derek/TheBandwidthBenchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Merge branch 'master' of github.com:RRZE-HPC/TheBandwidthBenchmark

Browse Source
This commit is contained in:
Jan Eitzinger
2019-03-13 15:27:44 +01:00
parent bd9c83d34b 101efd0a3b
commit d87d668dcd
1 changed files with 22 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
README.md
View File

@@ -1,19 +1,20 @@
# The Bandwidth Benchmark # The Bandwidth Benchmark
This is a collection of simple streaming kernels for teaching purposes. This is a collection of simple streaming kernels for teaching purposes.
It is heavily inspired by John McCalpin's https://www.cs.virginia.edu/stream/. It is heavily inspired by John McCalpin's https://www.cs.virginia.edu/stream/ benchmark.
It contains the following streaming kernels with corresponding data access pattern (Notation: S - store, L - load, WA - write allocate): It contains the following streaming kernels with corresponding data access pattern (Notation: S - store, L - load, WA - write allocate). All variables are vectors, s is a scalar:
* init (S1, WA): Initilize an array. Store only. * init (S1, WA): Initilize an array: `a = s`. Store only.
* sum (L1): Vector reduction. Load only. * sum (L1): Vector reduction: `s += a`. Load only.
* copy (L1, S1, WA): Classic memcopy. * copy (L1, S1, WA): Classic memcopy: `a = b`.
* update (L1, S1): Update a vector. Also load + store but without write allocate. * update (L1, S1): Update vector: `a = a * scalar`. Also load + store but without write allocate.
* triad (L2, S1, WA): Stream triad - `a = b + b * scalar`. * triad (L2, S1, WA): Stream triad - `a = b + b * scalar`.
* daxpy (L2, S1): Daxpy - `a = a + b * scalar`. * daxpy (L2, S1): Daxpy - `a = a + b * scalar`.
* striad (L3, S1, WA): Schoenauer triad - `a = b + c * d`. * striad (L3, S1, WA): Schoenauer triad - `a = b + c * d`.
* sdaxpy (L3, S1): Schoenauer triad without write allocate - `a = a + b * c`. * sdaxpy (L3, S1): Schoenauer triad without write allocate - `a = a + b * c`.
As added benefit the code is a blueprint for a minimal benchmarking application with a generic makefile and modules for aligned array allocation, accurate timing and affinity settings. Those components can be used standalone in your own project.
## Build ## Build
@@ -27,7 +28,19 @@ TAG = GCC # Supported GCC, CLANG, ICC
OPENMP = -fopenmp OPENMP = -fopenmp
``` ```
3. Build with: 3. Adjust options set in config.mk:
```
OPTIONS = -DSIZE=40000000ull
OPTIONS += -DNTIMES=10
OPTIONS += -DARRAY_ALIGNMENT=64
#OPTIONS += -DVERBOSE_AFFINITY
#OPTIONS += -DVERBOSE_DATASIZE
#OPTIONS += -DVERBOSE_TIMER
```
The verbosity options enable detailed output about affinity settings, allocation sizes and timer resolution.
4. Build with:
``` ```
make make
``` ```
@@ -39,7 +52,7 @@ To output the executed commands use:
make Q= make Q=
``` ```
4. Clean up with: 5. Clean up with:
``` ```
make clean make clean
``` ```
@@ -50,7 +63,7 @@ make distclean
``` ```
to clean intermediate build results and binary. to clean intermediate build results and binary.
5. (Optional) Generate assembler: 6. (Optional) Generate assembler:
``` ```
make asm make asm
``` ```