old CPU’s can compile with SSE2 support by removing the flag, without fiddling with the source.

I can also add a CMake property for that in the CMake build.

knightmb August 10, 2010 00:02 UTC Source ·
Quote from: dkaparis on August 09, 2010, 8:57:02 PM UTC

From what I understand, removing the offending define is required only for 64-bit builds and 64-bit architectures are already guaranteed to support SSE2 instructions. 32-bit builds can retain the flag and thus continue to disable SSE2 instructions and remain compatible with older computers.

Older CPU that don’t support SSE2 shouldn’t be excluded from coin generation in my opinion. They can still serve a vital role in block generation; I have many machines that don’t support the SSE2 enhancements, but still generate blocks all the time, even if they are only churning 300-400 khash/s

Ground Loop August 10, 2010 03:38 UTC Source ·

Hah! This sure helps explain a long-standing mystery.. why my 10,000 khash/sec Linux 64-bit machine is the only one that has never, in two weeks, generated a block. 😊

Ground Loop August 10, 2010 03:47 UTC Source ·

Just to confirm — this fix is in SVN 125?

mizerydearia August 10, 2010 06:18 UTC Source ·
Quote from: Ground Loop on August 09, 2010, 6:38:23 PM UTC

Hah! This sure helps explain a long-standing mystery.. why my 10,000 khash/sec Linux 64-bit machine is the only one that has never, in two weeks, generated a block.

You’re not the only one. I haven’t generated in over a month.

ArtForz August 10, 2010 09:37 UTC Source ·

Fix is in svn r124, r125 added -maxconnections.

Satoshi Nakamoto August 10, 2010 23:46 UTC Source ·

SVN rev 128: disable SSE2 on 32-bit.  This may only disable it for MSVC and GCC.  Other compilers might have different 64-bit defines.

SEN-5241 August 14, 2010 14:24 UTC Source ·

I think the client should have a “test coin generation” feature. It would essentially just generate some known block just to show everything is working properly.

NewLibertyStandard August 14, 2010 14:54 UTC Source ·
Quote from: SEN-5241 on August 14, 2010, 2:24:52 PM UTC

I think the client should have a “test coin generation” feature. It would essentially just generate some known block just to show everything is working properly.

I agree. I have a second computer with which I can test difficulty 1 block generation, but I never actually do it. If there was a button to do it, I would.