Abstract: A field programmable gate array (FPGA) is provided that includes a plurality of pads and a plurality of delay locked loops (DLLs). Programmable connections enable any one of the DLLs to have multiple pads as inputs. Programmable connections also enable the DLLs to be selectively connected to one another. Programmable connections further enable the pads to be selectively connected to general interconnect circuitry or global clock drivers of the FPGA. Programmable connections are also provided for selectively connecting the DLLs to the global clock drivers. This FPGA structure enables the pads to be configured to receive either clock or non-clock signals. This structure also enables the FPGA to operate as a clock mirror, and to generate one clock signal from another clock signal on the FPGA.

Abstract: The invention provides an FPGA comprising an array of identical tiles. Each tile comprises a logic block that includes a Configurable Logic Element (CLE). In one embodiment, the CLE is implemented in two similar portions called "slices". Each slice has a separate carry chain. In a CLE with four function generators, each carry chain incorporates the outputs of two function generators.

Abstract: A RAM block includes a circuit for causing the RAM to provide all 0's on the output when a Reset signal is active. The Reset signal does not change the RAM contents but causes all outputs of the block RAM to be 0. This is useful when the RAM block is configured as a state machine. Thus, in an FPGA or other programmable device, an application can start the state machine in a known state with all address bits equal to 0 and can reset the state machine to this startup state. When the reset signal is active, the state machine feeds back the state of 0 to the address inputs of the RAM block that receive state feedback data, regardless of the data actually in those locations.

Abstract: A clock distribution circuit and method for programmable ICs whereby the incoming clock frequency is optionally divided by two and distributed at the new, lower frequency. Programmable dual-edge/single-edge flip-flops are provided that optionally operate at twice the frequency of the distributed clock, being responsive to both rising and falling edges of the distributed clock. When the clock divider is enabled and the flip-flops are programmed as dual-edge, the operating frequency is the same as that of the incoming clock; however, the frequency of the distributed clock is reduced by one-half. This reduction halves the frequency at which the clock distribution circuits operate, and consequently approximately halves the power dissipated by the clock distribution circuit, thereby providing a programmable power-saving mode.

Abstract: The invention provides a Configurable Logic Element (CLE) preferably included in each of an array of identical tiles. A CLE according to the invention has four function generators. The outputs of two function generators are combined with a fifth independent input in a five-input-function multiplexer or function generator to produce an output that can be any function of five inputs, or some functions of up to nine inputs. The outputs of the other two function generators are similarly combined. The outputs of the two five-input-function multiplexers or function generators are then combined with a sixth independent input in a six-input-function multiplexer or function generator. The six-input-function multiplexer or function generator therefore produces an output that can be any function of up to six inputs. Some functions of up to nineteen inputs can also be generated in a single CLE.

Abstract: The invention provides an FPGA interconnect structure preferably included in an array of identical tiles. Each tile comprises a logic block that includes a Configurable Logic Element (CLE) and an output multiplexer. Fast feedback paths are provided within the logic block to connect the CLE outputs to the CLE inputs, either directly or through an input multiplexer. The fast feedback paths bypass the output multiplexer and therefore provide faster feedback than can be obtained in most conventional FPGA logic blocks. In one embodiment, the fast feedback paths provide the ability for all function generators in one CLE to drive each other through fast feedback paths, regardless of how logic is mapped into the function generators of the CLE.

Abstract: The invention provides an FPGA interconnect structure preferably included in an array of identical tiles. The interconnect structure includes both buffered and unbuffered interconnect lines. Some buffered interconnect lines are bidirectional, and others are unidirectional. A carefully selected mixture of unidirectional and bidirectional lines provides a balance of flexibility, silicon area, and performance.

Abstract: The invention provides a Configurable Logic Element (CLE) preferably included in each of an array of identical tiles. A CLE according to the invention has four function generators. The outputs of two function generators are combined with a fifth independent input in a five-input-function multiplexer or function generator to produce an output that can be any function of five inputs, or some functions of up to nine inputs. The outputs of the other two function generators are similarly combined. The outputs of the two five-input-function multiplexers or function generators are then combined with a sixth independent input in a first six-input-function multiplexer or function generator, and with a different sixth independent input in a second six-input-function multiplexer or function generator.

