diff --git a/.github/workflows/deploy.yaml b/.github/workflows/deploy.yaml
index ebcdfe9..ff4443e 100644
--- a/.github/workflows/deploy.yaml
+++ b/.github/workflows/deploy.yaml
@@ -46,8 +46,8 @@ jobs:
uses: Jimver/cuda-toolkit@v0.2.5
with:
cuda: '11.5.1'
- method: 'network'
- sub-packages: '["nvcc", "cudart"]'
+ #method: 'local'
+ #sub-packages: '["nvcc", "cudart"]'
- name: Build on Linux GNU
if: matrix.os == 'ubuntu-18.04'
diff --git a/Cargo.lock b/Cargo.lock
index 959dd7e..edf58b7 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -648,6 +648,25 @@ version = "1.0.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "b9e0384b61958566e926dc50660321d12159025e767c18e043daf26b70104c39"
+[[package]]
+name = "include_dir"
+version = "0.7.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "482a2e29200b7eed25d7fdbd14423326760b7f6658d21a4cf12d55a50713c69f"
+dependencies = [
+ "include_dir_macros",
+]
+
+[[package]]
+name = "include_dir_macros"
+version = "0.7.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "5e074c19deab2501407c91ba1860fa3d6820bfde307db6d8cb851b55a10be89b"
+dependencies = [
+ "proc-macro2",
+ "quote",
+]
+
[[package]]
name = "indexmap"
version = "1.7.0"
@@ -745,6 +764,7 @@ version = "0.1.0"
dependencies = [
"clap",
"env_logger",
+ "include_dir",
"kaspa-miner",
"log",
"opencl3",
diff --git a/plugins/cuda/src/lib.rs b/plugins/cuda/src/lib.rs
index 23e4cea..c7c2d2e 100644
--- a/plugins/cuda/src/lib.rs
+++ b/plugins/cuda/src/lib.rs
@@ -49,81 +49,82 @@ impl Plugin for CudaPlugin {
}
//noinspection RsTypeCheck
- fn process_option(&mut self, matches: &ArgMatches) -> Result<(), kaspa_miner::Error> {
+ fn process_option(&mut self, matches: &ArgMatches) -> Result<usize, kaspa_miner::Error> {
let opts: CudaOpt = CudaOpt::from_arg_matches(matches)?;
self._enabled = !opts.cuda_disable;
+ if self._enabled {
+ let gpus: Vec<u16> = match &opts.cuda_device {
+ Some(devices) => devices.clone(),
+ None => {
+ let gpu_count = Device::num_devices().unwrap() as u16;
+ (0..gpu_count).collect()
+ }
+ };
+
+ // if any of cuda_lock_core_clocks / cuda_lock_mem_clocks / cuda_power_limit is valid, init nvml and try to apply
+ if opts.cuda_lock_core_clocks.is_some()
+ || opts.cuda_lock_mem_clocks.is_some()
+ || opts.cuda_power_limits.is_some()
+ {
+ for i in 0..gpus.len() {
+ let lock_mem_clock: Option<u32> = match &opts.cuda_lock_mem_clocks {
+ Some(mem_clocks) if i < mem_clocks.len() => Some(mem_clocks[i]),
+ Some(mem_clocks) if !mem_clocks.is_empty() => Some(*mem_clocks.last().unwrap()),
+ _ => None,
+ };
- let gpus: Vec<u16> = match &opts.cuda_device {
- Some(devices) => devices.clone(),
- None => {
- let gpu_count = Device::num_devices().unwrap() as u16;
- (0..gpu_count).collect()
- }
- };
-
- // if any of cuda_lock_core_clocks / cuda_lock_mem_clocks / cuda_power_limit is valid, init nvml and try to apply
- if opts.cuda_lock_core_clocks.is_some()
- || opts.cuda_lock_mem_clocks.is_some()
- || opts.cuda_power_limits.is_some()
- {
- for i in 0..gpus.len() {
- let lock_mem_clock: Option<u32> = match &opts.cuda_lock_mem_clocks {
- Some(mem_clocks) if i < mem_clocks.len() => Some(mem_clocks[i]),
