Status of Pascal

[KC9UDX]

I would've only had time to do that during compilation...

I distinctly recall not using the mouse in the IDE. I also seem to remember that many keystrokes from TP4 were changed by TP6 which irritated me. I'm not sure what they were, probably Wordstar style block commands or something.

[Ruud]

If you have a measured example of that behavior, send me a PM, maybe I can figure out what's going on in your code.

I will look for an example.

[Ruud]

Hallo Trixter,

I will look for an example.

I'm sorry but I think those programs were my assemblers and disassemblers and these have moved FreePascal quite some time ago (10 years?). I clearly understand your arguments but in those days it was a fact. But adding more and more functions and supporting more CPUs the tools didn't only get more memory hungry but I ran against this 64 KB barrier again and again. So at the end I had to switch to FreePascal. But doing so I got another idea: it were multi-pass assemblers. So why not reading the source code into memory the first time and start the disassembly from there? That worked out fine. And two things: FPC improved in time and PCs became faster and faster in time. Source code that took "ages" now are handled in seconds.
Just my two cents: could it be that handling the files was handled better by TP in those days?

[Trixter]

Just my two cents: could it be that handling the files was handled better by TP in those days?

It depends on what very early versions of FP were doing. Without seeing your code, there's no way to tell.

For very early versions of FP, and for specific workloads, it is possible 32-bit protected-mode FP programs were slower when it came to heavy I/O via DOS calls. It is also possible that early versions of FP created much larger code that didn't fit inside the L2 cache of your system, whereas TP did. But those would be only two use cases, which isn't enough to make the blanket statement "a program run under TP7 is about three times as fast as the same program run under FP" as you did. For most compute- or memory-hungry workloads, even a suboptimal 32-bit pmode program should drastically outperform a 16-bit real-mode program

[Ruud]

Without seeing your code, there's no way to tell.

Not needed, IMHO. I (still) only use two type of files: TEXT for the source and log files and "file of byte" for the generated binaries. For every pass the lines were read and handled one by one. Three passes handle most type of files but if a weird nesting of variables was used, more passes were added. The binaries were generated after the last pass.

even a suboptimal 32-bit pmode program should drastically outperform a 16-bit real-mode program

It sounds more than reasonable but... no offense meant or taken but it was what I experienced, I even used my watch to verify numbers. Maybe there were other things that played a role like programs running in the background.

[Trixter]

Most likely it was your use of TEXT file types without buffers. FP probably had very basic handling of the TEXT type in the early days; it was likely unbuffered.