Stream
Instructions
The STREAM benchmark is a simple synthetic benchmark program that measures sustainable memory bandwidth (in MB/s) and the corresponding computation rate for simple vector kernels.
The stream consists of a single file, and to perform the test, you only need to compile stream.c:
git clone <https://github.com/microseyuyu/STREAM.git>
cd STREAM
gcc -O3 -fopenmp -DN=2000000 -DNTIMES=10 stream.c -o stream
export OMP_NUM_THREADS=8
./stream
Parameter descriptions:
- -O3: Specifies the highest compilation optimization level, that is, 3.
- fopenmp: Enables OpenMP, suitable for multi-processor environments, to better achieve the actual maximum memory bandwidth. When enabled, the program defaults to running the number of threads equal to the number of CPU threads.
DN=2000000: Specifies the size of the test arrays a[], b[], and c[] (Array size). This value has a significant impact on the test results (the default value for version 5.9 is 2000000. If stream.c is version 5.10, the parameter name changes to -DSTREAM_ARRAY_SIZE, with a default value of 10000000). Note: the size of the test array must be set much larger than the size of the CPU's highest-level cache (usually L3 Cache), otherwise it tests the throughput performance of the CPU cache, rather than the memory throughput performance. - -DNTIMES=10: The number of repetitions, from which the optimal value is selected. - OMP_NUM_THREADS=8: Number of threads.
Sample results:
debian@lpi4a:~/Desktop/STREAM$ ./stream
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
***** WARNING: ******
It appears that you set the preprocessor variable N when compiling this code.
This version of the code uses the preprocesor variable STREAM_ARRAY_SIZE to control the array size
Reverting to default value of STREAM_ARRAY_SIZE=10000000
***** WARNING: ******
Array size = 10000000 (elements), Offset = 0 (elements)
Memory per array = 76.3 MiB (= 0.1 GiB).
Total memory required = 228.9 MiB (= 0.2 GiB).
Each kernel will be executed 10 times.
The *best* time for each kernel (excluding the first iteration)
will be used to compute the reported bandwidth.
-------------------------------------------------------------
Number of Threads requested = 8
Number of Threads counted = 8
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 21622 microseconds.
(= 21622 clock ticks)
Increase the size of the arrays if this shows that
you are not getting at least 20 clock ticks per test.
-------------------------------------------------------------
WARNING -- The above is only a rough guideline.
For best results, please be sure you know the
precision of your system timer.
-------------------------------------------------------------
Function Best Rate MB/s Avg time Min time Max time
Copy: 8364.2 0.019258 0.019129 0.019508
Scale: 8291.0 0.019572 0.019298 0.020162
Add: 6223.6 0.038835 0.038563 0.040011
Triad: 6222.5 0.038776 0.038570 0.039470
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------