syncacide commented 6 years ago

Please provide as much as possible information to reproduce the issue.

Hwloc: Memory Pinned. miner will run for a short amount of time with 7700-7900 h/s eventurally is will freeze and windows crashes to blue screen and then restarts.

Basic information

intel 4400
1 vega 56 flashed to 64, 3x vega 64
- 8g of DD4 2133 MHz (2x 4gb stick) -120g ssd

Compile issues

Windows 10

/*

pool_address - Pool address should be in the form "pool.supportxmr.com:3333". Only stratum pools are supported.
wallet_address - Your wallet, or pool login.
pool_password - Can be empty in most cases or "x".
use_nicehash - Limit the nonce to 3 bytes as required by nicehash.
use_tls - This option will make us connect using Transport Layer Security.
tls_fingerprint - Server's SHA256 fingerprint. If this string is non-empty then we will check the server's cert against it.
pool_weight - Pool weight is a number telling the miner how important the pool is. Miner will mine mostly at the pool
with the highest weight, unless the pool fails. Weight must be an integer larger than 0.
We feature pools up to 1MH/s. For a more complete list see M5M400's pool list at www.moneropools.com */ "pool_list" : [ {"pool_address" : "pool.supportxmr.com:7777", "wallet_address" : "4BHg3KsmxZkWJTKpG53huM4eSutiuoZMXBigxEDwNeD7KuC6UiW6UJDQMYLMQXSWpgK3Ew3MQ4gW8UbskwubecNaDMDuySu", "pool_password" : "x", "use_nicehash" : false, "use_tls" : false, "tls_fingerprint" : "", "pool_weight" : 1 }, ],

/*

currency to mine
allowed values: 'monero' or 'aeon' */ "currency" : "monero",

/*

Network timeouts.
Because of the way this client is written it doesn't need to constantly talk (keep-alive) to the server to make
sure it is there. We detect a buggy / overloaded server by the call timeout. The default values will be ok for
nearly all cases. If they aren't the pool has most likely overload issues. Low call timeout values are preferable -
long timeouts mean that we waste hashes on potentially stale jobs. Connection report will tell you how long the
server usually takes to process our calls.
call_timeout - How long should we wait for a response from the server before we assume it is dead and drop the connection.
retry_time - How long should we wait before another connection attempt.
Both values are in seconds.
giveup_limit - Limit how many times we try to reconnect to the pool. Zero means no limit. Note that stak miners
don't mine while the connection is lost, so your computer's power usage goes down to idle. */ "call_timeout" : 10, "retry_time" : 30, "giveup_limit" : 0,

/*

Output control.
Since most people are used to miners printing all the time, that's what we do by default too. This is suboptimal
really, since you cannot see errors under pages and pages of text and performance stats. Given that we have internal
performance monitors, there is very little reason to spew out pages of text instead of concise reports.
Press 'h' (hashrate), 'r' (results) or 'c' (connection) to print reports.
verbose_level - 0 - Don't print anything.
1 - Print intro, connection event, disconnect event
2 - All of level 1, and new job (block) event if the difficulty is different from the last job
3 - All of level 1, and new job (block) event in all cases, result submission event.
4 - All of level 3, and automatic hashrate report printing
print_motd - Display messages from your pool operator in the hashrate result. */ "verbose_level" : 3, "print_motd" : true,

/*

Automatic hashrate report
h_print_time - How often, in seconds, should we print a hashrate report if verbose_level is set to 4.
This option has no effect if verbose_level is not 4. */ "h_print_time" : 60,

/*

Manual hardware AES override
Some VMs don't report AES capability correctly. You can set this value to true to enforce hardware AES or
to false to force disable AES or null to let the miner decide if AES is used.
WARNING: setting this to true on a CPU that doesn't support hardware AES will crash the miner. */ "aes_override" : null,

/*

LARGE PAGE SUPPORT
Large pages need a properly set up OS. It can be difficult if you are not used to systems administration,
but the performance results are worth the trouble - you will get around 20% boost. Slow memory mode is
meant as a backup, you won't get stellar results there. If you are running into trouble, especially
on Windows, please read the common issues in the README.
By default we will try to allocate large pages. This means you need to "Run As Administrator" on Windows.
You need to edit your system's group policies to enable locking large pages. Here are the steps from MSDN
1. On the Start menu, click Run. In the Open box, type gpedit.msc.
1. On the Local Group Policy Editor console, expand Computer Configuration, and then expand Windows Settings.
1. Expand Security Settings, and then expand Local Policies.
1. Select the User Rights Assignment folder.
1. The policies will be displayed in the details pane.
1. In the pane, double-click Lock pages in memory.
1. In the Local Security Setting – Lock pages in memory dialog box, click Add User or Group.
1. In the Select Users, Service Accounts, or Groups dialog box, add an account that you will run the miner on
1. Reboot for change to take effect.
Windows also tends to fragment memory a lot. If you are running on a system with 4-8GB of RAM you might need
to switch off all the auto-start applications and reboot to have a large enough chunk of contiguous memory.
On Linux you will need to configure large page support "sudo sysctl -w vm.nr_hugepages=128" and increase your
ulimit -l. To do do this you need to add following lines to /etc/security/limits.conf - "* soft memlock 262144"
and "* hard memlock 262144". You can also do it Windows-style and simply run-as-root, but this is NOT
recommended for security reasons.
Memory locking means that the kernel can't swap out the page to disk - something that is unlikely to happen on a
command line system that isn't starved of memory. I haven't observed any difference on a CLI Linux system between
locked and unlocked memory. If that is your setup see option "no_mlck". */

/*

use_slow_memory defines our behaviour with regards to large pages. There are three possible options here:
always - Don't even try to use large pages. Always use slow memory.
warn - We will try to use large pages, but fall back to slow memory if that fails.
no_mlck - This option is only relevant on Linux, where we can use large pages without locking memory.
It will never use slow memory, but it won't attempt to mlock
never - If we fail to allocate large pages we will print an error and exit. */ "use_slow_memory" : "warn",

/*

TLS Settings
If you need real security, make sure tls_secure_algo is enabled (otherwise MITM attack can downgrade encryption
to trivially breakable stuff like DES and MD5), and verify the server's fingerprint through a trusted channel.
tls_secure_algo - Use only secure algorithms. This will make us quit with an error if we can't negotiate a secure algo. */ "tls_secure_algo" : true,

/*

Daemon mode
If you are running the process in the background and you don't need the keyboard reports, set this to true.
This should solve the hashrate problems on some emulated terminals. */ "daemon_mode" : false,

/*

Buffered output control.
When running the miner through a pipe, standard output is buffered. This means that the pipe won't read
each output line immediately. This can cause delays when running in background.
Set this option to true to flush stdout after each line, so it can be read immediately. */ "flush_stdout" : false,

/*

Output file
output_file - This option will log all output to a file.
*/ "output_file" : "",

/*

Built-in web server
I like checking my hashrate on my phone. Don't you?
Keep in mind that you will need to set up port forwarding on your router if you want to access it from
outside of your home network. Ports lower than 1024 on Linux systems will require root.
httpd_port - Port we should listen on. Default, 0, will switch off the server. */ "httpd_port" : 0,

/*

HTTP Authentication
This allows you to set a password to keep people on the Internet from snooping on your hashrate.
Keep in mind that this is based on HTTP Digest, which is based on MD5. To a determined attacker
who is able to read your traffic it is as easy to break a bog door latch.
http_login - Login. Empty login disables authentication.
http_pass - Password. */ "http_login" : "", "http_pass" : "",

/*

prefer_ipv4 - IPv6 preference. If the host is available on both IPv4 and IPv6 net, which one should be choose?
This setting will only be needed in 2020's. No need to worry about it now. */ "prefer_ipv4" : true,
```
run `cmake -LA .` in the build folder and add the output here
```

Issue with the execution

Do you compiled the miner by our own?

run `./xmr-stak --version-long` and add the output here

AMD OpenCl issue

/*

GPU configuration. You should play around with intensity and worksize as the fastest settings will vary.
index - GPU index number usually starts from 0
intensity - Number of parallel GPU threads (nothing to do with CPU threads)
worksize - Number of local GPU threads (nothing to do with CPU threads)
affine_to_cpu - This will affine the thread to a CPU. This can make a GPU miner play along nicer with a CPU miner.
strided_index - switch memory pattern used for the scratch pad memory
true = use 16byte contiguous memory per thread, the next memory block has offset of intensity blocks
false = use a contiguous block of memory per thread
"gpu_threads_conf" :
[
{ "index" : 0, "intensity" : 1000, "worksize" : 8, "affine_to_cpu" : false, "strided_index" : true },
], */

"gpu_threads_conf" : [ { "index" : 0, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 0, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 1, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 1, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 2, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 2, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 3, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, { "index" : 3, "intensity" : 1925, "worksize" : 8, "affine_to_cpu" : 0, "strided_index" : true }, ],

/*

Platform index. This will be 0 unless you have different OpenCL platform - eg. AMD and Intel. */ "platform_index" : 1,

/*

Thread configuration for each thread. Make sure it matches the number above.
low_power_mode - This can either be a boolean (true or false), or a number between 1 to 5. When set to true, this mode will double the cache usage, and double the single thread performance. It will
consume much less power (as less cores are working), but will max out at around 80-85% of
the maximum performance. When set to a number N greater than 1, this mode will increase the
cache usage and single thread performance by N times.
no_prefetch - Some sytems can gain up to extra 5% here, but sometimes it will have no difference or make
things slower.
affine_to_cpu - This can be either false (no affinity), or the CPU core number. Note that on hyperthreading
systems it is better to assign threads to physical cores. On Windows this usually means selecting
even or odd numbered cpu numbers. For Linux it will be usually the lower CPU numbers, so for a 4
physical core CPU you should select cpu numbers 0-3.
On the first run the miner will look at your system and suggest a basic configuration that will work,
you can try to tweak it from there to get the best performance.
A filled out configuration should look like this:
"cpu_threads_conf" :
[
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },
], */

"cpu_threads_conf" : [ // { "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 0 }, { "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },

],

[2017-12-15 01:40:53] : Start mining: MONERO WARNING: NVIDIA Insufficient driver! [2017-12-15 01:40:53] : WARNING: backend NVIDIA disabled. [2017-12-15 01:40:53] : Compiling code and initializing GPUs. This will take a while... [2017-12-15 01:40:53] : Device 0 work size 8 / 32. [2017-12-15 01:40:58] : Device 0 work size 8 / 32. [2017-12-15 01:41:03] : Device 1 work size 8 / 32. [2017-12-15 01:41:08] : Device 1 work size 8 / 32. [2017-12-15 01:41:13] : Device 2 work size 8 / 32. [2017-12-15 01:41:18] : Device 2 work size 8 / 32. [2017-12-15 01:41:23] : Device 3 work size 8 / 32. [2017-12-15 01:41:28] : Device 3 work size 8 / 32. [2017-12-15 01:41:33] : Starting AMD GPU thread 0, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 1, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 2, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 3, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 4, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 5, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 6, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting AMD GPU thread 7, affinity: 0. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:33] : Starting 1x thread, affinity: 1. [2017-12-15 01:41:33] : hwloc: memory pinned [2017-12-15 01:41:59] : Result accepted by the pool. [2017-12-15 01:42:25] : New block detected. HASHRATE REPORT - CPU | ID | 10s | 60s | 15m | | 0 | 57.5 | (na) | (na) |

HASHRATE REPORT - AMD | ID | 10s | 60s | 15m | ID | 10s | 60s | 15m | | 0 | 851.1 | (na) | (na) | 1 | 862.1 | (na) | (na) | | 2 | 982.5 | (na) | (na) | 3 | 987.2 | (na) | (na) | | 4 | 994.1 | (na) | (na) | 5 | 980.4 | (na) | (na) | | 6 | 990.5 | (na) | (na) | 7 | 988.6 | (na) | (na) |

Totals: 7694.0 (na) (na) H/s Highest: 7756.3 H/s [2017-12-15 01:42:32] : Result accepted by the pool.

Stability issue

GPU 1100hz power tables, cpu no overclocked
Is the Main memory of the CPU or GPU undervolted?

JeffBarrowbank commented 6 years ago

I also got HWLOC:MEMORY PINNED issue

[2017-12-15 17:54:24] : Starting 1x thread, affinity: 22. [2017-12-15 17:54:24] : hwloc: memory pinned [2017-12-15 17:54:24] : Fast-connecting to pool.electroneum.space:3333 pool ... [2017-12-15 17:56:34] : SOCKET ERROR - [pool.electroneum.space:3333] CONNECT error: Connection timed out

Followed the Large pages recommendations on Lubunto 16 LTS. Tried everything I can think of to change the settings etc. Cannot get the miner to work.

JerichoJones commented 6 years ago

"hwloc: memory pinned" is not an error. It indicates that the memory was successfully reserved.

@syncacide Not every card wins the silicon lottery. Lower clocks and/or Intensity until it stabilizes. You may have different settings per card.

fireice-uk / xmr-stak

Hwloc: Memory Pinned. #557