- Some(mem_clocks) if !mem_clocks.is_empty() => Some(*mem_clocks.last().unwrap()),
- _ => None,
- };
-
- let lock_core_clock: Option<u32> = match &opts.cuda_lock_core_clocks {
- Some(core_clocks) if i < core_clocks.len() => Some(core_clocks[i]),
- Some(core_clocks) if !core_clocks.is_empty() => Some(*core_clocks.last().unwrap()),
- _ => None,
- };
-
- let power_limit: Option<u32> = match &opts.cuda_power_limits {
- Some(power_limits) if i < power_limits.len() => Some(power_limits[i]),
- Some(power_limits) if !power_limits.is_empty() => Some(*power_limits.last().unwrap()),
- _ => None,
- };
-
- let mut nvml_device: NvmlDevice = self.nvml_instance.device_by_index(gpus[i] as u32)?;
-
- if let Some(lmc) = lock_mem_clock {
- match nvml_device.set_mem_locked_clocks(lmc, lmc) {
- Err(e) => error!("set mem locked clocks {:?}", e),
- _ => info!("GPU #{} #{} lock mem clock at {} Mhz", i, &nvml_device.name()?, &lmc),
+ let lock_core_clock: Option<u32> = match &opts.cuda_lock_core_clocks {
+ Some(core_clocks) if i < core_clocks.len() => Some(core_clocks[i]),
+ Some(core_clocks) if !core_clocks.is_empty() => Some(*core_clocks.last().unwrap()),
+ _ => None,
};
- }
- if let Some(lcc) = lock_core_clock {
- match nvml_device.set_gpu_locked_clocks(lcc, lcc) {
- Err(e) => error!("set gpu locked clocks {:?}", e),
- _ => info!("GPU #{} #{} lock core clock at {} Mhz", i, &nvml_device.name()?, &lcc),
+ let power_limit: Option<u32> = match &opts.cuda_power_limits {
+ Some(power_limits) if i < power_limits.len() => Some(power_limits[i]),
+ Some(power_limits) if !power_limits.is_empty() => Some(*power_limits.last().unwrap()),
+ _ => None,
};
- };
- if let Some(pl) = power_limit {
- match nvml_device.set_power_management_limit(pl * 1000) {
- Err(e) => error!("set power limit {:?}", e),
- _ => info!("GPU #{} #{} power limit at {} W", i, &nvml_device.name()?, &pl),
+ let mut nvml_device: NvmlDevice = self.nvml_instance.device_by_index(gpus[i] as u32)?;
+
+ if let Some(lmc) = lock_mem_clock {
+ match nvml_device.set_mem_locked_clocks(lmc, lmc) {
+ Err(e) => error!("set mem locked clocks {:?}", e),
+ _ => info!("GPU #{} #{} lock mem clock at {} Mhz", i, &nvml_device.name()?, &lmc),
+ };
+ }
+
+ if let Some(lcc) = lock_core_clock {
+ match nvml_device.set_gpu_locked_clocks(lcc, lcc) {
+ Err(e) => error!("set gpu locked clocks {:?}", e),
+ _ => info!("GPU #{} #{} lock core clock at {} Mhz", i, &nvml_device.name()?, &lcc),
+ };
};
- };
+
+ if let Some(pl) = power_limit {
+ match nvml_device.set_power_management_limit(pl * 1000) {
+ Err(e) => error!("set power limit {:?}", e),
+ _ => info!("GPU #{} #{} power limit at {} W", i, &nvml_device.name()?, &pl),
+ };
+ };
+ }
}
- }
- self.specs = (0..gpus.len())
- .map(|i| CudaWorkerSpec {
- device_id: gpus[i] as u32,
- workload: match &opts.cuda_workload {
- Some(workload) if i < workload.len() => workload[i],
- Some(workload) if !workload.is_empty() => *workload.last().unwrap(),
- _ => DEFAULT_WORKLOAD_SCALE,
- },
- is_absolute: opts.cuda_workload_absolute,
- blocking_sync: !opts.cuda_no_blocking_sync,
- })
- .collect();
- Ok(())
+ self.specs = (0..gpus.len())
+ .map(|i| CudaWorkerSpec {
+ device_id: gpus[i] as u32,
