GitLab

Local Threads of multiples 32 were introduced in #1348
But LocalThreads that are not multiple of 64 are causing correctness issues.

Upgraded Dockerfiles and fixed tidy issues to make Ubuntu 20.04 and ROCm 5.3.0 the default 

updated GPU pad to now use JIT version.
added range functions for JIT kernels.

Fixes an observed regression error on certain Frozen Protobuf models due to PR 1280

Reiterate the assertion on the standard shape but relax it for the multibroadcast ops deliberately inserted to explicit the broadcast.

Refactor dynamic compute
- add a compute_output_shape object that implicitly converts to a new dyn_output or shape object
- dyn_output object can handle computing the static output shape of an operator given the input arguments shapes
  change an operator's compute function to argument compute(const dyn_output& dyn_out, std::vector<argument> args) to 
  use dyn_output object

Dynamic ref unary functions
-  Included these changes to have an example of the refactored dynamic compute being used
-  Changes to unary base class to handle dynamic shapes
-  Changed elu and leaky_relu to use unary base class and pointwise JIT

* use find2.0 for the convolution
Co-authored-by: Vasilii Filippov <DrizztDoUrden@users.noreply.github.com>
Co-authored-by: Chris Austen <causten@users.noreply.github.com>

* Enable non-standard shape
* Use perfdb for non xdlops
* Fix transpose+broadcast strides
Co-authored-by: jungpark-mlir <jungwook.park@amd.com>

Rewrites the TF batch norm like operators to other MIGX operators
Removes the code related to batch_norm_inference

Stream sync changes and associated API level changes

optimize the softmax operator

Adds two methods for the custom_ops virtual class.

bool runs_on_offload_target(), if the custom op runs directly on the gpu then it should be set to true. in this case, custom op expects its parameters to reside in GPU memory and writes output to the GPU memory. If it is set to false then, custom op expects it's parameter to reside on the host and puts back the result into the host memory.

output_alias, if output of the custom op is aliasing the input buffer. i.e. interpreting the same input buffer with differnet shape and strides.

Update as_vector() in C++ API to handle non-standard shapes. It required exposing element_index to space_index conversion method for the shape class.

Improvements/Additions to be made:

changes for the quant_convolution,
changes for the deconvolution,
Macros for MIOpen status checks

test_gpu_pack_int8_args fails on gfx908 machine, because it doesn't set compute_fp32 flag correctly. This PR fixes the test such that it checks for the device-name, and rocblas-versions and sets this flag accordingly.

Implement operator for CPU and GPU implementations

* Remove device functions
* Update tests

This PR allows for other values of epsilon to be matched when finding layernorm. Similarly, the calculation now uses the variable for epsilon.

Compute mean and variance in same reduction
Set block size to numbers divisible by 32 instead powers of 2
Global is also set exactly instead of being divisible by block size
More exact matching of global/local can help get rid of branching/loops
Reduce vectors first before doing dpp_reduce
Explicitly vectorize array operators since the compiler doesnt always vectorize them
Still uses old for loop when its computing at compile-time since the reinterpret_cast nor the all the vector types is supported

* fix typo for add_sigmoid

* Replaced `find_library` with `find_package` to locate MLIR static library
* Unified the include dir for headers and remove backward compatibility
* Embedded the external/include dir into the exported library

* Implement concat using jit compilation

Improves performance for 4/6 GEMMs used by huggingface BERT models with batch_size>1 by using a non-batched rocBLAS call for GEMMs where the B input has a broadcasted batch dimension.
The four verify tests added reflect the actual configurations used by bert-base-cased, with varied batch sizes.

Also adds a matcher to simplify_reshapes to move multibroadcasts after concats.

* Remove unused headers

* Fix accuracy bug when vectorizing slices

Using not and or improves readability. The cppcheck rule will help ensure we are doing it consistently.

Rewrite_gelu pass replaces the gelu formula of x * (1/2) * (1 + erf(x/sqrt(2))) with the sigmoid approximation of x * Sigmoid(x * 1.702)

* Track kernel time

This will rewrite dot operators like X(Y + b) to XY + Xb when b is constant as we can fold the add away.
This improves handling pointwise with broadcasted operators, this helps improves const propagation.
Improve gemm fusion with a mul_add
Improve support for broadcast shapes in gemm

Once
https://github.com/ROCmSoftwarePlatform/llvm-project-mlir/pull/690
lands, the ABI for MLIR-generated kernels will change. This commit
prepares MIGraphX for the change by conditionally selecting the new
ABI if MLIR reports a sufficiently high API version in its headers.

Currently, while copying a host buffer to the device, it first registers/maps the host buffer pointer to address space of the device.

If the host buffer has been allocated by the hipHostMalloc then, it is implicitly registered to the device's address space, and no need to register again. This PR adds a check for the same.

* Add in changes for onnx Mod operator

Initial operator for mod implementation and test cases for integer and floating based types.

Need to use fmod from stdlib for floating point types. half_float::half thankfully is specced to the use the existing std::fmod() call when looking at the half.hpp implementation.

fmod_flag should mirror the onnx fmod attribute. Right now using a floating point type without setting that on the user side to true will result in an exception.

Ref ticket #1283 

Changes to operator includes:

removed some includes that were not used
included argument.hpp where clang-tidy wanted it

* Only run __syncthreads when there is data to preload
* Improve loops
* Add const attribute to improve optimizations