personal/time-to-botec
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

reorganize C code

Browse Source
This commit is contained in:
NunoSempere 2023-06-02 16:00:49 -06:00
parent ef04e0349a
commit ff3685766b
17 changed files with 651 additions and 615 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
C-optimized/README.md
View File

@ -1,32 +0,0 @@
# C-Optimized
An optimized version of the original C implementation.
The main changes are:
- an optimization of the mixture function (it passes the functions instead of the whole arrays, reducing in great measure the memory usage and the computation time) and
- the implementation of multi-threading with OpenMP.
## Performance
The mean time of execution is 6 ms. With the following distribution:
![Time histogram](https://i.imgur.com/6iT2PkF.png)
The hardware used has been an AMD 5800x3D and 16GB of DDR4-3200 MHz.
Also, the time data has been collected by executing the interior of the main() function 1000 times in a for loop, not executing the program itself 1000 times.
## Multithreading
Take into account that the multi-threading introduces a bit of dispersion in the execution time due to the creation and destruction of threads.
In Nuño's machine, multithreading actually introduces a noticeable slowdown factor.
## To do
- [ ] Use proper profiling tool to capture timing with 1M samples.
- [ ] Update above with correct timing
- [ ] Add Windows/Powershell time-measuring commands
- [ ] Add CUDA?
- [ ] See if program can be reworded so as to use multithreading effectively, e.g., so that you see speed gains proportional to the number of threads used

82
C-optimized/makefile
View File

@ -1,82 +0,0 @@
# Interface:
# make
# make build
# make format
# make run
# Compiler
CC=gcc
# CC=tcc # <= faster compilation
# Main file
SRC=samples.c
OUTPUT=out/samples
SRC_ONE_THREAD=./samples-one-thread.c
OUTPUT_ONE_THREAD=out/samples-one-thread
## Dependencies
# Has no dependencies
MATH=-lm
## Flags
DEBUG= #'-g'
STANDARD=-std=c99
WARNINGS=-Wall
OPTIMIZED=-O3 #-O3 actually gives better performance than -Ofast, at least for this version
OPENMP=-fopenmp
## Formatter
STYLE_BLUEPRINT=webkit
FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT)
## make build
build: $(SRC)
$(CC) $(OPTIMIZED) $(DEBUG) $(SRC) $(OPENMP) $(MATH) -o $(OUTPUT)
$(CC) $(OPTIMIZED) $(DEBUG) $(SRC_ONE_THREAD) $(OPENMP) $(MATH) -o $(OUTPUT_ONE_THREAD)
format: $(SRC)
$(FORMATTER) $(SRC)
run: $(SRC) $(OUTPUT)
OMP_NUM_THREADS=1 ./$(OUTPUT) && echo
./$(OUTPUT_ONE_THREAD)
multi:
OMP_NUM_THREADS=1 ./$(OUTPUT) && echo
OMP_NUM_THREADS=2 ./$(OUTPUT) && echo
OMP_NUM_THREADS=4 ./$(OUTPUT) && echo
OMP_NUM_THREADS=8 ./$(OUTPUT) && echo
OMP_NUM_THREADS=16 ./$(OUTPUT) && echo
./$(OUTPUT_ONE_THREAD) && echo
time-linux-simple:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
OMP_NUM_THREADS=1 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=2 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=4 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=8 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=16 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
/bin/time -f "Time: %es" ./$(OUTPUT_ONE_THREAD) && echo
time-linux:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=1 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=1 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 1 thread: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=2 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=2 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 2 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=4 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=4 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time for 4 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=8 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=8 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 8 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=16 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=16 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 16 threads: |" | sed 's|$$|ms|' && echo
debian-install-dependencies:
sudo apt-get install libomp-dev

