Abstract: An apparatus and method for accessing a cache memory. In a cache memory, an address is received that includes a set field and a partial tag field, the set field and the partial tag field together including fewer bits than necessary to uniquely identify a region of memory equal in size to a cache line of the cache memory. The set field is decoded to select one of a plurality of storage units within the cache memory, each of the plurality of storage units including a plurality of cache lines of the cache memory. The partial tag field is compared to a plurality of previously stored partial tags that correspond to the plurality of cache lines within the selected one of the plurality of storage units to determine if the partial tag field matches one of the plurality of previously stored partial tags. If the one of the previously stored partial tags matches the partial tag field, one of the plurality of cache lines that corresponds to the one of the plurality of previously stored partial tags is output.

Abstract: A microprocessor configured to rapidly execute floating point store status word (FSTSW) type instructions that are immediately preceded by floating point compare (FCOM) type instructions is disclosed. FCOM-type instructions are modified to store their results to an architectural floating point status word and a temporary destination register. If an FSTSW-type instruction is detected immediately following an FCOM-type instruction, then the FSTSW-type instruction is transformed into a special fast floating point store status word (FSTSWEF) instruction. Unlike the FSTSW-type instruction, which is serializing and negatively impacts performance, the FSTSWEF instruction is not serializing and allows execution to continue without undue serialization. A computer system and method for rapidly executing FSTSW instructions immediately preceded by FCOM-type instructions are also disclosed.

Abstract: Method and apparatus for including in a processor, instructions for performing intra-add operations on packed data. In one embodiment, an execution unit is coupled to a storage area. The storeage area has stored therein a first packed data operand and a second packed data operand. The execution unit performs operations on data elements in the first packed data operand and the second packed data operand to generate a plurality of data elements in a packed data result in response to receiving a single instruction. At least two of the plurality of data elements in a packed data result store the result of an intra-add operation using the first packed data operand and the second packed data operand.

Abstract: A portion of an x86 microprocessor that supports MMX instructions provides a write tracking unit that tracks writes to a separately provided MMX register file, and updates a status register accordingly. A write control unit uses the contents of the status register to control transfers between the MMX register file and the FP register file, so as to only copy those registers that have changed. According to another aspect of the invention, the write control unit insures that architecturally required modifications to the exponent portion of FP registers corresponding to modified MMX registers are provided.

Abstract: An apparatus and method for executing floating-point store instructions in a microprocessor is provided. If store data of a floating-point store instruction corresponds to a tiny number and an underflow exception is masked, then a trap routine can be executed to generate corrected store data and complete the store operation. In response to detecting that store data corresponds to a tiny number and the underflow exception is masked, the store data, store address information, and opcode information can be stored prior to initiating the trap routine. The trap routine can be configured to access the store data, store address information, and opcode information. The trap routine can be configured to generate corrected store data and complete the store operation using the store data, store address information, and opcode information.

Abstract: A microprocessor with a floating point unit configured to rapidly execute floating point load control word (FLDCW) type instructions in an out of program order context is disclosed. The floating point unit is configured to schedule instructions older than the FLDCW-type instruction before the FLDCW-type instruction is scheduled. The FLDCW-type instruction acts as a barrier to prevent instructions occurring after the FLDCW-type instruction in program order from executing before the FLDCW-type instruction. Indicator bits may be used to simplify instruction scheduling, and copies of the floating point control word may be stored for instruction that have long execution cycles. A method and computer configured to rapidly execute FLDCW-type instructions in an out of program order context are also disclosed.

Abstract: For use in a processor having a floating point unit (FPU) capable of managing denormalized numbers in floating point notation, logic circuitry for, and a method of adding or subtracting two floating point numbers. In one embodiment, the logic circuitry includes: (1) an adder that receives the two floating point numbers and, based on a received instruction, adds or subtracts the two floating point numbers to yield a denormal sum or difference thereof, (2) a leading bit predictor that receives the two floating point numbers and performs logic operations thereon to yield predictive shift data denoting an extent to which the denormal sum or difference is required to be shifted to normalize the denormal sum or difference, the predictive shift data subject to being erroneous and (3) predictor corrector logic that receives the two floating point numbers and performs logic operations thereon to yield shift compensation data denoting an extent to which the predictive shift is erroneous.

Abstract: An apparatus and method for bi-directional format conversion and transfer of data between integer and floating point registers is provided. A floating point register is configured to store floating point data, and integer data, in a variety of numerical formats. Data is moved in and out of the floating point register as integer data, and is converted into floating point format as needed. Separate processor instructions are provided for format conversion and data transfer to allow conversion and transfer operations to be separated.

