rdmsr vs rdpmc

[Philippe E. <ph...@wa...>]

applying the joined patch I get this results for
overhead (Celeron P6 core)

# real	user	system

# bz2-test-300000.20021117-142631.out (3 run)
2 372.35 (1.33%) 368.82 (1.39%) 3.48 (-3.33%)
2 372.25 (0.84%) 368.57 (0.85%) 3.62 (-0.55%)  pmc

# bz2-test-100000.20021117-142631.out (3 run)
3 373.06 (1.52%) 369.39 (1.54%) 3.60 (-0.00%)
3 374.23 (1.37%) 370.57 (1.40%) 3.56 (-2.20%)  pmc

# bz2-test-25000.20021117-142631.out (3 run)
4 382.06 (3.97%) 378.19 (3.96%) 3.82 (6.11%)
4 381.93 (3.46%) 378.15 (3.47%) 3.74 (2.75%)   pmc


# kernel-compile-300000.20021117-135204.out (3 run)
2 467.60 (1.10%) 433.16 (0.70%) 32.85 (1.51%)
2 467.22 (1.10%) 433.34 (0.71%) 32.28 (1.29%)  pmc

# kernel-compile-100000.20021117-135204.out (3 run)
3 473.86 (2.45%) 437.39 (1.68%) 32.59 (0.71%)
3 473.52 (2.46%) 435.88 (1.30%) 33.83 (6.15%)  pmc

# kernel-compile-25000.20021117-135204.out (3 run)
4 498.04 (7.68%) 451.66 (5.00%) 34.93 (7.94%)
4 496.14 (7.35%) 452.01 (5.05%) 32.78 (2.86%)  pmc

bz2 test show clearly an improvment, kernel test
compile don't show any regresion. I dunno why we
don't see improvment for kernel compile.

regards,
Phil