281
C-optimized/samples.c
View File

@ -1,281 +0,0 @@
#include <math.h>
#include <omp.h>
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
const float PI = 3.14159265358979323846;
#define N 1000000
//Array helpers
void array_print(float* array, int length)
{
for (int i = 0; i < length; i++) {
printf("item[%d] = %f\n", i, array[i]);
}
printf("\n");
}
void array_fill(float* array, int length, float item)
{
int i;
#pragma omp private(i)
{
#pragma omp for
for (i = 0; i < length; i++) {
array[i] = item;
}
}
}
float array_sum(float* array, int length)
{
float output = 0.0;
for (int i = 0; i < length; i++) {
output += array[i];
}
return output;
}
void array_cumsum(float* array_to_sum, float* array_cumsummed, int length)
{
array_cumsummed[0] = array_to_sum[0];
for (int i = 1; i < length; i++) {
array_cumsummed[i] = array_cumsummed[i - 1] + array_to_sum[i];
}
}
float rand_float(float to, unsigned int* seed)
{
return ((float)rand_r(seed) / (float)RAND_MAX) * to;
// See: <https://stackoverflow.com/questions/43151361/how-to-create-thread-safe-random-number-generator-in-c-using-rand-r>
// rand() is not thread-safe, as it relies on (shared) hidden state.
}
float ur_normal(unsigned int* seed)
{
float u1 = rand_float(1.0, seed);
float u2 = rand_float(1.0, seed);
float z = sqrtf(-2.0 * log(u1)) * sin(2 * PI * u2);
return z;
}
inline float random_uniform(float from, float to, unsigned int* seed)
{
return ((float) rand_r(seed) / (float)RAND_MAX) * (to - from) + from;
}
inline float random_normal(float mean, float sigma, unsigned int* seed)
{
return (mean + sigma * ur_normal(seed));
}
inline float random_lognormal(float logmean, float logsigma, unsigned int* seed)
{
return expf(random_normal(logmean, logsigma, seed));
}
inline float random_to(float low, float high, unsigned int* seed)
{
const float NORMAL95CONFIDENCE = 1.6448536269514722;
float loglow = logf(low);
float loghigh = logf(high);
float logmean = (loglow + loghigh) / 2;
float logsigma = (loghigh - loglow) / (2.0 * NORMAL95CONFIDENCE);
return random_lognormal(logmean, logsigma, seed);
}
int split_array_get_my_length(int index, int total_length, int n_threads)
{
return (total_length % n_threads > index ? total_length / n_threads + 1 : total_length / n_threads);
}
//Old version, don't use it!! Optimized version is called mixture_f. This one is just for display
/*
void mixture(float* dists[], float* weights, int n_dists, float* results)
{
float sum_weights = array_sum(weights, n_dists);
float* normalized_weights = malloc(n_dists * sizeof(float));
for (int i = 0; i < n_dists; i++) {
normalized_weights[i] = weights[i] / sum_weights;
}
float* cummulative_weights = malloc(n_dists * sizeof(float));
array_cumsum(normalized_weights, cummulative_weights, n_dists);
//create var holders
float p1, p2;
int index_found, index_counter, sample_index, i;
#pragma omp parallel private(i, p1, p2, index_found, index_counter, sample_index)
{
#pragma omp for
for (i = 0; i < N; i++) {
p1 = random_uniform(0, 1);
p2 = random_uniform(0, 1);
index_found = 0;
index_counter = 0;
while ((index_found == 0) && (index_counter < n_dists)) {
if (p1 < cummulative_weights[index_counter]) {
index_found = 1;
} else {
index_counter++;
}
}
if (index_found != 0) {
sample_index = (int)(p2 * N);
results[i] = dists[index_counter][sample_index];
} else
printf("This shouldn't be able to happen.\n");
}
}
free(normalized_weights);
free(cummulative_weights);
}
*/
void mixture_f(float (*samplers[])(unsigned int* ), float* weights, int n_dists, float** results, int n_threads)
{
float sum_weights = array_sum(weights, n_dists);
float* normalized_weights = malloc(n_dists * sizeof(float));
for (int i = 0; i < n_dists; i++) {
normalized_weights[i] = weights[i] / sum_weights;
}
float* cummulative_weights = malloc(n_dists * sizeof(float));
array_cumsum(normalized_weights, cummulative_weights, n_dists);
//create var holders
float p1;
int sample_index, i, own_length;
unsigned int* seeds[n_threads];
for(unsigned int i=0; i<n_threads; i++){
seeds[i] = malloc(sizeof(unsigned int));
*seeds[i] = i;
}
#pragma omp parallel private(i, p1, sample_index, own_length)
{
#pragma omp for
for (i = 0; i < n_threads; i++) {
own_length = split_array_get_my_length(i, N, n_threads);
for (int j = 0; j < own_length; j++) {
p1 = random_uniform(0, 1, seeds[i]);
for (int k = 0; k < n_dists; k++) {
if (p1 < cummulative_weights[k]) {
results[i][j] = samplers[k](seeds[i]);
break;
}
}
}
}
}
free(normalized_weights);
free(cummulative_weights);
for(unsigned int i=0; i<n_threads; i++){
free(seeds[i]);
}
}
float sample_0(unsigned int* seed)
{
return 0;
}
float sample_1(unsigned int* seed)
{
return 1;
}
float sample_few(unsigned int* seed)
{
return random_to(1, 3, seed);
}
float sample_many(unsigned int* seed)
{
return random_to(2, 10, seed);
}
void split_array_allocate(float** meta_array, int length, int divide_into)
{
int own_length;
for (int i = 0; i < divide_into; i++) {
own_length = split_array_get_my_length(i, length, divide_into);
meta_array[i] = malloc(own_length * sizeof(float));
}
}
void split_array_free(float** meta_array, int divided_into)
{
for (int i = 0; i < divided_into; i++) {
free(meta_array[i]);
}
free(meta_array);
}
float split_array_sum(float** meta_array, int length, int divided_into)
{
int i;
float output;
float* partial_sum = malloc(divided_into * sizeof(float));
#pragma omp private(i) shared(partial_sum)
for (int i = 0; i < divided_into; i++) {
float own_partial_sum = 0;
int own_length = split_array_get_my_length(i, length, divided_into);
for (int j = 0; j < own_length; j++) {
own_partial_sum += meta_array[i][j];
}
partial_sum[i] = own_partial_sum;
}
for (int i = 0; i < divided_into; i++) {
output += partial_sum[i];
}
return output;
}
int main()
{
//initialize randomness
srand(time(NULL));
// clock_t start, end;
// start = clock();
// Toy example
// Declare variables in play
float p_a, p_b, p_c;
int n_threads = omp_get_max_threads();
// printf("Max threads: %d\n", n_threads);
// omp_set_num_threads(n_threads);
float** dist_mixture = malloc(n_threads * sizeof(float*));
split_array_allocate(dist_mixture, N, n_threads);
// Initialize variables
p_a = 0.8;
p_b = 0.5;
p_c = p_a * p_b;
// Generate mixture
int n_dists = 4;
float weights[] = { 1 - p_c, p_c / 2, p_c / 4, p_c / 4 };
float (*samplers[])(unsigned int* ) = { sample_0, sample_1, sample_few, sample_many };
mixture_f(samplers, weights, n_dists, dist_mixture, n_threads);
printf("Sum(dist_mixture, N)/N = %f\n", split_array_sum(dist_mixture, N, n_threads) / N);
// array_print(dist_mixture[0], N);
split_array_free(dist_mixture, n_threads);
// end = clock();
// printf("Time (ms): %f\n", ((double)(end - start)) / (CLOCKS_PER_SEC) * 1000);
// ^ Will only measure how long it takes the inner main to run, not the whole program,
// including e.g., loading the program into memory or smth.
// Also CLOCKS_PER_SEC in POSIX is a constant equal to 1000000.
// See: https://stackoverflow.com/questions/10455905/why-is-clocks-per-sec-not-the-actual-number-of-clocks-per-second
return 0;
}

53
C/C-01-simple/makefile Normal file
View File

@ -0,0 +1,53 @@
# Interface:
# make
# make build
# make format
# make run
# Compiler
CC=gcc
# CC=tcc # <= faster compilation
# Main file
SRC=samples.c
OUTPUT=samples
## Dependencies
DEPS='gsl'
## Flags
INCS=`pkg-config --cflags ${DEPS}`
LIBS=`pkg-config --libs ${DEPS}`
DEBUG= #'-g'
STANDARD=-std=c99
WARNINGS=-Wall
FAST=-Ofast
## Formatter
STYLE_BLUEPRINT=webkit
FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT)
## make build
build: $(SRC)
$(CC) $(DEBUG) $(INCS) $(PLUGS) $(SRC) -o samples $(LIBS)
fast: $(SRC)
$(CC) $(FAST) $(DEBUG) $(INCS) $(PLUGS) $(SRC) -o samples $(LIBS)
format: $(SRC)
$(FORMATTER) $(SRC)
run: $(SRC) $(OUTPUT)
echo "Increasing stack size limit, because we are dealing with 1M samples"
# ulimit: increase stack size limit
# -Ss: the soft limit. If you set the hard limit, you then can't raise it
# 256000: around 250Mbs, if I'm reading it correctly.
# Then run the program
ulimit -Ss 256000 && ./$(OUTPUT)
# Old:
# Link libraries, for good measure
# LD_LIBRARY_PATH=/usr/local/lib
# export LD_LIBRARY_PATH

