Difference between revisions of "KDJ11 CPUs"

From Computer History Wiki
Jump to: navigation, search
m (Proper cat)
(Add NXM from PARs)
Line 7: Line 7:
 
==Floating point==
 
==Floating point==
  
All the KDJ11 CPUs except early revisions of the KDJ11-A CPU have two choices for [[floating point]] support (full [[FP11 floating point]]): the base DCJ11 chip, which implements floating point using [[microcode]]; and an optional  higher-performance separate dedicated chip, the [[FPJ11]].
+
All the KDJ11 CPUs except early revisions of the KDJ11-A CPU have two choices for [[floating point]] support (full [[FP11 floating point]]): the base DCJ11 chip, which implements floating point using [[microcode]]; and an optional  higher-performance separate dedicated chip, the [[FPJ11 floating point accelerator]].
  
 
(The issue with the KDJ11-A is that early revisions have bugs in their [[gate array]] [[chip]]s which prevent correct operation of the FPJ11.)
 
(The issue with the KDJ11-A is that early revisions have bugs in their [[gate array]] [[chip]]s which prevent correct operation of the FPJ11.)
Line 13: Line 13:
 
==ODT==
 
==ODT==
  
The J-11 chip set includes microcode which provides 'front panel' functionality named 'ODT'; the ability to read and write to memory, start the process, etc. Unlike the ODT in the [[KDF11 CPUs]], which only supported 18-bit addressing, the KDJ11's do not have this limitation.
+
The J-11 chip set includes [[microcode]] which provides 'front panel' functionality named 'ODT'; the ability to read and write to [[main memory]], start the processor, etc. Unlike the ODT in the [[KDF11 CPUs]], which only supported 18-bit addressing, the KDJ11's do not have this limitation.
  
Note, however, that the KDJ11-A and KDJ11-B power up with the cache enabled, even for ODT, so if the user writes some data into a given location using ODT, and then reads it back, they will get the correct data even if that memory location is faulty - the CPU is getting the (correct) data from the cache.
+
Note, however, that the KDJ11-A and KDJ11-B power up with the CPU's [[cache]] enabled, even for ODT; so, if the user writes some data into a given location using ODT, and then reads it back, they will get the correct data even if that memory location is faulty - the CPU is getting the (correct) data from the cache.
  
 
To have 'memory' reads and writes actually go to the memory, the cache has to be turned off:
 
To have 'memory' reads and writes actually go to the memory, the cache has to be turned off:
  
 
   17777746/ 02000
 
   17777746/ 02000
 +
 +
or
 +
 +
  17777746/ 014
 +
 +
(It is not clear why there are two different ways to disable the cache.)
  
 
Note that starting the machine does an INIT, which will again enable the cache.
 
Note that starting the machine does an INIT, which will again enable the cache.
 +
 +
==Code in PARs==
 +
 +
One commonly used diagnostic 'trick' is to store very small test programs in the Page Address Registers of the [[PDP-11 Memory Management]] unit. This is typically used for short 'oscilliscope loops', either to debug a memory card when isn't any working memory plugged in, or when there is a desire to avoid 'contaminating' the bus with [[instruction]] fetch cycles.
 +
 +
However, this does not work on the KDJ11 processors. The CPU cannot execute [[object code|code]] from the PARs; it gets a [[NXM]] [[trap]] on attempts to do so.
  
 
==Further reading==
 
==Further reading==

Revision as of 14:51, 9 July 2018

There are several single-board PDP-11 CPUs which all use the 'Jaws' J-11 chipset:

Floating point

All the KDJ11 CPUs except early revisions of the KDJ11-A CPU have two choices for floating point support (full FP11 floating point): the base DCJ11 chip, which implements floating point using microcode; and an optional higher-performance separate dedicated chip, the FPJ11 floating point accelerator.

(The issue with the KDJ11-A is that early revisions have bugs in their gate array chips which prevent correct operation of the FPJ11.)

ODT

The J-11 chip set includes microcode which provides 'front panel' functionality named 'ODT'; the ability to read and write to main memory, start the processor, etc. Unlike the ODT in the KDF11 CPUs, which only supported 18-bit addressing, the KDJ11's do not have this limitation.

Note, however, that the KDJ11-A and KDJ11-B power up with the CPU's cache enabled, even for ODT; so, if the user writes some data into a given location using ODT, and then reads it back, they will get the correct data even if that memory location is faulty - the CPU is getting the (correct) data from the cache.

To have 'memory' reads and writes actually go to the memory, the cache has to be turned off:

 17777746/ 02000

or

 17777746/ 014

(It is not clear why there are two different ways to disable the cache.)

Note that starting the machine does an INIT, which will again enable the cache.

Code in PARs

One commonly used diagnostic 'trick' is to store very small test programs in the Page Address Registers of the PDP-11 Memory Management unit. This is typically used for short 'oscilliscope loops', either to debug a memory card when isn't any working memory plugged in, or when there is a desire to avoid 'contaminating' the bus with instruction fetch cycles.

However, this does not work on the KDJ11 processors. The CPU cannot execute code from the PARs; it gets a NXM trap on attempts to do so.

Further reading

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