diff --git a/doc.zih.tu-dresden.de/docs/software/cfd.md b/doc.zih.tu-dresden.de/docs/software/cfd.md
index 492cb96d24f3761e2820fdba34eaa6b0a35db320..186d7b3a5a97a2daf06d8618c7c91dc91d7ab971 100644
--- a/doc.zih.tu-dresden.de/docs/software/cfd.md
+++ b/doc.zih.tu-dresden.de/docs/software/cfd.md
@@ -42,7 +42,7 @@ marie@login$ # source $FOAM_CSH
module load OpenFOAM
source $FOAM_BASH
cd /scratch/ws/1/marie-example-workspace # work directory using workspace
- srun pimpleFoam -parallel > "$OUTFILE"
+ srun pimpleFoam -parallel > "$OUTFILE"
```
## Ansys CFX
@@ -62,7 +62,7 @@ geometry and mesh generator cfx5pre, and the post-processor cfx5post.
module load ANSYS
cd /scratch/ws/1/marie-example-workspace # work directory using workspace
- cfx-parallel.sh -double -def StaticMixer.def
+ cfx-parallel.sh -double -def StaticMixer.def
```
## Ansys Fluent
diff --git a/doc.zih.tu-dresden.de/docs/software/data_analytics_with_python.md b/doc.zih.tu-dresden.de/docs/software/data_analytics_with_python.md
index bc9ac622530f2b355adef7337fb5d49447d79be1..00ce0c5c4c3ddbd3654161bab69ee0a493cb4350 100644
--- a/doc.zih.tu-dresden.de/docs/software/data_analytics_with_python.md
+++ b/doc.zih.tu-dresden.de/docs/software/data_analytics_with_python.md
@@ -212,11 +212,11 @@ for the partition `alpha` (queue at the dask terms) on the ZIH system:
```python
from dask_jobqueue import SLURMCluster
-cluster = SLURMCluster(queue='alpha',
+cluster = SLURMCluster(queue='alpha',
cores=8,
- processes=2,
- project='p_marie',
- memory="8GB",
+ processes=2,
+ project='p_marie',
+ memory="8GB",
walltime="00:30:00")
```
@@ -235,15 +235,15 @@ from distributed import Client
from dask_jobqueue import SLURMCluster
from dask import delayed
-cluster = SLURMCluster(queue='alpha',
+cluster = SLURMCluster(queue='alpha',
cores=8,
- processes=2,
- project='p_marie',
- memory="80GB",
+ processes=2,
+ project='p_marie',
+ memory="80GB",
walltime="00:30:00",
extra=['--resources gpu=1'])
-cluster.scale(2) #scale it to 2 workers!
+cluster.scale(2) #scale it to 2 workers!
client = Client(cluster) #command will show you number of workers (python objects corresponds to jobs)
```
@@ -288,7 +288,7 @@ for the Monte-Carlo estimation of Pi.
uid = int( sp.check_output('id -u', shell=True).decode('utf-8').replace('\n','') )
portdash = 10001 + uid
- #create a Slurm cluster, please specify your project
+ #create a Slurm cluster, please specify your project
cluster = SLURMCluster(queue='alpha', cores=2, project='p_marie', memory="8GB", walltime="00:30:00", extra=['--resources gpu=1'], scheduler_options={"dashboard_address": f":{portdash}"})
@@ -309,12 +309,12 @@ for the Monte-Carlo estimation of Pi.
def calc_pi_mc(size_in_bytes, chunksize_in_bytes=200e6):
"""Calculate PI using a Monte Carlo estimate."""
