joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
August 31, 2016, 10:11:17 PM Last edit: August 31, 2016, 10:24:43 PM by joblo |
|
This was my reference: https://gcc.gnu.org/onlinedocs/gcc/x86-Options.htmlIf you have mingw installed you can confirm with: gcc -Q -march=native --help=target | grep march Edit: yours looks more like bdver2 or 3. Edit2: Upon further thought there's probably no need to build bins for AMD. There is no AMD specific code so as long as the feature set lines up the Intel builds should work just as well.
|
|
|
|
Fuzzbawls
|
 |
August 31, 2016, 10:35:48 PM |
|
The following change will be made to the share result output in the next release. There is a small change from the initial proposal.
[2016-08-31 10:21:44] Accepted 10/10 (100%), 5026 H, 215.37 H/s [2016-08-31 10:21:44] Rejected 10/11 (99.9%), 5026 H, 215.37 H/s
Small nit: With the proposed change to the vocabulary after the timestamp, the following metric could also be adjusted. In the case of an accepted share, you would see the total of accepted shares followed by the grand total of all shares submitted. in the case of a rejected share, you would see the total of rejected shares followed by the grand total of all shares submitted. could also change the percentage metric to display the rejection percentage when a share is rejected, instead of defaulting to the acceptance percentage for all shares irregardless of their status. as follows: [2016-08-31 10:21:44] Accepted 10/10 (100%), 5026 H, 215.37 H/s [2016-08-31 10:21:56] Rejected 1/11 (0.1%), 5026 H, 215.37 H/s ...
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
August 31, 2016, 10:39:44 PM |
|
The following change will be made to the share result output in the next release. There is a small change from the initial proposal.
[2016-08-31 10:21:44] Accepted 10/10 (100%), 5026 H, 215.37 H/s [2016-08-31 10:21:44] Rejected 10/11 (99.9%), 5026 H, 215.37 H/s
Small nit: With the proposed change to the vocabulary after the timestamp, the following metric could also be adjusted. In the case of an accepted share, you would see the total of accepted shares followed by the grand total of all shares submitted. in the case of a rejected share, you would see the total of rejected shares followed by the grand total of all shares submitted. could also change the percentage metric to display the rejection percentage when a share is rejected, instead of defaulting to the acceptance percentage for all shares irregardless of their status. as follows: [2016-08-31 10:21:44] Accepted 10/10 (100%), 5026 H, 215.37 H/s [2016-08-31 10:21:56] Rejected 1/11 (0.1%), 5026 H, 215.37 H/s ... It makes sense to do both, good idea.
|
|
|
|
felixbrucker
|
 |
September 01, 2016, 07:03:18 AM |
|
@joblo indeed, found this (originally japanese) text: A description of the GCC source code of "BDver4" before delving into the subject . AMD has publicly the development plan of the architecture than before , until "Excavator" has been revealed . "BDver1" is that of the first generation "Bulldozer" which appeared in 2011 . "BDver2" is "Piledriver" in , are those some of the improvements from the "Bulldozer" has been carried out , has been used in the current FX CPU Ya "Richland" APU. "BDver3" is a CPU core to be used in the "Kaveri" APU of AMD's plans to introduce at the end of this year in the "Steamroller". so it seems my target needed is bdver2 cheers
|
|
|
|
denrok
Newbie
Offline
Activity: 25
Merit: 0
|
 |
September 02, 2016, 09:22:40 AM |
|
Tried compiling on a non-aesni cpu (core2), but it tries to use aesni on cryptonight:
In file included from /usr/lib/gcc/x86_64-linux-gnu/5/include/x86intrin.h:43:0, from algo/cryptonight/cryptonight-aesni.c:1: algo/cryptonight/cryptonight-aesni.c: In function ‘ExpandAESKey256_sub2’: /usr/lib/gcc/x86_64-linux-gnu/5/include/wmmintrin.h:87:1: error: inlining failed in call to always_inline ‘_mm_aeskeygenassist_si128’: target specific option mismatch _mm_aeskeygenassist_si128 (__m128i __X, const int __C) ^ algo/cryptonight/cryptonight-aesni.c:26:7: error: called from here tmp4 = _mm_aeskeygenassist_si128(*tmp1, 0x00); ^
I'm getting the same error i'm using corei7 on windows 7 how do i compile with corei7-avx? i'm doing via cygwin.... ./autogen.sh Unescaped left brace in regex is deprecated, passed through in regex; marked by <-- HERE in m/\${ <-- HERE ([^ \t=:+{}]+)}/ at /usr/bin/automake-1.14 line 3930.
