"Bingfeng Mei" <[EMAIL PROTECTED]> writes: > I tried to use doloop_end pattern to reduce loop overhead for our target > processor, which features a dedicated loop instruction. Somehow even a > simple loop just cannot pass the test of doloop_condition_get, which > requires following canonical pattern.
You are looking at this at the wrong level. That comment is describing what your doloop insn must look like in the MD file. It does not mean that the RTL for the loop must look like that in order for the doloop optimization to apply. Look at existing targets which support doloops for examples of how to write the insn. Ian