Add Hidden Markov Model Dwarf Sample

[andrey4latyshev]

Description

The Hidden Markov Model DPC++ Sample (Dwarf). Implements Viterbi algorithm using DPC++.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Sample Migration (Moving sample from old repository after completing checklist established)

How Has This Been Tested?

• Command Line
• oneapi-cli
• Visual Studio
• Eclipse IDE
• VSCode

Checklist for Moving samples:

Links and Details can be found in the samples WG Teams Files.

• Review sample design with domain reviewers https://github.com/oneapi-src/oneAPI-samples/wiki/Reviewers-and-Domain-Experts
• Implement coding guidelines and ensure code quality. (more polishing can be done)
• Adhere to sample.json specification. https://github.com/oneapi-src/oneAPI-samples/wiki/sample-json-specification
• Run jsonlint on sample.json to verify json syntax. www.jsonlint.com
• Adhere to readme template
• Ensure/create CI test configurations for sample (ciTests field) https://github.com/oneapi-src/oneAPI-samples/wiki/sample-json-ci-test-object
• Enforce format via clang-format config file
• Review DPC++ code with Paul Peterson. (GitHub User: pmpeter1)
• Review readme with Tom Lenth or Joe Oster. (GitHub User: JoeOster)
• Tested using Dev Cloud when applicable
• Implement fixes for ONSAM Jiras (n/a)
• If you have new dependencies/binaries, inform Samples Project Manager Swapna R Dontharaju (@srdontha) or @JoeOster

[pmpeter1]

Need to align with samples style guidelines (mostly naming).

Should use dpc_common.hpp for async exception.

Shorter names that 'accessor' should be used. More aligned with the algorithm's purpose. You usually try and make the body of your kernel most descriptive and then choose your accessor, buffer, and host object names to match.

Given this is primarily a set of linear stages, how do you propose to optimize this to get good performance on an offload accelerator?

[pmpeter1]

It should be "dpc_common.hpp" not <dpc_common.hpp> This is a shared/common header not a system header file.

[anjgola]

Status 8/20 am : Pending on changes requested from pmpeter1

[andrey4latyshev]

Given this is primarily a set of linear stages, how do you propose to optimize this to get good performance on an offload accelerator?

The acceleration comes from the sequential filling the massive matrices/buffers in parallel and the memory that is always on device.

[JoeOster]

approving to launch ci

[mkitez]

Triggering CI

[anjgola]

@praveenkk123 , you can merge PR if you approve them and CI passes.