From mboxrd@z Thu Jan 1 00:00:00 1970 From: Christopher Li Subject: Re: [PATCH 13/29] add insert_phi_node() Date: Fri, 18 Aug 2017 17:17:12 -0400 Message-ID: References: <20170816153455.97693-1-luc.vanoostenryck@gmail.com> <20170816153455.97693-14-luc.vanoostenryck@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" Return-path: Received: from mail-yw0-f170.google.com ([209.85.161.170]:34354 "EHLO mail-yw0-f170.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751446AbdHRVRN (ORCPT ); Fri, 18 Aug 2017 17:17:13 -0400 Received: by mail-yw0-f170.google.com with SMTP id s143so65932063ywg.1 for ; Fri, 18 Aug 2017 14:17:13 -0700 (PDT) In-Reply-To: Sender: linux-sparse-owner@vger.kernel.org List-Id: linux-sparse@vger.kernel.org To: Luc Van Oostenryck Cc: Linux-Sparse , Linus Torvalds On Fri, Aug 18, 2017 at 4:52 PM, Luc Van Oostenryck wrote: > On Fri, Aug 18, 2017 at 10:39 PM, Christopher Li wrote: >> > > It's not needed because: > 1) phi-nodes need to have 'parallel-assignment' semantics anyway > like in languages where you can write 'a, b = b, a' to exchange > two variables. In other words, if the order would matter it would > be a bug. I need to think more about that. > 2) you can never have (the target of) a phi-node (OP_PHI) > as another phi-node' source because all such sources are > created by OP_PHISOURCEs. I see. So that means you can just insert into the first instruction? Chris