This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |
Other format: | [Raw text] |
On 21.10.2010 4:30, Jie Zhang wrote:Thanks. Then let's wait for review from a scheduler maintainer.On 10/21/2010 04:44 AM, Richard Sandiford wrote:Jie Zhang<jie@codesourcery.com> writes:Andrey and scheduler maintainers?Currently TARGET_SCHED_REORDER2 returns mips_issue_rate (), which will reset can_issue_more to the issue_rate after scheduling each instruction. See haifa-sched.c:sched_block for the details. Because of this, issue_rate is not honored when scheduling. This caused PR rtl-optimization/37360 since the issue_rate is set to a value smaller than the number of instructions that scheduler DFA can issue in one cycle for mips sb1.
The haifa-sched.c part looks good to me, but I can't approve it. Thanks for tracking down the underlying problem.
I'm all for the patch, life is easier for the scheduler when nobody lies. I cannot approve it though. As I wrote earlier, we just need to be on the lookout for possible fallout on other targets which may have the same problem.
Thanks for testing.For an extra caution, I have tried all tests from 45352 with all flag combinations and -fschedule-insns[2] instead of -fselective-scheduling[2], and they work fine. This is with both your other haifa-sched.c patch and this one applied.
Regards, -- Jie Zhang CodeSourcery
Index Nav: | [Date Index] [Subject Index] [Author Index] [Thread Index] | |
---|---|---|
Message Nav: | [Date Prev] [Date Next] | [Thread Prev] [Thread Next] |