On Thu, 2024-10-31 at 17:06 +0100, Jan Hubicka wrote: > Hi, > this patch lets us to remove > void *mem = __builtin_malloc (s); > if (!mem) > __builtin_abort (); > __builtin_free (mem); > > Where we previously stopped on "if (!mem)" marking __builtin_malloc > necessary. > This is done by matching such conditional in tree-ssa-dce.cc and > playing same game > as we do for free functions. We consider if itself necessary (since > it does > control flow) but we do not propagate necessity over the SSA name. > > In removal stage we check whether SSA name is needed and if not, we > replace > it by constant 1, which makes the conditonal to be folded as constant > in > cfgccleanup. > > While working on this, I refactored the code a bit to avoid duplicate > matching > of allocation functions which made me notice that we handled > strdup/strndup just > half way and this prevented us from removing it, which is also fixed > and tested > in the testcase. > > Note that testcase pr68078.f90 checks that failed allocation is > reported > when memory limits are set. We no longer report it since the memory > is > unused and that is why I needed to disable DCE. > > I alos tested that with Jakub's patch we now eliminate new operator > with > std::nothrow_t. > > Bootstrapped/regtested x86_64-linux, OK?
I have a vague memory that one of the tests in SPEC has a loop that tries to malloc, doubling the size each time, until it fails. Would the patch change the behavior of such a loop? Sorry, I can't remember further details. Dave
