???
If you want to use AVX2, use Sarath-hotspot’s. I’m only interested in Xenoncat’s AVX1.
Well, a lot of people don’t understand AVX1 vs AVX2 or just want a single binary they can deploy on a mix of CPUs without SIGILLs.
Sweet! Thanks! This is really good… let me test it out
looks like you forgot to include equihash_avx1.o in the repo
make[2]: *** No rule to make target ../equihash_avx1.o', needed by nheqminer’. Stop.
make[1]: *** [CMakeFiles/nheqminer.dir/all] Error 2
Oops… my bad…
Committed the file. Check now.
Awesome thanks - it’s working great on the systems I’ve tried.
works great now!
Thanks
Does not compile, latest HEAD:
ZcashStratum.o: In function `ZcashMinerThread(ZcashMiner*, int, int)':
ZcashStratum.cpp:(.text+0x6325): undefined reference to `EhPrepare_AVX1'
ZcashStratum.cpp:(.text+0x6333): undefined reference to `EhSolver_AVX1'
ZcashStratum.cpp:(.text+0x6547): undefined reference to `EhPrepare'
ZcashStratum.cpp:(.text+0x6555): undefined reference to `EhSolver'
collect2: error: ld returned 1 exit status
make: *** [Makefile:315: nheqminer] Error 1
csnt get this to run … tried installing it as per instructions for ubuntu 16.4
when i get to “make”
it gives me a fatal error
boost/predef/hardware/simd/x86/versions.h: no such file or directory
compilation terminated
Oh hah, I just did the same thing, looks like we crossed paths. Mine also just kludges AVX1 in - I add a mode setting in the code but had doubled up symbols so just swapped out the bin in CMakeList. If people are having problems w/ peakay’s feel free to give mine a try:
Performance went from:
5.6 sol/s → 9.4 sol/s w/ -t3 on an Intel(R) Core™ i3-2130 CPU @ 3.40GHz (4C) ; Ubuntu 16.04
12.2 sol/s → 20.2 sol/s w -t6 on an Intel(R) Core™ i7-3770S CPU @ 3.10GHz (4C/8HT) ; Arch
Note, even w/ @sarath-hotspot’s build w/ XC AVX2 running on an i7-5600U 2.6GHz, I get better hashing (11.2 sol/s vs 7.8 sol/s) running the latest 0.3a Windows miner so if you can choose to run Windows or Linux, you’ll definitely be getting more bang/buck using one of those.
1 Like
hi,
I’ am trying to compile nheqminer on debian wheezy, I installed the libboost 1.55 (compiled for wheezy) and then I tried to compile:
cmake …
– Boost version: 1.55.0
– Found the following Boost libraries:
– system
– log_setup
– log
– date_time
– filesystem
– thread
– Configuring done
– Generating done
– Build files have been written to: /root/sarath-hotspot/nheqminer/nheqminer/build
~/sarath-hotspot/nheqminer/nheqminer/build# make
Scanning dependencies of target nheqminer
[ 6%] Building CXX object CMakeFiles/nheqminer.dir/main.cpp.o
In file included from /root/sarath-hotspot/nheqminer/nheqminer/primitives/transaction.h:12:0,
from /root/sarath-hotspot/nheqminer/nheqminer/primitives/block.h:9,
from /root/sarath-hotspot/nheqminer/nheqminer/main.cpp:5:
/root/sarath-hotspot/nheqminer/nheqminer/uint256.h:22:21: error: ISO C++ forbids declaration of ‘alignas’ with no type [-fpermissive]
/root/sarath-hotspot/nheqminer/nheqminer/uint256.h:22:21: error: expected ‘;’ at end of member declaration
make[2]: *** [CMakeFiles/nheqminer.dir/main.cpp.o] Error 1
make[1]: *** [CMakeFiles/nheqminer.dir/all] Error 2
make: *** [all] Error 2
any hint?
You did a good job modifying the function name with _AVX1.
Regarding AMD CPU, I am curious to know which AMD CPU does not run the code. There should be no problem as long as it supports AVX/AVX2.
Oh… I was under impression that this assembly code works only on intel-x86 CPUs.
I did not verify it on AMD CPUs.
Nice, thank you very much. Just recompiled, Now my amd fx 8350 shows it’s using xenoncat’s avx1 solver. My sols increased to 16 using -t 7!!! 
Nice! My AMD FX-8350 goes from ~9 Sol/s, to ~16! 
[02:15:03][0x00007f4601ea4700] stratum | Submitting share #4, nonce 02000000000000000000000000000002
[02:15:03][0x00007f4603cdb700] stratum | Accepted share #4
[02:15:04][0x00007f4601ea4700] stratum | Submitting share #5, nonce 02000000000000000000000000000002
[02:15:04][0x00007f4603cdb700] stratum | Rejected share #5 (duplicate share)
Almost each share is duplicated, there is any way to fix that?
1 Like
Any way to add tromp’s Cuda solver into this? It would be great to mine on both my gpu and cpu on Linux, no need for annoying windows.
Hmm, I am seeing a lot of dupes now too w/ AVX2, appears to have been caused by the AVX1 commits somehow. For now you can try reverting them on AVX2 systems (git checkout c07ad755ee92cac025a96ee9eed95dfd2ef97775).
2 Likes