runtime: fix 32B backward copy on ppc64x

runtime: fix 32B backward copy on ppc64x

The test to enter the 32b copy loop always fails, and execution
falls back to a single 8B/iteration copy loop for copies of more
than 7 bytes. Likewise, the 32B loop has SRC/DST args mixed,
and fails to truncate DWORDS after completing.

Fix these, and unroll the 8B/iteration loop as it will only
execute 1-3 times if reached.

POWER10 benchmarks:

name                             old speed      new speed       delta
MemmoveOverlap/32                5.28GB/s ± 0%  10.37GB/s ± 0%   +96.22%
MemmoveOverlap/64                5.97GB/s ± 0%  18.15GB/s ± 0%  +203.95%
MemmoveOverlap/128               7.67GB/s ± 0%  24.35GB/s ± 0%  +217.41%
MemmoveOverlap/256               14.1GB/s ± 0%   25.0GB/s ± 0%   +77.48%
MemmoveOverlap/512               14.2GB/s ± 0%   30.9GB/s ± 0%  +118.19%
MemmoveOverlap/1024              12.3GB/s ± 0%   36.4GB/s ± 0%  +194.75%
MemmoveOverlap/2048              13.7GB/s ± 0%   48.8GB/s ± 0%  +255.24%
MemmoveOverlap/4096              14.1GB/s ± 0%   43.4GB/s ± 0%  +208.80%
MemmoveUnalignedDstOverlap/32    5.07GB/s ± 0%   3.78GB/s ± 0%   -25.33%
MemmoveUnalignedDstOverlap/64    6.00GB/s ± 0%   9.59GB/s ± 0%   +59.78%
MemmoveUnalignedDstOverlap/128   7.66GB/s ± 0%  13.51GB/s ± 0%   +76.42%
MemmoveUnalignedDstOverlap/256   13.4GB/s ± 0%   24.3GB/s ± 0%   +80.92%
MemmoveUnalignedDstOverlap/512   13.9GB/s ± 0%   30.3GB/s ± 0%  +118.29%
MemmoveUnalignedDstOverlap/1024  12.3GB/s ± 0%   37.3GB/s ± 0%  +203.07%
MemmoveUnalignedDstOverlap/2048  13.7GB/s ± 0%   45.9GB/s ± 0%  +235.39%
MemmoveUnalignedDstOverlap/4096  13.9GB/s ± 0%   41.2GB/s ± 0%  +196.34%
MemmoveUnalignedSrcOverlap/32    5.13GB/s ± 0%   5.18GB/s ± 0%    +0.98%
MemmoveUnalignedSrcOverlap/64    6.26GB/s ± 0%   9.53GB/s ± 0%   +52.29%
MemmoveUnalignedSrcOverlap/128   7.94GB/s ± 0%  18.40GB/s ± 0%  +131.76%
MemmoveUnalignedSrcOverlap/256   14.1GB/s ± 0%   25.5GB/s ± 0%   +81.40%
MemmoveUnalignedSrcOverlap/512   14.2GB/s ± 0%   30.9GB/s ± 0%  +116.76%
MemmoveUnalignedSrcOverlap/1024  12.4GB/s ± 0%   46.4GB/s ± 0%  +275.22%
MemmoveUnalignedSrcOverlap/2048  13.7GB/s ± 0%   48.7GB/s ± 0%  +255.16%
MemmoveUnalignedSrcOverlap/4096  14.0GB/s ± 0%   43.2GB/s ± 0%  +208.89%

Change-Id: I9fc6956ff454a2856d56077d1014388fb74c1f52
Reviewed-on: https://go-review.googlesource.com/c/go/+/384074
Trust: Paul Murphy <murp@ibm.com>
Run-TryBot: Paul Murphy <murp@ibm.com>
Reviewed-by: Lynn Boger <laboger@linux.vnet.ibm.com>
Reviewed-by: Cherry Mui <cherryyz@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>