Abstract: A method of selecting a cache design for a computer system begins with the making of a prototype module with a processor, a “seed” cache, and a trace detection module. The prototype module can be inserted within a system that includes main memory and peripherals. While an application program is run on the system, the communications between the processor and the seed cache are detected and compressed. The compressed detections are stored in a trace capture module and collectively define a trace of the program on the prototype module. The trace is then expanded and used to evaluate a candidate cache design. The expansion and evaluation can be iterated to evaluate many cache designs. The method can be used to pick the cache design with the best performance or as a foundation for performing a cost/performance comparison of the evaluated caches. In this method, a single prototype is used to generate an accurate trace that permits many alternative cache designs to be evaluated.

Abstract: A computer system includes a RAM-based FIFO for buffering communications between a host processor and a remote serial-communications device. The FIFO provides for quadlet, doublet, and singlet transfer widths depending on the memory-mapped IO address asserted by the processor. Quadlet transfers can be implemented until the amount of data remaining to be transferred is less than four bytes. For each quadlet transfer, the read pointer, in the case of a read operation, or the write pointer, in the case of a write operation, is incremented by four. If two or three bytes of data remain after the quadlet transfers, a doublet transfer can be implemented; in this case, the appropriate pointer is incremented by two. Finally, if a byte of data remains after the quadlet transfers and a possible doublet transfer, then a singlet transfer is effected. In this case, the appropriate pointer is incremented by one.

Abstract: A method of designing a mask-programmable random-access read-only memory device begins with a step of assigning weightings to addresses according to their expected frequency of access. These weighting are used in a second step of determining for each sense amplifier, what is the low-power sense (inverted or uninverted) of the stored bits using that sense amplifier as an output. The third step involves storing the data in the low-power sense. The fourth step involves inverting the outputs for the data that is stored inverted. This can involve using sense inverting sense amplifiers for inverted data and sense preserving amplifiers for uninverted data. The method can result in memories in which some outputs are sense inverting while others are sense preserving. The result is a memory device with reduced power consumption relative to a comparable design not taking advantage of the relationship between data values and power consumption.

Abstract: In one example embodiment, a high-speed parallel-data communication approach transfers digital data in parallel from a first module to a second module over a communication channel including a plurality of parallel data-carrying lines and a clock path. The parallel bus lines are arranged in a plurality of groups, each of the groups including a plurality of data-carrying lines and a clock path adapted to carry a clock signal for synchronizing digital data carried from the first module to the second module. The sets of data are concurrently transferred using the groups of lines of the parallel bus, and at the second module and for each group, the transferred digital data is synchronously collected via the clock signal for the group. At the second module, the data collected for each group is aligned.

Abstract: A method and arrangement passes data between two busses without needing conventional bridge-interface protocols. Consistent with one method embodiment of the present invention, data is passed between a first bus on a reference chip and an external bus using a two-way buffer arrangement between the external bus and the first bus. The method includes coupling a two-way buffer arrangement between the external bus and the first bus, determining which of the busses is the initiating bus, and in response to this determination, controlling the two-way buffer arrangement to asynchronously copy data through the two-way buffer arrangement from the initiating bus to the other bus, wherein data is passed automatically in response to its presence at the buffer arrangement without any clock cycle delays. An example application is directed to interfacing with a bus used for a rapid silicon processing chip.

Abstract: A rapid silicon processing arrangement significantly decreases the time from initial design to market introduction. Consistent with one embodiment of the present invention, rapid silicon processing arrangement uses a deconfigurable and extendible reference-chip development platform that includes a programmable device such as an electronically reconfigurable gate array and an off-platform bus for communicating with external devices. The reference-chip development platform can be deconfigured by deselecting communicative activity by one or more of functional block macros. The external devices can be used with the reference-chip development platform to test a hardware representation of the synthesized of the functional block macros in the programmable device within the reference-chip development platform as extended by the off-platform bus. The approach significantly decreases the development time, from initial design to market introduction.

Abstract: A gray-code counter system (AP1) for a RAM-based FIFO comprises a read pointer (10), a write pointer (20), and a detector (30). The read pointer includes a gray-code decoder (11), a binary incrementer (12), a gray-code encoder (13), and a register (14) that holds the pointer count). The binary incrementer increments by 1 except when the input is 0110 (decimal 6) or 1110 (decimal 14); in these cases, it increments by 3. The result is a 4-bit modulo-12 gray-code sequence with the twelve allowed gray-code values being distributed among the sixteen possible 4-bit gray code values with translational and reflective bilateral symmetry. The write pointer is similar. Because of the translational symmetry, detectors that work with counters with modulo numbers that are power of two work with the corresponding non-power-of-two counter to provide “full” and “empty” indications.

Abstract: Non-power-of-two Gray-code counters, including modulos 10, 12, 14, and 22 are disclosed, along with a sequencing method they employ. Each counter includes a register for storing an N-bit, e.g., 4-bit, Gray-code count. The count is converted to binary code by a Gray-to-binary-code counter. The resulting binary count is incremented by an N-bit incrementer that skips certain binary values by toggling least-significant bits in unison when indicated by certain most-significant binary bits. The result is converted to Gray-code by a binary-to-Gray-code translator. The translated result is stored in the register as the next count. An algorithm is disclosed for designing such a Gray-code counter for any even modulo. The modulo is expressed as a sum of positive and negative terms, each term being a power of two. The exponents of the terms determine the counter design.

Abstract: A method and arrangement passes data between two busses without needing conventional bridge-interface protocols. Consistent with one method embodiment of the present invention, data is passed between a first bus on a reference chip and an external bus using a two-way buffer arrangement between the external bus and the first bus. The method includes coupling a two-way buffer arrangement between the external bus and the first bus, determining which of the busses is the initiating bus, and in response to this determination, controlling the two-way buffer arrangement to asynchronously copy data through the two-way buffer arrangement from the initiating bus to the other bus, wherein data is passed automatically in response to its presence at the buffer arrangement without any clock cycle delays. An example application is directed to interfacing with a bus used for a rapid silicon processing chip.

Abstract: A floating-point processor nominally capable of single and double, but not extended, precision execution stores operands in extended-precision format. A format converter converts single and double precision source values to extended-precision format. Trap logic checks the apparent precision of the extended-precision operands and the requested result precision to determine whether the floating-point processor can execute the requested operation and yield the appropriate result. If the maximum of the requested precision and the maximum apparent precision of the operands is single or double, the requested operation is executed in hardware. Otherwise, a trap is issued to call an extended precision floating-point subroutine. This approach augments the class of operations that can be handled in hardware by a double-precision floating-point processor, and thus improves the floating-point computational throughput of an incorporating computer system.

Abstract: A counting system with a dynamic maximum count includes a counter, match logic, and a maximum count controller. The counter has a present count register, a clock (or event indicator) event input, and a reset input. The maximum count controller can be programmed with an adjustable maximum count stored in a maximum count register. The match logic includes a count-wide AND gate fed by NAND gates. Each NAND gate has an inverted input coupled to a respective bit position of the present count register and an uninverted input coupled to a respective bit position of the maximum count register. The function of the match logic is to indicate a match whenever the present count has a 1 at every bit position that the maximum count has a 1, irrespective of the present count values at bit positions at which the maximum count has 0s. Thus, imperfect matches are provided for. The values of the imperfect matches always exceed the values of perfect matches, so they are not usually encountered.