wip

src/06_optimization/clock/Makefile

@@ -0,0 +1,54 @@
+EXE=clock
+SRCS=$(wildcard *.c)
+
+ifeq ($(target),)
+target=nano
+endif
+
+CFLAGS=-Wall -Wextra -g -c -O0 -MD -std=gnu11 -D_GNU_SOURCE
+
+ifeq ($(target),nano)
+TOOLCHAIN_PATH=/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)aarch64-linux-
+CFLAGS+=-mcpu=cortex-a53 -funwind-tables
+##CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/nano
+EXEC=$(EXE)
+endif
+
+ifeq ($(target),host)
+EXEC=$(EXE)_h
+endif
+
+CC=$(TOOLCHAIN)gcc
+LD=$(TOOLCHAIN)gcc
+AR=$(TOOLCHAIN)ar
+STRIP=$(TOOLCHAIN)strip
+OBJDUMP=$(TOOLCHAIN)objdump
+
+OBJDIR=.obj/$(target)
+OBJS= $(addprefix $(OBJDIR)/, $(SRCS:.c=.o))
+
+$(OBJDIR)/%o: %c
+	$(CC) $(CFLAGS) $< -o $@
+
+all: $(OBJDIR)/ $(EXEC)
+
+$(EXEC): $(OBJS) $(LINKER_SCRIPT)
+	$(LD) $(OBJS) $(LDFLAGS) -o $@
+
+$(OBJDIR)/:
+	mkdir -p $(OBJDIR)
+
+clean:
+	rm -Rf $(OBJDIR) $(EXEC) $(EXEC)_s *~ t.txt
+
+clean_all: clean
+	rm -Rf .obj $(EXE) $(EXE)_s $(EXE)_a $(EXE)_a_s $(EXE)_h $(EXE)_h_s
+
+dump: all
+	$(OBJDUMP) -dS $(EXEC) > t.txt
+
+-include $(OBJS:.o=.d)
+
+.PHONY: all clean clean_all dump

src/06_optimization/clock/clock.c

@@ -0,0 +1,56 @@
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <time.h>
+#include <sched.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdbool.h>
+
+void measure (int mode, int samples)
+{
+	struct timespec start_time;
+	struct timespec stop_time;
+	clock_gettime (mode, &start_time);  // setup...
+	clock_gettime (mode, &start_time);
+	for (int i = 0; i<samples; i++)
+	{
+		clock_gettime (mode, &start_time);
+		clock_gettime (mode, &stop_time);
+		long long t = (stop_time.tv_nsec - start_time.tv_nsec) +
+		    (stop_time.tv_sec - start_time.tv_sec) * 1000000000;
+		printf ("%lld\n", t);
+	}
+
+}
+
+
+/**
+ * main program...
+ */
+int main(int argc, char* argv[])
+{
+	cpu_set_t my_set;
+	CPU_ZERO(&my_set);
+	CPU_SET(2, &my_set);
+	sched_setaffinity(0, sizeof(cpu_set_t), &my_set);
+
+	printf ("clocks_per_sec=%ld\n", CLOCKS_PER_SEC);
+
+	int samples = 1000;
+	if (argc == 2)
+		samples = atol(argv[1]);
+
+	struct timespec start_time;
+	clock_getres (CLOCK_MONOTONIC_RAW, &start_time);
+	long t = start_time.tv_sec * 1000000000 + start_time.tv_nsec;
+	printf ("time=%ld'%03ld'%03ld ns\n", t/1000000, (t/1000)%1000, t%1000);
+
+	measure (CLOCK_MONOTONIC_RAW, samples);
+
+	return 0;
+}

src/06_optimization/gcov/Makefile

@@ -0,0 +1,34 @@
+CC?=gcc
+CFLAGS=-std=c11 -Wall -g -fprofile-arcs -ftest-coverage
+SOURCES=$(wildcard *.c)
+OBJECTS=$(SOURCES:.c=.o)
+EXECUTABLE=example-gcov
+
+all: $(SOURCES) $(EXECUTABLE)
+
+$(EXECUTABLE): $(OBJECTS)
+	$(CC) $(CFLAGS) -o $@ $(OBJECTS)
+
+.c.o:
+	$(CC) -c $(CFLAGS) $< -o $@
+
+
+clean:
+	@rm -f $(OBJECTS)
+	@rm -f *.d
+
+clean_all: clean
+	@rm -f $(EXECUTABLE)
+	@rm -f perf.data perf.data.old
+	@rm -f *.gcno *.gcda *.gcov
+
+
+gcov-generate: all
+	./$(EXECUTABLE)
+	gcov main.c
+
+gcov-read: gcov-generate
+	less main.c.gcov
+
+
+-include *.d

src/06_optimization/gcov/main.c

@@ -0,0 +1,19 @@
+#include <stdio.h>
+
+int main (void)
+{
+    int res = 1;
+    for (int i=0; i<16; i++)
+    {
+        res *= 2;
+    }
+
+    if(res < 10)
+    {
+        printf("dead code\n");
+    }
+
+    printf("res=%d\n", res);
+
+    return 0;
+}

src/06_optimization/gpio/Makefile

@@ -0,0 +1,54 @@
+EXE=gpio
+SRCS=$(wildcard *.c)
+
+ifeq ($(target),)
+target=nano
+endif
+
+CFLAGS=-Wall -Wextra -g -c -O1 -MD -std=gnu11 -D_GNU_SOURCE
+
+ifeq ($(target),nano)
+TOOLCHAIN_PATH=/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)aarch64-linux-
+CFLAGS+=-mcpu=cortex-a53 -funwind-tables
+##CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/nano
+EXEC=$(EXE)
+endif
+
+ifeq ($(target),host)
+EXEC=$(EXE)_h
+endif
+
+CC=$(TOOLCHAIN)gcc
+LD=$(TOOLCHAIN)gcc
+AR=$(TOOLCHAIN)ar
+STRIP=$(TOOLCHAIN)strip
+OBJDUMP=$(TOOLCHAIN)objdump
+
+OBJDIR=.obj/$(target)
+OBJS= $(addprefix $(OBJDIR)/, $(SRCS:.c=.o))
+
+$(OBJDIR)/%o: %c
+	$(CC) $(CFLAGS) $< -o $@
+
+all: $(OBJDIR)/ $(EXEC)
+
+$(EXEC): $(OBJS) $(LINKER_SCRIPT)
+	$(LD) $(OBJS) $(LDFLAGS) -o $@
+
+$(OBJDIR)/:
+	mkdir -p $(OBJDIR)
+
+clean:
+	rm -Rf $(OBJDIR) $(EXEC) $(EXEC)_s *~ t.txt
+
+clean_all: clean
+	rm -Rf .obj $(EXE) $(EXE)_s $(EXE)_a $(EXE)_a_s $(EXE)_h $(EXE)_h_s
+
+dump: all
+	$(OBJDUMP) -dS $(EXEC) > t.txt
+
+-include $(OBJS:.o=.d)
+
+.PHONY: all clean clean_all dump

src/06_optimization/gpio/gpio.c

@@ -0,0 +1,87 @@
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <time.h>
+#include <sched.h>
+#include <sys/mman.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+#define GPIO_EXPORT		"/sys/class/gpio/export"
+#define GPIO_UNEXPORT	"/sys/class/gpio/unexport"
+#define GPIO_GPIOG11	"/sys/class/gpio/gpio203"
+#define GPIOG11			"203"
+
+static int open_gpio()
+{
+	// unexport pin out of sysfs (reinitialization)
+	int f = open (GPIO_UNEXPORT, O_WRONLY);
+	write (f, GPIOG11, strlen(GPIOG11));
+	close (f);
+
+	// export pin to sysfs
+	f = open (GPIO_EXPORT, O_WRONLY);
+	write (f, GPIOG11, strlen(GPIOG11));
+	close (f);	
+
+	// config pin
+	f = open (GPIO_GPIOG11 "/direction", O_WRONLY);
+	write (f, "out", 3);
+	close (f);
+
+	// open gpio value attribute
+ 	f = open (GPIO_GPIOG11 "/value", O_RDWR);
+	return f;
+}
+
+/**
+ * main program...
+ */
+int main()
+{
+	cpu_set_t my_set;
+	CPU_ZERO(&my_set);
+	CPU_SET(2, &my_set);
+	sched_setaffinity(0, sizeof(cpu_set_t), &my_set);
+
+	int gpio = open_gpio();
+	bool on = false;
+	struct timespec start_time;
+	struct timespec stop_time;
+	clock_gettime (CLOCK_MONOTONIC, &start_time);  // setup...
+
+
+	// --> measurement with internal timers
+	clock_gettime (CLOCK_MONOTONIC, &start_time);
+	for (int i=0; i<1000; i++) {
+		if (on) {
+			pwrite (gpio, "1", sizeof("1"), 0);
+		} else {
+			pwrite (gpio, "0", sizeof("0"), 0);
+		}
+		on = !on;
+	}
+	clock_gettime (CLOCK_MONOTONIC, &stop_time);
+	long long t = (stop_time.tv_nsec - start_time.tv_nsec) +
+		    (stop_time.tv_sec - start_time.tv_sec) * 1000000000;
+		printf ("pwrite (gpio, \"1\", sizeof(\"1\"), 0) -> %lld ns\n", t/1000);
+
+
+	// --> measurement with oscilloscope
+	while(true) {
+		if (on) {
+			pwrite (gpio, "1", sizeof("1"), 0);
+		} else {
+			pwrite (gpio, "0", sizeof("0"), 0);
+		}
+		on = !on;
+	}
+	
+	return 0;
+}

src/06_optimization/gprof/Makefile

@@ -0,0 +1,34 @@
+CC?=gcc
+CFLAGS=-std=c11 -Wall -pg
+SOURCES=$(wildcard *.c)
+OBJECTS=$(SOURCES:.c=.o)
+EXECUTABLE=example-gprof
+
+all: $(SOURCES) $(EXECUTABLE)
+
+$(EXECUTABLE): $(OBJECTS)
+	$(CC) $(CFLAGS) -o $@ $(OBJECTS)
+
+.c.o:
+	$(CC) -c $(CFLAGS) $< -o $@
+
+
+clean:
+	@rm -f $(OBJECTS)
+	@rm -f *.d
+
+clean_all: clean
+	@rm -f $(EXECUTABLE)
+	@rm -f perf.data perf.data.old
+	@rm -f gmon.out gprof.out
+
+
+gprof-generate: all
+	./$(EXECUTABLE)
+	gprof $(EXECUTABLE) > gprof.out
+
+gprof-read:
+	less gprof.out
+
+
+-include *.d

src/06_optimization/gprof/main.c

@@ -0,0 +1,21 @@
+void func1(void)
+{
+    for(int i=0; i<0xfffffff; i++); // wait...
+    return;
+}
+
+void func2(void)
+{
+    for(int i=0;i<0xfffffff;i++); // wait...
+    func1();
+    return;
+}
+
+int main(void)
+{
+    for(int i=0;i<0xfffffff;i++);
+    func1();
+    func2();
+
+    return 0;
+}

src/06_optimization/gprof/target.mk

@@ -0,0 +1,71 @@
+EXE=example-gprof
+SRCS=$(wildcard *.c) 
+
+ifeq ($(target),)
+target=nano
+endif
+
+CFLAGS=-Wall -Wextra -g -c -O1 -MD -std=gnu11 -D_GNU_SOURCE -pg
+
+ifeq ($(target),nano)
+TOOLCHAIN_PATH=/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)aarch64-linux-
+CFLAGS+=-mcpu=cortex-a53 -funwind-tables
+CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/nano
+EXEC=$(EXE)
+endif
+
+ifeq ($(target),host)
+EXEC=$(EXE)_h
+endif
+
+ifeq ($(target),xu3)
+TOOLCHAIN_PATH=~/workspace/xu3/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)arm-linux-gnueabihf-
+CFLAGS+=-mcpu=cortex-a15.cortex-a7 -funwind-tables
+##CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/xu3
+EXEC=$(EXE)_a
+endif
+
+CC=$(TOOLCHAIN)gcc
+LD=$(TOOLCHAIN)gcc
+AR=$(TOOLCHAIN)ar
+STRIP=$(TOOLCHAIN)strip
+OBJDUMP=$(TOOLCHAIN)objdump
+
+OBJDIR=.obj/$(target)
+OBJS= $(addprefix $(OBJDIR)/, $(SRCS:.c=.o))
+
+$(OBJDIR)/%o: %c
+	$(CC) $(CFLAGS) $< -o $@
+	
+all: $(OBJDIR)/ $(EXEC)
+	
+$(EXEC): $(OBJS) $(LINKER_SCRIPT)
+	$(LD) $(OBJS) $(LDFLAGS) -o $@
+
+$(OBJDIR)/:
+	mkdir -p $(OBJDIR)
+
+clean:
+	rm -Rf $(OBJDIR) $(EXEC) $(EXEC)_s *~ t.txt
+
+clean_all: clean
+	rm -Rf .obj $(EXE) $(EXE)_s $(EXE)_a $(EXE)_a_s $(EXE)_h $(EXE)_h_s
+	@rm -f perf.data perf.data.old gmon.out gprof.out
+
+dump: all
+	$(OBJDUMP) -dS $(EXEC) > t.txt
+
+gprof-generate: all
+	./$(EXECUTABLE)
+	gprof $(EXECUTABLE) > gprof.out
+
+gprof-read:
+	less gprof.out
+
+-include $(OBJS:.o=.d)
+
+.PHONY: all clean clean_all dump