Abstract: The present invention provides extended precision in SIMD arithmetic operations in a processor having a register file and an accumulator. A first set of data elements and a second set of data elements are loaded into a first vector register and a second vector register, respectively. Each data element comprises N bits. Next, an arithmetic instruction is fetched from memory. The arithmetic instruction is decoded. Then, a first vector register and a second vector register are read from the register file. The present invention then executes the arithmetic instruction on corresponding data elements in the first and second vector registers. The result of the execution is then written into the accumulator. Then, each element in the accumulator is transformed into an N-bit width element and stored into the memory.

Abstract: A microprocessor with a floating point unit configured to rapidly execute floating point compare (FCOMI) type instructions that are followed by floating point conditional move (FCMOV) type instructions is disclosed. FCOMI-type instructions, which normally store their results to integer status flag registers, are modified to store a copy of their results to a temporary register located within the floating point unit. If an FCMOV-type instruction is detected following an FCOMI-type instruction, then the FCMOV-type instruction's source for flag information is changed from the integer flag register to the temporary register. FCMOV-type instructions are thereby able to execute earlier because they need not wait for the integer flags to be read from the integer portion of the microprocessor. A computer system and method for rapidly executing FCOMI-type instructions followed by FCMOV-type instructions are also disclosed.

Abstract: An apparatus and method for tracking coherence between distinct floating point and MMX register files in a microprocessor is provided. The apparatus keeps track of the last time a floating point or MMX instruction was translated and what the instruction type of that previous instruction was by storing the previous instruction type in a register. When the current instruction is translated, the translator compares the current instruction type with the previous instruction type stored in the register to determine if they are different, i.e., if an instruction boundary (a change from MMX to floating point instruction or vice versa) was encountered. If so, the translator generates a signal to indicate that the two register files may be incoherent and need to be made consistent again.

Abstract: An optimized, superscalar microprocessor architecture for supporting graphics operations in addition to the standard microprocessor integer and floating point operations. A number of specialized graphics instructions and accompanying hardware for executing them are disclosed to optimize the execution of graphics instruction with minimal additional hardware for a general purpose CPU.

Abstract: A reorder buffer is configured into multiple lines of storage, wherein a line of storage includes sufficient storage for instruction results regarding a predefined maximum number of concurrently dispatchable instructions. A line of storage is allocated whenever one or more instructions are dispatched. A microprocessor employing the reorder buffer is also configured with fixed, symmetrical issue positions. The symmetrical nature of the issue positions may increase the average number of instructions to be concurrently dispatched and executed by the microprocessor. The average number of unused locations within the line decreases as the average number of concurrently dispatched instructions increases. One particular implementation of the reorder buffer includes a future file. The future file comprises a storage location corresponding to each register within the microprocessor.

Abstract: An apparatus and method for handling tiny numbers using a super sticky bit are provided. In response to detecting that a preliminary result of an instruction corresponds to a tiny number and an underflow exception is masked, an execution pipeline can be configured to store a value corresponding to the preliminary result and a super sticky bit in a destination register. Also, a destination register tag corresponding to the destination register and a denormal exception indicator corresponding to the tiny number and masked underflow exception can be stored. A trap handler can be initiated to generate a corrected result for the instruction. The trap handler can detect that the denormal exception indicator has been set and can read the value and the super sticky bit from the destination register using the destination register tag. The trap handler can generate a corrected result for the instruction based on the value and the super sticky bit.

Abstract: A floating-point unit of a computer includes a floating-point computation unit, floating-point registers and a floating-point status register. The floating-point status register may include a main status field and one or more alternate status fields. Each of the status fields contains flag and control information. Different floating-point operations may be associated with different status fields. Subfields of the floating-point status register may be updated dynamically during operation. The control bits of the alternate status fields may include a trap disable bit for deferring interruptions during speculative execution. A widest range exponent control bit in the status fields may be used to prevent interruptions when the exponent of an intermediate result is within the range of the register format but exceeds the range of the memory format. The floating-point data may be stored in big endian or little endian format.

Abstract: A microprocessor with a floating point unit configured to efficiently allocate multi-pipeline executable instructions is disclosed. Multi-pipeline executable instructions are instructions that are not forced to execute in a particular type of execution pipe. For example, junk ops are multi-pipeline executable. A junk op is an instruction that is executed at an early stage of the floating point unit's pipeline (e.g., during register rename), but still passes through an execution pipeline for exception checking. Junk ops are not limited to a particular execution pipeline, but instead may pass through any of the microprocessor's execution pipelines in the floating point unit. Multi-pipeline executable instructions are allocated on a per-clock cycle basis using a number of different criteria. For example, the allocation may vary depending upon the number of multi-pipeline executable instructions received by the floating point unit in a single clock cycle.

Abstract: The present invention generally relates to a hybrid VLIW-SIMD programming model for a digital signal processor. The hybrid programming model broadcasts a packet of information to a plurality of functional units or processing elements. Each packet contains several instructions having certain characteristics, such as instruction type and instruction length, among others. The hybrid programming model includes functional units which are reconfigurable based upon the instructions with an instruction packet and the availability of the functional units. The model groups the functional units such that the operations specified in the instructions can be efficiently executed and selects which functional units should be utilized for a given operation.

Abstract: A floating point multiply-add operating device in which a critical path in an addition process of continuous multiply-add operations of floating point numbers is shortened to improve operation efficiency is disclosed. This operating device includes: an exponent operating section for comparing an exponent of a floating point number of an operation result of a preceding multiply-add operation n with an exponent of a multiplication result of a subsequent multiply-add operation (n+1), and calculating an alignment shift count of the multiply-add operation (n+1) by the comparison result; and a mantissa operating section for aligning one mantissa of mantissas of two operands according to the alignment shift count inputted from the exponent operating section, calculating a sum of an aligned mantissa of the operand and the mantissa of the other operand, and normalizing a calculated addition result of the mantissas as needed, thereby calculating a mantissa of the multiply-add operation (n+1).

Abstract: A method and apparatus for obtaining a scalar value from a vector register for use in a mixed vector and scalar instruction, including providing a vector in a vector register file, and embedding a location identifier of the scalar value within the vector in the bits defining the mixed vector and scalar instruction. The scalar value can be used directly from the vector register without the need to load the scalar to a scalar register prior to executing the instruction. The scalar location identifier may be embedded in the secondary op code of the instruction, or the instruction may have dedicated bits for providing the location of the scalar within the vector.

Abstract: According to the invention, a processing core that executes a compare instruction is disclosed. The processing core includes a register file, comparison logic, decode logic, and a store path. Included in the register file are a number of general-purpose registers. The general-purpose registers include a first input operand register, a second input operand register and an output operand register. Comparison logic is coupled to the register file. The comparison logic tests for at least two of the following relationships: less than, equal to, greater than and no valid relationship. The decode logic selects the output operand register from the plurality of general-purpose registers. The store path extends between the comparison logic and the selected output operand register.

Abstract: An on-line reorganization method of an object-oriented database with physical references involves a novel fuzzy traversal of the database, or a partition thereof, to identify the approximate parents of all migrating objects. Where the entire database is traversed the process begins from its persistent root. For traversals of a partition the process begins from each object with a reference pointing to it from outside the partition. To facilitate the identification of these inter-partitional objects an External Reference Table (“ERT”) is maintained. During the fuzzy traversal all new inserted and deleted references are tracked in a Temporary Reference Table (“TRT”). After the fuzzy traversal is completed, for each migrating object, a lock is obtained on the identified approximate parents and on all new parents in which references to the object were inserted, as indicated by the TRT.

Abstract: A vector artchitecture processing unit according to the present invention comprises a vector scatter (VSC) address coincidence detection unit 3 that comprises registers in which an area start address and an area end address of an area specified by an area-specified vector scatter instruction are stored; and a circuit that checks if the addresses specified by the area-specified vector scatter instruction overlap with an address to be accessed by a memory access instruction following the area-specified vector scatter instruction, wherein an instruction issue control unit 1 comprises a hold control circuit that holds the following memory access instruction in response to an address conflict signal from the VSC address conflict detector.

Abstract: A dual register file MMX-type architecture comprises monitoring logic for identifying which registers in a register file have been written to. The monitoring logic is coupled to write-enable logic associated with each register. Detection logic indicates the occurrence of an instruction boundary event and asserts a signal indicating the possibility of data incoherence between the register files. Control logic coupled to the register files cause a transfer of data between the two register files in response to the asserted signal. The monitoring logic acts in conjunction with the write-enable logic to disable write operations to the receiving registers when the corresponding transferring registers have not been written to.

