GitLab

Conversion works, just issues with predicate right now.

I may need additional checks for this, or to somehow find the matching division
by zero, and cause a dangling reference so this gets flagged correctly at compile time.
Current attempt inserts a divzero instruction that would later get picked up at the verify stage
during compile. Not sure if this is correct incase we run into operator collisions down the road

This reverts commit fcc84214.

Making things more maintainable by splitting the unit tests that shared the baseline program for validation which resulted in an & case that was a bit more cumbersome to debug.

Throw an exception when this occurs to indicate our simpliciation passes resulted in a singularity somewhere. Related to #1236

Simplify addition zero multiplication and divide operations. Added approrpiate test cases with returns and replacing the instruction and operand to just return zero.

Using the unit/neg unit matchers to handle subtraction operations in the same steps. Added unit tests for both cases.

Part of changes that go wtih #1236. Reverts -1 divide operations to a simple negation of the parameter

Added test case and code to simplify zero additions between paremeters and literals during simplifications.  In reference to issue #1236

Done to satisfy simplifications specified by #1236 .  Just replace every  parameter divided by 1 with itself. It's assumed that the eliminate_identity() pass will handle generated identity operators in our run_pass()

Save a multiply operation with that of a negation of  input parameter x. Suggested improvement via #1236

Original use case of having a literal 1, instead of any other number in simplify_mul_add, resulted in the find_unit_mult_const function to optimize away the literal 1 causing this test to fail. on the final check. Switched the constant to a non zero & one value, and now correctly passes.

Commit for the day, work in progress as I'm failing one of our unit tests outside of the change

This is an extension to insert_module_instructions, but instead of just inserting from a module, it can insert a range or a vector of instructions.

Allows PyTorch converted version of SSD-resnet34 to work

* Jit contiguous

Used to determine what files contain a license and are stamped. If not we exit and return an error code that can be later ingested by another script, as well as a list of the outstanding files in questions.

Currently baked in the list of files we should support or not support with licenses in them a well as some stuff to quickly ignore

Adds compute_method for the experimental custom ops.
Adds a test for the same using HIP APIs.
Depends on #1183
Solves #1101

Updated each source file in the repo with the existing license.

* [#935] Update tf_parser to have add_common_op() for parse_relu6

Similar to that of the onnx_parser.cpp add a add_common_op template and functionality to support clip based operations. This is done so clip operations can be guarenteed to have the same dimensions.

* fixup! [#935] Update tf_parser to have add_common_op() for parse_relu6

* fixup! fixup! [#935] Update tf_parser to have add_common_op() for parse_relu6

* fixup! fixup! fixup! [#935] Update tf_parser to have add_common_op() for parse_relu6

* fixup! fixup! fixup! fixup! [#935] Update tf_parser to have add_common_op() for parse_relu6

* Formatting

* fixup! Formatting
Co-authored-by: Umang Yadav <29876643+umangyadav@users.noreply.github.com>
Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>

* add allocate op header

* formatting

* add replace_allocate pass

* formatting

* move output param to remove_allocate pass

* formatting

* fix bugs in replace_allocate pass

* formatting

* fix verify if tests

* formatting

* move if op logic

* formatting

* cleanup lowering

* cleanup lowering

* formatting

* fix tidy

* formatting

* fix tidy

* add cpu allocate check

* formatting

* change cpu allocate in pass

* formatting

* add some tests for replace_allocate pass

* formatting

* pass by ref

* fix run_pass

* formatting

* update variable name for module

* update dce to use contains() and fix tidy

* formatting

* update cppcheck

* add if test

* formatting

* add if test

* rename var to mod_output_names

* formatting

* remove conditional

* update allocate op and tests

* formatting

* update replace_allocate tests

* update create_output_names() and conditional in replace_allocate

* formatting

* remove extra variable in replace_allocate

* update tools script for allocation_model
Co-authored-by: Umang Yadav <29876643+umangyadav@users.noreply.github.com>
Co-authored-by: Chris Austen <causten@users.noreply.github.com>
Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>

* Use custom distance function

* Pass module, skip order check if other module

* Change other valid()

* Remove unnecessary declaration

* test multiple module dependency

* Refactor to make more clear

* Code cleanup

* Simplify fix

* Test EXPECT
Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>

prioritizing int8 over int8x4 when it is applicable
Amend return to continue in apply loop
Adding error handling in case int8x4 compilation failed
Co-authored-by: Paul Fultz II <pfultz2@yahoo.com>

* Upgrade to cppcheck 2.8

* Improve applicable batched gemms for bert

As described in #1196, the ONNX mean parser does not work correctly for integral types. This update fixes the issue by handling integral types separately, where summation is performed before division. Additional test cases have also been added for handling integral types.

Fuse layernorm and added triadd_layernorm fusion.  This is a prep performance booster

Expose add_literal method in C/C++ api

Add compile tests for gpu math functions

Helps avoid dangling references. This also deprecates the constructors that didnt take a lifetime annotation since its ambiguous the lifetime.

Add ref and gpu implementations for ONNX op GatherND

Resolves #1032

* expose get_queue method

Implements the ReverseSequence ONNX operator as a parser.

This parser can only handle a constant sequence_lens input. This is the same as what is handled for TensorRT as far as I can tell.
We could handle a variable sequence_lens input; that would require ref and GPU implementations of the operator.
The ONNX backend tests are disabled because this does not handle variable sequence_lens.

Refactored the reference implementation of pooling to something like what was done for roialign. Moved the reference implementation of pooling from targets/ref/lowering.cpp to pooling.hpp.
Removed cpu_pooling, instead using reference pooling in pooling.hpp
Added reference implementation of Lp Norm pooling and the global version
Added tests for the Lp Norm Pooling

There is significant improvement on larger tensors with half almost 50% faster:

lens: [1024, 384, 768]
gpu::code_object[code_object=13832,symbol_name=kernel,global=39321600,local=256,]: 1.16685ms
gpu::reduce_sum[axes={2}]: 1.73126ms
Also for non-trivial layouts this can sometimes be over 2x faster:

lens: [64, 1024, 768, 4]
gpu::code_object[code_object=13832,symbol_name=kernel,global=39321600,local=256,]: 1.1706ms
gpu::reduce_sum[axes={1}]: 2.63375ms
Of course if the stride becomes larger this speed improvement diminishes due to poor memory access patterns. A lane_reduce instead of a block_reduce is needed for such type of kernels. I plan to address that in a future PR.

Finally, this also includes a MIGRAPHX_GPU_DUMP_ASM env variable which will print out the assembly when the kernel compiles.