+ workload: match &opts.cuda_workload {
+ Some(workload) if i < workload.len() => workload[i],
+ Some(workload) if !workload.is_empty() => *workload.last().unwrap(),
+ _ => DEFAULT_WORKLOAD_SCALE,
+ },
+ is_absolute: opts.cuda_workload_absolute,
+ blocking_sync: !opts.cuda_no_blocking_sync,
+ })
+ .collect();
+ }
+ Ok(self.specs.len())
}
}
diff --git a/plugins/opencl/Cargo.toml b/plugins/opencl/Cargo.toml
index 74bbf22..bc1bd5a 100644
--- a/plugins/opencl/Cargo.toml
+++ b/plugins/opencl/Cargo.toml
@@ -11,6 +11,7 @@ env_logger = "0.9"
opencl3 = {version = "0.6", features = ["CL_VERSION_2_1", "CL_VERSION_2_2", "CL_VERSION_3_0"]}
log = "0.4"
rand = "0.8"
+include_dir = "0.7"
[lib]
crate-type = ["cdylib"]
diff --git a/plugins/opencl/src/lib.rs b/plugins/opencl/src/lib.rs
index 5396bda..ed94d7c 100644
--- a/plugins/opencl/src/lib.rs
+++ b/plugins/opencl/src/lib.rs
@@ -45,7 +45,7 @@ impl Plugin for OpenCLPlugin {
}
//noinspection RsTypeCheck
- fn process_option(&mut self, matches: &ArgMatches) -> Result<(), kaspa_miner::Error> {
+ fn process_option(&mut self, matches: &ArgMatches) -> Result<usize, kaspa_miner::Error> {
let opts: OpenCLOpt = OpenCLOpt::from_arg_matches(matches)?;
self._enabled = opts.opencl_enable;
@@ -77,38 +77,39 @@ impl Plugin for OpenCLPlugin {
}
None => &platforms[0],
};
- info!(
- "Chose to mine on {}: {}.",
- &_platform.vendor().unwrap_or_else(|_| "Unk".into()),
- &_platform.name().unwrap_or_else(|_| "Unk".into())
- );
-
- let device_ids = _platform.get_devices(CL_DEVICE_TYPE_ALL).unwrap();
- let gpus = match opts.opencl_device {
- Some(dev) => {
- self._enabled = true;
- dev.iter().map(|d| device_ids[*d as usize]).collect::<Vec<cl_device_id>>()
- }
- None => device_ids,
- };
-
- self.specs = (0..gpus.len())
- .map(|i| OpenCLWorkerSpec {
- _platform: *_platform,
- device_id: Device::new(gpus[i]),
- workload: match &opts.opencl_workload {
- Some(workload) if i < workload.len() => workload[i],
- Some(workload) if !workload.is_empty() => *workload.last().unwrap(),
- _ => DEFAULT_WORKLOAD_SCALE,
- },
- is_absolute: opts.opencl_workload_absolute,
- experimental_amd: opts.experimental_amd,
- use_amd_binary: !opts.opencl_no_amd_binary,
- random: opts.nonce_gen,
- })
- .collect();
-
- Ok(())
+ if self._enabled {
+ info!(
+ "Chose to mine on {}: {}.",
+ &_platform.vendor().unwrap_or_else(|_| "Unk".into()),
+ &_platform.name().unwrap_or_else(|_| "Unk".into())
+ );
+
+ let device_ids = _platform.get_devices(CL_DEVICE_TYPE_ALL).unwrap();
+ let gpus = match opts.opencl_device {
+ Some(dev) => {
+ self._enabled = true;
+ dev.iter().map(|d| device_ids[*d as usize]).collect::<Vec<cl_device_id>>()
+ }
+ None => device_ids,
+ };
+
+ self.specs = (0..gpus.len())
+ .map(|i| OpenCLWorkerSpec {
+ _platform: *_platform,
+ device_id: Device::new(gpus[i]),
+ workload: match &opts.opencl_workload {
+ Some(workload) if i < workload.len() => workload[i],
+ Some(workload) if !workload.is_empty() => *workload.last().unwrap(),
+ _ => DEFAULT_WORKLOAD_SCALE,
+ },
+ is_absolute: opts.opencl_workload_absolute,
+ experimental_amd: opts.experimental_amd,
+ use_amd_binary: !opts.opencl_no_amd_binary,
+ random: opts.nonce_gen,
+ })
+ .collect();
