Closed hikettei closed 3 weeks ago
I will merge the changes once as it is unlikely to get much development time for a while T_T. (plus, the IR specifications I have formulated are so bad that it is difficult to add new features. I need to take enough time and redevelop the entire back-end from scratch. but is it worth it?...)
Refactors
cl-waffe2-simd
, with keeping AVX/Neon/SLEEF intrinsic supports.JITCPUTensor
instead of addingbackends/aten
.AbstractNodes
base-impl
!view
.ScalarAdd/ScalarSub/ScalarMul/ScalarDiv/InverseTensorNode
.!inv
->!reciprocal
!leaky-relu
Command Line Tool
./roswell/waffe2.ros
ShapeTracker
[ ]
to express ScalarTensor, instead of using[scal]
andout-scalar-p=t
.Frontend
Backend
[x] Ready for implementing AbstractTensor backend
[ ] Remove
cl-waffe2-simd
[ ] Remove obsolete SIMD dependencies
[x] Merge the aten-runtime branch.
APIs
[ ] Implement
AbstractListTensor
(Shape)
Tensor.!concatenate
, it creates(LazyXXX ~ Shape) where LazyXXX = (length x)
tensor[ ] Control-FLow
IfNode ~ EndIfNode
LoopNode ~ EndLoopNode