ADAPTIVE_GIANT benchmark

I reran the 12-concurrent tarball extraction with adaptive giant
enabled/disabled, on the 12-processor e4500:

real time:

x adaptive-giant
+ no-adaptive-giant
+--------------------------------------------------------------------------+
|              x    +   x                                                  |
|           +  x +  +   x                                                  |
|   x       + xx *x ++ xx+        +                                        |
|   x       + ** ** *+ xx*+ +     +                                        |
|   x      x+ **+** ** *x**** *   * +x  +                                  |
|   x x x  x+x*****************x*x* +x+ *   +                              |
|x  xxx **+x*x*********************x+x* * ++++  + + +x  +                 +|
|           |________A_________|                                           |
|             |_________M_A__________|                                     |
+--------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x 100         77.48         89.74         82.17       82.2382     2.2768285
+  99         79.02         94.62         82.77      83.29596     2.6837291
Difference at 95.0% confidence
        1.05776 +/- 0.691261
        1.28621% +/- 0.840559%
        (Student's t, pooled s = 2.48758)

system time:

x adaptive-giant
+ no-adaptive-giant
+--------------------------------------------------------------------------+
|                                  +x                                      |
|                      x           +x +                                    |
|                   x  x   xxx*  x +* +                                    |
|                   xx x+  xx**  * +* +   +   +                            |
|                   xx x*  x***x * ** ++ x+** +  +       +                 |
|              x    xxxx*x+x******x****++*+** +++++   x  ++                |
|x     x    x x*xx****x**x******************* *++**++***+*+      *+    +  +|
|                   |_________MA__________|                                |
|                          |__________MA___________|                       |
+--------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x 100        428.87        515.95        468.41      469.8407     15.138894
+  99        448.08        528.59        478.97     480.89152     15.913392
Difference at 95.0% confidence
        11.0508 +/- 4.31528
        2.35203% +/- 0.918455%
        (Student's t, pooled s = 15.529)

user time

x adaptive-giant
+ no-adaptive-giant
+--------------------------------------------------------------------------+
|                                   xx     +                               |
|                      x            xx     +         +                     |
|                      x        xxx *x *x *++  +   +++                     |
|                     xx +xxx   xx*+*x+**+*++ +* x *+++                    |
|                  x xx*x+xx*+*+*x***x******* **+* *++++        +          |
|x+    x   *  +  +x* xx*x*x***********************+*+**+*++xx  x+ +x+     +|
|                       |___________A__________|                           |
|                            |____________A___________|                    |
+--------------------------------------------------------------------------+
    N           Min           Max        Median           Avg        Stddev
x 100         27.66         28.96         28.35       28.3449    0.22686932
+  99         27.67          29.1         28.47     28.459899    0.24268269
Difference at 95.0% confidence
        0.114999 +/- 0.0652666
        0.405713% +/- 0.230259%
        (Student's t, pooled s = 0.234869)

So adaptive giant is in fact a small net win in all 3 measurements.
I'm not sure whether my previous measurements with generic build were
incorrect or have some other explanation.

Kris