Bnea
bnea $2,$3, addr
branch not equal, meanwhile, add SPR BR_INCR
minor tweak reduces some loop
overheads by %50
tagswi $0,$5, 3
tagswi $0,$5, 2
tagswi $0,$5, 1
tagswi $0,$5, 0
li
$4, kCacheSize - (kCacheLineSize << 2)
mtsri BR_INCR, (kCacheLineSize << 2)
bnea+ $5, $4,
___cache_invalidate_loop
___cache_invalidate_loop:
li $5, 0
BR_INCR is callee-saved