sha1-armv4-large.pl: add performance data for Cortex A8 core.

#	modes are limited. As result it takes more instructions to do

#	the same job in Thumb, therefore the code is never twice as

#	small and always slower.

# [***]	which is also ~35% better than compiler generated code.

# [***]	which is also ~35% better than compiler generated code. Dual-

#	issue Cortex A8 core was measured to process input block in

#	~990 cycles.

while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}

open STDOUT,">$output";