Skip to contentSkip to conversation

Introduce in kernel benchmarks

RFC In review Christoph Lameter 2 participants 5 messages Oct 13, 2009 – Oct 15, 2009

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Introduce in kernel benchmarks

In kernel benchmarks are benchmarks that directly run in kernel space.

The test modules are regular kernel modules that can be loaded with

modprobe slab_test

or so.

The initialization function of the module will then run the test and fail so that the module is not loaded. The output of the test results will be written to the system log.

This is a rough draft. I have used these in the past to benchmark page allocator, slab allocator and vmstat performance.

#1 lore raw

RFC 1/3 Add slab performance tests

This is a test of slab allocator performance tests.

Run

modprobe slab_test

to execute. Output will be written to the syslog.

---
 Makefile          |    2 
 include/Kbuild    |    1 
 lib/Kconfig.debug |    1 
 tests/Kconfig     |   19 ++
 tests/Makefile    |    2 
 tests/slab_test.c |  373 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 397 insertions(+), 1 deletion(-)
 create mode 100644 tests/Makefile
 create mode 100644 tests/slab_test.c
Makefile +1 -1
Index: linux-2.6/Makefile
===================================================================
@@ -686,7 +686,7 @@ export mod_strip_cmd
 
 
 ifeq ($(KBUILD_EXTMOD),)
-core-y		+= kernel/ mm/ fs/ ipc/ security/ crypto/ block/
+core-y		+= kernel/ mm/ fs/ ipc/ security/ crypto/ block/ tests/
 
 vmlinux-dirs	:= $(patsubst %/,%,$(filter %/, $(init-y) $(init-m) \
 		     $(core-y) $(core-m) $(drivers-y) $(drivers-m) \
include/Kbuild +1 -0
Index: linux-2.6/include/Kbuild
===================================================================
@@ -10,3 +10,4 @@ header-y += video/
 header-y += drm/
 header-y += xen/
 header-y += scsi/
+header-y += tests/
tests/Makefile new file +2 -0
Index: linux-2.6/tests/Makefile
===================================================================
@@ -0,0 +1,2 @@
+obj-$(CONFIG_BENCHMARK_SLAB) += slab_test.o
+
tests/slab_test.c new file +373 -0

373 lines not on this page · raw message

lib/Kconfig.debug +1 -0
Index: linux-2.6/lib/Kconfig.debug
===================================================================
@@ -920,6 +920,7 @@ config SYSCTL_SYSCALL_CHECK
 
 source mm/Kconfig.debug
 source kernel/trace/Kconfig
+source tests/Kconfig
 
 config PROVIDE_OHCI1394_DMA_INIT
 	bool "Remote debugging over FireWire early on boot"
tests/Kconfig new file +19 -0
Index: linux-2.6/tests/Kconfig
===================================================================
@@ -0,0 +1,19 @@
+menuconfig BENCHMARKS
+	bool "In kernel benchmarks"
+	def_bool n
+	help
+	  Includes in kernel benchmark modules in the build. These modules can
+	  be loaded later to trigger benchmarking kernel subsystems.
+	  Output will be generated in the system log.
+
+if BENCHMARKS
+
+config BENCHMARK_SLAB
+	tristate "Slab allocator Benchmark"
+	depends on m
+	default m
+	help
+	  A benchmark that measures slab allocator performance.
+
+endif # BENCHMARKS
+
#2 lore raw

RFC 2/3 Vm statistics performance test

In reply to Christoph Lameter

Tests the performance of some key vm statistics functions.

---
 tests/Kconfig       |    7 +++
 tests/Makefile      |    1 
 tests/vmstat_test.c |   97 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 105 insertions(+)
 create mode 100644 tests/vmstat_test.c
tests/Makefile +1 -0
Index: linux-2.6/tests/Makefile
===================================================================
@@ -1,2 +1,3 @@
 obj-$(CONFIG_BENCHMARK_SLAB) += slab_test.o
+obj-#(CONFIG_BENCHMARK_VMSTAT) += vmstat_test.o
 
tests/vmstat_test.c new file +97 -0

97 lines not on this page · raw message

tests/Kconfig +7 -0
Index: linux-2.6/tests/Kconfig
===================================================================
@@ -15,5 +15,12 @@ config BENCHMARK_SLAB
 	help
 	  A benchmark that measures slab allocator performance.
 
+config BENCHMARK_VMSTAT
+	tristate "VM statistics Benchmark"
+	default m
+	depends on m
+	help
+	  A benchmark measuring the performance of vm statistics.
+
 endif # BENCHMARKS
 
#3 lore raw

RFC 3/3 Page allocator test

In reply to Christoph Lameter

Implement a test of the page allocator speed in the same way as for the slab allocators.

---
 tests/Kconfig          |    7 +
 tests/Makefile         |    3 
 tests/pagealloc_test.c |  334 +++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 343 insertions(+), 1 deletion(-)
 create mode 100644 tests/page_test.c
tests/Makefile +2 -1
Index: linux-2.6/tests/Makefile
===================================================================
@@ -1,3 +1,4 @@
 obj-$(CONFIG_BENCHMARK_SLAB) += slab_test.o
-obj-#(CONFIG_BENCHMARK_VMSTAT) += vmstat_test.o
+obj-$(CONFIG_BENCHMARK_VMSTAT) += vmstat_test.o
+obj-$(CONFIG_BENCHMARK_PAGEALLOC) += pagealloc_test.o
 
tests/pagealloc_test.c new file +334 -0

334 lines not on this page · raw message

tests/Kconfig +7 -0
Index: linux-2.6/tests/Kconfig
===================================================================
@@ -22,5 +22,12 @@ config BENCHMARK_VMSTAT
 	help
 	  A benchmark measuring the performance of vm statistics.
 
+config BENCHMARK_PAGEALLOC
+	tristate "Page Allocator Benchmark"
+	default m
+	depends on m
+	help
+	  A benchmark measuring the performance of the page allocator.
+
 endif # BENCHMARKS
 
#4 lore raw
In reply to Christoph Lameter
Quoted 6 lines · Christoph Lameter

In kernel benchmarks are benchmarks that directly run in kernel space.

The test modules are regular kernel modules that can be loaded with

modprobe slab_test

Quoted 9 lines

or so.

The initialization function of the module will then run the test and fail so that the module is not loaded. The output of the test results will be written to the system log.

This is a rough draft. I have used these in the past to benchmark page allocator, slab allocator and vmstat performance.

I like the addition of these benchmarks to the kernel source, I've been adding Mathieu's slab benchmark to my tree for a while. It looks like they all need some degree of cleanup, though: they have checkpatch warnings, they frequently do not handle memory allocation failures, and it would be nice to have a consistent loglevel for all the results.