Async DMA support for Transpose Data Layout

[ramjana]

Support to move directly to LDS support for Transpose Data layout Added feature constraints for GLVW>DWORD

[nakajee]

I quickly ran DGEMM NN cases. It fails with GlobalReadVectorWidth=1 and VectorWidth=2 and ThreadSeparateGlobalReadB=0.

[nakajee]

Thanks for your update. With your latest change, dgemm NN + ThreadSeparateGlobalReadB tests passed, but dgemm TN + ThreadSeparateGlobalReadA still failed.

[nakajee]

Also, please merge the latest change in develop branch.

[nakajee]

Thank you for your update. Please add some test cases for ThreadSeparateGlobalReadA as well.

[nakajee]

Your test case failed in CI test. I think you need to add the following line as other precheckin tests to avoid failing with other architectures.

TestParameters: marks: [skip-gfx900, skip-gfx906, skip-gfx908, skip-gfx1010, skip-gfx1011, skip-gfx1012, skip-gfx1030] # not supported by arch

[nakajee]

Thanks for your update. Please resolve the conflicts so that CI tests can run. By the way, it seems like you have not added Test parameters to skip unsupported architectures (skip-gfx???). Some of CI tests on unsupported arch will fail without this line.

[nakajee]

CI test failed again. Please update your test case. WIthout skip, it will not pass.

[nakajee]

Move to #1574