Arithmetic processing device and methods thereof

Abstract

A device and methods are disclosed for communicating an unrounded result from one arithmetic calculation for use in a second, subsequent calculation. For example, an unrounded result of a first calculation can be forwarded to provide a multiplier, a multiplicand or an addend operand for the subsequent operation. The operand can be forwarded to the input of the same fused multiply addition module (FMAM) that supplied the result, or to another FMAM, and do so without regard to the precision of the forwarded operand, the precision of the subsequent operation, or the native precision of the FMAM.

1 - 20 . (canceled) 
     
     
         21 . A method, comprising:
 responsive to a first instruction, determining at a first multiply-addition module of a processor a first unrounded result based on a first operand in a register, the first unrounded result having a first precision; and   communicating the first unrounded result to a second multiply-addition module of the processor to perform an arithmetic operation having a second precision different than the first precision.   
     
     
         22 . The method of  claim 21 , further comprising:
 determining at the first multiply-addition module a first rounded result based on the first operand; and   communicating the first rounded result to the second multiply-addition module.   
     
     
         23 . The method of  claim 22 , further comprising:
 determining a second unrounded result at the second multiply-addition module based on the first unrounded result and a second operand.   
     
     
         24 . The method of  claim 23 , wherein determining the second unrounded result comprises determining the second unrounded result based on the first rounded result. 
     
     
         25 . The method of  claim 23 , further comprising:
 communicating the second unrounded result to the first multiply-addition module in response to determining the second unrounded result.   
     
     
         26 . The method of  claim 23 , further comprising:
 communicating the second unrounded result to an input of the second multiply-addition module in response to determining the second unrounded result.   
     
     
         27 . The method of  claim 23 , further comprising:
 communicating the second unrounded result to a third multiply-addition module in response to determining the second unrounded result.   
     
     
         28 . The method of  claim 21 , wherein the first precision is a single-precision and the second precision is a double-precision. 
     
     
         29 . The method of  claim 21 , wherein the first precision is a single-precision type and the second precision is an extended-precision. 
     
     
         30 . The method of  claim 21 , wherein the first precision is a double-precision and the second precision is an extended-precision. 
     
     
         31 . A method, comprising:
 determining a first multiply-addition result at a first multiply-addition module of a processor, the first multiply-addition result based on a unrounded result in a register and a first operand, the unrounded result being from one of a plurality of multiply-addition modules of the processor and being a first precision type, and the first operand being a second precision type different from the first precision type.   
     
     
         32 . The method of  claim 31 , wherein the first multiply-addition module corresponds to the one of the plurality of multiply-addition modules. 
     
     
         33 . The method of  claim 31 , wherein the one of the plurality of multiply-addition modules corresponds to a second multiply-addition module of the plurality of multiply-addition modules. 
     
     
         34 . The method of  claim 31 , wherein the first operand is a single-precision type and the unrounded result is a double-precision type. 
     
     
         35 . The method of  claim 31 , wherein the first operand is a single-precision type and the unrounded result is an extended-precision type. 
     
     
         36 . The method of  claim 31 , wherein the first operand is a double-precision type and the unrounded result is an extended-precision type. 
     
     
         37 . A device comprising:
 a first multiply-addition module comprising an input to receive a first operand responsive to a first instruction, a first output configured to provide a first unrounded result based on the first operand, and a second output configured to provide a first rounded result based on information received at the input, the first unrounded result having a first precision; and   a second multiply-addition module comprising an input coupled to the first output of the first multiply-addition module and a first output configured to provide a second rounded result based on information received at the input, the second rounded result having a second precision different from the first precision.   
     
     
         38 . The device of  claim 37 , wherein the second multiply-addition module further comprises a second input coupled to the first output of the first multiply-addition module, and wherein the first output of the second multiply-addition module is configured to provide a second rounded result based on information received at the second input. 
     
     
         39 . The device of  claim 38 , wherein the second multiply-addition module further comprises a second output coupled to the input of the first multiply-addition module, the second output configured to provide an unrounded result based on information received at the input of the second multiply-addition module. 
     
     
         40 . The device of  claim 37 , wherein the first operand is a single-precision type and the unrounded result is a double-precision type.