linux-trace-devel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v3] kernel-shark: Multi-thread the computaion of stream/combo plots
@ 2025-03-27  0:21 Libo Chen
  2025-04-03 17:35 ` Yordan Karadzhov
  0 siblings, 1 reply; 2+ messages in thread
From: Libo Chen @ 2025-03-27  0:21 UTC (permalink / raw)
  To: y.karadz; +Cc: linux-trace-devel

Parallelize _newCPUGraph() and _newTaskGraph() calls to dramatically
speed up graph rendering particularly for traces from very large systems.

OpenMP technically is a new dependency here, but it's part of GCC, so long
as your GCC >= v4.9, the libgomp library will make the code compiled.

Signed-off-by: Libo Chen <libo.chen@oracle.com>
---
 CMakeLists.txt     |  6 ++++++
 src/KsGLWidget.cpp | 30 ++++++++++++++++++++++++++----
 2 files changed, 32 insertions(+), 4 deletions(-)

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 988bfd6..7847177 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -84,6 +84,12 @@ set(EXECUTABLE_OUTPUT_PATH "${KS_DIR}/bin")
 set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS} -Wall -Wextra -pthread -fPIC -fno-common")
 set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wall -Wextra -pthread -fPIC -fno-common")
 
+find_package(OpenMP 3.2.5)
+if (OPENMP_FOUND)
+    set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS}   ${OpenMP_C_FLAGS}")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} ${OpenMP_CXX_FLAGS}")
+endif(OPENMP_FOUND)
+
 set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
diff --git a/src/KsGLWidget.cpp b/src/KsGLWidget.cpp
index 9311d98..7f2001c 100644
--- a/src/KsGLWidget.cpp
+++ b/src/KsGLWidget.cpp
@@ -13,6 +13,9 @@
 #include <GL/glut.h>
 #include <GL/gl.h>
 
+// OpenMP
+#include <omp.h>
+
 // KernelShark
 #include "libkshark-plugin.h"
 #include "KsGLWidget.hpp"
@@ -54,6 +57,7 @@ KsGLWidget::KsGLWidget(QWidget *parent)
 
 	connect(&_model,	&QAbstractTableModel::modelReset,
 		this,		qOverload<>(&KsGLWidget::update));
+	omp_set_num_threads(omp_get_num_procs());
 }
 
 void KsGLWidget::_freeGraphs()
@@ -690,23 +694,41 @@ void KsGLWidget::_makeGraphs()
 
 	for (auto it = _streamPlots.begin(); it != _streamPlots.end(); ++it) {
 		sd = it.key();
+		size_t nCpus = it.value()._cpuList.count();
+		size_t nTasks = it.value()._taskList.count();
+		QVector<KsPlot::Graph *> cpuGraphs(nCpus);
+		QVector<KsPlot::Graph *> taskGraphs(nTasks);
+
 		/* Create CPU graphs according to the cpuList. */
 		it.value()._cpuGraphs = {};
-		for (auto const &cpu: it.value()._cpuList) {
-			g = lamAddGraph(sd, _newCPUGraph(sd, cpu), _vSpacing);
+		#pragma omp parallel for
+		for (size_t idx = 0; idx < nCpus; ++idx) {
+			int cpu = it.value()._cpuList[idx];
+			cpuGraphs[idx] = _newCPUGraph(sd, cpu);
+		}
+		QVectorIterator<KsPlot::Graph *> itCpuGraphs(cpuGraphs);
+		while (itCpuGraphs.hasNext()) {
+			g = lamAddGraph(sd, itCpuGraphs.next(), _vSpacing);
 			it.value()._cpuGraphs.append(g);
 		}
 
 		/* Create Task graphs according to the taskList. */
 		it.value()._taskGraphs = {};
-		for (auto const &pid: it.value()._taskList) {
-			g = lamAddGraph(sd, _newTaskGraph(sd, pid), _vSpacing);
+		#pragma omp parallel for
+		for (size_t idx = 0; idx < nTasks; ++idx) {
+			int pid = it.value()._taskList[idx];
+			taskGraphs[idx] = _newTaskGraph(sd, pid);
+		}
+		QVectorIterator<KsPlot::Graph *> itTaskGraphs(taskGraphs);
+		while (itTaskGraphs.hasNext()) {
+			g = lamAddGraph(sd, itTaskGraphs.next(), _vSpacing);
 			it.value()._taskGraphs.append(g);
 		}
 	}
 
 	for (auto &c: _comboPlots) {
 		int n = c.count();
+		#pragma omp parallel for
 		for (int i = 0; i < n; ++i) {
 			sd = c[i]._streamId;
 			if (c[i]._type & KSHARK_TASK_DRAW) {
-- 
2.43.5


^ permalink raw reply related	[flat|nested] 2+ messages in thread
end of thread, other threads:[~2025-04-03 17:35 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2025-03-27  0:21 [PATCH v3] kernel-shark: Multi-thread the computaion of stream/combo plots Libo Chen
2025-04-03 17:35 ` Yordan Karadzhov

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).