> Just noticed, not a problem of the revert but wasn't that wrong before > without the early-clobber? > vwsll.vx v0, v0, a0 for LMUL = 2 would be allowed but should not? > We probably don't have tests for this, in particular runtime?
Yes, you are right according to the spec "the overlap is in the highest-numbered part of the destination register group" when lmul >= 1. We need to take care of this carefully when support overlap in GCC-15, specifically making sure every scenarios are well tested as you mentioned. Pan -----Original Message----- From: Robin Dapp <rdapp....@gmail.com> Sent: Wednesday, April 24, 2024 10:12 PM To: Li, Pan2 <pan2...@intel.com>; gcc-patches@gcc.gnu.org Cc: rdapp....@gmail.com; juzhe.zh...@rivai.ai; kito.ch...@gmail.com Subject: Re: [PATCH v1] Revert "RISC-V: Support highpart register overlap for vwcvt" > (define_insn "@pred_vwsll<mode>_scalar" > - [(set (match_operand:VWEXTI 0 "register_operand" "=vd, vr, vd, vr, vd, vr, > vd, vr, vd, vr, vd, vr, ?&vr, ?&vr") > + [(set (match_operand:VWEXTI 0 "register_operand" "=vr, vr") Just noticed, not a problem of the revert but wasn't that wrong before without the early-clobber? vwsll.vx v0, v0, a0 for LMUL = 2 would be allowed but should not? We probably don't have tests for this, in particular runtime? Regards Robin
