tensortiger idea: implement a tinygrad uir compiler that compiles to cuda/tileir/ptx on my 3080. use c++20 history left a lot of todos. fix the current example, then go through each todo and redesign the system. basic const folding of addition. links All cuda docs Tinygrad Tinygrad class TileIR Halide TVM LLVM Mirage Luminal ThunderKittens Romou Egg microegg MLIR LoopStack TileLang Cute Layout Algebra TASO PET Ansor Mojo kgen Gluon Mosaic GPU Cutlass, Cute DSL Cute layouts Linear Layouts