-
+
size = int(size_in_bytes / 8)
chunksize = int(chunksize_in_bytes / 8)
-
+
xy = da.random.uniform(0, 1, size=(size / 2, 2), chunks=(chunksize / 2, 2))
-
+
in_circle = ((xy ** 2).sum(axis=-1) < 1)
pi = 4 * in_circle.mean()
@@ -327,11 +327,11 @@ for the Monte-Carlo estimation of Pi.
f"\tErr: {abs(pi - np.pi) : 10.3e}\n"
f"\tWorkers: {num_workers}"
f"\t\tTime: {time_delta : 7.3f}s")
-
+
#let's loop over different volumes of double-precision random numbers and estimate it
for size in (1e9 * n for n in (1, 10, 100)):
-
+
start = time()
pi = calc_pi_mc(size).compute()
elaps = time() - start
@@ -339,7 +339,7 @@ for the Monte-Carlo estimation of Pi.
print_pi_stats(size, pi, time_delta=elaps, num_workers=len(cluster.scheduler.workers))
#Scaling the Cluster to twice its size and re-run the experiments
-
+
new_num_workers = 2 * len(cluster.scheduler.workers)
print(f"Scaling from {len(cluster.scheduler.workers)} to {new_num_workers} workers.")
@@ -349,11 +349,11 @@ for the Monte-Carlo estimation of Pi.
sleep(120)
client
-
+
#Re-run same experiments with doubled cluster
- for size in (1e9 * n for n in (1, 10, 100)):
-
+ for size in (1e9 * n for n in (1, 10, 100)):
+
start = time()
pi = calc_pi_mc(size).compute()
elaps = time() - start
diff --git a/doc.zih.tu-dresden.de/docs/software/distributed_training.md b/doc.zih.tu-dresden.de/docs/software/distributed_training.md
index bd45768f67c862b2a0137bd2a1656723fa6dfd91..1008e33f6a60ba3b4b189deeae2d0f2b14066ffd 100644
--- a/doc.zih.tu-dresden.de/docs/software/distributed_training.md
+++ b/doc.zih.tu-dresden.de/docs/software/distributed_training.md
@@ -183,7 +183,7 @@ DDP uses collective communications in the
[torch.distributed](https://pytorch.org/tutorials/intermediate/dist_tuto.html) package to
synchronize gradients and buffers.
-The tutorial can be found [here](https://pytorch.org/tutorials/intermediate/ddp_tutorial.html).
+Please also look at the [official tutorial](https://pytorch.org/tutorials/intermediate/ddp_tutorial.html).
To use distributed data parallelism on ZIH systems, please make sure the `--ntasks-per-node`
parameter is equal to the number of GPUs you use per node.
@@ -234,7 +234,7 @@ marie@compute$ module spider Horovod # Check available modules
Horovod/0.19.5-fosscuda-2019b-TensorFlow-2.2.0-Python-3.7.4
Horovod/0.21.1-TensorFlow-2.4.1
[...]
-marie@compute$ module load Horovod/0.19.5-fosscuda-2019b-TensorFlow-2.2.0-Python-3.7.4
+marie@compute$ module load Horovod/0.19.5-fosscuda-2019b-TensorFlow-2.2.0-Python-3.7.4
```
Or if you want to use Horovod on the partition `alpha`, you can load it with the dependencies:
diff --git a/doc.zih.tu-dresden.de/docs/software/ngc_containers.md b/doc.zih.tu-dresden.de/docs/software/ngc_containers.md
index 835259ce9d6ff5bb48912911f5f02bae7d449596..f19612d9a3310f869a483c20328d51168317552a 100644
--- a/doc.zih.tu-dresden.de/docs/software/ngc_containers.md
+++ b/doc.zih.tu-dresden.de/docs/software/ngc_containers.md
@@ -53,7 +53,7 @@ Create a container from the image from the NGC catalog.
