@ -30,9 +30,9 @@ Each directory has the name of the computational kernel routines described in th
In this test suite were considered only computational routines implemented by PSBLAS, according to the version 3.9 of the documentation. In the following table are reported all the kernels, their implementation and wheter or not they were tested yet.
|**Kernel**| **PSBLAS Subroutine**|**Description**|**Single Process Test**|**Multi-Process Test**|**Complex Test**|**GPU Test**|
|**General Dense Matrix Sum**| `psb_geaxpby`| This subroutine is an interface to the computational kernel for dense matrix sum: $Y \leftarrow \alpha X + \beta Y$ |Work in progress :hammer_and_wrench:|Work in progress :hammer_and_wrench:|No ❌|No ❌|
|**General Dense Matrix Sum**| `psb_geaxpby`| This subroutine is an interface to the computational kernel for dense matrix sum: $Y \leftarrow \alpha X + \beta Y$ |Yes ✅|Yes ✅|No ❌|No ❌|
| **Dot product**|`psb_gedot`|This function computes dot product between two vectors x and y. $dot \leftarrow x^T y$ If x and y are real vectors it computes dot-product as: $dot \leftarrow x^H y$ |Yes ✅|Yes ✅|No ❌|No ❌|
| **Generalized Dot Product** |`psb_gedots`|This subroutine computes a series of dot products among the columns of two dense matrices x and y:$res(i) \leftarrow x(:,i)^T y(:,i)$If the matrices are complex, then the usual convention applies, i.e. the conjugate transpose of x is used. If x and y are of rank one, then res is a scalar, else it is a rank one array.|No ❌|No ❌|No ❌|No ❌|
| **Generalized Dot Product** |`psb_gedots`|This subroutine computes a series of dot products among the columns of two dense matrices x and y:$res(i) \leftarrow x(:,i)^T y(:,i)$If the matrices are complex, then the usual convention applies, i.e. the conjugate transpose of x is used. If x and y are of rank one, then res is a scalar, else it is a rank one array.|No ❌|No ❌|No ❌|No ❌|
|**Infinity-Norm of Vector**|`psb_normi`/`psb_geamax`|This function computes the infinity-norm of a vector x. If x is a real vector it computes infinity norm as:$amax \leftarrow max \mid x_i \mid$else if x is a complex vector then it computes the infinity-norm as: $amax \leftarrow max(\mid re(x_i) \mid + \mid im(x_i) \mid)$ |No ❌|No ❌|No ❌|No ❌|
|**Generalized Infinity Norm**|`psb_geamaxs`|This subroutine computes a series of infinity norms on the columns of a dense matrix x: $res(i) \leftarrow max_k \mid x(k,i) \mid$ |No ❌|No ❌|No ❌|No ❌|
| **1-Norm of Vector**| `psb_norm1` / `psb_geasums`|This function computes the 1-norm of a vector x. If x is a real vector it computes 1-norm as: $asum \leftarrow \mid \mid x_i \mid \mid$ else if x is a complex vector then it computes 1-norm as: $asum \leftarrow \mid \mid re(x) \mid \mid_1 + \mid \mid im(x) \mid \mid_1$ |No ❌|No ❌|No ❌|No ❌|
@ -50,10 +50,6 @@ In this test suite were considered only computational routines implemented by PS
## Developer Notes
In order to keep compliant the excecution of the bash script used to automate the teest excecution, remember to create a new directory to put new tests and to use the name convention of psb_test_ signature for utilities functions and psb_kernel_test for tests used for new routines.
## TODO
- Finish the directories description
- Check memory occupancy of parallel/ serial/ and vectors/ directories (Maybe not the best way for lots of rputines?)
## References
<aid="testing">[1].</a> Higham, Nicholas J. Testing linear algebra software. Springer US, 1997
- Remove file generation in order to save up memory [OK]
- Clean main log at each run [OK] (Log aggregation was deleted)
- Fix log aggregation in main directory, see number of total tests [OK] (Log aggregation was deleted)
- Fix parallel and serial, using a fortran routine instead of the diff between files [OK]
TODO:
- Fix parallel and serial, using a fortran routine instead of the diff between files
- Force recompilation in main script (A flag should be added)
- Generate input vectors only if vectors/ directory doesn't exist to save up time (It is really dependant on the kernel analyzed, it is not always possible)
@ -6,7 +6,7 @@ This is a directory developed by Luca Pepè Sciarria and Simone Staccone froma T
Steps to reproduce the tests:
- Compile the code using ``` make ``` (Optional)
- Launch the script ./autotest.sh or with source ./autotest.sh if you want to add modules to the .bashrc file permenently.
- Check the output log file psblas_geaxpby_test.log to collect results
- Check the output log file psb_geaxpby_test.log to collect results
NOTE: If the code is changed and a new compilation is needed to show the changes, the autotest.sh script isn't aware of this scenario, therefore it is necessary to manually recompile the code.
@ -18,6 +18,8 @@ The ```psb_geaxpby```. The signature of the function is:
call psb_geaxpby(alpha, x, beta, y, desc_a, info)
```
The strategy to validate the correctness of the computation is to compare single precision result and double precision result in the test cases in which the test should not give an error. In this way it is possible to have a correctness check of the computation comparing the two results considering a number of significant digits which is tuned on the single precision computation.
### Parameters Values
**x** vectors are located in the vectors/ directory. They are generated randomly using the same seed and then saved on different files based on their characteristics. The size of the vector is choosen accordingly to the size of the matrix column space considered for the single test instance.
**$\alpha$** real coefficient multiplied by vector $x$
|$\alpha$|Value|Coefficients Description|
|:-:|:-:|:-:|
|$\alpha_1$|1.0|Positive value|
|$\alpha_2$|-1.0|Negative value|
|$\alpha_3$|0.0|Null value|
**$\beta$**
**$\beta$** real coefficient multiplied by vector $y$
|$\alpha$|Value|Coefficients Description|
|:-:|:-:|:-:|
|$\beta_1$|1.0|Positive value|
|$\beta_2$|-1.0|Negative value|
|$\beta_3$|0.0|Null value|
## Output
The ouput files generated by the test are automatically compared by the autotest.sh script, but if it is needed to manually run the test here it is the naming convenction used.
The results of the computation will be saved on different files based on the instance of the test considered. In particular the naming conventiona format the output file as sol_x#_y#_a#_b#.mtx, where each # is a number choosen w.r.t. the test instance. (Ex. sol_x1_y1_a1_b1.mtx is the solution computed using the first x vector file , the first y vector file, alpha = 1.0 and beta = 1.0). Moreover, the files will be saved in the serial/ directory if the program is launched using 1 process or in parrallel/ directory if the program is launched with more than one process.
## TODO
- Add computation with broken descriptor and catch the errore result
Stack-1
10 months ago