[SYSTEMDS-2839] IntelMKL Integration Extension #1887
Conversation
…pen blas, img_translate implemented
There was a problem hiding this comment.
I like the PR, but there are some things to fix before we can merge it.
Have you verified if the automated tests actually verify behavior?
if not can you add this test case to the file: .github/workflows/javaTests.yml
simply add a new line on something like line 72.
There was a problem hiding this comment.
i would prefer if we do not throw these binaries out. They are there to avoid having to compile on new installs.
There was a problem hiding this comment.
Hi @Baunsgaard ,
the binaries are thrown out automatically when I run build.sh
There was a problem hiding this comment.
and build, does not replace them with new ones?
There was a problem hiding this comment.
CMake Error at CMakeLists.txt:56 (find_package):
By not providing "FindSEAL.cmake" in CMAKE_MODULE_PATH this project has
asked CMake to find a package configuration file provided by "SEAL", but
CMake did not find one.
Could not find a package configuration file provided by "SEAL" (requested
version 3.7) with any of the following names:
SEALConfig.cmake
seal-config.cmake
Add the installation prefix of "SEAL" to CMAKE_PREFIX_PATH or set
"SEAL_DIR" to a directory containing one of the above files. If "SEAL"
provides a separate development package or SDK, be sure it has been
installed.
I get these errors and the whole he folder inc. the flies inside it get removed after build
| // ------------------------------------------------------------------- | ||
| JNIEXPORT void JNICALL Java_org_apache_sysds_utils_NativeHelper_setMaxNumThreads | ||
| /*JNIEXPORT void JNICALL Java_org_apache_sysds_utils_NativeHelper_setMaxNumThreads |
There was a problem hiding this comment.
I understand that you could not make it work without commenting it out.
But can we do a second try? or in the PR at least not comment it out.
One alternative strategy is to compile a separate file systemds_img.cpp ? maybe.
| @@ -416,4 +416,21 @@ public static native boolean conv2dSparse(int apos, int alen, int[] aix, double[ | |||
| // different tradeoffs. In current implementation, we always use GetPrimitiveArrayCritical as it has proven to be | |||
| // fastest. We can revisit this decision later and hence I would not recommend removing this method. | |||
| private static native void setMaxNumThreads(int numThreads); | |||
|
|
|||
| //test interface for native binding | |||
| public static native void testNativeBindingWithDgemm(char transa, char transb, int m, int n, int k, double alpha, double[] A, int lda, double[] B, int ldb, double beta, double[] C, int ldc); | |||
There was a problem hiding this comment.
if you go with the separate cpp file idea, then we can move these lines to another file.
| } | ||
| System.out.println(); | ||
| }*/ | ||
| System.out.println("Execution time: " + executionTime + " ms"); |
There was a problem hiding this comment.
remove print statements in the tests.
|
Closed and continue on #1843 |
This is where I intend to explore if intel MKL -> oneApi is a viable update.