added results for rayon git, more crates to check

README.md

@@ -21,6 +21,7 @@ Supported command line options:
         --img_size <IMAGE_SIZE>              size of image in pixel (square, default: 1024, must be a power of two)
         --img1 <IMAGINARY1>                  lower part (default: -1.50)
         --img2 <IMAGINARY2>                  upper part (default: 1.50)
+        --write_meta_data                    write meta data like run time into the ppm file (default: off)
         --max_iter <MAX_ITER>                maximum number of iterations (default: 2048)
         --num_threads <NUMBER_OF_THREADS>    number of threads to use (default: 2)
         --re1 <REAL1>                        left real part (default: -2.0)
@@ -62,17 +63,21 @@ simple parallel | 1 | 2508.58119
 simple parallel | 8 | 389.50966
 simple parallel | 24 | 161.75248
 rayon* | 24 | 139.04431
+rayon git* | 24 | 127.69423
 
 (*) Note that rayon uses whatever number of cores are available at the moment.
 
 With just using one thread the overhead for both scoped thread pool and simple parallel is too high and thus both are slower than the serial version.
 Using all cores (including virtual one due to hyper threading) rayon is the fastest method. Is uses explicit work stealing to utilize all the cores more efficiently.
 
 # TODO:
-- [ ] Check ArrayFire
-- [ ] Check Collenchyma
-- [ ] Check Timely Dataflow
-- [ ] Check Crossbeam
+- [ ] Check [ArrayFire](https://github.com/arrayfire/arrayfire-rust)
+- [ ] Check [Collenchyma](https://github.com/autumnai/collenchyma)
+- [ ] Check [Timely Dataflow](https://github.com/frankmcsherry/timely-dataflow)
+- [ ] Check [Crossbeam](https://github.com/aturon/crossbeam)
+- [ ] Check [rust-threadpool](https://github.com/frewsxcv/rust-threadpool)
+- [ ] Check [jobsteal](https://github.com/rphmeier/jobsteal)
+- [ ] Check [forkjoin](https://github.com/faern/forkjoin)
 - [ ] Use rust-fmt on source code (Thanks to matklad)
 - [ ] Check docopt (instead of clap ? Thanks to matklad)
 

plot/plot.sh

@@ -14,6 +14,8 @@ gnuplot <<PLOT
     plot "serial.txt" title "serial", \
          "scoped_thread_pool.txt" title "scoped thread pool", \
          "simple_parallel.txt" title "simple parallel", \
-         "rayon.txt" title "rayon"
+         "rayon.txt" title "rayon", \
+         "rayon_git.txt" title "rayon git"
 
 PLOT
+

plot/rayon_git.txt

@@ -0,0 +1 @@
+24 127.69423

src/main.rs

@@ -195,6 +195,7 @@ fn rayon_mandel(mandel_config: &MandelConfig, image: &mut [u32]) {
     rayon_mandel_helper(mandel_config, image, 0);
 }
 
+// Rayon helper function for recursive divide-and-conquer call
 fn rayon_mandel_helper(mandel_config: &MandelConfig, slice: &mut [u32], y: u32) {
     if slice.len() == (mandel_config.img_size as usize) { // just process one scanline of the mandelbrot image
         for x in 0..mandel_config.img_size {
@@ -214,6 +215,8 @@ fn rayon_mandel_helper(mandel_config: &MandelConfig, slice: &mut [u32], y: u32)
     }
 }
 
+
+
 // Prepares and runs one version of the mandelbro set calculation
 fn do_run(file_name_prefix: &str, mandel_func: &Fn(&MandelConfig, &mut [u32]) -> (),
     mandel_config: &MandelConfig, image: &mut [u32], time_now: &str) {