Age | Commit message (Collapse) | Author |
|
matrix of GEMM/GEMMLowp
Change-Id: I77f2bfcc5d170bcc2428a2f27104942c1ec877d7
Reviewed-on: https://review.mlplatform.org/375
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
matrix of GEMM/GEMMLowp
Change-Id: I8c5fd4c8bcdffda1522c83158981ed92baa045f4
Reviewed-on: https://review.mlplatform.org/364
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Change-Id: Ia7fb128e1f3944d0d831e1d125a6db3e1d257106
Reviewed-on: https://review.mlplatform.org/355
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Isabella Gottardi <isabella.gottardi@arm.com>
Reviewed-by: Anthony Barbier <Anthony.barbier@arm.com>
|
|
Change-Id: I8b59b9b94cbd132e1ff5157a4c59882719e12e3b
Reviewed-on: https://review.mlplatform.org/327
Reviewed-by: Anthony Barbier <Anthony.barbier@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Reduced the binary size of NEPoolingLayerKernel.o form 266k to 95K
Change-Id: Ia1e79849430a5f34f5c1fa3fb15f23a61555a7f0
Reviewed-on: https://review.mlplatform.org/344
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Change-Id: I827b26239043a9e90d26c2583122648d2a45303a
Reviewed-on: https://review.mlplatform.org/317
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Change-Id: Id0d4a07af24e2331161996083b0c1bab072bd405
Reviewed-on: https://review.mlplatform.org/322
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Change-Id: I9e6e43a5839d04c2e4b4552c05446efb0a5074cf
Reviewed-on: https://review.mlplatform.org/232
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Change-Id: Ic6a1f55f14d53896725afe426bc2e2acb1546589
Reviewed-on: https://review.mlplatform.org/343
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Change-Id: I13f6e4c600f39355f69e015409bf30dafdc5e3aa
Reviewed-on: https://review.mlplatform.org/332
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
|
|
Change-Id: Ie0d5387c0546045e14e62c84c03894a9b0339585
Reviewed-on: https://review.mlplatform.org/335
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Change-Id: I3de6bb33746d52f8d8c337ab7776eccee8c205fb
Reviewed-on: https://review.mlplatform.org/328
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Pablo Marquez <pablo.tello@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
NELSTM, NEFullyConnectedLayer(For quantised types only), NERNN and NEWinogradLayer were all defaulting to on-the-fly reshaping of B
Fixed a bug in GemmInterleaved: it was ignoring the 'multis' dimension of the tensor to allocate the memory for B reshaped
Change-Id: I7b30f7f57fc65d6a03cccde0bf5515a811f17b54
Reviewed-on: https://review.mlplatform.org/323
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Change-Id: Ice653e48211053bd3cd20a693bd76de6b4efc370
Reviewed-on: https://review.mlplatform.org/270
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Tested-by: Arm Jenkins <bsgcomp@arm.com>
|
|
Change-Id: I7eae2e55cc0b0b7bbebb7617299daaca6f75f40c
Reviewed-on: https://review.mlplatform.org/292
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
INESimpleFunctionNoBorder
Change-Id: Ia9fdc75b23e9a6208058f8406fb7b5fcd917de2c
Reviewed-on: https://review.mlplatform.org/311
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Michele Di Giorgio <michele.digiorgio@arm.com>
|
|
Adds support for Equal,NotEqual,Less,LessEqual,Greater,GreaterEqual
Change-Id: If0cdf4aae7f95c94709b195eee485f6663f45909
|
|
Change-Id: I2a18f0acea382960a8bc71a8f56928a5998f0dd6
|
|
Change-Id: Id74cc7ba8e5cabee6acd3798d4779f88b1f00a9b
|
|
Change-Id: I49b2e8b4200c9ed654736d9451e4ab9c073b4b10
|
|
Change-Id: I29e35024e29781a6b943b568abec9c73649215e6
|
|
Change-Id: I6ee2c0b670727fc808fa636c53ddfaec3a0036c9
|
|
Change-Id: I49f1d865f5e7562f1d80db849353a89ef77e6a9e
|
|
Output of Priorbox should be independent of the input
data layout and should always be in NCHW format
Change-Id: Ie80cd4e51c78945b158c0db1af1923bdf8d7ea7b
|
|
Fixes for:
- ReduceMean, reduction on the X axis for FP16 with 8 elements was
performed only up to a certain point. The fix now takes into account the
number of elements of the vector and does as many reductions as
necessary.
- YOLOLayer, activation for FP16 has to be performed on 32 bits until
the FP16 approximations is fixed.
Change-Id: I75373f4edd37de476e6fe1a56de3ef386b65c619
|
|
-Simplifies import memory interface
-Changes the used of void** handles with appropriate interfaces.
Change-Id: I5918c855c11f46352058864623336b352162a4b7
|
|
-Adds NHWC support for FP16
Change-Id: I61addf8efecf511ac8cd5f8aa9afc3e09c476aaf
|
|
Change-Id: I6e7dee8bd615a5eff01c523f208a218574ee5eab
|
|
kernels
Change-Id: I98183f95814442b6f3dbb67a1bdae99df05b9b01
|
|
Change-Id: I5d2ed5dcc342abff8124762f7bdee587cdf20032
|
|
- Fixing a bug for which we did not scale the boxes before transforming them
- Adding the correct_transform_coords option to BoundingBoxTransformInfo
Change-Id: I40281254bcf87e7c8583c119e99562414fe59822
|
|
Introduced F32 accumulation for F16 winograd gemm and output transform
WinogradConvolution will be available for F16 only if fast math flag is enabled
Change-Id: I215593c205236a0f9669218437bb40b184ec6a4f
|
|
Change-Id: I99e1c3939cfea4b9cb0ddfa313706f31b213ca89
|
|
AccessWindowRectangle::update_window_if_needed()
Change-Id: I56426cc9c9688a0aa0acdd439d5887c7ef208cd2
Note: The code to shrink the window hasn't been fixed yet.
|
|
Change-Id: I69e995973597ba3927d29e4f6ed5438560e53d77
|
|
Change-Id: Ib0798cc17496b7817f5b5769b25d98913a33a69d
|
|
Change-Id: I5bf5d751ec7c02d96c26a769f49d03ea23a248b7
|
|
Change-Id: Ibab049f09413258c99335b7da6b151530a1bd136
|
|
and 8 tensors (Part 1)
Creating special cases for concatening 2 and 4 tensors.
Change-Id: I6a739a494ae45011acb65369e353f9ef96970b90
|
|
NEGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint
Change-Id: I1d5bc4d24059917f9ddef0873dd3043b1f2320a8
|
|
Some systems don't have enough memory to run the VGG networks, for example
on systems with only 2GB memory the VGG example fails throwing a bad_alloc exception.
This patch introduces the concept of global memory policy in ACL, the policy
is a mechanism which could be used by the library's functions to try to reduce
memory consumption on systems with limited memory.
In this specific case the VGG examples set the policy to MINIMIZE. The GEMM
function checks if the policy is MINIMIZE and in this case does not use the
pretransposed weights path as this requires considerable more memory.
Change-Id: I53abc3c9c64d045d8306793ffc9d24b28e228b7b
|
|
Change-Id: If5be77602e37b14aea63d7ec6d8adab324628f04
|
|
CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFloat
Since we perform an element-wise operation, it is not necessary to pass the output_depth3d.
Change-Id: Ibfa07a0706e902acf59b444aa61e18a348162ea9
|
|
The issue was related to CLIm2Col when the number of input channels was less than
the number of elements processed by each thread.
The bug has been fixed in the validate_and_configure_window() function setting the correct number of elements accessed
in the output tensor.
Also fixed an issue GEMM3D when we have a single output channel
Change-Id: I094292d0c7662599c4a4c3916ec5f5821df5faef
|
|
Change-Id: I6d5f91579850906e1eb973ff6c5612195255e631
|
|
Change-Id: I5aae537372bf797fbb2a2bae81038f8963b041a9
|
|
Fixed a typo that caused compilation issues for ArmNN.
Change-Id: Iab22adaf163eb3d2978d264f0ecf1238de98a67e
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/156483
Reviewed-by: Francis Murtagh <francis.murtagh@arm.com>
Reviewed-by: Michele DiGiorgio <michele.digiorgio@arm.com>
Tested-by: bsgcomp <bsgcomp@arm.com>
|
|
OpenCL
COMPMID-1424 - Add dot product support for CLDepthwise QASYMM8 3x3 NHWC non-unit stride
With this patch we are able to improve the performance of MobileNet v1-qasymm8 by 37 %
Tried to use the dot product instruction in CLDepthwise QASYMM8 3x3 NHWC non-unit stride
but I have not seen any benefit (maybe because we have few arithemtic operation and we
do not have more load instructions). However Depthwise convolution has been improved by
30%
Change-Id: Id768a99c2e53a04276707e427af5d0ec93419ada
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/155082
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
|
|
Change-Id: Ic8312a5b6790aa7cd4468d42f08d557ad40e9441
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/154570
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
|
|
Change-Id: I91865506166951b3bf7f06a0b2d4cde925cfefb6
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/153447
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
|