0
C/samples → C/C-01-simple/samples
View File

157
C/C-01-simple/samples.c Normal file
View File

@ -0,0 +1,157 @@
#include <gsl/gsl_randist.h>
#include <gsl/gsl_rng.h>
#include <math.h>
#include <stdio.h>
#include <stdlib.h>
#include <time.h>
#define N 1000000
/*
* For very high values of N, you will want to increase the maximum stack trace, otherwise you will suffer a segmentation fault
* In Ubuntu/bash you can do this with $ ulimit -Ss 256000 ## ~256Mbs
* And confirm it with $ ulimit -a
*/
/* Helpers */
void print(double* ys)
{
for (int i = 0; i < N; i++) {
printf("%f\n", ys[i]);
}
printf("\n");
}
void fill(double* ys, float f)
{
for (int i = 0; i < N; i++) {
ys[i] = f;
}
}
double sum(double* ps, int n)
{
double result = 0;
for (int i = 0; i < n; i++) {
result += ps[i];
}
return (result);
}
void cumsum(double* ps, double* rs, int n)
{
double counter = 0;
for (int i = 0; i < n; i++) {
counter += ps[i];
rs[i] = counter;
}
}
/* Distributions*/
void normal(gsl_rng* r, double* ys, double mean, double std)
{
for (int i = 0; i < N; i++) {
ys[i] = mean + gsl_ran_gaussian(r, std);
}
}
void lognormal(gsl_rng* r, double* ys, double zeta, double sigma)
{
for (int i = 0; i < N; i++) {
ys[i] = gsl_ran_lognormal(r, zeta, sigma);
}
}
void to(gsl_rng* r, double* ys, double low, double high)
{
double normal95confidencePoint = 1.6448536269514722;
double log_low = log(low);
double log_high = log(high);
double zeta = (log_low + log_high) / 2;
double sigma = (log_high - log_low) / (2.0 * normal95confidencePoint);
lognormal(r, ys, zeta, sigma);
}
/* Mixture of distributions */
void mixture(gsl_rng* r, double* dists[], double* weights, int n, double* results)
{
/* Get cummulative, normalized weights */
double sum_weights = sum(weights, n);
double normalized_weights[n];
for (int i = 0; i < n; i++) {
normalized_weights[i] = weights[i] / sum_weights;
}
double cummulative_weights[n];
cumsum(normalized_weights, cummulative_weights, n);
/* Get N samples, drawn from the different distributions in proportion to their weights. */
for (int i = 0; i < N; i++) {
double p_1 = gsl_rng_uniform(r);
double p_2 = gsl_rng_uniform(r);
int index_found = 0;
int index_counter = 0;
while ((index_found == 0) && (index_counter < n)) {
if (p_1 < cummulative_weights[index_counter]) {
index_found = 1;
} else {
index_counter++;
}
}
if (index_found == 0) {
printf("\nThis shouldn't be able to happen");
// gsl_rng_free (r);
// abort(); // this shouldn't have happened.
} else {
int sample_index = (int)floor(p_2 * N);
results[i] = dists[index_counter][sample_index];
}
}
}
/* Main */
int main(void)
{
// Start clock
clock_t start, end;
start = clock();
/* Initialize GNU Statistical Library (GSL) stuff */
const gsl_rng_type* T;
gsl_rng* r;
// gsl_rng_env_setup();
T = gsl_rng_default;
r = gsl_rng_alloc(T);
/* Toy example */
/* Declare variables in play */
double p_a, p_b, p_c;
double dist_none[N], dist_one[N], dist_few[N], dist_many[N], dist_mixture[N];
/* Initialize variables */
p_a = 0.8;
p_b = 0.5;
p_c = p_a * p_b;
fill(dist_none, 0);
fill(dist_one, 1);
to(r, dist_few, 1, 3);
to(r, dist_many, 2, 10);
/* Generate mixture */
int n = 4;
double weights[] = { 1 - p_c, p_c / 2, p_c / 4, p_c / 4 };
double* dists[] = { dist_none, dist_one, dist_few, dist_many };
mixture(r, dists, weights, n, dist_mixture);
printf("%f\n", sum(dist_mixture, N) / N);
/* Clean up GSL */
gsl_rng_free(r);
// End clock
end = clock();
printf("Total time (ms): %f\n", ((double)(end - start)) / CLOCKS_PER_SEC * 1000);
/* Return success*/
return EXIT_SUCCESS;
}

53
C/C-02-better-algorithm-one-thread/makefile Normal file
View File

@ -0,0 +1,53 @@
# Interface:
# make
# make build
# make format
# make run
# Compiler
CC=gcc
# CC=tcc # <= faster compilation
# Main file
SRC_ONE_THREAD=./samples-one-thread.c
OUTPUT_ONE_THREAD=out/samples-one-thread
## Dependencies
# Has no dependencies
MATH=-lm
## Flags
DEBUG= #'-g'
STANDARD=-std=c99
WARNINGS=-Wall
OPTIMIZED=-O3 #-O3 actually gives better performance than -Ofast, at least for this version
OPENMP=-fopenmp
## Formatter
STYLE_BLUEPRINT=webkit
FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT)
## make build
build: $(SRC_ONE_THREAD)
mkdir -p out
$(CC) $(OPTIMIZED) $(DEBUG) $(SRC_ONE_THREAD) $(OPENMP) $(MATH) -o $(OUTPUT_ONE_THREAD)
format: $(SRC_ONE_THREAD)
$(FORMATTER) $(SRC_ONE_THREAD)
run: $(SRC_ONE_THREAD) $(OUTPUT_ONE_THREAD)
./$(OUTPUT_ONE_THREAD)
time-linux:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
@echo "Running 100x and taking avg time: $(OUTPUT_ONE_THREAD)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do $(OUTPUT_ONE_THREAD); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time: |" | sed 's|$$|ms|' && echo
time-linux-simple:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
/bin/time -f "Time: %es" ./$(OUTPUT_ONE_THREAD) && echo
debian-install-dependencies:
sudo apt-get install libomp-dev

BIN
C/C-02-better-algorithm-one-thread/out/samples-one-thread Executable file
View File

Binary file not shown.

2
C-optimized/samples-one-thread.c → C/C-02-better-algorithm-one-thread/samples-one-thread.c
View File

@ -5,7 +5,7 @@
const float PI = 3.14159265358979323846; const float PI = 3.14159265358979323846;
#define N 10000000 #define N 1000000
//Array helpers //Array helpers

15
C/README.md Normal file
View File

@ -0,0 +1,15 @@
# Time to BOTEC in C
This repository contains a few implementations of a simple botec (back-of-the-envelope) calculation in C:
- In the folder C-01-simple/, you can see a simple implementation, which passes large arrays
- In the folder C-02-better-algorithm-one-thread/ you can see a better implementations, that passes around pointers to functions, which makes the implementation more efficient
- In the top level, you can see an implementation that uses the better implementation in C-02..., and that also implements multithreading using OpenMP
## To do
- [ ] Update repository with correct timing
- [ ] Add Windows/Powershell time-measuring commands
- [ ] Add CUDA?
- [x] Use better profiling approach to capture timing with 1M samples.
- [x] See if program can be reworded so as to use multithreading effectively, e.g., so that you see speed gains proportional to the number of threads used

135
C/makefile
View File

@ -1,53 +1,82 @@
# Interface: # Interface:
# make # make
# make build # make build
# make format # make format
# make run # make run
# Compiler # Compiler
CC=gcc CC=gcc
# CC=tcc # <= faster compilation # CC=tcc # <= faster compilation
# Main file # Main file
SRC=samples.c SRC=samples.c
OUTPUT=samples OUTPUT=out/samples
## Dependencies SRC_ONE_THREAD=./samples-one-thread.c
DEPS='gsl' OUTPUT_ONE_THREAD=out/samples-one-thread
## Flags ## Dependencies
INCS=`pkg-config --cflags ${DEPS}` # Has no dependencies
LIBS=`pkg-config --libs ${DEPS}` MATH=-lm
DEBUG= #'-g'
STANDARD=-std=c99 ## Flags
WARNINGS=-Wall DEBUG= #'-g'
FAST=-Ofast STANDARD=-std=c99
## Formatter WARNINGS=-Wall
STYLE_BLUEPRINT=webkit OPTIMIZED=-O3 #-O3 actually gives better performance than -Ofast, at least for this version
FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT) OPENMP=-fopenmp
## make build ## Formatter
build: $(SRC) STYLE_BLUEPRINT=webkit
$(CC) $(DEBUG) $(INCS) $(PLUGS) $(SRC) -o samples $(LIBS) FORMATTER=clang-format -i -style=$(STYLE_BLUEPRINT)
fast: $(SRC) ## make build
$(CC) $(FAST) $(DEBUG) $(INCS) $(PLUGS) $(SRC) -o samples $(LIBS) build: $(SRC)
$(CC) $(OPTIMIZED) $(DEBUG) $(SRC) $(OPENMP) $(MATH) -o $(OUTPUT)
format: $(SRC) $(CC) $(OPTIMIZED) $(DEBUG) $(SRC_ONE_THREAD) $(OPENMP) $(MATH) -o $(OUTPUT_ONE_THREAD)
$(FORMATTER) $(SRC)
format: $(SRC)
run: $(SRC) $(OUTPUT) $(FORMATTER) $(SRC)
echo "Increasing stack size limit, because we are dealing with 1M samples"
# ulimit: increase stack size limit run: $(SRC) $(OUTPUT)
# -Ss: the soft limit. If you set the hard limit, you then can't raise it OMP_NUM_THREADS=1 ./$(OUTPUT) && echo
# 256000: around 250Mbs, if I'm reading it correctly. ./$(OUTPUT_ONE_THREAD)
# Then run the program
ulimit -Ss 256000 && ./$(OUTPUT) multi:
OMP_NUM_THREADS=1 ./$(OUTPUT) && echo
OMP_NUM_THREADS=2 ./$(OUTPUT) && echo
OMP_NUM_THREADS=4 ./$(OUTPUT) && echo
# Old: OMP_NUM_THREADS=8 ./$(OUTPUT) && echo
# Link libraries, for good measure OMP_NUM_THREADS=16 ./$(OUTPUT) && echo
# LD_LIBRARY_PATH=/usr/local/lib ./$(OUTPUT_ONE_THREAD) && echo
# export LD_LIBRARY_PATH
time-linux:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=1 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=1 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 1 thread: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=2 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=2 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 2 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=4 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=4 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time for 4 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=8 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=8 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 8 threads: |" | sed 's|$$|ms|' && echo
@echo "Running 100x and taking avg time: OMP_NUM_THREADS=16 $(OUTPUT)"
@t=$$(/usr/bin/time -f "%e" -p bash -c 'for i in {1..100}; do OMP_NUM_THREADS=16 $(OUTPUT); done' 2>&1 >/dev/null | grep real | awk '{print $$2}' ); echo "scale=2; 1000 * $$t / 100" | bc | sed "s|^|Time using 16 threads: |" | sed 's|$$|ms|' && echo
time-linux-simple:
@echo "Requires /bin/time, found on GNU/Linux systems" && echo
OMP_NUM_THREADS=1 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=2 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=4 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=8 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
OMP_NUM_THREADS=16 /bin/time -f "Time: %es" ./$(OUTPUT) && echo
/bin/time -f "Time: %es" ./$(OUTPUT_ONE_THREAD) && echo
debian-install-dependencies:
sudo apt-get install libomp-dev

