[gpu/enhance] Shared memory implementation for Blas kernels

[s-debadri]

Implementation of shared memory for Blas kernels.

Updates made:

• Added data pointer for clCreateBuffer calls using wrapper constructor
• Removed all WriteData calls
• Used CL_MEM_USE_HOST_PTR flag
• Used MapBuffer and UnMapBuffer wrappers to ensure cache consistency

Note: GPU can cache buffer data on device memory which might not get flushed to host after kernel execution. To enforce this MapBuffer or ReadData wrapper should be used.

Self evaluation:

1. Build test: [X]Passed [ ]Failed [ ]Skipped
2. Run test: [X]Passed [ ]Failed [ ]Skipped

Signed-off-by: Debadri Samaddar [email protected]

[taos-ci]

📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2784. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/.

[taos-ci]

@s-debadri, 💯 All CI checkers are successfully verified. Thanks.

[baek2sm]

LGTM.

I have a simple question. All of these codes are running within a do-while statement, but why is it necessary to use a do-while loop even though the current while condition is always false?

[s-debadri]

LGTM.

I have a simple question. All of these codes are running within a do-while statement, but why is it necessary to use a do-while loop even though the current while condition is always false?

We want the code to return at the point of failure and don't proceed any further. Hence, to avoid multiple nested if-else conditions we are using a do-while loop to make the code readable and clean.

[github-actions]

This PR is stale because it has been open 14 days with no activity. Remove stale label or comment or this will be closed in 3 days.