Show Posts
I took the time to finally get xptMiner to compile with GCC for Windows.
It was a nightmare for me, but what does that mean for you? Christmas may have come early.
I've compiled my optimized code for each of the following architectures: generic, core2, corei7, corei7-avx, k8, k8-sse3, and barcelona (k10).
The Windows x64 binaries are available from here:
https://mega.co.nz/#F!YsZSHYKA!IC8LK_MBGwqC-gWOpO7zoQ
https://www.dropbox.com/sh/wtxvxvkirxax2vj/9P_Rxb9V1y (dropbox mirror)
Use whichever corresponds to your processor architecture.
If you're unsure, here's a quick guide based on your CPU manufacturer and age:
Intel: [Older] core2 -> corei7 -> corei7-avx [Newer]
AMD: [Older] k8 -> k8-sse3 -> barcelona (k10) [Newer]
Worst case scenario, use generic.
I will not be releasing any 32-bit builds. "High performance 32-bit" is an oxymoron.
Let me know what sort of improvements you see!
Some builds may be slightly slower than the stock jh00 miner. It all depends on your CPU.
It was a nightmare for me, but what does that mean for you? Christmas may have come early.
I've compiled my optimized code for each of the following architectures: generic, core2, corei7, corei7-avx, k8, k8-sse3, and barcelona (k10).
The Windows x64 binaries are available from here:
https://mega.co.nz/#F!YsZSHYKA!IC8LK_MBGwqC-gWOpO7zoQ
https://www.dropbox.com/sh/wtxvxvkirxax2vj/9P_Rxb9V1y (dropbox mirror)
Use whichever corresponds to your processor architecture.
If you're unsure, here's a quick guide based on your CPU manufacturer and age:
Intel: [Older] core2 -> corei7 -> corei7-avx [Newer]
AMD: [Older] k8 -> k8-sse3 -> barcelona (k10) [Newer]
Worst case scenario, use generic.
I will not be releasing any 32-bit builds. "High performance 32-bit" is an oxymoron.
Let me know what sort of improvements you see!
Some builds may be slightly slower than the stock jh00 miner. It all depends on your CPU.
I've added another 5-10% optimization to the code and hopefully fixed the issue with AMD builds.
It seems that mingw-gcc thinks that AMD processors don't support SSE, so it left them disabled. I've fixed that.
The code optimizations have been pushed to the github repo.
The new binaries have been uploaded. (same URLs)
I've added a few more performance tweaks and pushed the kHash overflow fix to github.
The other performance tweaks I have not pushed because they hard code 64-bit mode for some of the hash algorithms. For the sake of openness, here's the changes:
keccak.c
Code:
#define SPH_KECCAK_64 1
#define SPH_KECCAK_UNROLL 0
#define SPH_KECCAK_NOCOPY 0
metis.c
Code:
#define SPH_KECCAK_NOCOPY 0
#define SPH_64 1
I've also compiled the code through Visual Studio 2010 with performance guided optimization and the windows builds now regularly meet or exceeds the jh00 speeds.
I've built two versions: one with the /favor:INTEL64 flag and one with /favor:AMD64. Pick whichever matches your processor.
edit: AMD users! Try the favorAMD build first, then the favorBLEND build. Apparently /favor:AMD64 prefers older AMD processors (Athlon, Opteron, and Sempron). See which works best for you!
Download Links:
https://mega.co.nz/#F!YsZSHYKA!IC8LK_MBGwqC-gWOpO7zoQ
https://www.dropbox.com/sh/wtxvxvkirxax2vj/9P_Rxb9V1y (dropbox mirror)
(My AMD Phenom II x4 955 @ stock clock)
Tested favorIntel and it gave me 1100kH/s compared to 1150kH/s for the ypool suggested one.. It was on a i7-4930K.
Thank you very much!
☺