(For this example, the alpha is used):
```console
-marie@login$ srun --partition=alpha --nodes=1 --ntasks-per-node=1 --ntasks=1 --gres=gpu:1 --time=08:00:00 --pty --mem=50000 bash
+marie@login$ srun --partition=alpha --nodes=1 --ntasks-per-node=1 --ntasks=1 --gres=gpu:1 --time=08:00:00 --pty --mem=50000 bash
marie@compute$ cd /scratch/ws/<name_of_your_workspace>/containers #please create a Workspace
diff --git a/doc.zih.tu-dresden.de/docs/software/perf_tools.md b/doc.zih.tu-dresden.de/docs/software/perf_tools.md
index 16007698726b0430f84ef20acc80cb9e1766d64d..83398f49cb68a3255e051ae866a3679124559bef 100644
--- a/doc.zih.tu-dresden.de/docs/software/perf_tools.md
+++ b/doc.zih.tu-dresden.de/docs/software/perf_tools.md
@@ -1,8 +1,8 @@
# Introduction
`perf` consists of two parts: the kernel space implementation and the userland tools. This wiki
-entry focusses on the latter. These tools are installed on taurus, and others and provides support
-for sampling applications and reading performance counters.
+entry focusses on the latter. These tools are installed on ZIH systems, and others and provides
+support for sampling applications and reading performance counters.
## Configuration
@@ -34,18 +34,18 @@ Run `perf stat <Your application>`. This will provide you with a general
overview on some counters.
```Bash
-Performance counter stats for 'ls':=
- 2,524235 task-clock # 0,352 CPUs utilized
- 15 context-switches # 0,006 M/sec
- 0 CPU-migrations # 0,000 M/sec
- 292 page-faults # 0,116 M/sec
- 6.431.241 cycles # 2,548 GHz
- 3.537.620 stalled-cycles-frontend # 55,01% frontend cycles idle
- 2.634.293 stalled-cycles-backend # 40,96% backend cycles idle
- 6.157.440 instructions # 0,96 insns per cycle
- # 0,57 stalled cycles per insn
- 1.248.527 branches # 494,616 M/sec
- 34.044 branch-misses # 2,73% of all branches
+Performance counter stats for 'ls':=
+ 2,524235 task-clock # 0,352 CPUs utilized
+ 15 context-switches # 0,006 M/sec
+ 0 CPU-migrations # 0,000 M/sec
+ 292 page-faults # 0,116 M/sec
+ 6.431.241 cycles # 2,548 GHz
+ 3.537.620 stalled-cycles-frontend # 55,01% frontend cycles idle
+ 2.634.293 stalled-cycles-backend # 40,96% backend cycles idle
+ 6.157.440 instructions # 0,96 insns per cycle
+ # 0,57 stalled cycles per insn
+ 1.248.527 branches # 494,616 M/sec
+ 34.044 branch-misses # 2,73% of all branches
0,007167707 seconds time elapsed
```
@@ -142,10 +142,10 @@ If you added a callchain, it also gives you a callchain profile.\<br /> \*Discla
not an appropriate way to gain exact numbers. So this is merely a rough overview and not guaranteed
to be absolutely correct.\*\<span style="font-size: 1em;"> \</span>
-### On Taurus
+### On ZIH systems
-On Taurus, users are not allowed to see the kernel functions. If you have multiple events defined,
-then the first thing you select in `perf report` is the type of event. Press right
+On ZIH systems, users are not allowed to see the kernel functions. If you have multiple events
+defined, then the first thing you select in `perf report` is the type of event. Press right
```Bash
Available samples
@@ -165,7 +165,7 @@ If you'd select cycles, you would get such a screen:
```Bash
Events: 96 cycles
+ 49,13% test_gcc_perf test_gcc_perf [.] main.omp_fn.0
-+ 34,48% test_gcc_perf test_gcc_perf [.]
++ 34,48% test_gcc_perf test_gcc_perf [.]
+ 6,92% test_gcc_perf test_gcc_perf [.] omp_get_thread_num@plt
+ 5,20% test_gcc_perf libgomp.so.1.0.0 [.] omp_get_thread_num
+ 2,25% test_gcc_perf test_gcc_perf [.] main.omp_fn.1