All examples are using Bazel as a main build system.
Let's generate 10,000,000 random numbers.
$ bazel run :random -- --n 2500000 --k 4 --output $(pwd)/out/10m_4t
INFO: Found 1 target...
Target //:random up-to-date:
bazel-bin/random
INFO: Elapsed time: 0.041s, Critical Path: 0.00s
INFO: Running command line: bazel-bin/random --n 2500000 --k 4 --output /home/lionell/dev/labs/algo/out/10m_4t
I0318 13:38:44.870348 3702 random.cc:58] R=4294967295, M=10019, M^k=262120981
I0318 13:38:44.872545 3702 random.cc:31] [0]: u0=1234
I0318 13:38:44.872555 3705 random.cc:31] [3]: u0=3487891871
I0318 13:38:44.872545 3704 random.cc:31] [2]: u0=3610281214
I0318 13:38:44.872545 3703 random.cc:31] [1]: u0=12363446
I0318 13:38:45.205608 3702 benchmark.h:37] Generation took 1.304133s cpu, 0.335000s wall.
Now, we have generated 4 chunks of random data.
$ ls -lh out/
total 103M
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.0
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.1
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.2
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.3
Let's concatenate them into one huge file containing 10M random numbers.
$ bazel run :concat $(pwd)/out/10m_4t 4
INFO: Found 1 target...
Target //:concat up-to-date:
bazel-bin/concat
INFO: Elapsed time: 0.050s, Critical Path: 0.00s
INFO: Running command line: bazel-bin/concat /home/lionell/dev/labs/algo/out/10m_4t 4
Now we have one file containing all generated random numbers.
$ ls -lh out/
total 205M
-rw-r--r-- 1 lionell lionell 103M бер 18 13:40 10m_4t
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.0
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.1
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.2
-rw-r--r-- 1 lionell lionell 26M бер 18 13:39 10m_4t.3
Let's generate another one file running :generate in just one thread.
$ bazel run :random -- --n 10000000 --k 1 --output $(pwd)/out/10m_1t
INFO: Found 1 target...
Target //:random up-to-date:
bazel-bin/random
INFO: Elapsed time: 0.061s, Critical Path: 0.00s
INFO: Running command line: bazel-bin/random --n 10000000 --k 1 --output /home/lionell/dev/labs/algo/out/10m_1t
I0318 13:46:10.093510 13557 random.cc:58] R=4294967295, M=10019, M^k=10019
I0318 13:46:10.093595 13557 random.cc:31] [0]: u0=1234
I0318 13:46:10.783151 13557 benchmark.h:37] Generation took 0.689168s cpu, 0.689000s wall.
Finally, let's make sure that files are the same. We can just run diff out/10m_4t out/10m_1t.
Sometimes we wan't to check acceleration of algorithm on some pretty huge amount of data, but it
can overuse some resources available on machine it's running. Eg. this algorithm uses filesystem to
store generated data. Having n = 10,000,000,000 we are going to generate 10B 32-bit random numbers.
It's about 37GB of disk space. Fortunatelly there is a dummy mode, where we can tell program not to
store generated data. To get into dummy mode, just run it without --output parameter.
$ bazel run :random -- --n 250000000 --k 4
INFO: Found 1 target...
Target //:random up-to-date:
bazel-bin/random
INFO: Elapsed time: 0.196s, Critical Path: 0.01s
INFO: Running command line: bazel-bin/random --n 250000000 --k 4
I0318 21:07:38.078176 13114 random.cc:58] R=4294967295, M=10019, M^k=262120981
W0318 21:07:38.085633 13114 io.h:15] Switching to dummy mode.
I0318 21:07:38.085650 13114 random.cc:31] [0]: u0=1234
W0318 21:07:38.085674 13117 io.h:15] Switching to dummy mode.
I0318 21:07:38.085685 13117 random.cc:31] [3]: u0=3487891871
W0318 21:07:38.085703 13115 io.h:15] Switching to dummy mode.
I0318 21:07:38.085710 13115 random.cc:31] [1]: u0=12363446
W0318 21:07:38.086269 13116 io.h:15] Switching to dummy mode.
I0318 21:07:38.086273 13116 random.cc:31] [2]: u0=3610281214
I0318 21:07:41.660795 13114 benchmark.h:37] Generation took 12.589956s cpu, 3.582000s wall.
MIT