Abstract: The invention provides an FPGA interconnect structure preferably included in an array of identical tiles. A combination of single-length lines connecting to adjacent tiles and intermediate-length lines connecting to tiles several tiles away creates an interconnect hierarchy which allows any logic block to be connected to any other logic block, yet also allows for fast paths to both adjacent tiles and tiles some distance away. Longer interconnect lines may be included as a third level of hierarchy to permit interconnection of widely separated tiles. In a preferred embodiment, from a given tile an intermediate-length line connects to the tile three tiles away, then continues and connects to the tile six tiles away. In this embodiment, the intermediate-length line does not connect to the intervening tiles one, two, four, and five tiles away.

Abstract: The invention provides an FPGA interconnect structure preferably included in an array of identical tiles. A combination of single-length lines connecting to adjacent tiles and intermediate-length lines connecting to tiles several tiles away creates an interconnect hierarchy which allows any logic block to be connected to any other logic block, yet also allows for fast paths to both adjacent tiles and tiles some distance away. Longer interconnect lines may be included as a third level of hierarchy to permit interconnection of widely separated tiles. In a preferred embodiment, from a given tile an intermediate-length line connects to the tile three tiles away, then continues and connects to the tile six tiles away. In this embodiment, the intermediate-length line does not connect to the intervening tiles one, two, four, and five tiles away. According to a second aspect of the invention, high fanout signals can be distributed to any tile in the array.

Abstract: A logic element for an FPGA which can be configured as any one of a random access memory, a shift register and a lookup table. The logic element includes a plurality of memory cells which are interconnected such that the data output of each cell can serve as the input to the next memory cell. Thus the logic element effectively functions as a shift register. Shift registers of arbitrary length can be created by using a lookup table address multiplexer to select any memory cell output (not necessarily the last memory cell output) of the lookup table, and by chaining lookup tables of plural logic elements in series.

Abstract: A programmable memory element clocks in new data on both rising and falling edges of the clock, thereby optionally operating at twice the frequency of the distributed clock. The circuit according to the invention comprises two latches, one rising edge triggered and one falling edge triggered. One of these latches, each time the clock changes state, latches in a new value. When configured as a dual-edge flip-flop, the output of the inactive latch is fed forward to drive the output of the memory element. In one embodiment, the outputs of the two latches are multiplexed together and the clock selects the active output. According to a first embodiment of the invention, the memory element is used in an FPGA and can be programmed to function as either a latch or a dual-edge flip-flop. A second embodiment of the invention comprises a third latch. Based on the contents of a configuration memory cell, two of the three latches are selected to form a flip-flop. One such flip-flop is dual-edge, the other is single-edge.

Abstract: A method and apparatus for loading memory within a reconfigurable programmable logic device including configuring the device as a RAM loader circuit, loading the RAM with data and then reconfiguring the device with a circuit utilizing the loaded RAM. The inventive method and apparatus allow use of the RAM as high density functional centers of the desired design immediately upon initialization of the circuit, without wasting valuable time or FPGA resources on a static, non-flexible RAM loader structure.

Abstract: The present invention is part of a Field Programmable Gate Array logic block which performs arithmetic functions as well as logic functions. The novel structure includes a small amount of extra hardware which can implement the XOR function as well as several other useful functions. With the invention, one-bit adders can be generated using only a single lookup table, a carry multiplexer, and the extra hardware. N-bit adders can be implemented with N lookup tables. Multipliers, adders, counters, loadable synchronous set-reset counters and many other common functions are all more efficiently implemented with the invention.

Abstract: The direction of a bidirectional buffer inserted along a bus line is dynamically controlled to be always away from the signal source. The signals which select a tristate buffer to turn on for driving the bus also affect or determine the direction of all buffers inserted in the bus line. Since only one tristate enable signal will be active, the direction which each bidirectional buffer should drive is dynamically determined by the presently active enable signal.