Abstract: Disclosed is a data processing system and a method for performing an arithmetic operation on a plurality of signed data values. In the data processing system and the method, there is a first step in which two or more signed data values are encoded into a composite value and an arithmetic operation is applied to the composite value to produce an encoded result. The encoded result can then be decoded to produce final results where each final result represents the application of the arithmetic operation to a corresponding signed data value. Thus, by using the encoded composite value, a single arithmetic operation can be applied simultaneously to multiple data values and the result then decoded. The decoded result represents the result of applying the arithmetic operation to each data value separately. The advantage of this operation is that operations can be formed on multiple data values without requiring the provision of dedicated hardware or new instructions as required by the prior art.

Abstract: In a processor, store instructions are divided or cracked into store data and store address generation portions for separate and parallel execution within two execution units. The address generation portion of the store instruction is executed within the load store unit, while the store data portion of the instruction is executed in an execution unit other than the load store unit. If the store instruction is a fixed point execution unit, then the store data portion is executed within the fixed point unit. If the store instruction is a floating point store instruction, then the store data portion of the store instruction is executed within the floating point unit.

Abstract: A method and apparatus for performing a move mask operation. The present invention provides a method and apparatus for performing operations on packed data values of a first size and format and conversion of the results to data of a second size and format by eliminating redundant data. The present invention is useful, for example, when comparisons are performed on floating point data that is typically larger (e.g., 64 bits) than integer data (e.g., 32 bits) and integer operations are preformed based on the result. Because many processors branch based on integer data, the comparison results stored as floating point data must be transferred to an integer register prior to branching. The present invention takes advantage of redundancy of the floating point comparison results to transfer enough data to convey the comparison result to integer registers with a single instruction.

Abstract: A method in a computer system which includes receiving a first instruction which indicates termination of execution of instructions which operate upon packed data stored in a first storage area. The first storage area is used for modifying data responsive to execution of floating point instructions. A plurality of tags is associated with the first storage area indicating that locations in the first storage area are either empty or non-empty responsive to the execution of the floating point instructions which modify data contained in the first storage area. Responsive to the receiving of the first instruction which indicates termination of execution of instructions which operate upon the packed data stored in the first storage area, the method sets only the plurality of tags to an empty state. In different embodiments, setting of the plurality of tags to a non-empty state occurs responsive to receiving a second instruction.

Abstract: 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.

Abstract: An apparatus and method for bi-directional format conversion and transfer of data between integer and floating point registers is provided. A floating point register is configured to store floating point data, and integer data, in a variety of numerical formats. Data is moved in and out of the floating point register as integer data, and is converted into floating point format as needed. Separate processor instructions are provided for format conversion and data transfer to allow conversion and transfer operations to be separated.

Abstract: A method and apparatus for performing a move mask operation. The present invention provides a method and apparatus for performing operations on packed data values of a first size and format and conversion of the results to data of a second size and format by eliminating redundant data. The present invention is useful, for example, when comparisons are performed on floating point data that is typically larger (e.g., 64 bits) than integer data (e.g., 32 bits) and integer operations are preformed based on the result. Because many processors branch based on integer data, the comparison results stored as floating point data must be transferred to an integer register prior to branching. The present invention takes advantage of redundancy of the floating point comparison results to transfer enough data to convey the comparison result to integer registers with a single instruction.

Abstract: An apparatus and method for performing a MOVHPS-MOVLPS operation on packed data using computer-implemented steps is described. In one embodiment, a first packed data operand having a pair of data elements is accessed. A second packed data operand having two pairs of data elements is then accessed. One of the two pairs of data elements in the second packed data operand is replaced with the pair of data elements in the first packed data operand.

Abstract: The present invention is generally directed to a system and method for supporting speculative execution of an instruction set for a central processing unit (CPU) including non-speculative and speculative instructions. In accordance with one aspect of the invention a method includes the steps of evaluating the instructions of the program to determine whether the individual instructions are speculative or non-speculative, and assessing each of the speculative instructions to determine whether it generates an exception. For each of the speculative instructions that generates an exception, the method then encode a deferred exception token (DET) into an unused register value of a register of the CPU. In accordance with another aspect of the invention, a system is provided, which system includes circuitry configured to evaluate the instructions of the instruction set to determine whether the individual instructions are speculative or non-speculative.

Abstract: Apparatus, methods, and computer program products are disclosed for determining how to compile a program at runtime. A bytecode instruction associated with the program that can be compiled in multiple ways is retrieved and compiled in a particular way, typically the default way. At runtime, a virtual machine determines whether another way of compiling the bytecode instruction is more desirable and, if so, the bytecode is then recompiled the other way. In some embodiments, the portion of the program that contains the bytecode instruction to be recompiled is placed in a queue with other instructions that are to be recompiled. The virtual machine may examine changing requirements of the program that have developed at the program's execution in which the requirements are derived from profile data on each of the multiple ways the program can be compiled. The bytecode instruction within the program may be recompiled in a more preferred way based upon the profile data.

