GitLab

Reduce memory footprint by std::move of temporary (potentially very large) containers.
Minor cleanup for performance optimization: e.g. of Index() calculation -- which can get repeated millions of times in large tensors/vectors in a single Visit.

This PR constitutes the MIGraphX-side changes needed to not break the build in the presence of
ROCmSoftwarePlatform/rocMLIR#1136 , and updates what data is sent in to MLIR during the kernel generation and tuning process.

* Add initial tuning support

* Format

* Add extra param

* Format

* Use exauhstive flag

* Format

* Set expected shapes

* Format

* Format

* Fix missing symbol

* Format

* Add missing license header

* Format

* Update src/targets/gpu/include/migraphx/gpu/mlir.hpp

* Updates to add_embed_library

* Fix warnings for extern arrays

* Improve performance of pointwise/reduction kernels when using NHWC layouts

* Format

* Add nhwc test

* Format

* Remove inline namespace

* Add reduce test

* Remove rocblas_gemm_flags_pack_int8x4 in rocblas v3.0 or higher (#2002)

* Update src/targets/gpu/gemm_impl.cpp

HIP requires global work items in multiple of local work items. If it is not it is not guaranteed to generate correct results all the time.
Fixes #1977
Fixes #1644
MIGraphX CI has moved to rocm-5.6 which doesn't require hipRTC workarounds

Fixes #1946

* add support for rocm 5.6 in CI
* Disable anonymous namespace check
* add default c'tors to avoid warnings

This commit removes the build options to disable threading and removes the mutex in compile_mlir.
The commit being tested is a draft PR on rocMLIR that'll get merged if this passes

Allows the rocMLIR CI (which builds rocMLIR tip against MIGraphX tip) to pass.

Renames deconvolution -> convolution_backwards to be more consistent with the literature
Note: this is not the cross-correlation operator (which is the adjoint of convolution). This is technically a standard convolution operator combined with an upsampling operator rather than a downsampling operator.
Adds unit tests for the padding, strides, dilations, and other op attributes.
Throws on auto_pad attribute since it has not been implemented
Previously it read the attribute and set it but then did nothing with it
Extended for dynamic shapes
Does not support using asymmetric padding (padding_L != padding_R) and output_shape with dynamic shapes.

* do not use int8x4 format for the rocblas

Export API symbols for migraphx, migraphx_ref, migraphx_cpu, migrphx_gpu, migraphx_device, migraphx_tf, and migraphx_onnx. There is a separate PR for migrahx_c.

API symbol exporting affects only Windows. It is transparent on Linux.

This will also annotate the function with the block size so the compiler can do a better job of optimizing.

Needed to run multi-targeted program where "main" isn't the only root module. There could be many root modules other than main.

Add a CI job to test CK
Add MIGRAPHX_TUNE_CK env variable to only do tuning for CK
Continue tuning even when there is invalid configs
Fix a bug with parallel compilation not using all available threads
Add additional test for gemms using half types
Removed int32 as supported type since it doesnt pass our test suite

Bump MLIR commit to include latest supported pointwise ops.
Expand the MLIR approve list 
Ensure that operations such as tanh() that don't have integer implementations (at least in MLIR) aren't used within MLIR modules.
Add additional tests.

Co-authored-by: Artur Wojcik <artur.wojcik@amd.com>

Update `mlir_program` to only create one dialect registry, and to call
registerRocMLIRPasses() (which is needed and may not be thread-safe)
exactly once. 

In addition, use a single thread pool across all contexts. This is
recommended practice upstream for libraries that perform a lot of
compile jobs, and saves on the overhead of creating and destroying a
lot of threads

Add mlir quant_dot operator support