Merge master into ascend branch

Author: Syllo

AppImage/make_appimage_ubuntu_1804_docker.sh

@@ -2,7 +2,7 @@
 
 install_deps() {
   apt-get update
-  apt-get install -y cmake gcc g++ libncurses5-dev libncursesw5-dev libdrm-dev wget file libudev-dev
+  apt-get install -y gcc g++ libncurses5-dev libncursesw5-dev libdrm-dev wget file libudev-dev
 }
 
 configure_nvtop() {
@@ -23,8 +23,15 @@ get_linuxdeploy() {
   ./linuxdeploy-x86_64.AppImage --appimage-extract
 }
 
+get_cmake() {
+  wget -nc https://github.com/Kitware/CMake/releases/download/v3.18.0/cmake-3.18.0.tar.gz
+  tar zxf cmake-3.18.0.tar.gz
+  ./cmake-3.18.0/bootstrap --prefix=/usr && make && make install
+}
+
 create_AppImage() {
   install_deps
+  get_cmake
   mkdir nvtop_build
   cd nvtop_build
   configure_nvtop

CMakeLists.txt

@@ -66,20 +66,26 @@ if(APPLE)
   set(AMDGPU_SUPPORT_DEFAULT OFF)
   set(INTEL_SUPPORT_DEFAULT OFF)
   set(MSM_SUPPORT_DEFAULT OFF)
+  set(PANFROST_SUPPORT_DEFAULT OFF)
+  set(PANTHOR_SUPPORT_DEFAULT OFF)
   set(ASCEND_SUPPORT_DEFAULT OFF)
 elseif(ASCEND_SUPPORT)
   set(APPLE_SUPPORT_DEFAULT OFF)
   set(NVIDIA_SUPPORT_DEFAULT OFF)
   set(AMDGPU_SUPPORT_DEFAULT OFF)
   set(INTEL_SUPPORT_DEFAULT OFF)
   set(MSM_SUPPORT_DEFAULT OFF)
+  set(PANFROST_SUPPORT_DEFAULT OFF)
+  set(PANTHOR_SUPPORT_DEFAULT OFF)
   set(ASCEND_SUPPORT_DEFAULT ON)
 else()
   set(APPLE_SUPPORT_DEFAULT OFF)
   set(NVIDIA_SUPPORT_DEFAULT ON)
   set(AMDGPU_SUPPORT_DEFAULT ON)
   set(INTEL_SUPPORT_DEFAULT ON)
   set(MSM_SUPPORT_DEFAULT ON)
+  set(PANFROST_SUPPORT_DEFAULT ON)
+  set(PANTHOR_SUPPORT_DEFAULT ON)
   set(ASCEND_SUPPORT_DEFAULT OFF)
 endif()
 
@@ -88,6 +94,8 @@ option(AMDGPU_SUPPORT "Build support for AMD GPUs through amdgpu driver" ${AMDGP
 option(INTEL_SUPPORT "Build support for Intel GPUs through i915 driver" ${INTEL_SUPPORT_DEFAULT})
 option(MSM_SUPPORT "Build support for Adreno GPUs through msm driver" ${MSM_SUPPORT_DEFAULT})
 option(APPLE_SUPPORT "Build support for Apple GPUs through Metal" ${APPLE_SUPPORT_DEFAULT})
+option(PANFROST_SUPPORT "Build support for Mali GPUs through panfrost driver" ${PANFROST_SUPPORT_DEFAULT})
+option(PANTHOR_SUPPORT "Build support for Mali GPUs through panthor driver" ${PANTHOR_SUPPORT_DEFAULT})
 option(ASCEND_SUPPORT "Build support for Ascend NPUs through Ascend DCMI" ${ASCEND_SUPPORT_DEFAULT})
 
 add_subdirectory(src)

Dockerfile

@@ -11,9 +11,13 @@ FROM ${IMAGE} as builder
 ENV DEBIAN_FRONTEND=noninteractive
 
 RUN apt-get update && \
-  apt-get install -yq build-essential cmake libncurses5-dev libncursesw5-dev \
+  apt-get install -yq build-essential wget libncurses5-dev libncursesw5-dev libssl-dev \
   pkg-config libdrm-dev libgtest-dev libudev-dev
 
+RUN wget https://github.com/Kitware/CMake/releases/download/v3.18.0/cmake-3.18.0.tar.gz
+RUN tar zxf cmake-3.18.0.tar.gz
+RUN ./cmake-3.18.0/bootstrap --prefix=/usr && make && make install
+
 COPY . /nvtop
 WORKDIR /nvtop
 RUN mkdir -p /nvtop/build && \

include/nvtop/extract_gpuinfo.h

@@ -38,6 +38,8 @@ bool gpuinfo_fix_dynamic_info_from_process_info(struct list_head *devices);
 
 bool gpuinfo_refresh_processes(struct list_head *devices);
 
+bool gpuinfo_utilisation_rate(struct list_head *devices);
+
 void gpuinfo_clean(struct list_head *devices);
 
 void gpuinfo_clear_cache(void);

include/nvtop/extract_gpuinfo_common.h

@@ -30,6 +30,9 @@
 
 #include "list.h"
 
+#define STRINGIFY(x) STRINGIFY_HELPER_(x)
+#define STRINGIFY_HELPER_(x) #x
+
 #define IS_VALID(x, y) ((y)[(x) / CHAR_BIT] & (1 << ((x) % CHAR_BIT)))
 #define SET_VALID(x, y) ((y)[(x) / CHAR_BIT] |= (1 << ((x) % CHAR_BIT)))
 #define RESET_VALID(x, y) ((y)[(x) / CHAR_BIT] &= ~(1 << ((x) % CHAR_BIT)))
@@ -55,6 +58,9 @@ enum gpuinfo_static_info_valid {
   gpuinfo_max_pcie_link_width_valid,
   gpuinfo_temperature_shutdown_threshold_valid,
   gpuinfo_temperature_slowdown_threshold_valid,
+  gpuinfo_n_shared_cores_valid,
+  gpuinfo_l2cache_size_valid,
+  gpuinfo_n_exec_engines_valid,
   gpuinfo_static_info_count,
 };
 
@@ -66,6 +72,9 @@ struct gpuinfo_static_info {
   unsigned max_pcie_link_width;
   unsigned temperature_shutdown_threshold;
   unsigned temperature_slowdown_threshold;
+  unsigned n_shared_cores;
+  unsigned l2cache_size;
+  unsigned n_exec_engines;
   bool integrated_graphics;
   unsigned char valid[(gpuinfo_static_info_count + CHAR_BIT - 1) / CHAR_BIT];
 };
@@ -146,6 +155,8 @@ enum gpuinfo_process_info_valid {
   gpuinfo_process_cpu_usage_valid,
   gpuinfo_process_cpu_memory_virt_valid,
   gpuinfo_process_cpu_memory_res_valid,
+  gpuinfo_process_gpu_cycles_valid,
+  gpuinfo_process_sample_delta_valid,
   gpuinfo_process_info_count
 };
 
@@ -154,10 +165,12 @@ struct gpu_process {
   pid_t pid;                           // Process ID
   char *cmdline;                       // Process User Name
   char *user_name;                     // Process User Name
+  uint64_t sample_delta;               // Time spent between two successive samples
   uint64_t gfx_engine_used;            // Time in nanoseconds this process spent using the GPU gfx
   uint64_t compute_engine_used;        // Time in nanoseconds this process spent using the GPU compute
   uint64_t enc_engine_used;            // Time in nanoseconds this process spent using the GPU encoder
   uint64_t dec_engine_used;            // Time in nanoseconds this process spent using the GPU decoder
+  uint64_t gpu_cycles;                 // Number of GPU cycles spent in the GPU gfx engine
   unsigned gpu_usage;                  // Percentage of GPU used by the process
   unsigned encode_usage;               // Percentage of GPU encoder used by the process
   unsigned decode_usage;               // Percentage of GPU decoder used by the process
@@ -184,6 +197,7 @@ struct gpu_vendor {
 
   void (*populate_static_info)(struct gpu_info *gpu_info);
   void (*refresh_dynamic_info)(struct gpu_info *gpu_info);
+  void (*refresh_utilisation_rate)(struct gpu_info *gpu_info);
 
   void (*refresh_running_processes)(struct gpu_info *gpu_info);
   char *name;
@@ -205,6 +219,8 @@ void register_gpu_vendor(struct gpu_vendor *vendor);
 
 bool extract_drm_fdinfo_key_value(char *buf, char **key, char **val);
 
+void gpuinfo_refresh_utilisation_rate(struct gpu_info *gpu_info);
+
 // fdinfo DRM interface names common to multiple drivers
 extern const char drm_pdev[];
 extern const char drm_client_id[];

include/nvtop/interface_internal_common.h

@@ -65,6 +65,9 @@ struct device_window {
   WINDOW *gpu_clock_info;
   WINDOW *mem_clock_info;
   WINDOW *pcie_info;
+  WINDOW *shader_cores;
+  WINDOW *l2_cache_size;
+  WINDOW *exec_engines;
   bool enc_was_visible;
   bool dec_was_visible;
   nvtop_time last_decode_seen;
@@ -140,6 +143,9 @@ enum device_field {
   device_power,
   device_pcie,
   device_clock,
+  device_shadercores,
+  device_l2features,
+  device_execengines,
   device_field_count,
 };
 

