- Add release notes for 4.0.0
- Add new config file additions to OPTIONS.MD description
- Add Regor debug CLI-options to OPTIONS.MD

Change-Id: Ie31a68e26076da401b9c24fca47579aa63d1faeb
Signed-off-by: Rickard Bolin <rickard.bolin@arm.com>

Tim Hall authored Aug 21, 2024 and Rickard Bolin committed Sep 04, 2024

 - Added Ethos-U85 section to supported operators
 - Regenerated SUPPORTED_OPS.md

Change-Id: Ia70fbdae89349d41490eb77280cf4cf1f668f617
Signed-off-by: Tim Hall <tim.hall@arm.com>
Signed-off-by: Rickard Bolin <rickard.bolin@arm.com>

 - ofmType is missing from ExecQuery in the reader
so not all contrainsts are checked

Change-Id: Ie8d2670fabfb82af32c7987bcc7942e820706b32
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

 - Changed default CLI behaviour of Ethos-U85 to be consistent with
Ethos-U55 and Ethos-U65
 - Changed default config values to be read from the default configuration
file Arm/vela.ini rather than being hard coded

Change-Id: I28df5513d091cca87f85f381f97173c9baa3e3eb
Signed-off-by: Tim Hall <tim.hall@arm.com>

Fixed undefined behavior when scale is negative.

Change-Id: Ic09be321403b109ecf6ea5e93493bdd1a1194a1e
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

Fixed incorrect decomposition of dilation greater than two which
caused out-of-bounds read.

Change-Id: Ibd4741a36b925829884ef49662dff6d5d7298ed9
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

Change-Id: I20dccc8c0f9db5797c3603e4bcd08033e54ac58d
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

Alexander Bengtsson authored Aug 07, 2024 and Johan Alfvén committed Aug 28, 2024

Change-Id: I8465d263bf86ebf2faf7ef92d5b6f9592aada9de
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

Improved ifm Rescale fusing for Ethos-U55/65.

Change-Id: Ia2297c59f4a5f966b0925021b22cac96f96d2524
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

Implemented fusing of RESCALE to previous/next op in
GraphIrOptimiser to improve performance for TOSA networks.

Change-Id: I47293e04d903217d4947712f2e6f20014be02cb8
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

Fixed TOSA AVG_POOL2D with no padding for Ethos U55/65.

Change-Id: I4d162208accf3317b46b61d7cc3fdcb5b5d51459
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

William Isaksson authored Aug 05, 2024 and Johan Alfvén committed Aug 27, 2024

deprecated built in codes

Regor can now pass through unsupported operators with deprecated builtin
codes.

Change-Id: Ia4e487ee585901f1a73392d863c683bb828e5613
Signed-off-by: William Isaksson <william.isaksson@arm.com>

- Implemented config selection that maximizes width and heigh
for FullyConnected

Change-Id: If53d20d7565df1094df1c7a5f79627a400d41474
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

 - Add operator type to error message when compilation fails due to
out of index
 - Add null pointer checks in ConstraintEmptyConstTensors

Change-Id: I997ffacb0a59a5f55d8fe0b91ee14b35923d5a98
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

 - Fix compiler crash due to compiling CAST operator
with data type float
 - Fix is to add arch constraint on the CAST operator

Change-Id: Ic1192f8a634294b062b2141e99c573a821136cfb
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

Johan Gunnarsson authored Aug 21, 2024 and Johan Alfvén committed Aug 26, 2024

* Add support for TOSA SLICE with fewer than 4 dimensions.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I7ec7bc689a48499782992a39ad4ba295445d0ed3

Verbose quantization print for Regor was missing ofm quantization
information.

Change-Id: I16e69b95cadf7e1104a7623cd949e21f0dce9fef
Signed-off-by: Fredrik Svedberg <fredrik.svedberg@arm.com>

William Isaksson authored Aug 20, 2024 and Johan Alfvén committed Aug 21, 2024

No longer rounds up to the ofm block to improve performance estimates.

Change-Id: I47bff892ee3f71623c83aea3796e13a978cc09b8
Signed-off-by: William Isaksson <william.isaksson@arm.com>

 - Extend heuristic with a bandwidth ratio to solve
use cases where it is beneficial to do buffering

Change-Id: I32121f73c523e3eae60e623a93aaf3255eab50c2
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

* Add support for TOSA NEGATE with fewer than 4 dimensions.
* Add asserts to ensure that both IFMs have same data type where
  that's required.
* Add helper to create const tensors with a certain value.
* Update some other places to not trigger the new asserts.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I8f572d3e754c3521e739bf146aa0a00358c758ec

Alexander Bengtsson authored Mar 19, 2024 and Alexander Bengtsson committed Aug 19, 2024

Change-Id: I5ad75ba62971d19f760408cbcfa0b0eac1967c67
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

* Move all CONCAT processing to graphir_optimiser.
* Convert TFLite PACK into TOSA CONCAT.
* Drop OpType::ConcatTFLite. Use OpType::Concat instead.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I757b0945f9ad28d0b1e3e6e18bcff9fcbd93398d

* Add support for TOSA CONCAT fewer than 4 dimensions.
* Add some missing RecordOptimisation.
* Fix handling of constant bool tensors.
* Fix TOSA validation.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I577f6abf31ef28c1ca128bf2bf3b007b86bec767

William Isaksson authored Aug 08, 2024 and Rickard Bolin committed Aug 16, 2024

Adds function for calculating per channel bws and all required inputs. Currently only used for m2m cycle estimation and weight transfer cycle estimation for weight decoder selection.

