possible nvml fixgpu_fix_pray

3 files changed, 66 insertions, 39 deletions

diff --git a/src/data_collection/nvidia.rs b/src/data_collection/nvidia.rs
index d463d2c4..c833b711 100644
--- a/src/data_collection/nvidia.rs
+++ b/src/data_collection/nvidia.rs
@@ -2,14 +2,17 @@ use std::sync::OnceLock;
 
 use hashbrown::HashMap;
 use nvml_wrapper::{
-    enum_wrappers::device::TemperatureSensor, enums::device::UsedGpuMemory, error::NvmlError, Nvml,
+    enum_wrappers::device::{PerformanceState, TemperatureSensor},
+    enums::device::UsedGpuMemory,
+    error::NvmlError,
+    Nvml,
 };
 
 use crate::{
     app::{filter::Filter, layout_manager::UsedWidgets},
     data_collection::{
         memory::MemHarvest,
-        temperature::{is_temp_filtered, TempHarvest, TemperatureType},
+        temperature::{TempHarvest, TemperatureType},
     },
 };
 
@@ -53,17 +56,42 @@ pub fn get_nvidia_vecs(
                                 ));
                             }
                         }
-                        if widgets_to_harvest.use_temp && is_temp_filtered(filter, &name) {
-                            if let Ok(temperature) = device.temperature(TemperatureSensor::Gpu) {
-                                let temperature = temp_type.convert_temp_unit(temperature as f32);
 
-                                temp_vec.push(TempHarvest {
-                                    name: name.clone(),
-                                    temperature: TemperatureReading::Value(temperature),
-                                });
+                        if widgets_to_harvest.use_temp
+                            && filter
+                                .as_ref()
+                                .map(|filter| filter.keep_entry(&name))
+                                .unwrap_or(true)
+                        {
+                            // Following https://docs.nvidia.com/gameworks/content/gameworkslibrary/coresdk/nvapi/group__gpupstate.html,
+                            // it seems like performance state 12 and lower are "minimum idle power consumption".
+                            match device.performance_state() {
+                                Ok(PerformanceState::Fifteen)
+                                | Ok(PerformanceState::Fourteen)
+                                | Ok(PerformanceState::Thirteen)
+                                | Ok(PerformanceState::Twelve) => {
+                                    temp_vec.push(TempHarvest {
+                                        name,
+                                        temperature: TemperatureReading::Off,
+                                    });
+                                }
+                                _ => {
+                                    if let Ok(temperature) =
+                                        device.temperature(TemperatureSensor::Gpu)
+                                    {
+                                        let temperature =
+                                            temp_type.convert_temp_unit(temperature as f32);
+
+                                        temp_vec.push(TempHarvest {
+                                            name,
+                                            temperature: TemperatureReading::Value(temperature),
+                                        });
+                                    }
+                                }
                             }
                         }
                     }
+
                     if widgets_to_harvest.use_proc {
                         let mut procs = HashMap::new();
                         if let Ok(gpu_procs) = device.process_utilization_stats(None) {
@@ -73,6 +101,7 @@ pub fn get_nvidia_vecs(
                                 procs.insert(pid, (0, gpu_util));
                             }
                         }
+
                         if let Ok(compute_procs) = device.running_compute_processes() {
                             for proc in compute_procs {
                                 let pid = proc.pid;
@@ -87,7 +116,8 @@ pub fn get_nvidia_vecs(
                                 }
                             }
                         }
-                        // Use the legacy API too but prefer newer API results
+
+                        // Use the legacy API too, but prefer newer API results
                         if let Ok(graphics_procs) = device.running_graphics_processes_v2() {
                             for proc in graphics_procs {
                                 let pid = proc.pid;
@@ -102,6 +132,7 @@ pub fn get_nvidia_vecs(
                                 }
                             }
                         }
+
                         if let Ok(graphics_procs) = device.running_graphics_processes() {
                             for proc in graphics_procs {
                                 let pid = proc.pid;
@@ -116,9 +147,11 @@ pub fn get_nvidia_vecs(
                                 }
                             }
                         }
+
                         if !procs.is_empty() {
                             proc_vec.push(procs);
                         }
+
                         // running total for proc %
                         if let Ok(mem) = device.memory_info() {
                             total_mem += mem.total;
diff --git a/src/data_collection/temperature.rs b/src/data_collection/temperature.rs
index 2726ab65..b7413244 100644
--- a/src/data_collection/temperature.rs
+++ b/src/data_collection/temperature.rs
@@ -13,8 +13,6 @@ cfg_if::cfg_if! {
     }
 }
 
