On Thu, Jul 5, 2012 at 11:23 AM, Richard Guenther <richard.guent...@gmail.com> wrote: > On Thu, Jul 5, 2012 at 11:08 AM, Dehao Chen <de...@google.com> wrote: >> Hi, >> >> This patch added block field to phi_arg_d to make sure the associated >> source locus is consistent with its block info. >> >> Bootstrapped and passed gcc regression tests. >> >> OK for trunk? > > Hum - makes me want a location like we have on RTL (which maps to > location plus block ...). But oh well ... > > struct GTY(()) phi_arg_d { > /* imm_use MUST be the first element in struct because we do some > pointer arithmetic with it. See phi_arg_index_from_use. */ > struct ssa_use_operand_d imm_use; > location_t locus; > + tree block; > }; > > please place block before locus though. > > Ok with that change.
Ehm, wait a moment, please -- does this have to be a tree pointer? Why not use BLOCK_NUMBER instead? Adding a pointer to such a frequently-used structure should be avoided if an int suffices. Ciao! Steven