[onert-micro] Compare Operation list to TFLM

[chunseoklee]

latest status, see below : #10933 (comment)

Details

tflite-micro	onert-micro	remark
Abs	✅　#10946	@BalyshevArtem
Add	✅
AddN	✅　#10947	@BalyshevArtem
ArgMax	✅　#10948	@BalyshevArtem
ArgMin	✅ 　 #10948	@BalyshevArtem
AssignVariable	🛑	NOT SUPPORTED BY CIRCLE
AveragePool2D	✅ #10961	@BalyshevArtem
BatchToSpaceNd	✅ #11620	@BalyshevArtem
BroadcastArgs	🛑	NOT SUPPORTED BY CIRCLE
BroadcastTo	✅ #11633	@BalyshevArtem
CallOnce	🛑	NOT SUPPORTED BY CIRCLE
Cast	✅ #11632	@BalyshevArtem
Ceil	✅ #11627	@BalyshevArtem
CircularBuffer	🛑	NOT SUPPORTED BY CIRCLE
Concatenation	✅
Conv2D	✅
Cos	✅ #11673	@SlavikMIPT
CumSum	🛑	NOT SUPPORTED BY CIRCLE
DepthToSpace	✅ #11622	@BalyshevArtem
DepthwiseConv2D	✅ #11556	@BalyshevArtem
Dequantize	✅ #11619	@BalyshevArtem
DetectionPostProcess	🛑	NOT SUPPORTED BY CIRCLE
Div	✅　#10962	@BalyshevArtem
Elu	✅　 #10963	@BalyshevArtem
Equal	✅　 #10978	@BalyshevArtem
EthosU	🛑	NOT SUPPORTED BY CIRCLE
Exp	✅　 #10964	@BalyshevArtem
ExpandDims	✅
Fill	✅
Floor	✅ #11571	@SlavikMIPT
FloorDiv	✅ #11616	@SlavikMIPT
FloorMod	✅ #11617	@SlavikMIPT
FullyConnected	✅
Gather	✅
GattherND	✅　#11634	@BalyshevArtem
Greater	✅ 　#10978	@BalyshevArtem
GreaterEqual	✅　#10978	@BalyshevArtem
AddHardSwish	🛑	NOT SUPPORTED BY CIRCLE
If	#11957	@SlavikMIPT
L2Normalization	✅　 #11590	@BalyshevArtem
L2Pool2D	✅ #11636	@BalyshevArtem
LeakyRelu	✅　#10965	@BalyshevArtem
Less	✅
LessEqual	✅ #10978	@BalyshevArtem
Log	✅　 #11602	@BalyshevArtem
LogicalAnd	✅ #10998	@BalyshevArtem
LogicalNot	✅ #11662	@SlavikMIPT
LogicalOr	✅　#10998	@BalyshevArtem
Logistic	✅
LogSoftmax	✅ #11624	@BalyshevArtem
MaxPool2D	✅
Maximum	✅ #11628	@SlavikMIPT
Mean	✅ #11661	@BalyshevArtem
Minimum	✅ #11626	@SlavikMIPT
MirrorPad	✅ #11653	@BalyshevArtem
Mul	✅
Neg	✅　 #11045	@Torrero
NotEqual	✅　 #10978	@BalyshevArtem
Pack	✅
Pad	✅　 #10999	@BalyshevArtem
PadV2	✅　 #10999	@BalyshevArtem
Prelu	✅　 #11559	@BalyshevArtem
Quantize	✅ 　 #11618	@BalyshevArtem
ReadVariable	🛑	NOT SUPPORTED BY CIRCLE
Reduced_Prod	✅	not supported by TFLM
ReduceMax	✅ #11635	@BalyshevArtem
Relu	✅　#10965	@BalyshevArtem
Relu6	✅　 #10965	@BalyshevArtem
Reshape	✅
ResizeBilinear	✅ #11149	@Torrero
ResizeNearestNeighbor	✅ #11682	@BalyshevArtem
Round	✅#11625	@BalyshevArtem
Rsqrt	✅ #11576	@BalyshevArtem
SelectV2	✅ #11601	@BalyshevArtem
Shape	✅
Sin	✅ 　 #11672	@SlavikMIPT
Slice	✅
SoftMax	✅
SpaceToBatchNd	✅ #11621	@BalyshevArtem
SpaceToDepth	✅#11623	@BalyshevArtem
Split	✅
SplitV	✅
Sqrt	✅　#11586	@BalyshevArtem
Square	✅ 　#11671	@SlavikMIPT
SquaredDifference	✅ #11587	@BalyshevArtem
Squeeze		✅ #11696
StridedSlice	✅
sub	✅
sum	✅ 　#11655	@BalyshevArtem
Svdf	✅　 #11680	@BalyshevArtem
Tanh	✅
Transpose	✅
TransposeConv	✅　 #11567	@BalyshevArtem
UnidirectionalSequenceLSTM	✅
Unpack	✅ 　#11715	@BalyshevArtem
VarHandle	🛑	NOT SUPPORTED BY CIRCLE
While	✅
ZerosLike	✅ 　#11615	@BalyshevArtem

[chunseoklee]

@lemmaa I and @BalyshevArtem estimates that additional (98-25=73) ops can be supported within a month. Should we start this now ?

[lemmaa]

@chunseoklee , TizenRT 4.0 support is a higher priority for now. Perhaps top priority. Expansion of operator kernel coverage may start after TizenRT 4.0 porting is completed.