Commit 299690fe authored by Jan Hoeckesfeld's avatar Jan Hoeckesfeld
Browse files

journalupdate

parent 03ce7a67
......@@ -21,3 +21,29 @@ This results in a expectation value for an unobserved kmer:
```
E = sum over all si in Spatype: a^{mutated_bases} * (1-a)^{unmutated bases}
```
## 25.11.20 - 2.12.20
Implemented the ideas of the
previous week. The code is not yet optimized. There is still room for caching.
## 3.12.20 - 8.12.20
This week I ran the code on the HPC and did a lot of bug resolving. Previously the ckmertools took with 8 cores and 15 GB of RAM less then 30 mins.
With the new implementation features I decreased the cores to 4 increase the memory to 40GB and it still did not finish after 12 hours.
My suspicion is, that the cpp-functions that I added are all call-by-value. This means very large Objects need to be copyed each call.
## 9.12.20 - 15.12.20
I updated the c code to now use call by reference, but this did not fix the runtime. Changing the iterationset (in the config) to "OnV" (instead of "V") also did not work.
The program takes 33.88 GB on 4 cores and 28.88 on 3 cores. Im guessing this is an issue.
- Pre changes (067a47c2c0c5b949ba3b694298d1ba7721da3044): 3 cores with 12.67GB time: 9 min
- fix2: 3f81795f963cd032ef8f014d1c48e2d0369e6e53: still slow
- fix3: a71b298c25b2def340c399ac40ade4b6b551ec0e: fast
This might be the final cause:
```
Test test = (*sharedPtr.get());
&(test.getxy()) != &((*sharedPtr.get()).getxy())
```
Also the provided ckmer-dockerfile is broken (also in master). This may be caused by an "apt update". Right now I am pulling an older ckmer-version from dockerhub and updating that one in a new docker-file as a workarround.
This cannot be a longtime solution and the Dockerfile needs to be fixed.
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment