diff options
| author | Georgios Pinitas <georgios.pinitas@arm.com> | 2017-12-13 12:48:44 +0000 |
|---|---|---|
| committer | Anthony Barbier <anthony.barbier@arm.com> | 2018-11-02 16:42:33 +0000 |
| commit | acacc324216f466a548a69c2d8cc6540dc2769be (patch) | |
| tree | 72203bea625b1af1abc0a4ddb664aad2748fe177 /src/runtime/CL/functions/CLGEMM.cpp | |
| parent | 9a33b5411fed53e2fa339957b4c0b8e31e402968 (diff) | |
| download | ComputeLibrary-acacc324216f466a548a69c2d8cc6540dc2769be.tar.gz | |
COMPMID-556: Add tolerance in CustomConvolution for NEON
NEON target multiplies convolution result with the reciprocal of scale,
while the reference directly divides with scale leading to mismatches of
range of 1.
Increases absolute tolerance to 1 for NEON target.
Change-Id: Ifb8328699479be0e0804aaf319d9c825dfb734cb
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/113084
Reviewed-by: Michalis Spyrou <michalis.spyrou@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Diffstat (limited to 'src/runtime/CL/functions/CLGEMM.cpp')
0 files changed, 0 insertions, 0 deletions