search for: construct_buf_ref

...ddressed via a normal pointer, then *maybe* we can paper over this, but if you're mixing pointer and non-pointer accesses to the same address, you have a potentially serious problem. As a concrete example, let's take the following: %mem = malloc() <-- returns a noalias pointer %buf = construct_buf_ref(%mem) <-- assume this is transparent and doesn't capture load %mem buffer_store(%buf) load %mem In this (psuedo code) example, basicaa will use capture tracking to *prove* that buffer_store can't have written to %mem. buffer_store is allowed to write to any memory it can address, bu...

On 19.03.2016 14:47, Philip Reames wrote: > I'm generally in support of this change. I haven't looked at the patch > yet, but the direction seems worthwhile. > > Note that we already have a writeonly predicate in a few places in the > code (BasicAA being one). If we do introduce the new intrinsic > property, we should refactor all of these places to use the new >