This is a summary of discussions relative to the merge request created by Alex
Coplan (acoplan) <[email protected]> titled
doc: Fix description of GET_MODE_MASK
since its creation.
Description: This updates the description of GET_MODE_MASK in the GCC internals
manual. Currently it says that the macro can only be used for modes
whose bitsize is less than or equal to HOST_BITS_PER_INT. That may have
once been correct, but these days it seems to work for modes up to
HOST_BITS_PER_WIDE_INT in size. The code in genmodes.cc:emit_mode_mask
seems to support this. It has:
print_maybe_const_decl ("%sunsigned HOST_WIDE_INT", "mode_mask_array",
"NUM_MACHINE_MODES", adj_nunits);
puts ("\
((m) >= HOST_BITS_PER_WIDE_INT) \\\n\
? HOST_WIDE_INT_M1U \\\n\
: (HOST_WIDE_INT_1U << (m)) - 1\n");
I did wonder whether the second sentence should be rephrased. The macro
_can_ be used for modes whose bitsize is greater than
HOST_BITS_PER_WIDE_INT, it's just that it doesn't tell you very much
(just that the mode is at least as big as a HOST_WIDE_INT). I've left
it as is for now, but would be happy to change it if reviewers think
it's worth doing.
Tested with make html and eyeballing the built docs.
gcc/ChangeLog:
* doc/rtl.texi (Machine Modes): Update description of
GET_MODE_MASK to use HOST_BITS_PER_WIDE_INT instead of
HOST_BITS_PER_INT as the upper limit of the input mode's bitsize.
The full and up to date discussion can be found at
https://forge.sourceware.org/gcc/gcc-TEST/pulls/157
The merge request has been closed without being merged directly on the forge
repository.
On 2026-05-13 09:22:27+00:00, Richard Earnshaw (rearnsha) <[email protected]>
approved the changes:
This is OK.
On 2026-05-14 13:06:25+00:00, Alex Coplan (acoplan) wrote:
Pushed upstream as
https://gcc.gnu.org/cgit/gcc/commit/?id=568e1b4823b076235497e8fc790ded7629e472a4