0
C-optimized/out/samples → C/out/samples
View File

0
C-optimized/out/samples-one-thread → C/out/samples-one-thread
View File

0
C-optimized/out/samples.exe → C/out/samples.exe
View File

0
C-optimized/out/samples.txt → C/out/samples.txt
View File

438
C/samples.c
View File

@ -1,157 +1,281 @@
#include <gsl/gsl_randist.h> #include <math.h>
#include <gsl/gsl_rng.h> #include <omp.h>
#include <math.h> #include <stdio.h>
#include <stdio.h> #include <stdlib.h>
#include <stdlib.h> #include <time.h>
#include <time.h>
const float PI = 3.14159265358979323846;
#define N 1000000
/* #define N 1000000
* For very high values of N, you will want to increase the maximum stack trace, otherwise you will suffer a segmentation fault
* In Ubuntu/bash you can do this with $ ulimit -Ss 256000 ## ~256Mbs //Array helpers
* And confirm it with $ ulimit -a
*/ void array_print(float* array, int length)
{
/* Helpers */ for (int i = 0; i < length; i++) {
void print(double* ys) printf("item[%d] = %f\n", i, array[i]);
{ }
for (int i = 0; i < N; i++) { printf("\n");
printf("%f\n", ys[i]); }
}
printf("\n"); void array_fill(float* array, int length, float item)
} {
int i;
void fill(double* ys, float f) #pragma omp private(i)
{ {
for (int i = 0; i < N; i++) { #pragma omp for
ys[i] = f; for (i = 0; i < length; i++) {
} array[i] = item;
} }
}
double sum(double* ps, int n) }
{
double result = 0; float array_sum(float* array, int length)
for (int i = 0; i < n; i++) { {
result += ps[i]; float output = 0.0;
} for (int i = 0; i < length; i++) {
return (result); output += array[i];
} }
return output;
void cumsum(double* ps, double* rs, int n) }
{
double counter = 0; void array_cumsum(float* array_to_sum, float* array_cumsummed, int length)
for (int i = 0; i < n; i++) { {
counter += ps[i]; array_cumsummed[0] = array_to_sum[0];
rs[i] = counter; for (int i = 1; i < length; i++) {
} array_cumsummed[i] = array_cumsummed[i - 1] + array_to_sum[i];
} }
}
/* Distributions*/
void normal(gsl_rng* r, double* ys, double mean, double std) float rand_float(float to, unsigned int* seed)
{ {
for (int i = 0; i < N; i++) { return ((float)rand_r(seed) / (float)RAND_MAX) * to;
ys[i] = mean + gsl_ran_gaussian(r, std); // See: <https://stackoverflow.com/questions/43151361/how-to-create-thread-safe-random-number-generator-in-c-using-rand-r>
} // rand() is not thread-safe, as it relies on (shared) hidden state.
} }
void lognormal(gsl_rng* r, double* ys, double zeta, double sigma) float ur_normal(unsigned int* seed)
{ {
for (int i = 0; i < N; i++) { float u1 = rand_float(1.0, seed);
ys[i] = gsl_ran_lognormal(r, zeta, sigma); float u2 = rand_float(1.0, seed);
} float z = sqrtf(-2.0 * log(u1)) * sin(2 * PI * u2);
} return z;
}
void to(gsl_rng* r, double* ys, double low, double high)
{ inline float random_uniform(float from, float to, unsigned int* seed)
double normal95confidencePoint = 1.6448536269514722; {
double log_low = log(low); return ((float) rand_r(seed) / (float)RAND_MAX) * (to - from) + from;
double log_high = log(high); }
double zeta = (log_low + log_high) / 2;
double sigma = (log_high - log_low) / (2.0 * normal95confidencePoint); inline float random_normal(float mean, float sigma, unsigned int* seed)
lognormal(r, ys, zeta, sigma); {
} return (mean + sigma * ur_normal(seed));
}
/* Mixture of distributions */
void mixture(gsl_rng* r, double* dists[], double* weights, int n, double* results) inline float random_lognormal(float logmean, float logsigma, unsigned int* seed)
{ {
/* Get cummulative, normalized weights */ return expf(random_normal(logmean, logsigma, seed));
double sum_weights = sum(weights, n); }
double normalized_weights[n];
for (int i = 0; i < n; i++) { inline float random_to(float low, float high, unsigned int* seed)
normalized_weights[i] = weights[i] / sum_weights; {
} const float NORMAL95CONFIDENCE = 1.6448536269514722;
double cummulative_weights[n]; float loglow = logf(low);
