STREAM and Comparison to Spatter

Supposedly the max BW for this part is ~119 GB/s so PRS seems a bit closer in this sense.

#Hive STREAM and nstream results #Xeon Gold 6226 Cascade Lake

Parallel Research Kernels version 2.17
OpenMP stream triad: A = B + scalar*C
Number of threads    = 24
Vector length        = 126000000
Offset               = 0
Number of iterations = 100
Allocation type      = dynamic
Solution validates
Rate (MB/s): 94358.455000 Avg time (s): 0.042731

STREAM results

-------------------------------------------------------------
Number of Threads requested = 24
Number of Threads counted = 24
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 21533 microseconds.
   (= 21533 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:           67576.0     0.031212     0.029833     0.034071
Scale:          67717.2     0.031358     0.029771     0.034963
Add:            76375.3     0.040747     0.039594     0.046324
Triad:          76309.6     0.040726     0.039628     0.043491
-------------------------------------------------------------
Solution Validates: avg error less than 1.000000e-13 on all three arrays
-------------------------------------------------------------

Recompiling STREAM using nstream compiler settings. Note that you will get closer to the theoretical "peak", although it's unclear whether you can report these numbers to McCalpin's STREAM database.

make
icc -O3 -pthread -O3 -qopenmp -DVERBOSE=0 -DMAXTHREADS=24 -DRESTRICT_KEYWORD=0 -DSTREAM_ARRAY_SIZE=20000000 -DNTIMES=100   -c -o mysecond.o mysecond.c
icc -O3 -pthread -O3 -qopenmp -DVERBOSE=0 -DMAXTHREADS=24 -DRESTRICT_KEYWORD=0 -DSTREAM_ARRAY_SIZE=20000000 -DNTIMES=100 -c mysecond.c
gfortran -O3 -fopenmp -c stream.f
gfortran -O3 -fopenmp stream.o mysecond.o -o stream_f.exe
icc -O3 -pthread -O3 -qopenmp -DVERBOSE=0 -DMAXTHREADS=24 -DRESTRICT_KEYWORD=0 -DSTREAM_ARRAY_SIZE=20000000 -DNTIMES=100 stream.c -o stream_c.exe
[jyoung9@atl1-1-01-011-10-l STREAM]$ ./stream_c.exe
-------------------------------------------------------------
STREAM version $Revision: 5.10 $
-------------------------------------------------------------
This system uses 8 bytes per array element.
-------------------------------------------------------------
Array size = 20000000 (elements), Offset = 0 (elements)
Memory per array = 152.6 MiB (= 0.1 GiB).
Total memory required = 457.8 MiB (= 0.4 GiB).
Each kernel will be executed 100 times.
 The *best* time for each kernel (excluding the first iteration)
 will be used to compute the reported bandwidth.
-------------------------------------------------------------
Number of Threads requested = 24
Number of Threads counted = 24
-------------------------------------------------------------
Your clock granularity/precision appears to be 1 microseconds.
Each test below will take on the order of 9058 microseconds.
   (= 9058 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:           81531.8     0.005402     0.003925     0.014771
Scale:          89062.9     0.005237     0.003593     0.014981
Add:            90978.6     0.006959     0.005276     0.019589
Triad:          91955.1     0.006990     0.005220     0.019815
-------------------------------------------------------------

System-specific STREAM Numbers

KNL repo for STREAM - run make stream.icc to get the right ICC settings. https://github.com/jeffhammond/STREAM/tree/knl