diff options
author | Eric Kunze <eric.kunze@arm.com> | 2020-11-03 11:23:09 -0800 |
---|---|---|
committer | Eric Kunze <eric.kunze@arm.com> | 2020-11-03 12:37:44 -0800 |
commit | 57e79c0cf2b37d37c4895ee08fd7c2534ae44cad (patch) | |
tree | 501e4aaeae3b849eb58839e78af0db2ae7628401 /chapters/introduction.adoc | |
parent | b0b4bf083b9e73e8f3e96a02a612fac53d3ad651 (diff) | |
download | specification-57e79c0cf2b37d37c4895ee08fd7c2534ae44cad.tar.gz |
Add int48 to supported number formats
Currently only enabled for the accumulator of a convolution and the
input to a RESCALE.
Signed-off-by: Eric Kunze <eric.kunze@arm.com>
Change-Id: Ic5474124b4a8735f12ef9c20c2be48e55767a3fc
Diffstat (limited to 'chapters/introduction.adoc')
-rw-r--r-- | chapters/introduction.adoc | 7 |
1 files changed, 6 insertions, 1 deletions
diff --git a/chapters/introduction.adoc b/chapters/introduction.adoc index 5134330..53a6511 100644 --- a/chapters/introduction.adoc +++ b/chapters/introduction.adoc @@ -121,13 +121,18 @@ The following number formats are defined in TOSA. See section 2.3 for details on |int16 | -32768 | +32768 -|Signed 16-bit twos-complement values. Symmetric per-tensor quantization. No zero point is used , scale factor is provided in the operation. +|Signed 16-bit twos-complement values. Symmetric per-tensor quantization. No zero point is used, scale factor is provided in the operation. |int32 | (1<<31)-1 | -(1<<31) |32-bit twos-complement value. No scale factor used. +|int48 +| (1<<47)-1 +| -(1<<47) +|48-bit twos-complement value. No scale factor used. + |float | -infinity | +infinity |