On Sun, 2 Jun 2024 21:38:11 +0000, Seymour J Metz wrote:
>    ...
>I've seen two different versions of the history posted here. One version 
>claimed that IBM intended on having paging from the beginning, but delayed it 
>due to engineering problems. The other version is that IBM decided after the 
>fact to add paging. I suspect that this is in the realm of if I knew they'd 
>have to shoot me.
> 
The legend I heard was that Gene Amdahl deemed the performance
cost of DAT, etc. unacceptable.

Storage consistency was long based on effective address which vould
be determined before DAT generated a physical address and no IBM
product wold map the same physical page to two different virtual
addresses.  UNIX shared memory violates that and required redesign
of the consistency logic.

The cost of false positives when two different address spaces mapped
the same effective address to different physical addresses was
deemed less than the cost of deferring the test to physical address.

-- 
gil

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to