Created
August 4, 2013 04:33
-
-
Save nebuta/6149137 to your computer and use it in GitHub Desktop.
Halide simple example
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
// Modified from tutorials/lesson_02.cpp and test/vectorize.cpp | |
// On linux, you can compile and run it like so: | |
// g++ -I ../include -L ../bin -lHalide -lpthread -ldl -lpng blur.cpp | |
// LD_LIBRARY_PATH=../bin ./lesson_02 | |
// On os x: | |
// g++ -I ../include -L ../bin -lHalide `libpng-config --cflags --ldflags` blur.cpp | |
// DYLD_LIBRARY_PATH=../bin ./lesson_02 | |
// The only Halide header file you need is Halide.h. It includes all of Halide. | |
#include <Halide.h> | |
using Halide::Image; | |
#ifdef _WIN32 | |
extern "C" bool QueryPerformanceCounter(uint64_t *); | |
extern "C" bool QueryPerformanceFrequency(uint64_t *); | |
double currentTime() { | |
uint64_t t, freq; | |
QueryPerformanceCounter(&t); | |
QueryPerformanceFrequency(&freq); | |
return (t * 1000.0) / freq; | |
} | |
#else | |
#include <sys/time.h> | |
double currentTime() { | |
timeval t; | |
gettimeofday(&t, NULL); | |
return t.tv_sec * 1000.0 + t.tv_usec / 1000.0f; | |
} | |
#endif | |
#include "../apps/support/image_io.h" | |
int main(int argc, char **argv) { | |
Halide::Image<uint8_t> image = load<uint8_t>("../apps/images/rgb.png"); | |
Halide::Func brighter,brighter_par; | |
Halide::Var x, y, xi, yi, c; | |
Halide::Func input; | |
int w = image.width(); | |
int h = image.height(); | |
int ch = image.channels(); | |
printf("Input: %d x %d x %d\n",w,h,ch); | |
//Define algorithm. | |
brighter(x, y, c) = Halide::cast<uint8_t>(min(image(x, y, c) * 1.5f, 255)); | |
brighter_par(x,y,c) = brighter(x,y,c); | |
// Specify a schedule. | |
brighter_par.split(y, y, yi, 8).parallel(y).vectorize(x, 8); | |
// brighter_par.vectorize(x,4).parallel(y); | |
//By invoking "realize" first, JIT is done before time measurement. (JIT takes a few seconds.) | |
Image<uint8_t> out = brighter.realize(w,h,ch); | |
Image<uint8_t> out_par = brighter_par.realize(w,h,ch); | |
printf("Start measurement.\n"); | |
//Execution time measurement. | |
double t1 = currentTime(); | |
const int repeat = 1000; | |
for (int i = 0; i < repeat; i++) { | |
brighter.realize(out); | |
} | |
double t2 = currentTime(); | |
for (int i = 0; i < repeat; i++) { | |
brighter_par.realize(out_par); | |
} | |
double t3 = currentTime(); | |
printf("Vectorized vs scalar (average over %d times): %.2f ms %.2f ms. Speedup = %1.3f\n", repeat, | |
((t3-t2)/repeat), ((t2-t1)/repeat), (t2-t1)/(t3-t2)); | |
save(out, "brighter.png"); | |
save(out_par, "brighter_par.png"); | |
printf("Success!\n"); | |
return 0; | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment