On 7/12/18, U.Mutlu <um@xxxxxxxxxxx> wrote: > Aeeh, what? :-) > Come on, isn't something else missing here? B/c the above code does nothing > but some printf() and calling exit(1), thereby interrupting or even aborting > the build... Well, I did say it was proof of concept, not the benchmarking I wanted! You can't say it isn't a short piece of code! :-) When I had this working back in February.2018 and in late 2017 there was a lot more going on instead of the printf: I was comparing choose_multiplier to my ideas for better versions, both for the actual values and for the average time taken to make the calculations for typical divisors. I was hoping to repeat that for slightly different newer versions of choose_multiplier. Unfortunately it seems I can't build GCC-8.1.0 using the GCC make files. The exit(1) was quite deliberate: actually building GCC was the last thing on my mind, because I already have GCC installed and working on my netbook. All I wanted to do was run those comparisons and then stop, hence the exit(1). This method of hooking into/hijacking choose_multiplier the first time the GCC make calls it to test it was the only way I could find to actually be able to use "wide int". Once I'd got the output I wanted from the "hook" I didn't need the make to continue, in fact from my perspective having the make continue would just have taken a longish and unnecessary time! :-)
