I used this script to run buildworld 31 times with each kernel configuration. I threw away the first data point from each case. The "base" case is the given Subversion revision. The "1ppq" case adds this patch. Get the raw data here.
CPU: Intel(R) Xeon(R) CPU E5-2420 0 @ 1.90GHz (1900.04-MHz K8-class CPU) FreeBSD/SMP: 1 package(s) x 6 core(s) x 2 hardware threads head r315335 make -j8 buildworld
x base_j8_c12 + 1ppq_j8_c12 +--------------------------------------------------------------------------+ | + | | + | | + + | | + ++ x x | | + ++ x x | | ++++++ + x x xx x x | |++ + + ++*++++* x*x x xxx x x xxxxxxxx x| | |______A_M_____| |______________A_______M______| | +--------------------------------------------------------------------------+ N Min Max Median Avg Stddev x 30 5403 5490 5472 5455.8333 27.752705 + 30 5354 5417 5404.5 5401.3667 13.694911 Difference at 95.0% confidence -54.4667 +/- 11.3118 -0.99832% +/- 0.207334% (Student's t, pooled s = 21.8834)
same machine as previous make -j16 buildworld
x base_j16_c12 + 1ppq_j16_c12 +--------------------------------------------------------------------------+ | + x | | + x | | + + x | | + + + x x | | ++ + + + xxxxx | | ++ + + * xxxxx | |+ + + *+*+*+* *++ x xxxxx x x x| | |________A_M_____| |____________A____M_______| | +--------------------------------------------------------------------------+ N Min Max Median Avg Stddev x 30 4877 4976 4955.5 4943.4667 29.205259 + 30 4812 4899 4885.5 4881.1333 18.63392 Difference at 95.0% confidence -62.3333 +/- 12.6627 -1.26092% +/- 0.256149% (Student's t, pooled s = 24.4966)
CPU: Intel(R) Xeon(R) CPU X7560 @ 2.27GHz (2261.05-MHz K8-class CPU) FreeBSD/SMP: 4 package(s) x 8 core(s) x 2 hardware threads head r315421 make -j32 buildworld
x base_j32_c64 + 1ppq_j32_c64 +--------------------------------------------------------------------------+ | + x | | + + x xx | | + + ++ +x+ * xxx x | |+ +x x + + + ++ +***+***+x* xxxxx ** x* x| | |__________|_A_M______A_M_|______| | +--------------------------------------------------------------------------+ N Min Max Median Avg Stddev x 30 1796 1857 1836 1833.3333 11.966428 + 30 1775 1849 1825.5 1823.2333 14.116958 Difference at 95.0% confidence -10.1 +/- 6.76431 -0.550909% +/- 0.368962% (Student's t, pooled s = 13.0859)
same machine as previous cpuset -l 32-63 make -j40 buildworld
x base_j40_c32 + 1ppq_j40_c32 +--------------------------------------------------------------------------+ | x + | | x + | | x x+* xx | | xxx***+x* x | |+ x +x + x + + ++ **x***+**x**++ + +| | |__|____________A__MM________|__| | +--------------------------------------------------------------------------+ N Min Max Median Avg Stddev x 30 2264 2342 2331.5 2327.3667 18.217997 + 30 2260 2363 2333 2327.2667 22.420639 No difference proven at 95.0% confidence