[CLML] Version compatibility and various test cases #13670
Conversation
Codegen verification test cases for all the ops (convolution, concat, pad, pool ..etc.) that are supported by clml BYOC path. Fix depthwise conv2d issue with layout
|
Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.
Generated by tvm-bot |
srkreddy1238
force-pushed
the
clml-base
branch
from
bc82a06
to
1b002e0
Compare
srkreddy1238
force-pushed
the
clml-base
branch
from
1b002e0
to
32ce3bf
Compare
srkreddy1238
changed the title
srkreddy1238
force-pushed
the
clml-base
branch
2 times, most recently
from
e84da6b
to
c82af19
Compare
srkreddy1238
force-pushed
the
clml-base
branch
from
c82af19
to
a85e777
Compare
| if(EXISTS ${CLML_VERSION_HEADER}) | ||
| file(READ ${CLML_VERSION_HEADER} ver) | ||
| string(REGEX MATCH "CL_QCOM_ML_OPS_H_MAJOR_VERSION ([0-9]*)" _ ${ver}) | ||
| set(CLML_VERSION_MAJOR ${CMAKE_MATCH_1}) |
There was a problem hiding this comment.
Is it possible situation that we build TVM with a new version of CLML but our target device doesn't support it? Maybe it is better to restrict version of CLML as it was done for OpenCL (we use OpenCL 1.2)?
There was a problem hiding this comment.
For codegen, OpenCL1.2 restriction is good as we don't have any requirement of using features from higher versions.
CLML has newer versions with new operators support and is essential to support to get perf improvements on new gerations.
| } | ||
| ICHECK(h_ClmlIntf != NULL) | ||
| << "clGetMLInterfaceVxQCOM:" << result | ||
| << " Perhaps there is mispatch between CLML SDK version to target supported version"; |
There was a problem hiding this comment.
Let's print target supported version
| # time_f = gen_module.module.time_evaluator("run", device.device.cl(0), number=1) | ||
| # cost = time_f().mean | ||
| # print("%g secs/iteration\n" % cost) |
There was a problem hiding this comment.
This affects CI time and also to get ideal numbers we need to set number to a higher value too.
For now, I am leaving it for developers to uncomment and use.
srkreddy1238
force-pushed
the
clml-base
branch
2 times, most recently
from
bd58646
to
a85e777
Compare
srkreddy1238
force-pushed
the
clml-base
branch
from
258f240
to
9469b05
Compare
* [CLML][TEST] Codegen test cases for ops Codegen verification test cases for all the ops (convolution, concat, pad, pool ..etc.) that are supported by clml BYOC path. Fix depthwise conv2d issue with layout * * lint errors * * version compatilibility changes. * * review comments * * Make the adreno container compatible w/ and w/o CLML SDK availability Co-authored-by: Siva Rama Krishna Reddy B <sivb@qti.qualcomm.com>
CLML version compatibility changes to support future versions.
Codegen verification test cases for all the ops (convolution, concat, pad, pool ..etc.) that are supported by clml BYOC path.
Fix depthwise conv2d issue with layout
Adreno container is now free from CLML SDK dependency. We can use it for texture only builds now.