Skip to main content

Optimizing C programs on z/OS

A few years ago I investigated the performance of our application on z/OS. The application is written in C and runs in batch and CICS. A customer reported that our latest version was not performing as well as the version they were currently running.

Testing showed there was a difference. The same tests did not show a difference on other platforms. z/OS has unique characteristics, not surprising given the hardware it runs on, and there are bound to be factors that affect performance that do not affect other platforms.

We build our application using unix system services and link the final modules into a Dataset (load library). What this means is that we can use standard unix shell utilities, make files and other tools just like on any other unix platform. This makes development much easier. The final link is the first time anything leaves the unix environment.

When investigating anything on z/OS the first place to go for information is the IBM Redbooks site. I found several books describing performance including one which used C to build several different types of applications. 

The result of two weeks work was a final set of modules that performed about twenty percent faster and an increased understanding of how to set up the Unix system to be able to compile and link using the maximum optimization options. The final version I delivered did not use maximum optimizations as I had achieved the aim of improving performance to match or exceed that achieved with the old version. Using the greater optimizations increased build times and exceeded the temporary work-space settings.  

For those of you building C programs using Unix System Services the default workspace sizes (for SYSLMOD for example) can be adjusted using the environment variable _CC_WORK_SPACE. The default is (32000,(30,30))

For example
export _CC_WORK_SPACE="(32000,(90,90))"

Configuring z/OS is not for the faint of heart and IBM support helped me with some of the issues. The final one I managed to figure out for myself. I used to administer our z/OS systems when we were running P/390 and FLEXES systems and I wish I had been able to access IBM support then! Searching through IBM manuals for something as complex as OS/390 (or z/OS as it has become) was never easy!

 (Originally posted here)

Comments

Post a Comment

Popular posts from this blog

Vising another office - What do you need when you arrive?

One of the challenges when traveling is being able to connect and charge all the different devices we take with us. These devices differ depending on the purpose of the trip (business or pleasure) but have the same challenges.When traveling for pleasure you may be staying with relatives, friends or at an AirBnB. What is available may be unknown. Many hotels cater for international travelers and have sockets that allow many different plugs to be used - The JW Marriott in Bangalore allows you to plug in any type of plug (all that I have seen!). One hotel in Sydney had USB charger ports available. The first challenge is being able to plug your charger in. When traveling for work I always carry two adapters that convert from the local plug standard to the plugs required for my devices and chargers. An alternative is to carry one adapter and a power board. The exception to the two adapters is when I am traveling to multiple countries which have different plugs! In those cases...
Post a Comment
Read more

The Art Of Debugging

My first taste of commercial software development came using Adabas / Natural - a data base with a 3GL programming language. There were no debuggers available but it had an interpreted development environment and you could write code and use print statements to debug your code. I was amazed when I was asked by another developer to help fix a bug in his code and observed that they did not use the same debugging technique. This developers debugging technique was to (apparently randomly) choose a line or two of code, change something, and rerun their application to see if the outcome changed. A couple of print statements quickly narrowed down the problem and had their program working. I found my productivity was many times better just because I had mastered that simple debugging technique. It took me minutes to fix a problem rather than days (or never!). One day I volunteered to port an imap implementation to OS/2. I thought that because it was written for Unix and on my OS/2 developmen...
Post a Comment
Read more

"No child processes" error

A problem was reported by a customer. They were getting a failure and in the logs it reported error → waitpid failed 'Reason: No child processes' The “No child processes” error came from waitpid() after using  fork/spawn to launch a utility to load data into a data base. Upon detailed investigation it appears it is possible that some other process that the user is running has changed the default handler for SIGCHLD - possibly the shell (e.g. bash!) used to launch our server processes.  If the signal handler is set to SIG_IGN then when a process is started using fork()/exec() the return code from the process is NOT returned and waitpid() cannot retrieve the response code. The most likely reason for "No child processes" error from waitpid() is that the signal handler for child processes (SIGCHLD) is not set to SIG_DFL. This should not be possible however it seems that on Linux a process run in the shell (or maybe a shell process) can set it to SIG_IG...
Post a Comment
Read more