src/06_optimization/mmio/Makefile

@@ -0,0 +1,54 @@
+EXE=mmio
+SRCS=$(wildcard *.c)
+
+ifeq ($(target),)
+target=nano
+endif
+
+CFLAGS=-Wall -Wextra -g -c -O1 -MD -std=gnu11 -D_GNU_SOURCE
+
+ifeq ($(target),nano)
+TOOLCHAIN_PATH=/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)aarch64-linux-
+CFLAGS+=-mcpu=cortex-a53 -funwind-tables
+##CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/nano
+EXEC=$(EXE)
+endif
+
+ifeq ($(target),host)
+EXEC=$(EXE)_h
+endif
+
+CC=$(TOOLCHAIN)gcc
+LD=$(TOOLCHAIN)gcc
+AR=$(TOOLCHAIN)ar
+STRIP=$(TOOLCHAIN)strip
+OBJDUMP=$(TOOLCHAIN)objdump
+
+OBJDIR=.obj/$(target)
+OBJS= $(addprefix $(OBJDIR)/, $(SRCS:.c=.o))
+
+$(OBJDIR)/%o: %c
+	$(CC) $(CFLAGS) $< -o $@
+
+all: $(OBJDIR)/ $(EXEC)
+
+$(EXEC): $(OBJS) $(LINKER_SCRIPT)
+	$(LD) $(OBJS) $(LDFLAGS) -o $@
+
+$(OBJDIR)/:
+	mkdir -p $(OBJDIR)
+
+clean:
+	rm -Rf $(OBJDIR) $(EXEC) $(EXEC)_s *~ t.txt
+
+clean_all: clean
+	rm -Rf .obj $(EXE) $(EXE)_s $(EXE)_a $(EXE)_a_s $(EXE)_h $(EXE)_h_s
+
+dump: all
+	$(OBJDUMP) -dS $(EXEC) > t.txt
+
+-include $(OBJS:.o=.d)
+
+.PHONY: all clean clean_all dump

src/06_optimization/mmio/mmio.c

@@ -0,0 +1,98 @@
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <fcntl.h>
+#include <unistd.h>
+#include <errno.h>
+#include <time.h>
+#include <sched.h>
+#include <sys/mman.h>
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <stdbool.h>
+#include <stdint.h>
+
+#define GPIO_EXPORT		"/sys/class/gpio/export"
+#define GPIO_UNEXPORT	"/sys/class/gpio/unexport"
+#define GPIO_GPIOG11	"/sys/class/gpio/gpio203"
+#define GPIOG11			"203"
+
+static int open_gpio()
+{
+	// unexport pin out of sysfs (reinitialization)
+	int f = open (GPIO_UNEXPORT, O_WRONLY);
+	write (f, GPIOG11, strlen(GPIOG11));
+	close (f);
+
+	// export pin to sysfs
+	f = open (GPIO_EXPORT, O_WRONLY);
+	write (f, GPIOG11, strlen(GPIOG11));
+	close (f);	
+
+	// config pin
+	f = open (GPIO_GPIOG11 "/direction", O_WRONLY);
+	write (f, "out", 3);
+	close (f);
+
+	// open gpio value attribute
+ 	f = open (GPIO_GPIOG11 "/value", O_RDWR);
+	return f;
+}
+
+/**
+ * main program...
+ */
+int main()
+{
+	cpu_set_t my_set;
+	CPU_ZERO(&my_set);
+	CPU_SET(2, &my_set);
+	sched_setaffinity(0, sizeof(cpu_set_t), &my_set);
+
+	int gpio = open_gpio();
+	pwrite (gpio, "1", sizeof("1"), 0);
+	pwrite (gpio, "0", sizeof("0"), 0);
+
+	int fd = open ("/dev/mem", O_RDWR);
+	if (fd == -1) return -1;
+	off_t psz = getpagesize();
+	volatile uint32_t*  pio = mmap (0, psz, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0x01c20000);
+	volatile uint32_t* dat_pio  =  pio + (0x810+6*0x24)/4;
+
+	struct timespec start_time;
+	struct timespec stop_time;
+	clock_gettime (CLOCK_MONOTONIC, &start_time);  // setup...
+
+	// --> measurement with internal timers on gpio 
+	clock_gettime (CLOCK_MONOTONIC, &start_time);
+	
+	for (int i=0; i<1000; i++)
+		*dat_pio ^= (1<<11);
+
+	clock_gettime (CLOCK_MONOTONIC, &stop_time);
+	long long t = (stop_time.tv_nsec - start_time.tv_nsec) +
+		    (stop_time.tv_sec - start_time.tv_sec) * 1000000000;
+	printf ("*dat_pio ^= (1<<11) --> %6lld ns\n", t/1000);
+
+
+	// --> measurement on internal data with internal timers
+	volatile uint32_t data = 0;
+	clock_gettime (CLOCK_MONOTONIC, &start_time);
+	
+	for (int i=0; i<1000; i++)
+		data ^= (1<<11);
+	
+	clock_gettime (CLOCK_MONOTONIC, &stop_time);
+	t = (stop_time.tv_nsec - start_time.tv_nsec) +
+		    (stop_time.tv_sec - start_time.tv_sec) * 1000000000;
+	printf ("data     ^= (1<<11) --> %6lld ns\n", t/1000);
+
+
+	// --> measurement with oscilloscope
+	while(true) {
+		*dat_pio ^= (1<<11);
+	}
+	
+	return 0;
+}

