From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.7.7) Gecko/20050416 Fedora/1.0.3-1.3.1 Firefox/1.0.3 Description of problem: PAPI/perfctr does not seem to work on Red Hat. Version-Release number of selected component (if applicable): How reproducible: Always Steps to Reproduce: 1. Install kernel patch from PAPI: http://icl.cs.utk.edu/papi/ 2. load perfctr module 3. run papi test suite Actual Results: tests fail. PAPI tries to make a /proc/self/perfctr file and does not succeed. Expected Results: PAPI test suite should pass. Additional info: dmesg contains the following, so something somewhere is partly working: PERFCTR INIT: vendor 0, family 15, model 2, stepping 4, clock 1799842 kHz PERFCTR INIT: NITER == 64 PERFCTR INIT: loop overhead is 512 cycles PERFCTR INIT: rdtsc cost is 82.8 cycles (5816 total) PERFCTR INIT: rdpmc cost is 372.1 cycles (24332 total) PERFCTR INIT: rdmsr (counter) cost is 251.7 cycles (16624 total) PERFCTR INIT: rdmsr (escr) cost is 166.8 cycles (11188 total) PERFCTR INIT: wrmsr (counter) cost is 849.8 cycles (54900 total) PERFCTR INIT: wrmsr (escr) cost is 921.0 cycles (59456 total) PERFCTR INIT: read cr4 cost is 5.1 cycles (840 total) PERFCTR INIT: write cr4 cost is 253.3 cycles (16728 total) PERFCTR INIT: rdpmc (fast) cost is 61.5 cycles (4448 total) PERFCTR INIT: rdmsr (cccr) cost is 167.6 cycles (11244 total) PERFCTR INIT: wrmsr (cccr) cost is 873.8 cycles (56436 total) PERFCTR INIT: write LVTPC cost is 33.6 cycles (2668 total) PERFCTR INIT: sync_core cost is 261.6 cycles (17260 total) perfctr: driver 2.6.13 DEBUG, cpu type Intel P4 at 1799842 kHz
Steven, what's this bug about? We don't support custom-built kernels. Are you trying to propose an enhancement? Or is there some problem with a stock RHEL3 kernel? Thanks in advance for more info. -ernie