LIBCURL="-lcurldll" ./configure CFLAGS="-O3" configure: error: Missing required libcurl >= 7.15.2 (I've installed libcurl?)
make from aes.c:2: aes.c: In function ‘ExpandAESKey256_sub2’: /usr/lib/gcc/x86_64-pc-cygwin/5.4.0/include/wmmintrin.h:87:1: error: inlining failed in call to always_inline ‘_mm_aeskeygenassist_si128’: target specific option mismatch _mm_aeskeygenassist_si128 (__m128i __X, const int __C) ^ aes.c:21:7: error: called from here tmp4 = _mm_aeskeygenassist_si128(*tmp1, 0x00); ^ make[2]: *** [Makefile:569: hodlminer-aes.o] Error 1 make[2]: Leaving directory '/home/Laptop/hodlminer-wolf' make[1]: *** [Makefile:632: all-recursive] Error 1 make[1]: Leaving directory '/home/Laptop/hodlminer-wolf' make: *** [Makefile:384: all] Error 2
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 02, 2016, 02:28:15 PM |
|
Tried compiling on a non-aesni cpu (core2), but it tries to use aesni on cryptonight:
In file included from /usr/lib/gcc/x86_64-linux-gnu/5/include/x86intrin.h:43:0, from algo/cryptonight/cryptonight-aesni.c:1: algo/cryptonight/cryptonight-aesni.c: In function ‘ExpandAESKey256_sub2’: /usr/lib/gcc/x86_64-linux-gnu/5/include/wmmintrin.h:87:1: error: inlining failed in call to always_inline ‘_mm_aeskeygenassist_si128’: target specific option mismatch _mm_aeskeygenassist_si128 (__m128i __X, const int __C) ^ algo/cryptonight/cryptonight-aesni.c:26:7: error: called from here tmp4 = _mm_aeskeygenassist_si128(*tmp1, 0x00); ^
I'm getting the same error i'm using corei7 on windows 7 how do i compile with corei7-avx? i'm doing via cygwin.... ./autogen.sh Unescaped left brace in regex is deprecated, passed through in regex; marked by <-- HERE in m/\${ <-- HERE ([^ \t=:+{}]+)}/ at /usr/bin/automake-1.14 line 3930.
LIBCURL="-lcurldll" ./configure CFLAGS="-O3" configure: error: Missing required libcurl >= 7.15.2 (I've installed libcurl?)
make from aes.c:2: aes.c: In function ‘ExpandAESKey256_sub2’: /usr/lib/gcc/x86_64-pc-cygwin/5.4.0/include/wmmintrin.h:87:1: error: inlining failed in call to always_inline ‘_mm_aeskeygenassist_si128’: target specific option mismatch _mm_aeskeygenassist_si128 (__m128i __X, const int __C) ^ aes.c:21:7: error: called from here tmp4 = _mm_aeskeygenassist_si128(*tmp1, 0x00); ^ make[2]: *** [Makefile:569: hodlminer-aes.o] Error 1 make[2]: Leaving directory '/home/Laptop/hodlminer-wolf' make[1]: *** [Makefile:632: all-recursive] Error 1 make[1]: Leaving directory '/home/Laptop/hodlminer-wolf' make: *** [Makefile:384: all] Error 2
Please provide more information such as CPU, and compile procedure. The post you quoted was withan AMD CPU and appeared to be a problem with the compiler detecting the correct CPU architecture.
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 03, 2016, 04:26:03 PM |
|
@Joblo would it be cool with you if I mine to your BTC donation address at nicehash as donation time for your work? I'll mine one full day per week as donation. Every other user can do the same as well.
