[hercules-390] 370 Instruction Timings

[[e-mail] [hercules-390]]

In the 70's I workedon a program that analyzed the assembly listings for COBOL programs to help programmers tune their code. To do that we had a document that detailed the cycle times for the various instructions. It was something like an addendum to the POP that was specific to the machine (a 370/158 as it happens). Does anyone have such a thing today?

[Ivan Warren [e-mail] [hercules-390]]

Post by ***@yahoo.com [hercules-390]
In the 70's I workedon a program that analyzed the assembly listings
for COBOL programs to help programmers tune their code. To do that we
had a document that detailed the cycle times for the various
instructions.  It was something like an addendum to the POP that was
specific to the machine (a 370/158 as it happens).  Does anyone have
such a thing today?

I doubt it...

Modern processor technology is such that instructions tend to have VERY
variable timings because of caching, pipelining, out of order execution,
predictive execution, etc...

--Ivan


[Non-text portions of this message have been removed]

[Martin Truebner [e-mail] [hercules-390]]

addind to Ivans stmt:

.... and it could be anywhere between 0 (at the right
spot executed in parallel to previous/next) and 50 or more micro-cycles
(ie. after an SPKA).

The hardware now has (since z10) extra instructions to get the number
of instructions executed and the number of stalls and the number of
cycles per instruction for a given time slot and a lot of other usefull
information. It is called measurement and counter facility and has
(aside of the various model-specifc as well as common counters)
instructions to sample the cpu 1.000.000 million times per second. so
can get a picture of where the code spends most of the time.

It is an extra book next to POP

Martin

[[e-mail] [hercules-390]]

Interesting -- thanks.

['\'Fish\' (David B. Trout)' [e-mail] [hercules-390]]

(RESENDING; original send attempt made on 2018-04-20)


Who are you replying to??
--
"Fish" (David B. Trout)
Software Development Laboratories
http://www.softdevlabs.com

-----Original Message-----
Sent: Friday, April 20, 2018 5:57 AM
Subject: [hercules-390] Re: 370 Instruction Timings
Interesting -- thanks.
________________________________
________________________________

[[e-mail] [hercules-390]]

sorry: I didn't mean for a modern processor, i meant for the /158 - a much simpler beast. As I recall it was an addendum to the POP and heavy with caveats about variability.

I see the /360 timings posted separately - i had not seen that.

[[e-mail] [hercules-390]]

http://bitsavers.trailing-edge.com/pdf/ibm/360/A22_6825-1_360instrTiming.pdf http://bitsavers.trailing-edge.com/pdf/ibm/360/A22_6825-1_360instrTiming.pdf

[[e-mail] [hercules-390]]

great - thank you.

['\'Fish\' (David B. Trout)' [e-mail] [hercules-390]]

(RESENDING; original attempt to send was made on 2018-04-20)


Who are you replying to??
--
"Fish" (David B. Trout)
Software Development Laboratories
http://www.softdevlabs.com

-----Original Message-----
Sent: Friday, April 20, 2018 5:57 AM
Subject: [hercules-390] Re: 370 Instruction Timings
great - thank you.
________________________________
________________________________