scripts: Reduce parallelism in "all" builds

Kernel builds are a bit asymmetrical in as far as CPU and memory
resources go.  Compilation itself tends to scale fairly linearly in
RAM usage per number of CPUs.  However, the link phase tends to use a
fairly constant large amount of RAM and generally only one CPU.

The existing attempt to build more in parallel has mostly resulted in
too many link steps running, which ends up swapping.

Instead of dividing the CPUs by the number of targets, just divide it
by two, and assume this will at least get us some compilation
happening during the link phase of one target build.

Change-Id: I0f8cac2d73600700d8ad6cb2a1b98e9529bb53a8
Signed-off-by: David Brown <davidb@codeaurora.org>

0 files changed, 0 insertions, 0 deletions