On 10/01/2016 06:24 AM, Robert Prins wrote:
> On 2016-10-01 01:59, CM Poncelet wrote:
>> 't is Friday, so here we go again.
>>
>> I wrote the following in Intel assembler in the 1990's to check the
>> CPU-cycles
>> performance of DOS vs Windows 3.1 (if memory serves), using a 486/33Mhz
>> processor and practically all executable code loaded in the
>> instruction cache:
>
> <snip code>
>
>> This took approx 1'45" (elapsed) to execute under native DOS with a
>> 80486/33Mhz
>> processor - i.e. there was a 1-to-1 correlation between the number of
>> machine
>> instructions vs the number of actual 33Mhz CPU cycles.
>>
>> I reran this under Windows XP from the 'command prompt' with a 3.2Ghz
>> processor
>> (approx 100 times faster than a 33Mhz one), and it then took 3'48" to
>> complete -
>> thus about 200 times longer (CPU-wise/elapsed) than when running
>> under native DOS.
>>
>> Moral: Use native DOS (preferably with a DOS extender) for fast PC
>> performance.
>
> NO, NO, NO!!!
>
> Moral, write programs that run natively. A 16-bit DOS program running
> in an emulated DOS box will never run as fast as a native windoze
> program... Ask for a native windoze version on
> <http://board.flatassembler.net/index.php> or
> <http://masm32.com/board/index.php> and the program will probably run
> faster than you can blink an eye.
>
> Robert
>
> PS: Decode the email address, you can send me copies of the .EXE's
> renamed to .QQQ and in a ZIP file!
Actually I think you may both be missing a significant point.
Native DOS only used text mode input output. Windows used graphical
output, which consumes a very significant amount of resources by itself,
besides forcing additional multitasking overhead. I would expect any
operating system that did that using the same CPU and memory to generate
and control graphic output would exhibit significantly less power
delivered to applications.
There is a reason why 1960's mainframes like a 360/65 with only a 0.5
MIP processor could run a major business -- they only had to deal with
record I/O and text data. The closest they came to graphics was 3270
full screen displays, and the translation from text streams to display
was handled wholly by external controllers, without adding overhead to
the main processor and memory. I remember there being management
resistance to even moving from monochrome monocase 3270's to multi-color
dual-case 3270's because of the "unnecessary" extra overhead of
supporting those features.
Joel C. Ewing
--
Joel C. Ewing, Bentonville, AR [email protected]
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN
