In the progress of getting hypertransport setup completely dynamic, I made the following patch that completely gets rid of the CONNECTION and link name macros in coherent_ht.c (UP, ACROSS, DOWN and CONNECTION_0_1, CONNECTION_0_2, CONNECTION_1_3) Therefore I added two simple helper functions that gain the needed information from generate_row(). This patch probably raises the register pressure during ht setup, but even on a 4p machine I could not see any problems with romcc. Dropping another couple of hardcodes seems like the right thing to do[tm].
If nobody is opposing, I am going to check this into CVS.
Stefan
On Wed, 24 Mar 2004, Stefan Reinauer wrote:
If nobody is opposing, I am going to check this into CVS.
I like it.
ron
Stefan Reinauer stepan@suse.de writes:
In the progress of getting hypertransport setup completely dynamic, I made the following patch that completely gets rid of the CONNECTION and link name macros in coherent_ht.c (UP, ACROSS, DOWN and CONNECTION_0_1, CONNECTION_0_2, CONNECTION_1_3) Therefore I added two simple helper functions that gain the needed information from generate_row(). This patch probably raises the register pressure during ht setup, but even on a 4p machine I could not see any problems with romcc. Dropping another couple of hardcodes seems like the right thing to do[tm].
If nobody is opposing, I am going to check this into CVS.
If that works and compiles it should certainly make things much less error prone.
Eric