GitLab

- Add output stream operators for LoopSched and PiplineVer

Pass pointer to whole workspace not the shifted one.

* adding files for F32 example

* adding functioning implementation with scalar multiplication and unary operator support

* added fp 16 type check in unary square

* updating scalar multiplication as an operator

* functioning version with scalar operator

* changing strides for col major

* updated column major implementation

* working column major implementation

* cleaned up comments, rearranged/renamed files

* small edits to 3d transpose profiler

* adding test/profiler/instance files for hipTensor permute unit test

* added more test instances

* cleaned up errors, randomized input tensor, added more instances

* turned off time printouts

* removed conflicting transpose profiler

* rearranged some files

* rename folder

* Add type string

* Remove typo

* Add deviceOp to backward x

* Add comment to describe the behavior of backward normalization

* Add kernel function, prepare to implement

* implement generic kernel

* Check vector size

* Add sweep once pipeline for small reduce size

* Fix bug of KRaw_ error

* Fix bug of dx stride

* sanity check for mean and rstd

* backward x for groupnorm

* Add bwd x instance

* add layernorm 2d bwd gamma beta instances

* Change save mean var type from f32 to f16 in f16 mode

* Change the example to f16

* Add groupnorm bwd gamma beta instance

* Add groupnorm bwd x instance

* Fix naming

* Add layernorm bwd x ckprofiler

* Add groupnorm bwd x profiler

* clang format

* Rename bwd x to bwd data

* Fix bug of verification in profiler

* Add test of layernorm and groupnorm bwd data

* Add missing cmake

* Add layernorm2d bwd data

* rename fwd example

* Add groupnorm client example

* Fix typo. replace Invarient with Invariant

* Add checking before running the best instance

This PR optimizes fp16 instances of direct load GEMM kernel introduced in #999 and #1052.

Measured the performance of new instances on CDNA2 GPU and compared it against the performance of the best non-direct-load GEMM instances. Used 76 different GEMM problems.
On average, this change improves the performance of the tested problems by 47%. For cases known as latency-bound, the speedup is around 126%.

* upgrade to rocm6.0 compiler

* move rocm6.0 from private to public repo

* switch to testing hipTensor mainline in CI

* Add files via upload

* fixed extra space typo

* add mi300 GPU architectures and rocm versions 5.6.1 and 6.0.0

---------
Co-authored-by: illsilin <Illia.Silin@amd.com>
Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>

* Add tensor structure to wrapper

* update changelog

* Fix names

* Comment fixes

Copied from the llvm-project LLVM_PARALLEL_*_JOBS

Concurrent linking can break the build as well as having too many
compile jobs for the avaiable memory.  These options allow the user
to fine tune the build to fit within their machines memory
constraints.

An example use on linux is
COMPILE_JOBS=`cat /proc/cpuinfo | grep -m 1 'cpu cores' | awk '{ print $4 }'`
if [ ${COMPILE_JOBS}x = x ]; then
  COMPILE_JOBS=1
fi
BUILD_MEM=4
MEM_KB=0
MEM_KB=`cat /proc/meminfo | grep MemTotal | awk '{ print $2 }'`
MEM_MB=`eval "expr ${MEM_KB} / 1024"`
MEM_GB=`eval "expr ${MEM_MB} / 1024"`
COMPILE_JOBS_MEM=`eval "expr 1 + ${MEM_GB} / ${BUILD_MEM}"`
if [ "$COMPILE_JOBS_MEM" -lt "$COMPILE_JOBS" ]; then
  COMPILE_JOBS=$COMPILE_JOBS_MEM
fi
LINK_MEM=32
LINK_JOBS=`eval "expr 1 + ${MEM_GB} / ${LINK_MEM}"`

cmake -G Ninja -DCK_PARALLEL_LINK_JOBS=$LINK_JOBS
               -DCK_PARALLEL_COMPILE_JOBS=$COMPILE_JOBS
Signed-off-by: Tom Rix <trix@redhat.com>

Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>

* fix Werror unused-parameter

* sync doc requirements

* fix blank space format

* fix dependency issue

* disabling some fp8 gemm instances to reduce build time

* disable fp8 gemm instances to reduce build time

* remove the unused variable

* build fp8 gemm default and padded instances separately

* fix include pathsc

Current implementation of IsSupported method in contraction ops does not cover a lot of possible cases in which ScalarPerVector cannot really be used to read A, B or D, or write E.

This PR extends both the regular and multiABD contraction ops with improved checks and also adds new instances with smaller values of ScalarPerVector to support instances that are not supported by other instances.

* Support broadcast for bias in grouped conv fwd

* Fix comment

* Comment fixes

* Remove GK layout

* switch from ROCmSoftwarePlatform to ROCm org

* replace ROCmSoftwarePlatform with ROCm in few more places

Co-authored-by: Jing Zhang <jizha@amd.com>
Co-authored-by: Illia Silin <98187287+illsilin@users.noreply.github.com>

Bumps [rocm-docs-core](https://github.com/RadeonOpenCompute/rocm-docs-core) from 0.29.0 to 0.30.1.
- [Release notes](https://github.com/RadeonOpenCompute/rocm-docs-core/releases)
- [Changelog](https://github.com/RadeonOpenCompute/rocm-docs-core/blob/develop/CHANGELOG.md)
- [Commits](https://github.com/RadeonOpenCompute/rocm-docs-core/compare/v0.29.0...v0.30.1

)

---
updated-dependencies:
- dependency-name: rocm-docs-core
  dependency-type: direct:production
  update-type: version-update:semver-minor
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* turn on -O3 compiler flag explicitly

* change cmake syntax for CI

* modify cmake line breaks in jenkinsfile

* prevent clearing c_thread_buffer between consecutive k-dim data tiles GEMM.
* Limit the number of launched thread blocks.

Enable passing in cmdline grouped gemm arguments.

* Introduce wrapper library

* Update cmake files

* Revert "Update cmake files"

This reverts commit c27f88b56590c11a88e26d5d0df7aca51a08133d.

* Fix comments

Relates to https://github.com/RadeonOpenCompute/rocm-docs-core/issues/330

* add daily build with mainline compiler

* fix the compiler paths for ci

* remove the -flto flag

* build with clang by default

This PR introduces support for double buffering in LDS into GEMM kernels that use direct load instructions.

Direct loads now use inline asm instead of intrinsics. Usage of intrinsics results in compiler adding additional waitcnt instructions what breaks possible load/compute overlap in case of double buffering.

Usage of inline asm results in the need to use sched_barrier in order to make sure that compiler cannot incorrectly reschedule instructions since it does not know the data dependencies between global->LDS and LDS->registers.

* StorePartials work on offseted pointer.
* Read flags as uint32_t value
* Accumulate partials only if there is more than one cooperating workgroup
* Add condition for waiting on reduction end, only when there is still work to do.
* Fix creation od a/b grid desc in CheckArgument.
* LaunchKernel will use preprocess lambda to set flags value to zero.
* Add condition in IsSupportedArgument to check if xdl is supported.

Update doc.

* fixed examples; add async_mem_set

* add stream to all deviceOp using SetWorkspace

---------
Co-authored-by: Jing Zhang <jizha@amd.com>

* Introduce wrapper for layout

* Extend functionality

* Fix for getLength

* Comment fixes

* Add comments and remove not needed getters