-use crate::app::filter::Filter;
-
 #[derive(Default, Debug, Clone)]
 pub enum TemperatureReading {
     Value(f32),
@@ -56,21 +54,6 @@ impl TemperatureType {
     }
 }
 
-pub fn is_temp_filtered(filter: &Option<Filter>, text: &str) -> bool {
-    if let Some(filter) = filter {
-        let mut ret = filter.is_list_ignored;
-        for r in &filter.list {
-            if r.is_match(text) {
-                ret = !filter.is_list_ignored;
-                break;
-            }
-        }
-        ret
-    } else {
-        true
-    }
-}
-
 #[cfg(test)]
 mod test {
     use crate::data_collection::temperature::TemperatureType;
diff --git a/src/data_collection/temperature/linux.rs b/src/data_collection/temperature/linux.rs
index c7608539..125fb0ca 100644
--- a/src/data_collection/temperature/linux.rs
+++ b/src/data_collection/temperature/linux.rs
@@ -8,7 +8,7 @@ use std::{
 use anyhow::Result;
 use hashbrown::{HashMap, HashSet};
 
-use super::{is_temp_filtered, TempHarvest, TemperatureReading, TemperatureType};
+use super::{TempHarvest, TemperatureReading, TemperatureType};
 use crate::{app::filter::Filter, utils::error::BottomError};
 
 const EMPTY_NAME: &str = "Unknown";
@@ -184,8 +184,12 @@ fn finalize_name(
 /// If neither are found, it will always return true and be treated as "awake".
 #[inline]
 fn is_device_awake(path: &Path) -> bool {
+    // XXX: Should we initialize all devices that support runtime_status_path here,
+    // in a map, and take `power/autosuspend_delay_ms` into account?
+
     // Try checking `power/runtime_status` if it exists! For more information, see
-    // https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-devices-power
+    // https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-devices-power and
+    // https://gitlab.com/mission-center-devs/mission-center/-/issues/30#note_1697130114
     let runtime_status_path = path.join("power/runtime_status");
     if runtime_status_path.exists() {
         if let Ok(status) = fs::read_to_string(runtime_status_path) {
@@ -337,15 +341,18 @@ fn hwmon_temperatures(temp_type: &TemperatureType, filter: &Option<Filter>) -> H
                 let name = finalize_name(hwmon_name, sensor_label, &sensor_name, &mut seen_names);
 
                 // TODO: It's possible we may want to move the filter check further up to avoid probing hwmon if not needed?
-                if is_temp_filtered(filter, &name) {
-                    if let Ok(temp_celsius) = parse_temp(&temp_path) {
-                        temperatures.push(TempHarvest {
-                            name,
-                            temperature: TemperatureReading::Value(
-                                temp_type.convert_temp_unit(temp_celsius),
-                            ),
-                        });
-                    }
+                if filter
+                    .as_ref()
+                    .map(|filter| filter.keep_entry(&name))
+                    .unwrap_or(true)
+                {
+                    let temperature = if let Ok(temp_celsius) = parse_temp(&temp_path) {
+                        TemperatureReading::Value(temp_type.convert_temp_unit(temp_celsius))
+                    } else {
+                        TemperatureReading::Unavailable
+                    };
+
+                    temperatures.push(TempHarvest { name, temperature });
                 }
             }
         }
@@ -388,7 +395,11 @@ fn add_thermal_zone_temperatures(
                     name
                 };
 
-                if is_temp_filtered(filter, &name) {
+                if filter
+                    .as_ref()
+                    .map(|filter| filter.keep_entry(&name))
+                    .unwrap_or(true)
+                {
                     let temp_path = file_path.join("temp");
                     if let Ok(temp_celsius) = parse_temp(&temp_path) {
                         let name = counted_name(&mut seen_names, name);