On Wed, 25 May 2016, Nathan Sidwell wrote: > This patch reimplements crt0 as C rather than assembly. That means it;ll be > good with 32 bit addresses. I've also completely moved abort and exit into > newlib (a fork on github), which had it's own copies that failed to set the > exit code. Things happened to work because they were never found, being > hidden by the ones provided in crt0.s. You'll need to update your newlib > sources for exit codes to continue to work.
Doesn't this break offloading compilation? When you link offloaded code against new newlib, nothing provides a definition of __exitval_ptr (crt0 is only linked in with -mmainkernel), and PTX does not support ELF semantics for undefined weak symbols where they get a null address. OpenACC testcases that call abort() should get a link error after this change. Also note that &__exitval_ptr is always non-null for that reason. > Alex, this should make it simple to add a check in __main for a single thread > soft stack allocation. I imagine a check on a weakly declared __soft_stack > symbol or something? For the above reason weak linkage is not usable in that fashion. Code on my branch simply sets up soft stack in __main regardless of whether it's used. > Or is that something nvptx-run should check for and initialize? The interface between nvptx-run and crt0 would need to change to allow that. Alexander
