On Tue, 29 Jul 2025, Jakub Jelinek wrote: > On Tue, Jul 29, 2025 at 11:02:08AM -0400, Hans-Peter Nilsson wrote: > > > If you have a target with bogus MAX_FIXED_MODE_SIZE definition, you should > > > just fix that. > > > > That macro should be avoided, for one because it's misleading > > and ends up being used for other than target tweaking (QED). > > No. > That macro is used in tons of places to mean exactly what is documented. > "An integer expression for the size in bits of the largest integer > machine mode that should actually be used."
It says "should", not "can". > If the target supports TImode as a generally usable integral mode > (e.g. it returns true from targetm.scalar_mode_supported_p (TImode), > rather than just have the mode but not really support say addition/moves > etc. on it), then DImode is definitely not the largest integer machine mode > that should actually be used. The documentation is currently missing "internally" there, like in what gcc generate by expanding operations. Empirically, it can be observed that what I write is true: it's only your generic-type-changes (which I amend in the patch) that cause breakage. brgds, H-P
