Merge pull request #903 from bjia56/intel

Intel GPU support
aristocratos · Sep 21, 2024 · 24fc97c · 24fc97c
2 parents 00c9088 + db5f895
commit 24fc97c
Show file tree

Hide file tree

Showing 19 changed files with 9,797 additions and 9 deletions.
diff --git a/.github/workflows/cmake-linux.yml b/.github/workflows/cmake-linux.yml
@@ -30,7 +30,7 @@ jobs:
       - uses: actions/checkout@v4
 
       - name: Install build tools
-        run: apk add --no-cache --update clang cmake lld ninja lowdown
+        run: apk add --no-cache --update clang cmake lld ninja lowdown linux-headers
 
       - name: Configure
         run: CXX=clang++ LDFLAGS=-fuse-ld=lld cmake -B build -G Ninja -DBTOP_STATIC=ON

diff --git a/.github/workflows/continuous-build-gpu.yml b/.github/workflows/continuous-build-gpu.yml
@@ -40,7 +40,7 @@ jobs:
       - uses: actions/checkout@v4
 
       - name: Install build tools
-        run: apk add --no-cache --update gcc g++ make
+        run: apk add --no-cache --update gcc g++ make linux-headers
 
       - name: Compile
         run: make CXX=g++ GPU_SUPPORT=true

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -13,7 +13,7 @@ endif()
 project("btop"
   DESCRIPTION "A monitor of resources"
   HOMEPAGE_URL "https://github.com/aristocratos/btop"
-  LANGUAGES CXX
+  LANGUAGES CXX C
 )
 
 include(CheckCXXCompilerFlag)
@@ -70,6 +70,14 @@ elseif(CMAKE_SYSTEM_NAME STREQUAL "NetBSD")
   target_sources(btop PRIVATE src/netbsd/btop_collect.cpp)
 elseif(LINUX)
   target_sources(btop PRIVATE src/linux/btop_collect.cpp)
+  if(BTOP_GPU)
+    target_sources(btop PRIVATE
+      src/linux/intel_gpu_top/intel_gpu_top.c
+      src/linux/intel_gpu_top/igt_perf.c
+      src/linux/intel_gpu_top/intel_device_info.c
+      src/linux/intel_gpu_top/intel_name_lookup_shim.c
+    )
+  endif()
 else()
   message(FATAL_ERROR "${CMAKE_SYSTEM_NAME} is not supported")
 endif()

diff --git a/Makefile b/Makefile
@@ -40,6 +40,7 @@ override PLATFORM_LC := $(shell echo $(PLATFORM) | tr '[:upper:]' '[:lower:]')
 ifeq ($(PLATFORM_LC)$(ARCH),linuxx86_64)
 	ifneq ($(STATIC),true)
 		GPU_SUPPORT := true
+		INTEL_GPU_SUPPORT := true
 	endif
 endif
 ifneq ($(GPU_SUPPORT),true)
@@ -208,13 +209,22 @@ endif
 
 SOURCES	:= $(sort $(shell find $(SRCDIR) -maxdepth 1 -type f -name *.$(SRCEXT)))
 
-SOURCES += $(sort $(shell find $(SRCDIR)/$(PLATFORM_DIR) -type f -name *.$(SRCEXT)))
-
-#? Setup percentage progress
-SOURCE_COUNT := $(words $(SOURCES))
+SOURCES += $(sort $(shell find $(SRCDIR)/$(PLATFORM_DIR) -maxdepth 1 -type f -name *.$(SRCEXT)))
 
 OBJECTS	:= $(patsubst $(SRCDIR)/%,$(BUILDDIR)/%,$(SOURCES:.$(SRCEXT)=.$(OBJEXT)))
 
+ifeq ($(GPU_SUPPORT)$(INTEL_GPU_SUPPORT),truetrue)
+	IGT_OBJECTS := $(BUILDDIR)/igt_perf.c.o $(BUILDDIR)/intel_device_info.c.o $(BUILDDIR)/intel_name_lookup_shim.c.o $(BUILDDIR)/intel_gpu_top.c.o
+	OBJECTS += $(IGT_OBJECTS)
+	SHOW_CC_INFO = false
+	CC_VERSION := $(shell $(CC) -dumpfullversion -dumpversion || echo 0)
+else
+	SHOW_CC_INFO = true
+endif
+
+#? Setup percentage progress
+SOURCE_COUNT := $(words $(OBJECTS))
+
 ifeq ($(shell find $(BUILDDIR) -type f -newermt "$(DATESTAMP)" -name *.o >/dev/null 2>&1; echo $$?),0)
 	ifneq ($(wildcard $(BUILDDIR)/.*),)
 		SKIPPED_SOURCES := $(foreach fname,$(SOURCES),$(shell find $(BUILDDIR) -type f -newer $(fname) -name *.o | grep "$(basename $(notdir $(fname))).o" 2>/dev/null))
@@ -248,7 +258,8 @@ info:
 	@printf "\033[1;96mARCH         \033[1;93m?| \033[0m$(ARCH)\n"
 	@printf "\033[1;95mGPU_SUPPORT  \033[1;94m:| \033[0m$(GPU_SUPPORT)\n"
 	@printf "\033[1;93mCXX          \033[1;93m?| \033[0m$(CXX) \033[1;93m(\033[97m$(CXX_VERSION)\033[93m)\n"
-	@printf "\033[1;94mTHREADS      \033[1;94m:| \033[0m$(THREADS)\n"
+	@$(SHOW_CC_INFO) || printf "\033[1;93mCC           \033[1;93m?| \033[0m$(CC) \033[1;93m(\033[97m$(CC_VERSION)\033[93m)\n"
+	@printf "\033[1;94mTHREADS      \033[1;94m:| \033[0m$(THREADS)\n" gcc -dumpfullversion -dumpversion
 	@printf "\033[1;92mREQFLAGS     \033[1;91m!| \033[0m$(REQFLAGS)\n"
 	@printf "\033[1;91mWARNFLAGS    \033[1;94m:| \033[0m$(WARNFLAGS)\n"
 	@printf "\033[1;94mOPTFLAGS     \033[1;94m:| \033[0m$(OPTFLAGS)\n"
@@ -409,5 +420,16 @@ $(BUILDDIR)/%.$(OBJEXT): $(SRCDIR)/%.$(SRCEXT) | rocm_smi directories config.h
 	@$(CXX) $(CXXFLAGS) $(INC) -MMD -c -o $@ $< || exit 1
 	@printf "\033[1;92m$$($(PROGRESS))$(P)\033[10D\033[5C-> \033[1;37m$@ \033[100D\033[38C\033[1;93m(\033[1;97m$$(du -ah $@ | cut -f1)iB\033[1;93m) \033[92m(\033[97m$$($(DATE_CMD) -d @$$(expr $$($(DATE_CMD) +%s 2>/dev/null || echo "0") - $${TSTAMP} 2>/dev/null) -u +%Mm:%Ss 2>/dev/null | sed 's/^00m://' || echo '')\033[92m)\033[0m\n"
 
+#? Compile intel_gpu_top C sources for Intel GPU support
+.ONESHELL:
+$(BUILDDIR)/%.c.o: $(SRCDIR)/$(PLATFORM_DIR)/intel_gpu_top/%.c | directories
+	@sleep 0.3 2>/dev/null || true
+	@TSTAMP=$$(date +%s 2>/dev/null || echo "0")
+	@$(QUIET) || printf "\033[1;97mCompiling $<\033[0m\n"
+	@$(VERBOSE) || printf "$(CC) $(INC) -c -o $@ $<\n"
+	@$(CC) $(INC) -w -c -o $@ $< || exit 1
+	@printf "\033[1;92m$$($(PROGRESS))$(P)\033[10D\033[5C-> \033[1;37m$@ \033[100D\033[38C\033[1;93m(\033[1;97m$$(du -ah $@ | cut -f1)iB\033[1;93m) \033[92m(\033[97m$$($(DATE_CMD) -d @$$(expr $$($(DATE_CMD) +%s 2>/dev/null || echo "0") - $${TSTAMP} 2>/dev/null) -u +%Mm:%Ss 2>/dev/null | sed 's/^00m://' || echo '')\033[92m)\033[0m\n"
+
+
 #? Non-File Targets
 .PHONY: all config.h msg help pre
diff --git a/src/linux/btop_collect.cpp b/src/linux/btop_collect.cpp
@@ -47,6 +47,14 @@ tab-size = 4
 #include "../btop_config.hpp"
 #include "../btop_tools.hpp"
 
+#if defined(GPU_SUPPORT)
+	#define class class_
+extern "C" {
+	#include "./intel_gpu_top/intel_gpu_top.h"
+}
+	#undef class
+#endif
+
 using std::clamp;
 using std::cmp_greater;
 using std::cmp_less;
@@ -210,6 +218,19 @@ namespace Gpu {
 		template <bool is_init> bool collect(gpu_info* gpus_slice);
 		uint32_t device_count = 0;
 	}
+
+
+	//? Intel data collection
+	namespace Intel {
+		const char* device = "i915";
+		struct engines *engines = nullptr;
+
+		bool initialized = false;
+		bool init();
+		bool shutdown();
+		template <bool is_init> bool collect(gpu_info* gpus_slice);
+		uint32_t device_count = 0;
+	}
 #endif
 }
 
@@ -276,6 +297,7 @@ namespace Shared {
 	#ifdef GPU_SUPPORT
 		Gpu::Nvml::init();
 		Gpu::Rsmi::init();
+		Gpu::Intel::init();
 		if (not Gpu::gpu_names.empty()) {
 			for (auto const& [key, _] : Gpu::gpus[0].gpu_percent)
 				Cpu::available_fields.push_back(key);
@@ -1570,7 +1592,117 @@ namespace Gpu {
 		}
 	}
 
-	// TODO: Intel
+	namespace Intel {
+		bool init() {
+			if (initialized) return false;
+
+			char *gpu_path = find_intel_gpu_dir();
+			if (!gpu_path) {
+				Logger::debug("Failed to find Intel GPU sysfs path, Intel GPUs will not be detected");
+				return false;
+			}
+
+			char *gpu_device_id = get_intel_device_id(gpu_path);
+			if (!gpu_device_id) {
+				Logger::debug("Failed to find Intel GPU device ID, Intel GPUs will not be detected");
+				return false;
+			}
+
+			char *gpu_device_name = get_intel_device_name(gpu_device_id);
+			if (!gpu_device_name) {
+				Logger::warning("Failed to find Intel GPU device name in internal database");
+			}
+
+			free(gpu_device_id);
+
+			engines = discover_engines(device);
+			if (!engines) {
+				Logger::debug("Failed to find Intel GPU engines, Intel GPUs will not be detected");
+				return false;
+			}
+
+			int ret = pmu_init(engines);
+			if (ret) {
+				Logger::warning("Intel GPU: Failed to initialize PMU");
+				return false;
+			}
+
+			pmu_sample(engines);
+
+			device_count = 1;
+
+			gpus.resize(gpus.size() + device_count);
+			gpu_names.resize(gpus.size() + device_count);
+
+			if (gpu_device_name) {
+				gpu_names[Nvml::device_count + Rsmi::device_count] = string(gpu_device_name);
+			} else {
+				gpu_names[Nvml::device_count + Rsmi::device_count] = "Intel GPU";
+			}
+
+			free(gpu_device_name);
+
+			initialized = true;
+			Intel::collect<1>(gpus.data() + Nvml::device_count + Rsmi::device_count);
+
+			return true;
+		}
+
+		bool shutdown() {
+			if (!initialized) return false;
+			if (engines) {
+				free_engines(engines);
+				engines = nullptr;
+			}
+			initialized = false;
+			return true;
+		}
+
+		template <bool is_init> bool collect(gpu_info* gpus_slice) {
+			if (!initialized) return false;
+
+			if constexpr(is_init) {
+				gpus_slice->supported_functions = {
+					.gpu_utilization = true,
+					.mem_utilization = false,
+					.gpu_clock = true,
+					.mem_clock = false,
+					.pwr_usage = true,
+					.pwr_state = false,
+					.temp_info = false,
+					.mem_total = false,
+					.mem_used = false,
+					.pcie_txrx = false
+				};
+			}
+
+			pmu_sample(engines);
+			double t = (double)(engines->ts.cur - engines->ts.prev) / 1e9;
+
+			double max_util = 0;
+			for (unsigned int i = 0; i < engines->num_engines; i++) {
+				struct engine *engine = &(&engines->engine)[i];
+				double util = pmu_calc(&engine->busy.val, 1e9, t, 100);
+				if (util > max_util) {
+					max_util = util;
+				}
+			}
+			gpus_slice->gpu_percent.at("gpu-totals").push_back((long long)round(max_util));
+
+			double pwr = pmu_calc(&engines->r_gpu.val, 1, t, engines->r_gpu.scale); // in Watts
+			gpus_slice->pwr_usage = (long long)round(pwr * 1000);
+			if (gpus_slice->pwr_usage > 0) {
+				gpus_slice->gpu_percent.at("gpu-pwr-totals").push_back(100);
+			} else {
+				gpus_slice->gpu_percent.at("gpu-pwr-totals").push_back(0);
+			}
+
+			double freq = pmu_calc(&engines->freq_act.val, 1, t, 1); // in MHz
+			gpus_slice->gpu_clock_speed = (unsigned int)round(freq);
+
+			return true;
+		}
+	}
 
 	//? Collect data from GPU-specific libraries
 	auto collect(bool no_update) -> vector<gpu_info>& {
@@ -1581,6 +1713,7 @@ namespace Gpu {
 		//* Collect data
 		Nvml::collect<0>(gpus.data()); // raw pointer to vector data, size == Nvml::device_count
 		Rsmi::collect<0>(gpus.data() + Nvml::device_count); // size = Rsmi::device_count
+		Intel::collect<0>(gpus.data() + Nvml::device_count + Rsmi::device_count); // size = Intel::device_count
 
 		//* Calculate average usage
 		long long avg = 0;