2 * arch/alpha/lib/ev67-strlen.S
3 * 21264 version by Rick Gorton <rick.gorton@alpha-processor.com>
5 * Finds length of a 0-terminated string. Optimized for the
8 * - memory accessed as aligned quadwords only
9 * - uses bcmpge to compare 8 bytes in parallel
11 * Much of the information about 21264 scheduling/coding comes from:
12 * Compiler Writer's Guide for the Alpha 21264
13 * abbreviated as 'CWG' in other comments here
14 * ftp.digital.com/pub/Digital/info/semiconductor/literature/dsc-library.html
15 * Scheduling notation:
17 * U - upper subcluster; U0 - subcluster U0; U1 - subcluster U1
18 * L - lower subcluster; L0 - subcluster L0; L1 - subcluster L1
28 ldq_u $1, 0($16) # L : load first quadword ($16 may be misaligned)
30 insqh $2, $16, $2 # U :
31 andnot $16, 7, $0 # E :
34 cmpbge $31, $1, $2 # E : $2 <- bitmask: bit i == 1 <==> i-th byte == 0
38 $loop: ldq $1, 8($0) # L :
39 addq $0, 8, $0 # E : addr += 8
40 cmpbge $31, $1, $2 # E :
46 subq $0, $16, $0 # E :