include/nvtop/interface_options.h

@@ -49,7 +49,8 @@ typedef struct nvtop_interface_option_struct {
                                                     // process list are displayed
   bool show_startup_messages;                       // True to show the startup messages
   bool filter_nvtop_pid;                            // Do not show nvtop pid in the processes list
-  bool has_monitored_set_changed; // True if the set of monitored gpu was modified through the interface
+  bool has_monitored_set_changed;                   // True if the set of monitored gpu was modified through the interface
+  bool has_gpu_info_bar;                            // Show info bar with additional GPU parametres
 } nvtop_interface_option;
 
 inline bool plot_isset_draw_info(enum plot_information check_info, plot_info_to_draw to_draw) {

src/CMakeLists.txt

@@ -90,6 +90,7 @@ if(AMDGPU_SUPPORT OR MSM_SUPPORT)
       target_sources(nvtop PRIVATE extract_gpuinfo_msm.c)
       target_sources(nvtop PRIVATE extract_gpuinfo_msm_utils.c)
     endif()
+
   else()
     message(FATAL_ERROR "libdrm not found; This library is required for AMDGPU and MSM support")
   endif()
@@ -109,6 +110,20 @@ if(APPLE_SUPPORT)
     "-framework IOKit")
 endif()
 
+if (PANFROST_SUPPORT)
+  target_sources(nvtop PRIVATE extract_gpuinfo_panfrost.c)
+  target_sources(nvtop PRIVATE extract_gpuinfo_panfrost_utils.c)
+endif()
+
+if (PANTHOR_SUPPORT)
+  target_sources(nvtop PRIVATE extract_gpuinfo_panthor.c)
+  target_sources(nvtop PRIVATE extract_gpuinfo_panthor_utils.c)
+endif()
+
+if ((PANFROST_SUPPORT) OR (PANTHOR_SUPPORT))
+  target_sources(nvtop PRIVATE extract_gpuinfo_mali_common.c)
+endif()
+
 target_include_directories(nvtop PRIVATE
   ${PROJECT_SOURCE_DIR}/include
   ${PROJECT_BINARY_DIR}/include)

src/extract_gpuinfo.c

@@ -253,6 +253,17 @@ bool gpuinfo_refresh_processes(struct list_head *devices) {
   return true;
 }
 
+bool gpuinfo_utilisation_rate(struct list_head *devices) {
+  struct gpu_info *device;
+
+  list_for_each_entry(device, devices, list) {
+    if (device->vendor->refresh_utilisation_rate)
+      device->vendor->refresh_utilisation_rate(device);
+  }
+
+  return true;
+}
+
 void gpuinfo_clear_cache(void) {
   if (cached_process_info) {
     struct process_info_cache *pid_cached, *tmp;
@@ -311,3 +322,40 @@ unsigned nvtop_pcie_gen_from_link_speed(unsigned linkSpeed) {
   }
   return pcieGen;
 }
+
+void gpuinfo_refresh_utilisation_rate(struct gpu_info *gpu_info) {
+  /*
+   * kernel Documentation/gpu/drm-usage-stats.rst:
+   *  - drm-maxfreq-<keystr>: <uint> [Hz|MHz|KHz]
+   *  Engine identifier string must be the same as the one specified in the
+   *  drm-engine-<keystr> tag and shall contain the maximum frequency for the given
+   *  engine. Taken together with drm-cycles-<keystr>, this can be used to calculate
+   *  percentage utilization of the engine, whereas drm-engine-<keystr> only reflects
+   *  time active without considering what frequency the engine is operating as a
+   *  percentage of it's maximum frequency.
+   *
+   */
+
+  uint64_t gfx_total_process_cycles = 0;
+  uint64_t total_delta = 0;
+  unsigned int utilisation_rate;
+  uint64_t max_freq_hz;
+  double avg_delta_secs;
+
+  for (unsigned processIdx = 0; processIdx < gpu_info->processes_count; ++processIdx) {
+    struct gpu_process *process_info = &gpu_info->processes[processIdx];
+
+    gfx_total_process_cycles += process_info->gpu_cycles;
+    total_delta += process_info->sample_delta;
+  }
+
+  if (!gfx_total_process_cycles)
+    return;
+
+  avg_delta_secs = ((double)total_delta / gpu_info->processes_count) / 1000000000.0;
+  max_freq_hz = gpu_info->dynamic_info.gpu_clock_speed_max * 1000000;
+  utilisation_rate = (unsigned int)((((double)gfx_total_process_cycles) / (((double)max_freq_hz) * avg_delta_secs * 2)) * 100);
+  utilisation_rate = utilisation_rate > 100 ? 100 : utilisation_rate;
+
+  SET_GPUINFO_DYNAMIC(&gpu_info->dynamic_info, gpu_util_rate, utilisation_rate);
+}