Abstract: 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.

Abstract: A method and instruction for converting a number from a floating point format to an integer format are described. Numbers are stored in the floating point format in a register of a first set of architectural registers in a packed format. At least one of the numbers in the floating point format is converted to at least one 8-bit number in the integer format. The 8-bit number in the integer format is placed in a register of a second set of architectural registers in the packed format.

Abstract: Rapid thread processing is performed by associating thread contexts stored in a remote memory with interrupts for controlling the operation of a hardware-accelerated processor. This both minimizes the use of registers in the processor and provides a flexible, remotely accessible storage medium for the thread contexts.

Abstract: A method in a computer system which includes receiving a first instruction which indicates indicates termination of execution of instructions which operate upon packed data stored in a first storage area. The first storage area is used for modifying data responsive to execution of floating point instructions. A plurality of tags is associated with the first storage area indicating that locations in the first storage area are either empty or non-empty responsive to the execution of the floating point instructions which modify data contained in the first storage area. Responsive to the receiving of the first instruction which indicates termination of execution of instructions which operate upon the packed data stored in the first storage area, the method sets only the plurality of tags to an empty state. In different embodiments, setting of the plurality of tags to a non-empty state occurs responsive to receiving a second instruction.

Abstract: An apparatus and method are provided for concurrently loading single-precision operands into registers in a microprocessor floating point register file. The apparatus includes translation logic, data logic, and write back logic. The translation logic receives a load macro instruction prescribing an address, and decodes the load macro instruction into a double load micro instruction. The double load micro instruction directs the microprocessor to retrieve the two single-precision operands from the address and to load the two single-precision operands into the two floating point registers. The data logic, coupled to the translation logic, executes the double load micro instruction and retrieves the two single-precision operands from the address. The write back logic, coupled to the data logic, loads the two single-precision operands into the two floating point registers during a single write cycle.

Abstract: A structure and method for processing data comprises a processing unit having a base cache, base registers having a base width and being operatively connected to the processing unit, and virtual cache registers having a virtual width and being located in the base cache and operatively connected to the processing unit, wherein a base processing precision of the processing system is determined by the base width of the base registers and a selectable enhanced processing precision is determined by the virtual width of the virtual cache registers, wherein the base registers store base instructions and data and the virtual cache registers store enhanced data, the virtual width being greater than the base width, and wherein the base cache includes tags identifying a portion of the base cache as the virtual registers, the virtual cache registers being accessible by the processing unit only for execution of enhanced instructions for providing the enhanced processing precision.

Abstract: An apparatus and method for bi-directional format conversion and transfer of data between integer and floating point registers is provided. A floating point register is configured to store floating point data, and integer data, in a variety of numerical formats. Data is moved in and out of the floating point register as integer data, and is converted into floating point format as needed. Separate processor instructions are provided for format conversion and data transfer to allow conversion and transfer operations to be separated.

Abstract: The use of checking instructions to detect special and exceptional cases of a defined data format in a microprocessor is disclosed. Generally speaking, a checking instruction is included with the microcode of floating-point instructions to detect special and exceptional cases of operand values for the floating-point instructions. A checking instruction is configured to set one or more flags in a flags register if it detects a special or exceptional case for an operand value. A checking instruction may also set the result or results of a floating-point instruction to a result value if a special or exceptional case is detected. In addition, a checking instruction may be configured to set one or more bits in status register if a special or exceptional case is detected. After a checking instruction completes execution, a subsequent microcode instruction can be executed to determine if one or more flags were set by the checking instruction.

Abstract: A floating point unit is provided which conveys the rounding mode in effect upon dispatch of a particular instruction with that particular instruction into the execution pipeline of the floating point unit. Upon dispatch of a control word update instruction into the execution pipeline, the rounding mode is updated according to the updated control word provided for the control word update instruction. Instructions subsequent to the control word update instruction thereby receive the updated rounding mode as those instructions are dispatched. The updated rounding mode is available to the subsequent instructions prior to retiring the control word update instruction. The rounding mode is therefore updated without serializing the update. If the control word update instruction modifies the value in a field other than the rounding mode, the instructions subsequent to the control word update instruction may be discarded and re-executed subsequent to updating the control word register with the updated control word.

