G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
AMD produced what is probably the first floating point "coprocessor" for microprocessors, the AMD 9511 "arithmetic circuit" (1979), which performed 32 bit (23 + 7 bit floating point) RPN-style operations (4 element stack) under CPU control.
________________________________________________________________
AM9511 APU (arithmetic processing unit) performs floating point calculations that would be too time-consuming or cumbersome if done on Z80 microprocessor.
In addition to the four basic operations, the 9511 can implement transcendental functions (sin, log, etc.), and it can also perform conversion from floating point to fixed point and vice versa. It can also perform operations in 16 or 32 bit fixed point two's complement integer representation.
The fastest instruction is 16 bit fixed point addition which takes 4 microsec, the slowest instruction is arc-tangent which takes almost 2 milisec; 32 bit floating point addition takes 14-87 microsec.
The 9511 is an autonomous processor: it acts like a peripheral and it in not tied to any particular microcrocessor. AM9511 is configured as a pair of I/O ports with the capability of causing the Z80 to pause for completion of the processor's operation if it has not been completed by the time the Z80 code requires the results.
The 9511 is a stack orientaded processor. Operands are first pushed onto the internal stack, and then a comand is issued to the processor to perform an operation on the stack. Results are removed from the stack.
All transfers, including operands, results, status and command information, take place over an 8-bit bidirectional data bus. The internal data paths and the internal ALU are 16 bits wide, even though some operations are on 32 bits operands (this fact indicates that extending the 9511 to handle double precision format could have been implemented internally quite easily). The transcendental functions are evaluated using Chebyshev Polynomials which provide an even distribution of errors within the selected data representation.
The 9511 was in 1980's an high cost device: the APU was far more complex than the CPU and the volume of sales was not so high. As a result this was reflected in the price: in 1983 the Am9511@2Mhz cost £90 (considering inflation it would cost £240 now in 2009), the 3Mhz £123.75, the 4Mhz £146.25 ..
G.G. (Italy) posted a photo:
This board:
1) Second 64K (8kx8) SRAM: CDM6264E3 (datacode 1982)
2) Serial RS232 Interface with 8251 USART
3) Memory mapped 8 bit parallel IN/OUT with LS374 latched outputs and LS244 inputs with one interrupt line
4) Memory mapped co-processor C8231A: 16 bit floating point arithmetic unit running @ 2.5 Mhz
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
◦ Co-processor manufacturer: AMD (AM9511) or Rochester Electronics replacement, licensed by Intel (c) 1980 (C8231A).
Rochester Electronics is licensed and authorized by semiconductor manufacturers to provide a continuing manufacturing source for mature/discontinued products.
AM9511 APU performs floating point calculations that would be too time-consuming or cumbersome if done on Z80. In addition to the four basic operations, the 9511 can implement transcendental functions (sin, log, etc.), and it can also perform conversion from floating point to fixed point and vice versa. It can also perform operations in 16 or 32 bit fixed point two's complement integer representation.
G.G. (Italy) posted a photo:
G.G. (Italy) posted a photo:
