Execution unit for evaluating functions using newton raphson iterations

A method of operating an execution unit in a processor, the method comprising:

• receiving from a computer readable instruction, an op code defining a nature of the instruction;
• accessing one or more operands defined by the instruction, wherein the op code defines that the instruction evaluates a function of the one or more operands to produce a function result, the one or more operands including one or more input values in floating point format;
• storing, in a look up table, a plurality of entries, each of the plurality of entries comprising an initial estimate for a result of an operation;
• searching the look up table using an index value dependent upon the one or more operands to locate an entry of the plurality of entries comprising a first initial estimate for a result of the operation when performed on at least one of the one or more operands;
• providing the first initial estimate to at least one multiplier circuit of a plurality of processing circuits so as to perform processing, by the plurality of processing circuits, of the first initial estimate to generate the function result, said processing comprising:
• applying one or more Newton Raphson iterations to the first initial estimate to determine a preliminary function result;
• applying a reversal of the function to the preliminary function result to determine an error in the preliminary function result, including performing a subtraction between a first one of the input values and an approximation of the first one of the input values generated from the preliminary function result;
• and
• performing rounding of one or more bits of the preliminary function result in dependence upon the determined error, including applying a unit in the last place (ULP) circuit to determine a ULP.

An execution unit for a processor, the execution unit configured to receive from a computer readable instruction an op code defining the nature of the instruction and to access one or more operands defined by the instruction, wherein the op code defines that the instruction evaluates a function of one or more operands to produce a function result, the one or more operands being in floating point format, the execution unit comprising:

• a look up table having a plurality of entries, each of the plurality of entries comprising an initial estimate for a result of an operation;
• a preparatory circuit configured to search the look up table using an index value dependent upon the one or more operands to locate an entry of the plurality of entries comprising a first initial estimate for a result of the operation when performed on at least one of the one or more operands;
• a processing pipeline comprising a plurality of processing circuits comprising a plurality of multiplier circuits, wherein the processing pipeline is operable to evaluate a plurality of different functions;
• and
• control circuitry configured to, in dependence upon the op code of the instruction, control multiplexers of the processing pipeline so as to control to which of the multiplier circuits of the pipeline, values from a preceding processing stage of the pipeline are directed so as to evaluate the function defined by the op code, wherein the evaluation of the function defined by the op code comprises the control circuitry providing the first initial estimate to at least one of the multiplier circuits so as to perform processing, by the plurality of processing circuits, of the first initial estimate to generate the function result, said processing comprising applying one or more Newton Raphson iterations to the first initial estimate.

A method of operating an execution unit in a processor, the execution unit comprising a processing pipeline comprising a plurality of processing circuits comprising a plurality of multiplier circuits, wherein the processing pipeline is operable to evaluate a plurality of different functions, the method comprising:

• receiving from a computer readable instruction, an op code defining the nature of the instruction;
• accessing one or more operands defined by the instruction, wherein the op code defines that the instruction evaluates a function of the one or more operands to produce a function result, the one or more operands being in floating point format;
• storing, in a look up table, a plurality of entries, each of the plurality of entries comprising an initial estimate for a result of an operation;
• searching the look up table using an index value dependent upon the operand to locate an entry of the plurality of entries comprising a first initial estimate for a result of the operation when performed on at least one of the one or more operands;
• and
• in dependence upon the op code of the instruction, controlling multiplexers of the processing pipeline so as to control to which of the multiplier circuits of the pipeline, values from a preceding processing stage of the pipeline are directed so as to evaluate the function defined by the op code,
• wherein the evaluation of the function defined by the op code comprises providing the first initial estimate to the at least one multiplier circuit of the plurality of processing circuits so as perform processing, by the plurality of processing units, of the first initial estimate to generate the function result, said processing comprising applying one or more Newton Raphson iterations to the first initial estimate.