[ blas/OpenCL ] Added multiply OpenCL kernel and unit test #2630
Conversation
s-debadri
requested review from
myungjoo,
jijoongmoon,
again4you,
jaeyun-jung,
leemgs,
wooksong,
helloahn,
kparichay,
gichan-jang,
anyj0527,
zhoonit and
lhs8928
as code owners
|
📝 TAOS-CI Version: 1.5.20200925. Thank you for submitting PR #2630. Please a submit 1commit/1PR (one commit per one PR) policy to get comments quickly from reviewers. Your PR must pass all verificiation processes of cibot before starting a review process from reviewers. If you are new member to join this project, please read manuals in documentation folder and wiki page. In order to monitor a progress status of your PR in more detail, visit http://ci.nnstreamer.ai/. |
s-debadri
requested review from
songgot,
jihochu,
DonghakPark,
SeoHyungjun,
baek2sm,
skykongkong8,
djeong20,
EunjuYang and
a team
as code owners
There was a problem hiding this comment.
@s-debadri, 💯 All CI checkers are successfully verified. Thanks.
s-debadri
force-pushed
the
blas_kernel_enhance
branch
from
9b1e33b to
8baf1c6
Compare
There was a problem hiding this comment.
@s-debadri, 💯 All CI checkers are successfully verified. Thanks.
s-debadri
changed the title
| @@ -662,6 +662,10 @@ std::string RunLayerContext::getKernelName(LayerKernel layerKernel) { | |||
| return "dot_cl_fp16"; | |||
| case LayerKernel::SGEMM_FP16: | |||
| return "sgemm_cl_fp16"; | |||
| case LayerKernel::SSCAL: | |||
| return "sscal_cl"; | |||
| case LayerKernel::SSCAL_FP16: | |||
There was a problem hiding this comment.
please keep in mind that sscal_fp16 should later be hscal!
| * @brief Process data and dimensions for OpenCL dot operation | ||
| * @param[in] input Tensor | ||
| * @param[in] m Tensor | ||
| * @param[in] result Tensor |
There was a problem hiding this comment.
| * @param[in] result Tensor |
There was a problem hiding this comment.
Updated in latest commit.
|
|
||
| /** | ||
| * @brief fp16 sscal value element by element immediately | ||
| * @param[in] X float * input |
There was a problem hiding this comment.
| * @param[in] X float * input | |
| * @param[in] X __fp16 * input |
There was a problem hiding this comment.
Updated in latest commit.
There was a problem hiding this comment.
@s-debadri, 💯 All CI checkers are successfully verified. Thanks.
| @@ -662,6 +662,10 @@ std::string RunLayerContext::getKernelName(LayerKernel layerKernel) { | |||
| return "dot_cl_fp16"; | |||
| case LayerKernel::SGEMM_FP16: | |||
| return "sgemm_cl_fp16"; | |||
| case LayerKernel::SSCAL: | |||
There was a problem hiding this comment.
This cl kernel information should be handled in cl context. Otherwise, we cannot support the custom cl layer. We do need CLLernel Class and it should be managed in Cl_Context. and if the layer is created, then N Cl kernel should be mapped to the layer object. and also, if there is a custom cl layer, the cl kernel must be registered to the cl context. I guess the basic cl kernels for dot, sgemm, etc. could be managed as an basic cl math kernels we provide and CLKernel has the list of them.
There was a problem hiding this comment.
As discussed, this will be done in future PRs.
| @@ -30,6 +30,14 @@ void dotBatchedCl(Tensor const &input, Tensor const &m, Tensor &result, | |||
| } | |||
| } | |||
|
|
|||
| Tensor dotCl(Tensor const &input, Tensor const &m, RunLayerContext &context, | |||
There was a problem hiding this comment.
RunLayerContext should be used as a data provider. It shouldn't have Kernel related information.
This dotCL level could be the CLKernel Class. I guess. and managed in clAppContext.
There was a problem hiding this comment.
As discussed, this will be done in future PRs.
Added sscal equivalent kernel and multiply function. Added unit test setup to test standalone kernels. Signed-off-by: Debadri Samaddar <[email protected]>
Modified doxygen docs as required Signed-off-by: Debadri Samaddar <[email protected]>
s-debadri
force-pushed
the
blas_kernel_enhance
branch
from
cb9c323 to
bce13bd
Compare
There was a problem hiding this comment.
@s-debadri, 💯 All CI checkers are successfully verified. Thanks.
sscalequivalent kernel added for multiply.unittest_blas_kernels_cl.cppfor easily testing OpenCL kernels.dotCl.Self evaluation:
Signed-off-by: Debadri Samaddar [email protected]