cumsum(normalized_weights, cummulative_weights, n); float loghigh = logf(high);
float logmean = (loglow + loghigh) / 2;
/* Get N samples, drawn from the different distributions in proportion to their weights. */ float logsigma = (loghigh - loglow) / (2.0 * NORMAL95CONFIDENCE);
for (int i = 0; i < N; i++) { return random_lognormal(logmean, logsigma, seed);
double p_1 = gsl_rng_uniform(r); }
double p_2 = gsl_rng_uniform(r);
int split_array_get_my_length(int index, int total_length, int n_threads)
int index_found = 0; {
int index_counter = 0; return (total_length % n_threads > index ? total_length / n_threads + 1 : total_length / n_threads);
while ((index_found == 0) && (index_counter < n)) { }
if (p_1 < cummulative_weights[index_counter]) {
index_found = 1; //Old version, don't use it!! Optimized version is called mixture_f. This one is just for display
} else { /*
index_counter++; void mixture(float* dists[], float* weights, int n_dists, float* results)
} {
} float sum_weights = array_sum(weights, n_dists);
if (index_found == 0) { float* normalized_weights = malloc(n_dists * sizeof(float));
printf("\nThis shouldn't be able to happen"); for (int i = 0; i < n_dists; i++) {
// gsl_rng_free (r); normalized_weights[i] = weights[i] / sum_weights;
// abort(); // this shouldn't have happened. }
} else { float* cummulative_weights = malloc(n_dists * sizeof(float));
int sample_index = (int)floor(p_2 * N); array_cumsum(normalized_weights, cummulative_weights, n_dists);
results[i] = dists[index_counter][sample_index];
} //create var holders
} float p1, p2;
} int index_found, index_counter, sample_index, i;
/* Main */ #pragma omp parallel private(i, p1, p2, index_found, index_counter, sample_index)
int main(void) {
{ #pragma omp for
// Start clock for (i = 0; i < N; i++) {
clock_t start, end; p1 = random_uniform(0, 1);
start = clock(); p2 = random_uniform(0, 1);
/* Initialize GNU Statistical Library (GSL) stuff */ index_found = 0;
const gsl_rng_type* T; index_counter = 0;
gsl_rng* r;
// gsl_rng_env_setup(); while ((index_found == 0) && (index_counter < n_dists)) {
T = gsl_rng_default; if (p1 < cummulative_weights[index_counter]) {
r = gsl_rng_alloc(T); index_found = 1;
} else {
/* Toy example */ index_counter++;
/* Declare variables in play */ }
double p_a, p_b, p_c; }
double dist_none[N], dist_one[N], dist_few[N], dist_many[N], dist_mixture[N]; if (index_found != 0) {
sample_index = (int)(p2 * N);
/* Initialize variables */ results[i] = dists[index_counter][sample_index];
p_a = 0.8; } else
p_b = 0.5; printf("This shouldn't be able to happen.\n");
p_c = p_a * p_b; }
}
fill(dist_none, 0); free(normalized_weights);
fill(dist_one, 1); free(cummulative_weights);
to(r, dist_few, 1, 3); }
to(r, dist_many, 2, 10); */
void mixture_f(float (*samplers[])(unsigned int* ), float* weights, int n_dists, float** results, int n_threads)
/* Generate mixture */ {
int n = 4; float sum_weights = array_sum(weights, n_dists);
double weights[] = { 1 - p_c, p_c / 2, p_c / 4, p_c / 4 }; float* normalized_weights = malloc(n_dists * sizeof(float));
double* dists[] = { dist_none, dist_one, dist_few, dist_many }; for (int i = 0; i < n_dists; i++) {
normalized_weights[i] = weights[i] / sum_weights;
mixture(r, dists, weights, n, dist_mixture); }
printf("%f\n", sum(dist_mixture, N) / N);
float* cummulative_weights = malloc(n_dists * sizeof(float));
/* Clean up GSL */ array_cumsum(normalized_weights, cummulative_weights, n_dists);
gsl_rng_free(r);
//create var holders
// End clock float p1;
end = clock(); int sample_index, i, own_length;
printf("Total time (ms): %f\n", ((double)(end - start)) / CLOCKS_PER_SEC * 1000); unsigned int* seeds[n_threads];
/* Return success*/ for(unsigned int i=0; i<n_threads; i++){
return EXIT_SUCCESS; seeds[i] = malloc(sizeof(unsigned int));
} *seeds[i] = i;
}
#pragma omp parallel private(i, p1, sample_index, own_length)
{
#pragma omp for
