On Tue, Nov 27, 2012 at 12:20 AM, Steven Bosscher <stevenb....@gmail.com> wrote: > Hello, > > This patch performs some necessary TLC on slim RTL printing in sched-vis.c: > > * Make it independent of the scheduler. Actually it already was, > mostly. This patch completes the job. > > * Harmonize dumping templates for INSN_UID. > > * Always print the pattern of a CALL_INSN. > > * Don't print "jump" for sched dumps only. > > * Harmonize function names: > - print_* print to a "char *" buffer > - dump_* print to a given "FILE *" > - debug_* print to stderr. > > Some of the ideas come from Richard S.'s > http://gcc.gnu.org/ml/gcc-patches/2012-11/msg00284.html. > > The next patch addresses a bigger problem: The print_* variants should > use something better than a fixed-size buffer. I plan to use > pretty-print. An obstack would also work; but I'd like to keep one > layer of abstraction between the file output and the printers, so that > users can post-process the string if necessary (the case I'm thinking > of, is the graph dumpers, which need to escape some characters before > writing to a file). Any thoughts on this from anyone? > > Bootstrapped&tested on powerpc64-unknown-linux-gnu. OK for trunk? > (Yes, I know it's stage3, but look at it this way: the patch does no > harm, and good dumps help for debugging real bugs ;-)
Ok! Thanks, Richard. > Ciao! > Steven
