Timings

[Mihail Konev]

In mesurements below, caches were warm (i.e. configure; make; make; touch; time make).
Speedups are not purely linking ones, as they include non-recursifying the make.

  CCLD     libxservertest.la
real	0m2.192s
user	0m1.367s
sys	0m0.758s

  CCLD     protocol-xchangedevicecontrol
real	0m0.682s
user	0m0.588s
sys	0m0.079s

  CCLD     tests
real	0m2.103s
user	0m1.508s
sys	0m0.450s

Worst case: make -j12, i.e. 6 physical CPU cores

  Recursive test/:
        test/libxservertest.la + test/* + test/xi1/* + test/xi2/* ~=
                  2            +  0.5   +    0.5     +    0.5     ~=
        3.5

  Speedup:
        Recursive test/ - tests ~=
              3.5       -   2   ~=
        1.5 sec

make -j4, Xorg-only rebuild with recursive test/, built in-source:

  # touch os/backtrace.c
  # make
      0m11.25s real

make -j4, Xorg-only rebuild with nonrecursive test/, built out-of-source:

  # touch os/backtrace.c
  # make
      0m07.30s real

  Speedup: (11 - 7) = 4 sec = 100% * 4 / 11 = 36%

make -j4, non-DMX, non-KDrive rebuild with recursive test/, built in-source:

  # touch os/backtrace.c
  # make
    0m20.34s real

make -j4, non-DMX, non-KDrive rebuild with nonrecursive test/, built out-of-source:

  # touch os/backtrace.c
  # make
    0m12.66s real # cache too warm?
    0m12.65s real # maybe it's srcdir != builddir

  Speedup: (20 - 12) = 8 sec = 100% * 8 / 20 = 40%

make -j4, non-DMX, non-KDrive rebuild with nonrecursive test/, built in-source:

  # touch os/backtrace.c
  # make
    0m12.27s real

  Speedup: 40%