Show Posts
|
Pages: [1] 2 3 »
|
Got +10 MH/s on 5850 by increasing the worksize from 128 to 256 @ 2 vectors, keeping the RAM @ 300, because @ 200 = -10 MH/s. Thanks for the update! cgminer is better than ever...
|
|
|
(except I use none of the crappy tools you describe, especially not SNMP (ugh) ) Well-well-well...
|
|
|
Oh yeah, and also it needs an SNMP backend so that stats from the rigs can be aggregated into something like Zabbix or cacti + icinga. I see it: mining in the enterprise.
|
|
|
ck, thanks for the %/RPM display fix!~ ^_^
|
|
|
You don't seem to get it: here nobody owes you anything. You might ask nicely, but if you are refused you have to shut it up. This is what DiabloD3 is politely implying. If you can't comprehend it, maybe ban is the best thing for everybody.
|
|
|
ck, did you know that you are like crazy cool, man? I am running 2.0.1 and well, I don't know how did it manage to achieve it, but it down-clocked my memory to 300 Mhz, which was not possible using pplib commands on my old FirePro V5700 (sic!). This saves me like 8C or so. It doesn't seem to be able to read RPMs though, Fan Speed: 65% (-1 RPM). But well, it's already something! Also, I am sick of the curses bug, so here is how to compile cgminer statically against latest curses: 1) Go to http://ftp.gnu.org/gnu/ncurses/ and download the latest version 2) Unpack, do ./configure --prefix=$HOME/opt/ncurses --enable-static --disable-shared && make && make install 3) Go to cgminer, do CFLAGS="-O2 -Wall -march=native" CPPFLAGS="-I$HOME/opt/ncurses/include" LDFLAGS="-L$HOME/opt/ncurses/lib" ./configure --prefix=/opt/cgminer 4) Enjoy! Might be worth adding to the readme.
|
|
|
FYI as I migrated to 1.6.2 from 1.6.1 and 1.6.0, my U went up from 4.0 to 4.2 and 4.4 respectively. Great job ck, I am really impressed highly appreciate your efforts (in spite of the persistent curses bug lol). There is something I find weird in 1.6.1/2 though. Does this E indicator actually make sense? The total E indicator seems to work properly. [P]ool management [G]PU management [S]ettings [D]isplay options [Q]uit GPU 0: [313.7 / 313.8 Mh/s] [Q:2 A:858 R:8 HW:0 E:42900% U:4.39/m]
|
|
|
New release, version 1.6.2
Builds on RHEL5 with stock libcurl again!
|
|
|
which sounds a lot like problems with older curl to me... Yes, yours are the problems that I mentioned above. I am preparing an RPM that you can rebuild with rpmbuild if you need GPU mining or use it as is if you only need CPU mining this evening, which features a static build against curl, so which version is in the base system repos would be irrelevant.
|
|
|
Obviously, I need it. It's not because I care so much about other people that might need it that I am asking for it. Having to do it myself is a logical answer though.
|
|
|
Ok, got it, I'll try to make you some working patch tomorrow as I'll have to spend the time on train... The problem is as always the lack of time, so I'm hoping till the very end that someone else will do it for me :-)
P.S. I've checked on the CPU mining memleak and it's still leaking albeit slower. Like 200 mb per day or so.