I just received an anonymous payment at my donation address. Thank you and any others who followed your suggestion.
|
|
|
|
Epsylon3
Legendary
Offline
Activity: 1484
Merit: 1122
ccminer/cpuminer developer
|
 |
September 03, 2016, 05:53:36 PM |
|
me too  maybe the same source...
|
|
|
|
pallas
Legendary
Offline
Activity: 2716
Merit: 1094
Black Belt Developer
|
 |
September 03, 2016, 06:24:50 PM |
|
Me nothing.... nobody mining M7 on nvidia? :-/
|
|
|
|
Epsylon3
Legendary
Offline
Activity: 1484
Merit: 1122
ccminer/cpuminer developer
|
 |
September 03, 2016, 06:29:37 PM |
|
i think i was 2 years ago 
|
|
|
|
ryen123
|
 |
September 04, 2016, 06:27:58 AM |
|
@Joblo would it be cool with you if I mine to your BTC donation address at nicehash as donation time for your work? I'll mine one full day per week as donation. Every other user can do the same as well.
I just received an anonymous payment at my donation address. Thank you and any others who followed your suggestion.  That was from nicehash payout. Cryptonight mining has been quite good especially with the fixed rejection issues.
|
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 04, 2016, 01:46:55 PM |
|
|
|
|
|
maxfunky
|
 |
September 05, 2016, 03:16:33 AM |
|
thanks for the great work!  Is it possible to opt the cryptolight algo 
|
|
|
|
aquelol
Newbie
Offline
Activity: 2
Merit: 0
|
 |
September 05, 2016, 11:43:37 AM |
|
hi, just a report on notebook haswell cpu (4700MQ @ 3.4ghz 47W) with espers mining (hmq1725) v3.3.x-v3.3.8 Start mining with AES-AVX optimizations =150/155 Kh/s ; 3.4ghz core/cache and ~46W no throttle (stable optimised voltage/temp 76° max ) since v3.4.x Start mining with SSE2 AES AVX AVX2 = 140/145 Kh/s ; 3.4ghz core / 3.2ghz cache and +47w with some cpu throttle down the hashrate to 130 Kh/s and +85° i have a question, i wanna try to solo mine but your client crash when request/receiving getwork (tested with all 3.3.x/3.4.x) log 3.4.5 crash http://pastebin.com/ynJiLnsP
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 05, 2016, 12:37:21 PM |
|
hi, just a report on notebook haswell cpu (4700MQ @ 3.4ghz 47W) with espers mining (hmq1725) v3.3.x-v3.3.8 Start mining with AES-AVX optimizations =150/155 Kh/s ; 3.4ghz core/cache and ~46W no throttle (stable optimised voltage/temp 76° max ) since v3.4.x Start mining with SSE2 AES AVX AVX2 = 140/145 Kh/s ; 3.4ghz core / 3.2ghz cache and +47w with some cpu throttle down the hashrate to 130 Kh/s and +85° i have a question, i wanna try to solo mine but your client crash when request/receiving getwork (tested with all 3.3.x/3.4.x) log 3.4.5 crash http://pastebin.com/ynJiLnsPThanks for testing. cpuminer-opt only supports stratum mining. I am curious about your performance results. Can you provide more info such as whether you compiled you own, Linux or Windows, or used precompiled binaries. Thanks.
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 05, 2016, 12:48:10 PM Last edit: September 05, 2016, 07:07:58 PM by joblo |
|
thanks for the great work!  Is it possible to opt the cryptolight algo  Cryptolight is not a very popular algo and until recently I didn't even know there was a coin using it. Therefore it is very low on my priority list. I haven't analysed the code so I don't know if the cryptonight optimizations can be ported for use on cryptolight. If they can it might be possible for me to implement it. Otherwise you will need to pay someone like Wolf to build one for you. I am pretty busy keeping up with new algos, while trying to opitimize the more popular existing algos. It is unlikely I will get to cryptolight any time soon. Edit: I took a quick look at cryptolight and it appears to already have AES optimizartions implemented. You can confirm this by testing on a CPU with AES_NI using the SSE2 build of cpuminer. If it is slower than the AES version for your CPU then you are already using AES_NI.
|
|
|
|
|
joblo (OP)
Legendary
Offline
Activity: 1470
Merit: 1114
|
 |
September 06, 2016, 01:38:55 AM |
|
I guess that means it is working and I can take out the warning about it being untested? If your tests show the AES build is faster than SSE2 I'll change the display to indicate the algo has AES. I haven't look at the Arux-BTT fork yet, let me know how it performs compared with cpuminer-opt.
|
|
|
|
maxfunky
|
 |
September 06, 2016, 02:17:10 AM |
|
yes you can take out the warning
|
|
|
|
|