+ }
+ Ok(self.specs.len())
}
}
diff --git a/plugins/opencl/src/worker.rs b/plugins/opencl/src/worker.rs
index da2a7e9..87a4174 100644
--- a/plugins/opencl/src/worker.rs
+++ b/plugins/opencl/src/worker.rs
@@ -1,8 +1,9 @@
use crate::cli::NonceGenEnum;
use crate::Error;
+use include_dir::{include_dir, Dir};
use kaspa_miner::xoshiro256starstar::Xoshiro256StarStar;
use kaspa_miner::Worker;
-use log::info;
+use log::{info, warn};
use opencl3::command_queue::{CommandQueue, CL_QUEUE_OUT_OF_ORDER_EXEC_MODE_ENABLE};
use opencl3::context::Context;
use opencl3::device::Device;
@@ -18,6 +19,7 @@ use std::ffi::c_void;
use std::ptr;
use std::sync::Arc;
+static BINARY_DIR: Dir = include_dir!("./plugins/opencl/resources/bin/");
static PROGRAM_SOURCE: &str = include_str!("../resources/kaspa-opencl.cl");
pub struct OpenCLGPUWorker {
@@ -197,70 +199,21 @@ impl OpenCLGPUWorker {
let program = match use_binary {
true => {
- let device_name = device.name().unwrap_or_else(|_| "Unknown".into()).to_lowercase();
+ let mut device_name = device.name().unwrap_or_else(|_| "Unknown".into()).to_lowercase();
+ if device_name.contains(':') {
+ device_name = device_name.split_once(':').expect("We checked for `:`").0.to_string();
+ }
info!("{}: Looking for binary for {}", name, device_name);
- match device_name.as_str() {
- /*"ellesmere" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/ellesmere_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|e| {
- panic!("{}::Program::create_and_build_from_binary failed: {}", name, String::from(e))
- }),*/
- "gfx906" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx906_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx908" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx908_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1010" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1010_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1011" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1011_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1012" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1012_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1030" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1030_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1031" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1031_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|_| panic!("{}::Program::create_and_build_from_binary failed", name)),
- "gfx1032" => Program::create_and_build_from_binary(
- &context,
- &[include_bytes!("../resources/bin/gfx1032_kaspa-opencl.bin")],
- "",
- )
- .unwrap_or_else(|e| panic!("{}::Program::create_and_build_from_binary failed: {}", name, e)),
- other => {
- panic!(
- "{}: Found device {} without prebuilt binary. Trying to run without --opencl-amd-binary.",
- name, other
- );
+ match BINARY_DIR.get_file(format!("{}_kaspa-opencl.bin", device_name)) {
+ Some(binary) => {
+ Program::create_and_build_from_binary(&context, &[binary.contents()], "").unwrap_or_else(|e|{
+ warn!("{}::Program::create_and_build_from_source failed: {}. Reverting to compiling from source", name, e);