|
|
|
Also, it is not clear to me what is the purpose of the following check in configure.ac: AC_CHECK_LIB(curl, curl_easy_init, , <------>AC_MSG_ERROR([Could not find curl library - please install libcurl-dev])) it comes *before* PKGCONFIG detection, so unless curl is installed in a standard location it just can't succeed. Also, LIBCURL_CPPFLAGS / LIBCURL_LIBS are substituted after the pkgconfig check, and if the check fails, then configure fails. So they don't make much sense, because if pkgconfig is non-existant or doesn't find the lib, it will AC_MSG_ERROR. LIBCURL_CFLAGS C compiler flags for LIBCURL, overriding pkg-config LIBCURL_LIBS linker flags for LIBCURL, overriding pkg-config
My suggestions: 1) Only pkgconfig if LIBCURL_CPPFLAGS / LIBCURL_LIBS are NOT set 2) Move AC_CHECK_LIB *after* pkgconfig 3) Add LIBCURL_CPPFLAGS / LIBCURL_LIBS to CPPFLAGS / LDFLAGS before AC_CHECK_LIB and restore afterwards Right now I basically can't compile cgminer against a statically linked libcurl.a
|
|
|
Sorry for being a PITA, but the ncurses bug is still there. The memleak seems to be gone in 1.6.1 though, it's been an hour already and the process is at 370M. Let's see if it leaks overnight or not. Also, it still doesn't build with older curl (at least make it fail at configure stage :-( ): util.c:249: error: expected declaration specifiers or '...' before 'curlsocktype' make[2]: *** [cgminer-util.o] Error 1 Thanks!
|
|
|
Another couple:
1) So what does more exactly longpoll do and how to enable it with cgminer? I was under impression that it should be enabled by default, unless I specify --no-longpoll, but somehow it doesn't work for me on Slush pool. Shall I add LP to the pool URL?
2) So what's the best kernel for Radeon 5xxx ATM, shall I add -k phatk for optimum performance?
P.S. I really want to kill AMD, upgraded to latest Debian testing and the 10.7 or 10.8 or whatever they are consume 100% CPU. Argh...
|
|
|
Couple of observations:
1) Since cgminer is unusable with older curses, maybe make a compile-time switch to disable them completely? Especially now that there is a monitor option etc. 2) ./configure doesn't fail if no curses development files were detected, but compilation fails subsequently.
|
|
|
PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 682 user 39 19 1915m 37m 11m S 389 0.5 4944:21 /opt/cgminer/bin/cgminer --algo auto --cpu-threads 4 --gpu-threads 1 ... I noticed this already with 1.5.3, which I left running for awhile and got like 29g VIRT at some point, but now the history repeats itself with 1.6.0. It really looks like if there were a memleak somewhere that is not fixed. It's not a ncurses thing, I am still running the text interface, because the WINCH bug is back again in 1.6.0. Also, this doesn't manifest itself if CPU mining is disabled and doesn't depend on which algorithm exactly one does use.
|
|
|
What is wrong with leaving in the manual override option? What is crappy about leaving people the option to not install more abstraction layers on things they don't want? If I want enough abstraction layers I may as well just go use windows and autoupdates. Did you actually read ./configure --help at some point? Is it your vax experience that did not prevent you from blaming pkg-config that is preventing you from exporting LIBCURL_LIBS and LIBCURL_CFLAGS before running ./configure if so you wish? --with-libcurl option was broken and hence removed. If you don't want to use pkg-config, fine, override the flags and there you go. I have just checked it by uninstalling pkg-config and it works fine.
|
|
|
Why? You seem to miss the point completly. I don't have pkgconfig. I don't want pkgconfig, so why should I read up on it? I have been building software from source for a few decades now. Since usenet was the way to distribute the sources for most. The all the worlds linux folks are as annoying today as the all the worlds a vax foks where then. If you have been building software for that long, it should have taught you that building complex software with many dependencies without an unified system for discovering library paths, compile, link flags etc. (which is exactly what pkg-config is) is a living hell of hooking up each and every library to the build system via crappy third-party autotools macros for which no decent unified and maintained archive exists. And in this particular instance, these macros were even unable to resolve version requirements for which reason (I guess) they were finally nuked in favor of pkg-config. It could be that your system doesn't ship pkg-config (which I would doubt, unless it's Windows), but since you've been building software since vax days, it would certainly take you near-zero effort to build pkg-config in addition to cgminer. I bet there are even third-party repositories with pre-build binaries somewhere around.
|
|
|
I don't want to talk about the reasons why I cpu-mine. It leads to nothing but arguments.
All I want to know is how to get cgminer to work without an openCl-device.
1.5.3 works for me for CPU mining, did you specify -g 0?
|
|
|
|