Difference between revisions of "FPJ11 floating point accelerator"

From Computer History Wiki
Jump to: navigation, search
m (External links: typo)
m (Fixed typo)
Line 5: Line 5:
 
Although the DCJ11 contains a full implementation of the FP11 floating point [[instruction]]s, in [[microcode]], if a J-11 CPU board has an FPJ11 installed, the DCJ111 will automatically notice the presence of the FPJ11, and allow it to handle all floating-point instructions.
 
Although the DCJ11 contains a full implementation of the FP11 floating point [[instruction]]s, in [[microcode]], if a J-11 CPU board has an FPJ11 installed, the DCJ111 will automatically notice the presence of the FPJ11, and allow it to handle all floating-point instructions.
  
Addition of the FPJ11 improves floating-point peformance by a factor of 5 to 8. The interface between the DCJ11 and the FPJ11 is designed to allow overlap between execution of floating point instructions in the FPJ11, and execution of non-floating point instructions in the DCJ11.
+
Addition of the FPJ11 improves floating-point performance by a factor of 5 to 8. The interface between the DCJ11 and the FPJ11 is designed to allow overlap between execution of floating point instructions in the FPJ11, and execution of non-floating point instructions in the DCJ11.
  
 
==Versions==
 
==Versions==

Revision as of 21:30, 3 June 2023

FPJ11 chip (plugged into KDJ11-A CPU)

The FPJ11 floating point accelerator is an optional floating point co-processor for the DCJ11 J-11 PDP-11 CPU chip, which implements the full FP11 floating point instruction set. The FPJ11 is a single chip which is plugged into a socket on a KDJ11 CPU card.

Although the DCJ11 contains a full implementation of the FP11 floating point instructions, in microcode, if a J-11 CPU board has an FPJ11 installed, the DCJ111 will automatically notice the presence of the FPJ11, and allow it to handle all floating-point instructions.

Addition of the FPJ11 improves floating-point performance by a factor of 5 to 8. The interface between the DCJ11 and the FPJ11 is designed to allow overlap between execution of floating point instructions in the FPJ11, and execution of non-floating point instructions in the DCJ11.

Versions

Two versions of the FPJ11 exist; the FPJ11-AA and the FPJ11-AB. DEC had a number of problems with 'corner cases' (e.g. DMA during certain floating point instructions), and had to issue revised versions of the FPJ11 and some J-11 CPU boards (e.g. the KDJ11-A) to fix them all.

External links

Further reading

  • μNote #025, "FPJ11-AA Compatibility with the LSI-11/73 (KDJ11-A)", 28 April 1985