This patch is version 9 of my patches to add -mcpu=future support to the PowerPC.
The version 8 patches were here: * https://gcc.gnu.org/pipermail/gcc-patches/2025-November/700537.html In particular, it answers questions from Surya Kumari Jangala * https://gcc.gnu.org/pipermail/gcc-patches/2025-November/701351.html Like the version 8 patches, there are two patches in this group. The first patch is revised as documented below. The second patch is unchanged. In terms of changes to the first patch: I added a commit log message in the first patch. I changed macros to macro in the ChangeLog message. I added an entry for the cpu=future entry in rs6000-cpus.def. The entry was actually there but in the wrong place, and I rewrote the message. I simplified the ChangeLog message for PROCESSOR_FUTURE. I added a ChangeLog message for rs6000.cc. In terms of: > @@ -24463,6 +24465,7 @@ static struct rs6000_opt_mask const rs6000_opt_masks[] = > { "float128", OPTION_MASK_FLOAT128_KEYWORD, false, true }, > { "float128-hardware", OPTION_MASK_FLOAT128_HW, false, true }, > { "fprnd", OPTION_MASK_FPRND, false, true }, > + { "future", OPTION_MASK_FUTURE, false, false }, The last entry should be 'true'. And better to have this entry after the entry for Power 11, so that t will be simpler when we replace 'future' with the new future processor's name. Historically the entries were alphabetical, but I moved it to be after power11. However, I kept the last entry (valid_target) to be "false" (just like it is for "power11"). The rs6000_opt_masks table has two functions. The first function is to print out the options that are set with -mdebug=reg. That function does not check valid_target. The section function is whether the target attribute and target pragma support allows the user to change that option. If valid_target is false, it will not allow the user to set this option. I.e. if they want to enable future, they have to use cpu=future. I removed the verbage about what -mcpu=future is for in the texinfo files, but I left the line saying future is a valid -mcpu= option. -- Michael Meissner, IBM PO Box 98, Ayer, Massachusetts, USA, 01432 email: [email protected]
