[ Tensor ] Refactor blas/math related files into cpu backend considering arch-dep @open sesame 10/02 13:19 #2549

skykongkong8 · 2024-04-18T07:18:19Z

While during the process of implementing additional features in NEON, I found myself making unnecessary code blocks.
This is a suggestion-draft of refactorization for current blas/math related files.

DONE

build on x86 & unittest (w/ & w/o fp16)
build on android & unittest
for all fallbacks
Temporarily, cpu_backend.h itself is done by now.
Following step should be syncing this work with TensorV2 : [Tensor] Refactorize Tensor Class to TensorV2 @open sesame 03/26 12:26 #2500
Remove CBLAS params in Tensor to get rid of unnecessary dependency
Use valid params for functions like : sgemv, sgemm (StorageOrder, data addr, ... )
This is already done for tensor.cpp in this PR, while intentionally eliminating ALL TensorV2 related files
Replace all #include <blas_interface.h> to #include <cpu_backend.h>
Replace all #include <blas_neon.h> to #include <neon_impl.h> and manage fp16 functions in .cpp files
Include all paths for clean builds for Tizen

Final form of this PR would be like:

...
tensor
├─── cpu_backend
│   └─── cpu_backend.h (has all external functions from previous `blas_interface.h`)
│   └─── arm
│      └─── arm_compute_backend : copy of `cpu_backend.h`. selectively uses functions from `cblas_interface`, `neon_impl`, `fallback_internal`
│      └─── neon_impl_fp16 (For armv8.2+)
│      └─── neon_impl
│           ...
│   └─── x86
│      └─── x86_compute_backend : copy of `cpu_backend.h`. selectively uses functions from `cblas_interface`, `blas_avx`, `fallback_internal`
│      └─── blas_avx : AVX2_implemented (vcvt, cblas) 
│           ...
│   └─── fallback
│      └─── fallback ( !x86 & !arm ) : copy of `cpu_backend.h`. uses ALL functions from `fallback_internal`
│      └─── fallback_internal : all raw implementations without SIMD or external lib
│           ...
│   └─── cblas_interface
│      └─── cblas_interface : all cblas-related function interfaces, and params
...

and removing blas_interface.h

In short,

Substitute blas_interface.h to cpu_backend.h which has virtual functions of blas_interface.h
Actual implementations are implemented at arm_compute_backend, x86_compute_backend, and fallback, and they are included considering target cpu architecture. cblas.h is used for both of them for fp32 computation.
There are some differences (unsupported intrinsics, or dataTypes ) along the versions, and they are managed under each arm or x86 directory.

taos-ci · 2024-04-18T07:18:21Z

📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2549. 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 · 2024-04-18T07:37:51Z

cibot: @skykongkong8, A builder checker could not be completed because one of the checkers is not completed. In order to find out a reason, please go to http://ci.nnstreamer.ai/nntrainer/ci/repo-workers/pr-checker/2549-202404181618220.29115009307861-c94059c0dfb6b54f2021eb42c4aa56adbc256ee3/.

taos-ci · 2024-04-23T05:48:38Z

cibot: @skykongkong8, nntrainer/tensor/cpu_backend/x86/x86_compute_library.cpp does not include Doxygen tags such as @file @brief @author @bug. You must include the Doxygen tags in the source code. Please refer to a Doxygen manual at http://github.com/nnstreamer/TAOS-CI/blob/main/ci/doc/doxygen-documentation.md

taos-ci · 2024-04-23T05:48:39Z

cibot: @skykongkong8, The last line of a text file must have a newline character. Please append a new line at the end of the line in nntrainer/tensor/cpu_backend/fallback/cblas_fallback.h.

taos-ci