Abstract: A method for performing fast multiplication in a microprocessor is disclosed. The method comprises detecting multiplication operations that have a floating point operand and an integer operand, wherein the integer operand is an integer power of two. Once detected, a multiplication operation meeting these criteria may be executed by using an integer adder to sum the integer power and the floating point operand's exponent to from a product exponent. The bias of the integer operand's exponent may also be subtracted. A product mantissa is simply copied from the floating point operand's mantissa. The floating point operand's sign bit may be inverted to form the product's sign bit if the integer operand is negative. Advantageously, the product is generated using integer addition which is faster than floating point multiplication. The method may be implemented in hardware or software.

Abstract: An apparatus and method for performing single precision multiplication in a microprocessor are provided. The apparatus includes translation logic and extended precision floating point execution logic. The translation logic decodes a single precision multiply instruction into an associated micro instruction sequence directing the microprocessor to fetch a single precision operand from memory and convert it to extended precision format. In addition, the associated micro instruction sequence directs floating point execution logic employing a dual pass multiplication unit to skip a pass associated with computing an insignificant partial product. This insignificant partial product would otherwise result from multiplication of a multiplicand by zeros which are appended to the significand of the fetched operand when it is converted to extended precision format.

Abstract: The present invention is a method and apparatus for rounding a result operand of a floating-point (FP) operation which causes an underflow. The FP operation is recomputed using a truncate rounding mode to generate an underflowed operand. The underflowed operand is denormalized and providing characteristic bits. A rounding bit is generated based on the characteristic bits. The rounding bit is merged with the denormalized operand to generate the rounded result operand.

Abstract: A data processing apparatus and method is provided, the apparatus comprising an execution unit having a plurality of pipelined stages for executing instructions, such that a maximum of ‘n’ instructions can be being executed simultaneously within the execution unit. Further, a set of at least ‘n’ logical exception registers are provided, each exception register being capable of storing a number of exception attributes associated with an instruction for which an exception has been detected during execution by the execution unit. In the event of an exception being detected during execution of a first instruction, the execution unit is arranged to: (i) store in a first of said exception registers said exception attributes associated with said first instruction; and (ii) to continue executing any remaining instructions already in the pipelined stages at the time the exception was detected.

Abstract: A floating-point unit of a computer includes a floating-point computation unit, floating-point registers and a floating-point status register. The floating-point status register may include a main status field and one or more alternate status fields. Each of the status fields contains flag and control information. Different floating-point operations may be associated with different status fields. Subfields of the floating-point status register may be updated dynamically during operation. The control bits of the alternate status fields may include a trap disable bit for deferring interruptions during speculative execution. A widest range exponent control bit in the status fields may be used to prevent interruptions when the exponent of an intermediate result is within the range of the register format but exceeds the range of the memory format. The floating-point data may be stored in big endian or little endian format.

Abstract: A method and apparatus for including in a processor, instructions for performing multiply-intra-add operations on packed data is described. In one embodiment, a processor is coupled to a memory. The memory has stored therein a first and a second packed data. The processor performs operations on data elements in the first packed data and the second packed data to generate a plurality of data elements in a third packed data in response to receiving an instruction. At least two of the plurality of data elements in the third packed data store the result of multiply-intra-add operations.

Abstract: A method and apparatus for converting a packed integer data item having first and second data elements, to a packed floating-point data item. In one embodiment, a method includes moving the first data element of the integer data item to a first data element of a first intermediate data item and extending a sign of the first data element into all bit positions of a second data element of the first intermediate data item. The method further includes moving the second data element of the integer data item to a first data element of a second intermediate data item and extending a sign of the second data element into all bit positions of a second data element of the second intermediate data item. The first and second intermediate data items are converted from integer data items to respective floating-point data items, and the first and second intermediate floating-point data items are packed to first and second data elements of a result.

Abstract: An FPU configured to operate in normal and fast modes. In normal mode, floating point instructions are stalled in an address calculation unit of the processor until the previously issued floating point instruction has cleared the FPU, thereby indicating that the previous floating point instruction will not have an exception. In fast mode, the address calculation unit will issue a next floating point instruction to the FPU, where it is held in a 4-deep instruction queue, regardless of whether a prior instruction has cleared. By eliminating stalls in the instruction execution pipeline caused by floating point instructions being held in the address calculation unit pending clearance of the prior floating point instruction, the instruction execution pipeline may issue floating point instructions to the FPU at a faster rate.