GitLab

Was debugging/ trying to figure out why indexing was incorrect. Used a bunch of
prints and such.

These work in tandem to create a shape via the calculate strides() call. Seemed
to introduce more issues than fix, since we don't have access to resize()

Right now this is cleanup but I had used rev_partial_sum and the multiplies()
template operator created in algorithm to achieve this during debugging for
gather. The idea here would be we would statically create array() with calculate_strides()
to fix the empty stride dimensions.

Was added to implement calculate strides() for array creation via shape() but
needed to use forward iterators instead since we didn't implement reverse iterators.

Removing since it's not needed but could be used still if needed.

This was added during debugging when attempting to add in partial_sum()
to replicate device behavior when using calculate_strides(). In this case
this isn't needed

This is needed to get multi(i) below to work correctly when indexing.

Originally thought this was with the output_t.

Tried to get a proper templated shape of out_comp but right now this
seems to break as I can't just update the length of a shape and get a proper
output of the strides. Currently this breaks/asserts.

I think this is the cause of axis > 0 failing since we're not getting proper gathering
for the other axes as a result and get repeated rows with the wrong data.

Currently failing negative indices and negative axis tests.

All others "seem" to work

Noticed an oddball case that the cases that fail pass, if the sizes of a dimension
of a container is even instead of odd...

Add stride based multi-index similar to device functions. Between the device
gather and what's available for jit it looks like we're using lens instead of
strides to calculate indicies.

Seems to fix the 1d case of indices for this jit gather.

Pair programming with Paul

Taken from gatherND.cpp and modified so we include the axis parameter as opposed
to the batch_dims attribute.

Should always exist since we default this to zero when no axis is provided from
the instruction

Work in progress for the .hpp jit side.

HIP had change in previous rocm releases to use --offload-arch instead of --cuda-gpu-arch.

This should be backwards compatbile. hipRTC also supports --offload-arch.

* Add implicit_conversion

Need this for when we debug and use MIGRAPHX_TRACE_EVAL() to show tuples
Without this we break when reading our buffer due to the use of visit()
This came up as part of #1283 debugging.

Update dialect registration interface
Update 2nd build pipeline call and use full arch name

Merging #1391 caused an extra adjust allocation pass for GPU targets. This removes that merge error.

Disabling it untill int8 fix is in mainline from MIOpen and also so that QA tests could run migraphx-driver and unittests from MIGraphX.

Can be enabled via environment variable MIGRAPHX_ENABLE_NHWC

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