On 11/11/24 1:30 PM, Jeff Law wrote:
+
+void
+emit_crc (machine_mode crc_mode, rtx* crc, rtx* op0)
+{
+ if (word_mode != crc_mode)
+ {
+ rtx tgt = simplify_gen_subreg (word_mode, *op0, crc_mode, 0);
Can CRC_MODE ever be wider than WORD_MODE?
If so, then that last argument needs adjustment to cope with endianness.
On the optimistic assumption that we're always generating a
paradoxical, I'm putting a gcc_assert in my local copy for testing
purposes.
Answering my own question. There's an assert in patch #2 which should
keep us safe on this issue.
jeff
