Re: Remove obfuscation from the scheduler

[Jeff Law <law at redhat dot com>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 05/19/11 07:37, Bernd Schmidt wrote:
> In the scheduler initialization code, we want to set up the h_i_d array.
> So we call haifa_init_h_i_d, which has a table of function pointers
> which it passes to sched_scan, which copies the function pointer table
> pointer to a global variable and then calls functions like init_insn,
> which checks the global variable whether one of the function pointers is
> nonnull, and if not calls the function we actually want (init_h_i_d).
> 
> Half of the arguments to sched_scan are never actually used. sel-sched
> pretends to use the insns vector, but it actually passes in a global
> variable that's always NULL. Another argument can be eliminated by
> simply calling the necessary function(s) directly rather than going
> through sched_scan.
> 
> This patch removes most of the craziness. The forest of indirections is
> gone, and sched_scan is moved (in a much reduced form) to sel-sched-ir.c
> where the sel-sched maintainers can decide whether to eliminate it
> altogether.
> 
> Bootstrapped on i686-linux; regression tests running. Ok?
Ok.
jeff

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJN2ojZAAoJEBRtltQi2kC7hyYIAJ2IZ91/I4o0fqNljWPQy33o
7hgnV6Uy790QAlrfIVhQ/K4ceJRyo/0oHe/L/R87lemqronnUZiSSvA917G1mXEb
LbjzjK0meGEl9pZZ3kwiCkJIWyUGqpEhDlRlVICqabpnFyu6fsOPJVrG9ZI8IIbz
4t+BwrawtB4+kAcsDZR2XvCEH4mte9an7nRlWUrHLvXuNWR8hxUS6T2Mx8dTtRH9
tC4BEodedbMGHeg8Sz3nAAftZ6SvGdS3K/Huwpl+VZ1uL2d2w+bTkgoLxobvNxtU
hqrRgKlYGaeDuqgUyDmayPR9mNwM9GC7OXsJ8n12vmVe6p9FlFPV5yN/d8hCfpU=
=aYbv
-----END PGP SIGNATURE-----