Intel Corporation

Institute for the Development of Emerging Architectures, L.l.c.

Idea Corporation

Sivakumar Makineni

Roger A Golliver

Gautam B Doshi

Alan Hersh Karp

Michael J Morrison

Jerome C Huck

Sunnhyuk Kimn

Vinod Sharma

Glenn T Colon-Bonet

Vivek Tiwari

Suri B Medapati

David N Harris

Gautam Doshi

Michael Y Chow

Mike Morrison

Suri Babu Medapati

Brian L Hughes

Albert Lo

Sunhyuk Kimn

Thomas Grutkowski

Guatam Bhawandas Doshi

Glen Colon-Bonet

A method and system are provided for performing soft error detection for integer addition and subtraction operations without the use of redundant logic. For integer addition and subtraction, compensate logic produces a compensate value utilizing arithmetic logic unit (ALU) result and operands. The compensate value is validated by the validate logic against a predetermined value to determine whether a soft error has occurred. Such compensate logic and validate logic operate on the integer operands and on the result produced by the ALU without redundant carry-propagate hardware.

Reduced-hardware soft error detection

A processor includes a pipeline having first and second stages and a shift register having first and second latches. An interface circuit is used to provide a clock signal from a clock signal line to the first and second stages based, at least in part, on first and second bits to be stored in the first and second latches, respectively.

Method and apparatus for reducing the power consumed by a processor by gating the clock signal to pipeline stages

A method and apparatus to handle exceptions. The method receives and prioritizes exceptions resulting from executing an instruction on different elements of an operand. The exceptions are reported to an interrupt service register which communicates the exceptions to an exception handler to effectively process the exceptions.

Exception reporting architecture for SIMD-FP instructions

A method is provided for loading a packed floating-point operand into a register file entry having one or more associated implicit bits. The packed floating point operand includes multiple component operands. Significand and exponent bits for each component operand are copied to corresponding fields of the register entry, and the exponent bits are tested to determine whether the component operand is normalized. An implicit bit corresponding to the component operand is set when the component operand is normalized.

Method for setting a bit associated with each component of packed floating-pint operand that is normalized in SIMD operations

A system for processing SIMD operands in a packed data format includes a scalar FMAC and a vector FMAC coupled to a register file through an operand delivery module. For vector operations, the operand delivery module bit steers a SIMD operand of the packed operand into an unpacked operand for processing by the first execution unit. Another SIMD operand is processed by the vector execution unit.

Scalar hardware for performing SIMD operations

An apparatus, a processor, a computer system and a method may be used to directly transfer and translate data between a memory format in an integer processing unit and a floating point format in a floating point processing unit. Data is stored in integer registers of the integer processing unit in a memory format and is stored in floating point registers of the floating point processing unit in a floating point format. A direct data link is provided between the integer register file of the integer processing unit and the floating point register file of the floating point processing unit. The direct data link includes a logic circuit which translates data between the memory format and the floating point format.

Method, apparatus and computer system for directly transferring and translating data between an integer processing unit and a floating point processing unit

An emulating agent and method is provided that receives numbers having si, exponents and significands of varying lengths and possibly configured in a variety of incompatible formats and to reformat the numbers into a standard uniform format for uniform arithmetic computations in processors operating with different architectures. In one embodiment, the emulating agent has a three-field superset register configured to receive the sign of a number in a first field, the exponent of a number in a second field and the significand of a number in a third field, regardless of the original format of the number, resulting in a number represented in a standard uniform format for computation. The embodiment also allows high level access to the fields to allow users to control the size of the numbers inserted into the fields.

Emulating agent and method for reformatting computer instructions into a

A microprocessor having a pipelined floating point unit operable to bypass pre-rounded results at clock cycle i and provide the pre-rounded results as an operand for a second instruction at clock cycle i+2. In one embodiment, the pipelined execution unit includes at least a first execution step at clock cycle i, and a second execution step at a clock cycle i+1 and clock cycle i+2. The unit includes a bypass leading from the first execution step at clock cycle i, however, there is no bypass leading from the second execution step at clock cycle i+1. The bypass carries the pre-rounded results from the end of the first execution step to the front end of the pipeline via a latched data path which delays the pre-rounded result one clock cycle.

Apparatus for bypassing intermediate results from a pipelined floating

In order to multiply operands of different binary lengths using a common combined array, for example to do both 8 bit by 8 bit and 16 bit by 16 bit multiplications, 2.sup.m-1 multiplications are performed, where m is equal to the number of different bit lengths it is desired to multiply. For example, where 8.times.8 bit and 16.times.16 bit multiplications are done, 2 different multiplications are done. Each multiplication is an n.times.n/2.sup.m-1 multiplication, e.g., a 16.times.8 bit multiplication. Sign correction is performed by adding a correction vector or by modifying one of the partial products. The results of the multiplications are added together to obtain a 2 n bit result. Groups of bits from said 2 n result are selected depending on the length of the operands being multiplied.

Efficient combined array for 2n bit n bit multiplications

A saturating alignment shifter for use in the multiply and accumulate unit in a floating point arithmetic unit of a microprocessor that mimics the ideal model of an infinitely wide shifter. A saturation alignment shifter is provided that, in the case of the operation of A*B.+-.C, saturates at a predetermined shifting increment and, in the case where the mantissa of C is saturated, places the mantissa C in the left most significant bits. After adding C to A*B in a summation unit, the mantissa of A*B ends up in the right most significant bits and a single intervening bit, termed the bubble bit, remains between the mantissa of C and the mantissa of A*B. The bubble bit acts to mimic any intervening bits that would have occurred in the case of an ideal infinitely wide shifter. The new shifter eliminates hardware required for special cases and treats all operations with a single alignment shifter giving a system that is faster and more simple than conventional systems.

Saturating alignment shifter

Provided is a multi-stage shifter for use in both alignment and normalization shifters that provides faster implementation of the shifting process, requires less hardware and is less complex. One embodiment of the present invention provides a shift controller generating shift controller signals and a multi-stage normalization shifter having distinct multiple stages, coupled to a shift controller and receiving a number for normalization. The multi-stage shifter shifts the input number at predetermined increments at each stage and producing a sticky bit output at each stage that participates in the final sticky bit. These sticky bit outputs are ORed together to produce the final sticky bit. The present invention provides for production of the sticky bit with less hardware without sacrificing performance.

Apparatus and method for computation of sticky bit in a multi-stage

An apparatus and method for performing variable precision floating point rounding operations is provided that accomplishes rounding of a number that is faster, less complex and requires less hardware than conventional devices. The apparatus and method provides for a single uniform incrementer located at the zero position that can add a logic 1 to that position. After the Round, Guard and Sticky bits are produced, logic 1s are inserted in the bit positions to the right of the significant bits of the result with a check unit. The check unit then outputs the intermediate result to an incrementer and a zero-out unit. The incrementer adds a 1 to the zero position and the carry function of the incrementer by operation ripples through the series of 1s up to the rightmost position of the resultant mantissa leaving a trail of zeroes behind producing a first output. Finally, a multiplexer is provided to choose between the first output and the second output as the final result.

Growing community of inventors

Sunnyvale, CA, United States of America

Sivakumar Makineni