src/06_optimization/trace/Makefile

@@ -0,0 +1,54 @@
+EXE=tracing
+SRCS=$(wildcard *.c)
+
+ifeq ($(target),)
+target=nano
+endif
+
+CFLAGS=-Wall -Wextra -g -c -O1 -MD -std=gnu11 -D_GNU_SOURCE
+
+ifeq ($(target),nano)
+TOOLCHAIN_PATH=/buildroot/output/host/usr/bin/
+TOOLCHAIN=$(TOOLCHAIN_PATH)aarch64-linux-
+CFLAGS+=-mcpu=cortex-a53 -funwind-tables
+##CFLAGS+=-O2 -fno-omit-frame-pointer
+OBJDIR=.obj/nano
+EXEC=$(EXE)
+endif
+
+ifeq ($(target),host)
+EXEC=$(EXE)_h
+endif
+
+CC=$(TOOLCHAIN)gcc
+LD=$(TOOLCHAIN)gcc
+AR=$(TOOLCHAIN)ar
+STRIP=$(TOOLCHAIN)strip
+OBJDUMP=$(TOOLCHAIN)objdump
+
+OBJDIR=.obj/$(target)
+OBJS= $(addprefix $(OBJDIR)/, $(SRCS:.c=.o))
+
+$(OBJDIR)/%o: %c
+	$(CC) $(CFLAGS) $< -o $@
+
+all: $(OBJDIR)/ $(EXEC)
+
+$(EXEC): $(OBJS) $(LINKER_SCRIPT)
+	$(LD) $(OBJS) $(LDFLAGS) -o $@
+
+$(OBJDIR)/:
+	mkdir -p $(OBJDIR)
+
+clean:
+	rm -Rf $(OBJDIR) $(EXEC) $(EXEC)_s *~ t.txt
+
+clean_all: clean
+	rm -Rf .obj $(EXE) $(EXE)_s $(EXE)_a $(EXE)_a_s $(EXE)_h $(EXE)_h_s
+
+dump: all
+	$(OBJDUMP) -dS $(EXEC) > t.txt
+
+-include $(OBJS:.o=.d)
+
+.PHONY: all clean clean_all dump

src/06_optimization/trace/example1.sh

@@ -0,0 +1,7 @@
+#!/bin/sh
+# 1. example
+echo 1 > /sys/kernel/debug/tracing/events/sched/sched_switch/enable
+./tracing  &
+pidof tracing > /sys/kernel/debug/tracing/set_event_pid
+echo 1 > /sys/kernel/debug/tracing/tracing_on ; sleep 2 ; echo 0 > /sys/kernel/debug/tracing/tracing_on
+trace-cmd show

src/06_optimization/trace/example2.sh

@@ -0,0 +1,10 @@
+#!/bin/sh
+# 2. example
+echo function_graph > /sys/kernel/debug/tracing/current_tracer
+echo *i2c* >  /sys/kernel/debug/tracing/set_ftrace_filter
+echo 1 >  /sys/kernel/debug/tracing/tracing_on
+i2cdetect -y 0
+echo 0 >  /sys/kernel/debug/tracing/tracing_on
+trace-cmd show
+echo nop > /sys/kernel/debug/tracing/current_tracer
+

src/06_optimization/trace/main.c

@@ -0,0 +1,8 @@
+#include <unistd.h>
+
+int main()
+{
+    while(1)
+        sleep(1);
+    return 0;
+}

src/06_optimization/trace/setup.sh

@@ -0,0 +1 @@
+mount -t debugfs nodev /sys/kernel/debug