Change-Id: I5de9678f09b5d3cfa7bb97d07fd780b72777f397
Signed-off-by: William Isaksson <william.isaksson@arm.com>

Alexander Bengtsson authored Aug 09, 2024 and Alexander Bengtsson committed Aug 16, 2024

- Fix implicit narrowing assert in python-bindings
- Fix unused parameter warning in graphir_optimiser

Change-Id: Iaaa2c1109d4e813d41f0b95a8e7569999a600d1b
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

 - Fixed a case where consecutive ops have the same weights and
it is beneficial to share the same weight buffering.
 - Added detection of this use case in the weight buffering
algorithm so it will force a full ofm depth slice and hence
the same weight buffer can be reused and only one dma transer is
needed.

Change-Id: I132f8bdc8d2c692541351e84441ad6c47a5c477f
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

- Remove Small 4x4 IFM constraint to allow larger blocks
  for Matmul operations.
- Consider OFM-coverage instead of IFM-fetch in relativeCost
  calculations.

Change-Id: I5aee67f35bd979cc77328afb22b254dbdd0cabdd
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

 - Add simple heuristic when to enable weight buffering
since dram configurations does not always benefit from it
 - Solution to be refined based on architecture and memory
configuration

Change-Id: I501c180dd846e6886520e0a2d25c2425d7df8592
Signed-off-by: Johan Alfven <johan.alfven@arm.com>

Alexander Bengtsson authored Aug 08, 2024 and Alexander Bengtsson committed Aug 13, 2024

Change-Id: I08f236996768794907d0fc44b09c29daa2509556
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

- Update mypy in pre-commit to v0.981

Change-Id: I5e7d067584e3ad0a977f70980ac157ac97737d47
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

Johan Gunnarsson authored Aug 12, 2024 and Johan Alfvén committed Aug 13, 2024

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I611ee58eb6e60041edcd20513ad79b86d0f4040a

William Isaksson authored Aug 10, 2024 and Johan Alfvén committed Aug 13, 2024

Ifm shape used in per block estimate instead of ifm block. Currently this is not correct as we are rounding up to the ifm shape later.

Change-Id: Ie2237056e7cde9fae8783eac95c91a46bcb1a9c2
Signed-off-by: William Isaksson <william.isaksson@arm.com>

Alexander Bengtsson authored Aug 09, 2024 and Alexander Bengtsson committed Aug 12, 2024

- Rename RESCALE_ROUND to RESCALE_DOUBLE_ROUND to match TOSA
  specification

Change-Id: Ic7818dcb559615fe71ab40220139a81e117a3496
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

Alexander Bengtsson authored Aug 09, 2024 and Fredrik Svedberg committed Aug 12, 2024

Change-Id: Ifacfe1da39c75a9a4ff033c4249cdc50b1f747b0
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

All CONV-like ops use the same validation paths, but
FULLY_CONNECTED has different shape sizes.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: Ic9a1008a9151001c0cf192cdf59345a274023acb

* Add support for TOSA CAST with fewer than 4 dimensions.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I6adb420e94ccfb5581635351e7db25efad91ece3

Alexander Bengtsson authored Aug 06, 2024 and Alexander Bengtsson committed Aug 09, 2024

- Fix names of Rescale entries in graphapi OpAttr
  This will map rescale attributes to the correct
  attribute struct: rescale_attr_t

Change-Id: I6788fde6c1b1813bf877bbc9b95896cbbca1cbd0
Signed-off-by: Alexander Bengtsson <Alexander.Bengtsson@arm.com>

In the hillclimb bottleneck fix, include all non-direct neighbours
to the bottleneck when making the last attempt at swapping live
ranges.

Signed-off-by: Johan Gunnarsson <johan.gunnarsson@arm.com>
Change-Id: I4f963c03118aff9d296ddb2cd5445d69816af931

William Isaksson authored Jul 30, 2024 and Rickard Bolin committed Aug 08, 2024

Changes default perf query shapes to the op slice rather than the op shape.

Change-Id: Ib3819df85e5c274efd98a92942a403573cdf00b7
Signed-off-by: wilisa01 <william.isaksson@arm.com>

 - Fixed output diff in regression tests
 - MoveSplitSliceToConsumer must run after any other optimisation
that can change ifm/ofm shapes
 - Since graphir_optimiser has functions that can change shapes
MoveSplitSliceToConsumer must be run in this step instead of
tflite_graph_optimiser

Change-Id: Ifa0f42d977e719111099e16d8bc98311e3bb1d62
Signed-off-by: Johan Alfven <johan.alfven@arm.com>