GCC was often preferred even at a performance cost because it wasn't worth figuring out some quirk of the HP, Data General, SCO, etc. C toolchain, especially because the pricing models of that era often meant that it cost tens of thousands of dollars just to setup the same environment that a single customer had.
The Sun machines were much more common and a lot of software was tested/built on SunOS.
I worked at HP during this time helping software vendors port to HP-UX. I also spent some time on the side compiling open source projects for fun on HP-UX and then pushing the required patches back to the project's owner.
Anyone remember Pointcast?! https://en.wikipedia.org/wiki/PointCast_(dotcom)
I was assigned to them briefly.
I think this might be something HP people were telling each other at the time while getting smacked around by Sun. The Sun machines were still ridiculously expensive and while common, I don't think they were the default. SGI Indies were also quite popular and the Pentium Pro quickly started eating the overpriced Unix workstation market. By 2000, I doubt there were that many SV startup developers sitting in front of a Sun box. There were piles of them in server rooms, of course.
This meant that code written on Sun that performed well performed poorly on HP-UX occasionally -- not because HP-UX was a worse platform, but because corner cases in architecture could bite you. On your tier1 development platform you tend to weed those issues out early.
But on HP-UX 11 we still had a version of aCC that was a mix between ANSI and K&R, so it was the variant with more pre-processor logic.
GCC was only accepted by customers with Linux installs.
Asking GCC to be available in any sort was not an option for other customers. In fact was a deciding factor why we got some contracts back then, as we always cope with whatever they had.
I stopped following their smartphone efforts though