+ from_source(&context, &device, options).unwrap_or_else(|e| panic!("{}::Program::create_and_build_from_binary failed: {}", name, e))
+ })
+ },
+ None => {
+ warn!("Binary file not found for {}. Reverting to compiling from source.", device_name);
+ from_source(&context, &device, options).unwrap_or_else(|e| panic!("{}::Program::create_and_build_from_binary failed: {}", name, e))
}
}
}
diff --git a/src/cli.rs b/src/cli.rs
index 2cf1cbb..6a0b47a 100644
--- a/src/cli.rs
+++ b/src/cli.rs
@@ -70,8 +70,12 @@ impl Opt {
}
if !self.kaspad_address.contains("://") {
- let port = self.port();
- self.kaspad_address = format!("grpc://{}:{}", self.kaspad_address, port);
+ let port_str = self.port().to_string();
+ let (kaspad, port) = match self.kaspad_address.contains(':') {
+ true => self.kaspad_address.split_once(':').expect("We checked for `:`"),
+ false => (self.kaspad_address.as_str(), port_str.as_str()),
+ };
+ self.kaspad_address = format!("grpc://{}:{}", kaspad, port);
}
log::info!("kaspad address: {}", self.kaspad_address);
diff --git a/src/lib.rs b/src/lib.rs
index 5b049a8..556c9c7 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -65,13 +65,17 @@ impl PluginManager {
Ok(specs)
}
- pub fn process_options(&mut self, matchs: &ArgMatches) -> Result<(), Error> {
+ /**
+ Process the options for a plugin, and reports how many workers are available
+ */
+ pub fn process_options(&mut self, matchs: &ArgMatches) -> Result<usize, Error> {
+ let mut count = 0usize;
self.plugins.iter_mut().for_each(|plugin| {
- plugin
+ count += plugin
.process_option(matchs)
.unwrap_or_else(|_| panic!("Could not process option for plugin {}", plugin.name()))
});
- Ok(())
+ Ok(count)
}
pub fn has_specs(&self) -> bool {
@@ -83,7 +87,7 @@ pub trait Plugin: Any + Send + Sync {
fn name(&self) -> &'static str;
fn enabled(&self) -> bool;
fn get_worker_specs(&self) -> Vec<Box<dyn WorkerSpec>>;
- fn process_option(&mut self, matchs: &ArgMatches) -> Result<(), Error>;
+ fn process_option(&mut self, matchs: &ArgMatches) -> Result<usize, Error>;
}
pub trait WorkerSpec: Any + Send + Sync {
@@ -116,7 +120,7 @@ pub fn load_plugins<'help>(
for path in paths {
app = unsafe {
factory.load_single_plugin(app, path.as_str()).unwrap_or_else(|(app, e)| {
- eprintln!("Failed loading plugin {}: {}", path, e);
+ eprintln!("WARNING: Failed loading plugin {} (ignore if you do not intend to use): {}", path, e);
app
})
};
diff --git a/src/main.rs b/src/main.rs
index aec3195..8e40553 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -117,11 +117,16 @@ async fn main() -> Result<(), Error> {
let matches = app.get_matches();
- plugin_manager.process_options(&matches)?;
+ let worker_count = plugin_manager.process_options(&matches)?;
let mut opt: Opt = Opt::from_arg_matches(&matches)?;
opt.process()?;
env_logger::builder().filter_level(opt.log_level()).parse_default_env().init();
info!("Found plugins: {:?}", plugins);
+ info!("Plugins found {} workers", worker_count);
+ if worker_count == 0 && opt.num_threads.unwrap_or(0) == 0 {
+ error!("No workers specified");
+ return Err("No workers specified".into());
+ }
let block_template_ctr = Arc::new(AtomicU16::new((thread_rng().next_u64() % 10_000u64) as u16));
if opt.devfund_percent > 0 {