Wipsel 128 upgrade asume no moving gc to latest version (#130)

* Upgrade asume-no-moving-gc.

* Updated versions to run thests for

* Fixed so govet won't mistaken the format vebs in the templates for actual format verbs

Co-authored-by: wisse <[email protected]>
Co-authored-by: chewxy <[email protected]>

Fixed a major bug when reusing an operand (#126)

* Fixed the main bug where a reuse tensor is the same as one of the operands.
This was fixed by means of generating new internal generic functions

* Added more tests, which found more issues which is now fixed

* Because Pow is such a PITA, I'm removing it from the generated tests

Merge branch 'optimizations'

Added Min/Max Between (#117)

* Fixed #90

* Added MinBetween and MaxBetween engine def

* Added code to generate Min/MaxBetween

* Moved example out from the generated file

* Generated MinBetween and MaxBetween methods for StdEng

* Added some compile time assertions

* Added API for Min/Max between

* Added better prep for min/max between of engine

Fix#111 (#112)

* Fixed #90

* Fixed #111

* Boyscout commit to fix python/numpy testing

Fix by indices bug (#106)

There was a subtle bug in `ByIndices`. The tests have also been updated to detect a wider class of bugs.

Fixed #90 (#105)

Clarifysemantics 2 (#100)

* Added a parentTensor check, as well as unsafe.go which will go into v0.10.0 as well

* Some weirdness from the master branch is fixed

* Added more tests for dense_format

* Added some more formatting niceness

Clarifysemantics 2 (#100)

* Added a parentTensor check, as well as unsafe.go which will go into v0.10.0 as well

* Some weirdness from the master branch is fixed

* Added more tests for dense_format

* Added some more formatting niceness

Cuda11 (#99)

* Fixed #90

* Removed everything that is unsafe.Pointer to use uintptr.
Any thing that requires a unsafe.Pointer to remain will have to use a refcounter.

* genlib'd the RC stuff into scalarHeaders

* Fixed so -race will not complain

* Updated travis to make sure things test with race as well

* Added some tests for Float64Engine and Float32Engine

* Moved to using raw byte slices as per Bryan C Mills' suggestion

* More fixed from just moving to raw byte slices

* Fixed more things for array

* Fixed tests

* Fixed all syntax errors

* removed .v from array

* Fixed some off that scalar business

* Fixed the slice bits

* tests pass

* Added benchmark script

* Fixed eng_arith_manual

* Fixed inplace transpose as well