home *** CD-ROM | disk | FTP | other *** search
-
- CORSPEED 6.2* Released October 19th, 1994.
-
-
-
- RELEASE UPDATE INFORMATION
-
- This special Sofwin Forum version of CORSPEED provides the same performance
- measurement facilities as the full commercial version. It is designed
- expressly for recording performance information in a format suitable for
- uploading to the message section of the Sofwin Forum.
-
- CHANGES APPLICABLE TO 6.2 VERSION
-
- The most important change in this version of CORSPEED is the way the
- effective running speeds are determined. New processor types which can
- accommodate not only a wider ratio between CPU clocking speed and bus speed,
- but even fractional relationships, required that the computational model
- used in prior versions be replaced with a measured curve system.
-
- EFFECTIVE RUNNING SPEED
- Users also told us that the instantaneous running speed model used in
- previous versions didn t meet their expectations. In its place, this
- version of CORSPEED uses an average running speed model that better depicts
- the expected execution speed model preferred by most users.
-
- As a result, this version of CORSPEED will report significantly different
- numbers which better reflect software execution speeds. This new methodology
- eliminates the need to know memory universe size, whether high performance
- memory is in use, the exact ratio between processor clocking rate and memory
- bus speed, and even whether any L2 cache exists. That's because the new
- model is entirely based on measured performance curves.
-
- There are several significant operational changes in the 6.2 version of
- CORSPEED to make it more useful as a means of predicting PC system
- performance under widely varying loads and conditions.
-
- ADDRESS COUNTS
- The total number of memory addresses accessed during any given period of time
- plays an important role in cache core engine architectures. While it is
- accurate to consider the number of addresses in use in terms of a memory
- window size, it is more accurately described as the number of addresses
- accessed during a specific time interval.
-
- In general, the greater the number of addresses accessed per unit of time,
- the lower the cache system hit rates. Higher numbers of cache misses slow
- down processing due to cache system upload operations which usually require
- the processor to stop during some or all of the memory management activities.
-
- ADDRESSES PER DCM
- While the address activity measures are the same as for prior versions of
- CORSPEED, the term used to describe the size has been changed from MEMORY
- WINDOW to ADDRESSES PER DECIMINUTE to better reflect that the number
- represents the total number of discreet addresses which are accessed during
- a 6 second ( 1/10th minute ) period.
-
- ADDRESS SIZE MULTIPLIERS
- To contribute to a better understanding of how memory activity contributes
- to perceived performance, this version supports continuously adjustable
- multipliers which can be used to vary the addresses per DCM over a wide range.
- The command line switch /* xx.yy can be used to set the multiplier within the
- range of 0.25 to 8.0. The default multiplier is 1.0.
-
- ***********************
- Commercial Version Only
- Address count multiplier can be adjusted using the L/R arrow keys. Each key
- stroke changes the multiplier in intervals of 1/8th.
- ***********************
-
- TASK SIMULATION
- This version supports full time-sliced multitasking simulation. While the
- default is to a single task environment, the number of concurrently running
- tasks can be set from 1.0 to 8.0 from the command line using the option /t
- xx.yy.
-
- Fractional tasks are supported to permit fine tuning adjustment where many
- less than full time slice tasks may be running in the background. For
- example, if 2 full time tasks were running in the foreground and a 1/2 time
- background printing job was also running, this condition could be simulated
- by setting the concurrent task level to 2.5.
-
-
- ***********************
- Commercial Version Only
- The task simulation count can be adjusted using the + and - keys on the
- keypad. Each keystroke changes the task simulation count in intervals of
- 1/8th.
- ***********************
-
-
- TIME SLICE CONTROL
- When more than a single task is running, each task is sequentially dispatched
- for a pre-defined period of time which is called the time slice. While
- multi-tasking operating systems usually provide an adjustable time slice,
- the default usually falls between 25 and 60 milliseconds. CORSPEED defaults
- to a 40 millisecond time slice, but it can be set to any value between 5.0
- and 250.0 milliseconds by the user.
-
- Use the command line option /s xx.yy to set the time slice to some other
- interval.
-
- ***********************
- Commercial Version Only
- Time Slice adjustment can also be made from the keyboard using the TAB and
- SHIFT+TAB keys so long as the number of tasks is greater than 1. Time slice
- control is not permitted whenever the task count is 1.
- ***********************
-
- GRAPHICS DISPLAY
- The Professional version of CORSPEED supports extensive graphics display of
- the performance curves on a new 2nd panel. A separate graph can be generated
- for each of the operating system environments under investigation.
-
- ***********************
- Commercial Version Only
- REAL-TIME CONTROLS
- To make it easier to see the impact of address counts, concurrent tasks and
- time slice ( when more than a single task is active ), these parameters can
- be directly manipulated from the keyboard in both the tabular and graphical
- displays. All of the real-time control keys are located on the numeric
- keypad. Address count multiplier can be adjusted using the L/R arrow keys.
- Each key stroke changes the multiplier in intervals of 1/8th.
-
- The total number of tasks being simulated can be adjusted using the + & -
- keys. Each key stroke changes the multiplier in 1/8th task intervals.
-
- To reset both the number of tasks and the address count multipliers to
- unity, press the HOME key.
-
- Time Slice adjustment can also be made from the keyboard using the TAB and
- SHIFT+TAB keys so long as the number of tasks is greater than 1.
-
- To toggle between the Tabular and Graphical displays, use either the PgUP or
- PgDN keys.
-
- As a result, the effective running speeds reported by this version are
- significantly smaller than prior releases
-
- +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
-
- HIGHLY DISCOUNTED COMMERCIAL RELEASE AVAIALBLE TO FORUM MEMBERS
-
- The commercial version of CORSPEED is $350 in quantity one.
- Sofwin Forum members can upgrade to the commercial version for only $95
- ( US funds ) by calling Sofwin's order desk at ( 800 ) 339-2579.
- This price includes printed documentation for the full SST tool package
- which is $50 additional if sold separately.
-
- For information on the full Sofwin SST Tool package, call the order desk
- at the number above, or email SOFWIN at 74431,1071.
-
- Sofwin Laboratories
-
-