This is a typo/thinko introduced with one of my recent scheduler
patches. Rather than flag_sched_pressure we have to use
sched_pressure_p, which accurately is only set during sched1.
Bootstrapped and tested on i686-linux, and committed as obvious.
Bernd
Index: ChangeLog
===================================================================
--- ChangeLog (revision 172524)
+++ ChangeLog (working copy)
@@ -1,3 +1,9 @@
+2011-04-16 Bernd Schmidt <ber...@codesourcery.com>
+
+ PR target/48629
+ * haifa-sched.c (prune_ready_list, schedule_block): Use
+ sched_pressure_p rather than flag_sched_pressure.
+
2011-04-15 Pat Haugen <pthau...@us.ibm.com>
* config/rs6000/rs6000.c (call_ABI_of_interest): Call
Index: haifa-sched.c
===================================================================
--- haifa-sched.c (revision 172368)
+++ haifa-sched.c (working copy)
@@ -2898,7 +2898,7 @@ prune_ready_list (state_t temp_state, bo
cost = 1;
reason = "asm";
}
- else if (flag_sched_pressure)
+ else if (sched_pressure_p)
cost = 0;
else
{
@@ -3232,7 +3232,7 @@ schedule_block (basic_block *target_bb)
{
memcpy (temp_state, curr_state, dfa_state_size);
cost = state_transition (curr_state, insn);
- if (!flag_sched_pressure)
+ if (!sched_pressure_p)
gcc_assert (cost < 0);
if (memcmp (temp_state, curr_state, dfa_state_size) != 0)
cycle_issued_insns++;
