Skip to content

Add RFFT kernel to MicroMutableOpResolver. #3265

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.

Sign up for GitHub

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

Open
shlmregev wants to merge 5 commits into
base: main
Choose a base branch
from
Open

Add RFFT kernel to MicroMutableOpResolver. #3265

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
2679957
Add RFFT kernel to MicroMutableOpResolver.
shlmregev Dec 9, 2025
79ecdd8
Fix include order
shlmregev Dec 10, 2025
16b2490
Add IRFFT as well
shlmregev Dec 10, 2025
0fa5048
Undo changes that were committed by mistake
shlmregev Dec 10, 2025
d8007e2
Merge branch 'main' into add-rfft-to-resolver
shlmregev Dec 10, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions tensorflow/lite/micro/micro_mutable_op_resolver.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,8 @@ limitations under the License.
#include <cstdio>
#include <cstring>

#include "signal/micro/kernels/irfft.h"
#include "signal/micro/kernels/rfft.h"
Copy link
Collaborator

@veblush veblush Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think the standard approach here is to add the register functions to

tflite-micro/tensorflow/lite/micro/kernels/micro_ops.h

Lines 145 to 158 in 9cbf6e0

namespace tflm_signal {
TFLMRegistration* Register_DELAY();
TFLMRegistration* Register_FFT_AUTO_SCALE();
TFLMRegistration* Register_FILTER_BANK();
TFLMRegistration* Register_FILTER_BANK_LOG();
TFLMRegistration* Register_FILTER_BANK_SPECTRAL_SUBTRACTION();
TFLMRegistration* Register_FILTER_BANK_SQUARE_ROOT();
TFLMRegistration* Register_ENERGY();
TFLMRegistration* Register_FRAMER();
TFLMRegistration* Register_OVERLAP_ADD();
TFLMRegistration* Register_PCAN();
TFLMRegistration* Register_STACKER();
TFLMRegistration* Register_WINDOW();
} // namespace tflm_signal
instead of including new headers.

Copy link
Member Author

@shlmregev shlmregev Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The reason why I need to include this file, is so that I can do this:
AddRfft(::tflite::tflm_signal::Register_RFFT_INT16());

Should I add ::tflite::tflm_signal::Register_RFFT_INT16() to micro_ops.h?

micro_ops.h doesn't define any other Register_ function for specialized types. And looking at tensorflow/lite/micro/micro_mutable_op_resolver.h, there are multiple kernel headers included there, presumably for the same purpose.

Copy link
Member

@ddavis-2015 ddavis-2015 Dec 10, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@shlmregev @veblush Ah, I understand now. This is how CONV_2D, FULLY_CONNECTED, MUL, etc. expose their additional registrations as well. So micro_mutable_op_resolver.h is the right place to put the include. I'm thinking that signal/micro/kernels/irfft.h should also be included here, as it also has multiple registrations.

#include "tensorflow/lite/c/common.h"
#include "tensorflow/lite/core/api/flatbuffer_conversions.h"
#include "tensorflow/lite/kernels/internal/compatibility.h"
Expand Down
Loading