TorchLean API

FloatLib.Floats.ExecFloat.Backends.TinyTable.Generic.Runtime

Direct byte-table execution #

The functions in this module are the warm execution path. They force a memoized table, compute a native row-major index, and perform one byte load. Proofs attached to Code and array bounds are erased. Once the table has been generated, lookup does not evaluate the model operation again.

@[inline]
def FloatLib.Floats.ExecFloat.Backend.TinyTable.runBinaryCode {Model : Type u} (encoding : Encoding Model) (op : ModelModelModel) (table : Thunk ByteArray) (table_eq : table.get = binaryTotal encoding op) (left right : Code encoding) :

Read the encoded result from an arbitrary certified binary-table thunk.

Instances For
    @[inline]
    def FloatLib.Floats.ExecFloat.Backend.TinyTable.runBinary {Model : Type u} (encoding : Encoding Model) (op : ModelModelModel) (table : Thunk ByteArray) (table_eq : table.get = binaryTotal encoding op) (left right : Code encoding) :
    Code encoding

    Execute one generated binary table directly on byte codes.

    Instances For
      @[inline]
      def FloatLib.Floats.ExecFloat.Backend.TinyTable.runUnaryCode {Model : Type u} (encoding : Encoding Model) (op : ModelModel) (table : Thunk ByteArray) (table_eq : table.get = unaryTotal encoding op) (value : Code encoding) :

      Read the encoded result from an arbitrary certified unary-table thunk.

      Instances For
        @[inline]
        def FloatLib.Floats.ExecFloat.Backend.TinyTable.runUnary {Model : Type u} (encoding : Encoding Model) (op : ModelModel) (table : Thunk ByteArray) (table_eq : table.get = unaryTotal encoding op) (value : Code encoding) :
        Code encoding

        Execute one generated unary table directly on a byte code.

        Instances For
          @[inline]
          def FloatLib.Floats.ExecFloat.Backend.TinyTable.runTernaryCode {Model : Type u} (encoding : Encoding Model) (op : ModelModelModelModel) (table : Thunk ByteArray) (table_eq : table.get = ternaryTotal encoding op) (left right addend : Code encoding) :

          Read the encoded result from an arbitrary certified ternary-table thunk.

          Instances For
            @[inline]
            def FloatLib.Floats.ExecFloat.Backend.TinyTable.runTernary {Model : Type u} (encoding : Encoding Model) (op : ModelModelModelModel) (table : Thunk ByteArray) (table_eq : table.get = ternaryTotal encoding op) (left right addend : Code encoding) :
            Code encoding

            Execute one generated ternary table directly on byte codes.

            Instances For