for (i = 0; i < n_threads; i++) {
own_length = split_array_get_my_length(i, N, n_threads);
for (int j = 0; j < own_length; j++) {
p1 = random_uniform(0, 1, seeds[i]);
for (int k = 0; k < n_dists; k++) {
if (p1 < cummulative_weights[k]) {
results[i][j] = samplers[k](seeds[i]);
break;
}
}
}
}
}
free(normalized_weights);
free(cummulative_weights);
for(unsigned int i=0; i<n_threads; i++){
free(seeds[i]);
}
}
float sample_0(unsigned int* seed)
{
return 0;
}
float sample_1(unsigned int* seed)
{
return 1;
}
float sample_few(unsigned int* seed)
{
return random_to(1, 3, seed);
}
float sample_many(unsigned int* seed)
{
return random_to(2, 10, seed);
}
void split_array_allocate(float** meta_array, int length, int divide_into)
{
int own_length;
for (int i = 0; i < divide_into; i++) {
own_length = split_array_get_my_length(i, length, divide_into);
meta_array[i] = malloc(own_length * sizeof(float));
}
}
void split_array_free(float** meta_array, int divided_into)
{
for (int i = 0; i < divided_into; i++) {
free(meta_array[i]);
}
free(meta_array);
}
float split_array_sum(float** meta_array, int length, int divided_into)
{
int i;
float output;
float* partial_sum = malloc(divided_into * sizeof(float));
#pragma omp private(i) shared(partial_sum)
for (int i = 0; i < divided_into; i++) {
float own_partial_sum = 0;
int own_length = split_array_get_my_length(i, length, divided_into);
for (int j = 0; j < own_length; j++) {
own_partial_sum += meta_array[i][j];
}
partial_sum[i] = own_partial_sum;
}
for (int i = 0; i < divided_into; i++) {
output += partial_sum[i];
}
return output;
}
int main()
{
//initialize randomness
srand(time(NULL));
// clock_t start, end;
// start = clock();
// Toy example
// Declare variables in play
float p_a, p_b, p_c;
int n_threads = omp_get_max_threads();
// printf("Max threads: %d\n", n_threads);
// omp_set_num_threads(n_threads);
float** dist_mixture = malloc(n_threads * sizeof(float*));
split_array_allocate(dist_mixture, N, n_threads);
// Initialize variables
p_a = 0.8;
p_b = 0.5;
p_c = p_a * p_b;
// Generate mixture
int n_dists = 4;
float weights[] = { 1 - p_c, p_c / 2, p_c / 4, p_c / 4 };
float (*samplers[])(unsigned int* ) = { sample_0, sample_1, sample_few, sample_many };
mixture_f(samplers, weights, n_dists, dist_mixture, n_threads);
printf("Sum(dist_mixture, N)/N = %f\n", split_array_sum(dist_mixture, N, n_threads) / N);
// array_print(dist_mixture[0], N);
split_array_free(dist_mixture, n_threads);
// end = clock();
// printf("Time (ms): %f\n", ((double)(end - start)) / (CLOCKS_PER_SEC) * 1000);
// ^ Will only measure how long it takes the inner main to run, not the whole program,
// including e.g., loading the program into memory or smth.
// Also CLOCKS_PER_SEC in POSIX is a constant equal to 1000000.
// See: https://stackoverflow.com/questions/10455905/why-is-clocks-per-sec-not-the-actual-number-of-clocks-per-second
return 0;
}

18
README.md
View File

@ -29,15 +29,15 @@ As of now, it may be useful for checking the validity of simple estimations. The
## Comparison table ## Comparison table
| Language | Time | Lines of code | | Language | Time | Lines of code |
|--------------------------|-----------|---------------| |-----------------------------|-----------|---------------|
| C (optimized, 1 thread) | 30ms | 183 | | C (optimized, 16 threads) | 6ms | 183 |
| Nim | 68ms | 84 | | Nim | 68ms | 84 |
| C | 292ms | 149 | | C (naïve implementation) | 292ms | 149 |
| Javascript (NodeJS) | 732ms | 69 | | Javascript (NodeJS) | 732ms | 69 |
| Squiggle | 1,536s | 14 | | Squiggle | 1,536s | 14 |
| R | 7,000s | 49 | | R | 7,000s | 49 |
| Python (CPython) | 16,641s | 56 | | Python (CPython) | 16,641s | 56 |
Time measurements taken with the [time](https://man7.org/linux/man-pages/man1/time.1.html) tool, using 1M samples: Time measurements taken with the [time](https://man7.org/linux/man-pages/man1/time.1.html) tool, using 1M samples: