GitLab

Signed-off-by: Artur Wojcik <artur.wojcik@amd.com>

* Disable -Wunsafe-buffer-usage when compiling gpu code

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

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

Enable with MIGRAPHX_ENABLE_CK=1 and --exhaustive-tune tune flag

back out changes for rocm-5.5

* use half hip functions to compute max and min
* add verify test for min and max

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

This fixes #1700

Automatically fuse multiple reductions and pointwise operations.

Avoids double global loads.  Strided loops are unrolled which lets store results in array which compiler will use registers for since the index access is constant.   Updated to handle large reductions so which results with a better stable diffusion result

* deprecate HCC

Added CMakeFlag for hipRTC. MIGRAPHX_USE_HIPRTC.
Added stages in Jenkins for hipRTC.
Fixes for some of the pending issues from hipRTC.

* Use cosine to compute half sin

JIT implementation of the gather operator
Added a few more unit tests to this one as well since I saw some odd behavior during bring up.

* Add implicit_conversion

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

optimize the softmax operator

Implement operator for CPU and GPU implementations

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

* Implement concat using jit compilation

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

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

* Add softmax kernel

* Jit contiguous

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

Consolidate the vectorize and preload
Add vectorization to reduction
Co-authored-by: kahmed10 <15948690+kahmed10@users.noreply.github.com>

Remove std references in runtime compilation since these are not available when using hiprtc and the headers may not be available on the system