Make strcmp do four byte comparisons on aligned strings.
Used libMicro's strcmp test to evaluate performance:
OPTS="-E -C 200 -S -W"
bin/strcmp $OPTS -N "strcmp_10" -s 10 -I 10
bin/strcmp $OPTS -N "strcmp_100" -s 100 -I 20
bin/strcmp $OPTS -N "strcmp_1k" -s 1k -I 50
bin/strcmp $OPTS -N "strcmp_10k" -s 10k -I 800
bin/strcmp $OPTS -N "strcmp_1m" -s 1m -I 500000
bin/strcmp $OPTS -N "strcmp_10m" -s 10m -I
5000000
Before:
prc thr usecs/call samples errors cnt/samp size
strcmp_10 1 1 0.02510 201 0 10000 10
strcmp_100 1 1 0.17520 169 0 5000 100
strcmp_1k 1 1 1.67700 177 0 2000 1024
strcmp_10k 1 1 17.24800 194 0 125 10240
strcmp_1m 1 1 17892.00000 160 0 1
1048576
strcmp_10m 1 1 183136.00000 201 0 1
10485760
After:
strcmp_10 1 1 0.01800 194 0 10000 10
strcmp_100 1 1 0.13540 190 0 5000 100
strcmp_1k 1 1 1.24950 188 0 2000 1024
strcmp_10k 1 1 12.85600 190 0 125 10240
strcmp_1m 1 1 12930.00000 170 0 1
1048576
strcmp_10m 1 1 134382.00000 195 0 1
10485760