Preparation for #1.4.0 (#30)
- Fixed CPU affinity on Windows for NUMA and CPUs with lot of cores - Implemented per thread configurable Multihash mode (double, triple, quadruple, quintuple) - Rebased from XMRig 2.4.4
This commit is contained in:
Ben Gräf
GitHub
parent
990bf8d963
commit
cf868666d4
41 changed files with 2575 additions and 1104 deletions
52
README.md
52
README.md
|
|
@ -106,12 +106,13 @@ xmrigDaemon -o pool.minemonero.pro:5555 -u YOUR_WALLET -p x -k --cc-url=IP_OF_CC
|
|||
-O, --userpass=U:P username:password pair for mining server
|
||||
-u, --user=USERNAME username for mining server
|
||||
-p, --pass=PASSWORD password for mining server
|
||||
-t, --threads=N number of miner threads
|
||||
-v, --av=N algorithm variation, 0 auto select
|
||||
-t, --threads=N number of miner threads (0 enables automatic selection of optimal number of threads, default: 0)
|
||||
-m, --multihash-factor=N number of hash blocks per thread to process at a time (0 enables automatic selection of optimal number of hash blocks, default: 0)
|
||||
-A, --aesni=N selection of AES-NI mode (0 auto, 1 on, 2 off, default: 0)
|
||||
-k, --keepalive send keepalived for prevent timeout (need pool support)
|
||||
-r, --retries=N number of times to retry before switch to backup server (default: 5)
|
||||
-R, --retry-pause=N time to pause between retries (default: 5)
|
||||
--doublehash-thread-mask for av=2/4 only, limits doublehash to given threads (mask), (default: all threads)
|
||||
--multihash-thread-mask for multihash-factor > 1 only, limits multihash to given threads (mask), (default: all threads)
|
||||
--cpu-affinity set process affinity to CPU core(s), mask 0x3 for cores 0 and 1
|
||||
--cpu-priority set process priority (0 idle, 2 normal to 5 highest)
|
||||
--no-huge-pages disable huge pages support
|
||||
|
|
@ -135,33 +136,51 @@ xmrigDaemon -o pool.minemonero.pro:5555 -u YOUR_WALLET -p x -k --cc-url=IP_OF_CC
|
|||
-l, --log-file=FILE log all output to a file
|
||||
-h, --help display this help and exit
|
||||
-V, --version output version information and exit
|
||||
-v, --av=N DEPRECATED - algorithm variation, 0 auto select
|
||||
--doublehash-thread-mask DEPRECATED - same as multihash-thread-mask
|
||||
|
||||
```
|
||||
|
||||
Also you can use configuration via config file, default **[config.json](https://github.com/Bendr0id/xmrigCC/wiki/Config-XMRigDaemon)**. You can load multiple config files and combine it with command line options.
|
||||
|
||||
## Algorithm variations
|
||||
* `--av=1` For CPUs with hardware AES.
|
||||
* `--av=2` Lower power mode (double hash) of `1`.
|
||||
* `--av=3` Software AES implementation.
|
||||
* `--av=4` Lower power mode (double hash) of `3`.
|
||||
## Multihash (multihash-factor)
|
||||
With this option it is possible to increase the number of hashblocks calculated by a single thread in each round.
|
||||
Selecting multihash-factors greater than 1 increases the L3 cache demands relative to the multihash-factor.
|
||||
E.g. at multihash-factor 2, each Cryptonight thread requires 4MB and each Cryptonight-lite thread requires 2 MB of L3 cache.
|
||||
With multihash-factor 3, they need 6MB or 3MB respectively.
|
||||
|
||||
## Doublehash thread Mask (only for low power mode (av=2 and av=4))
|
||||
With this option you can limit doublehash to the given threads (mask). This can significantly improve your hashrate by using unused l3 cache. The default is to run av2/av4 mode on all threads.
|
||||
Setting multihash-factor to 0 will allow automatic detection of the optimal value.
|
||||
Xmrig will then try to utilize as much of the L3 cache as possible for the selected number of threads.
|
||||
If the threads parameter has been set to auto, Xmrig will detect the optimal number of threads first.
|
||||
After that it finds the greatest possible multihash-factor.
|
||||
|
||||
### Multihash for low power operation
|
||||
Depending the CPU and its L3 caches, it can make sense to replace multiple single hash threads with single multi-hash counterparts.
|
||||
This change might come at the price of a minor drop in effective hash-rate, yet it will also reduce heat production and power consumption of the used CPU.
|
||||
|
||||
### Multihash for optimal CPU exploitation
|
||||
In certain environments (e.g. vServer) the system running xmrig can have access to relatively large amounts of L3 cache, but may has access to only a few CPU cores.
|
||||
In such cases, running xmrig with higher multihash-factors can lead to improvements.
|
||||
|
||||
|
||||
## Multihash thread Mask (only for multihash-factor > 1)
|
||||
With this option you can limit multihash to the given threads (mask).
|
||||
This can significantly improve your hashrate by using unused l3 cache.
|
||||
The default is to run the configured multihash-factor on all threads.
|
||||
|
||||
|
||||
```
|
||||
{
|
||||
...
|
||||
|
||||
"av":2,
|
||||
"doublehash-thread-mask":"0x5", // in binary -> 0101
|
||||
"multihash-factor":2,
|
||||
"multihash-thread-mask":"0x5", // in binary -> 0101
|
||||
"threads": 4,
|
||||
|
||||
...
|
||||
}
|
||||
```
|
||||
This will limit doublehash mode (av=2,av=4) to thread 0 and thread 2, thread 1 and thread 3 will run in single hashmode (av=1,av=3).
|
||||
This will limit multihash mode (multihash-factor = 2) to thread 0 and thread 2, thread 1 and thread 3 will run in single hashmode.
|
||||
|
||||
|
||||
## Common Issues
|
||||
|
|
@ -193,13 +212,18 @@ This will limit doublehash mode (av=2,av=4) to thread 0 and thread 2, thread 1 a
|
|||
|
||||
|
||||
### CPU mining performance
|
||||
Please note performance is highly dependent on system load. The numbers above are obtained on an idle system. Tasks heavily using a processor cache, such as video playback, can greatly degrade hashrate. Optimal number of threads depends on the size of the L3 cache of a processor, 1 thread requires 2 MB (Cryptonight) or 1MB (Cryptonigh-Lite) of cache.
|
||||
Please note performance is highly dependent on system load.
|
||||
The numbers above are obtained on an idle system.
|
||||
Tasks heavily using a processor cache, such as video playback, can greatly degrade hashrate.
|
||||
Optimal number of threads depends on the size of the L3 cache of a processor, 1 thread requires 2 MB (Cryptonight) or 1MB (Cryptonigh-Lite) of cache.
|
||||
|
||||
### Maximum performance checklist
|
||||
* Idle operating system.
|
||||
* Do not exceed optimal thread count.
|
||||
* Use modern CPUs with AES-NI instruction set.
|
||||
* Try setup optimal cpu affinity.
|
||||
* Try decreasing number of threads while increasing multihash-factor.
|
||||
Allocate unused cores and L3 cache with the help of multihash-thread-mask.
|
||||
* Enable fast memory (Large/Huge pages).
|
||||
|
||||
## Donations
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue