On 10/18/21 20:33, Richard Henderson wrote: > On 10/18/21 11:18 AM, WANG Xuerui wrote: >> On 10/19/21 01:29, Richard Henderson wrote: >>> On 10/18/21 8:38 AM, WANG Xuerui wrote: >>>> >>>> For now any implementation would suffice, and I already saw one or >>>> two bugs in the output during my TCG host work, but it surely would >>>> be nice to switch to generated decoder in the future. The >>>> loongarch-opcodes tables could be extended to support peculiarities >>>> as exhibited in the v1.00 ISA manual and binutils implementation, >>>> via additional attributes, and I'm open to such contributions. >>> >>> Perhaps it would be easiest to re-use the decodetree description? >>> See e.g. target/openrisc/disas.c. >>> >> Indeed; I didn't thought of disassemblers in target/ instead of >> disas/. That would be the most elegant way forward! > > > The one quirk will be that so far using decodetree for disas is limited > to the target, whereas you'll want this for host as well. It shouldn't > be a big deal, just a small matter of the correct build rules.
Oh, good to know. OTOH I expect very few developers to look at host disas.
