Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

F16 Support? #3784

Closed
4 of 11 tasks
myungjoo opened this issue Jun 8, 2022 · 1 comment
Closed
4 of 11 tasks

F16 Support? #3784

myungjoo opened this issue Jun 8, 2022 · 1 comment
Assignees
Labels
low priority Low Priority / Hobby Projects

Comments

@myungjoo
Copy link
Member

myungjoo commented Jun 8, 2022

There are a few frameworks with float16 support.
We may add one more tensor type for f16.

PR #3794

TODO

  • tensor-transform
  • unit test cases
  • actual tests (arm64/arm32)
  • API updates
  • More TCs with calculation result verification & data size verification.

Issue for later development

  • x64 tests (cannot build) (GCC12 required. postpone it)
  • x86_32 tests (cannot build) (GCC12 required. postpone it)
  • Enable ORC or SIMD w/ tensor-transform
  • More unit test cases (transpose, ...)
  • x86/x64 compatibility
  • Should we support BF16, too?
@myungjoo myungjoo added the low priority Low Priority / Hobby Projects label Jun 8, 2022
@taos-ci
Copy link
Collaborator

taos-ci commented Jun 8, 2022

:octocat: cibot: Thank you for posting issue #3784. The person in charge will reply soon.

myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 17, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 21, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 21, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 22, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 22, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jun 22, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
@myungjoo myungjoo self-assigned this Jun 23, 2022
myungjoo added a commit to myungjoo/nnstreamer that referenced this issue Jul 1, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed nnstreamer#3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
myungjoo added a commit that referenced this issue Jul 3, 2022
Add FLOAT16 type.
Fortunately, gcc for x64, aarch64, and arm has extension for
float16.

@todo: update meson script so that arch-specific compiler
options for float16 can be added if float16 is enabled.

Addressed #3784

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
@myungjoo myungjoo added this to the NNStreamer-23-PoC milestone Sep 19, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
low priority Low Priority / Hobby Projects
Projects
None yet
Development

No branches or pull requests

2 participants