The IMP-16, by National Semiconductor, was the first multi-chip 16-bit microprocessor, released in 1973. It consisted of five PMOS integrated circuits: four identical RALU chips, short for register and ALU, providing the data path, and one CROM, Control and ROM, providing control sequencing and microcode storage. The IMP-16 is a bit-slice processor; each RALU chip provides a 4-bit slice of the register and arithmetic that work in parallel to produce a 16-bit word length.
The IMP-16, by National Semiconductor, was the first multi-chip 16-bit microprocessor, released in 1973. It consisted of five PMOS integrated circuits: four identical RALU chips, short for register and ALU, providing the data path, and one CROM, Control and ROM, providing control sequencing and microcode storage. The IMP-16 is a bit-slice processor; each RALU chip provides a 4-bit slice of the register and arithmetic that work in parallel to produce a 16-bit word length.
==Configuration== Each RALU chip stores its own 4 bits of the program counter, several registers, the ALU, a 16-word LIFO stack, and status flags. There are four 16-bit accumulators, two of which could be used as index registers. The instruction set architecture was similar to that of the Data General Nova. Like the Nova, the smallest unit of addressable and writable memory is the 16-bit word; byte-level writes are not supported. IMP-16 can address up to 64K words (128K bytes). The chip set can be extended with the CROM II chip (IMP-16A / 522D) that implements 16-bit multiply and divide routines plus byte operations. Byte support is limited to the lower 32K words of memory. The chipset is driven by a 715 kHz four-phase non-overlapping clock with a +5 to -12 voltage swing. An integral part of the architecture is a 16-bit input mux that provided various condition bits from the ALUs such as zero, carry, overflow along with general purpose inputs.
Discovered by embedding cosine similarity (sentence-transformers MiniLM, 384-dim).