ia64/lib/strlen.S

1 /* SPDX-License-Identifier: GPL-2.0 */
14  * Copyright (C) 1999, 2001 Hewlett-Packard Co
30 //	  The goal is to look at the string in chunks of 8 bytes.
32 //	  string may not be 8-byte aligned. In this case we load the 8byte
55 //	- the cmp r0,r0 is used as a fast way to initialize a predicate
56 //	  register to 1. This is required to make sure that we get the parallel
59 //	- we don't use the epilogue counter to exit the loop but we need to set
62 //	- after the loop we must test for Nat values because neither the
68 //	- Clearly performance tuning is required.
86 	alloc saved_pfs=ar.pfs,11,0,0,8 // rotating must be multiple of 8
88 	.rotr v[2], w[2]	// declares our 4 aliases
92 	dep src=0,in0,0,3	// src=8byte-aligned in0 address
99 	ld8 v[1]=[src],8	// must not speculate: can fail here
100 	shl tmp=tmp,3		// multiply by 8bits/byte
101 	mov mask=-1		// our mask
103 	ld8.s w[1]=[src],8	// speculatively load next
107 	shr.u	mask=mask,tmp	// zero enough bits to hold v[1] valuable part
110 	add base=-16,src	// keep track of aligned base
111 	or v[1]=v[1],mask	// now we have a safe initial byte pattern
113 1:
114 	ld8.s v[0]=[src],8	// speculatively load next
115 	czx1.r val1=v[1]	// search 0 byte from right
116 	czx1.r val2=w[1]	// search 0 byte from right following 8bytes
118 	ld8.s w[0]=[src],8	// speculatively load next to next
119 	cmp.eq.and p6,p0=8,val1	// p6 = p6 and val1==8
120 	cmp.eq.and p6,p0=8,val2	// p6 = p6 and mask==8
121 (p6)	br.wtop.dptk 1b		// loop until p6 == 0
125 	// val1_is_nat || (val1==8 && val2_is_nat)
128 	//	- there must be a better way of doing the test
130 	cmp.eq  p8,p9=8,val1	// p6 = val1 had zero (disambiguate)
137 	cmp.eq.and  p7,p0=8,val1// val1==8?
142 (p8)	adds src=-16,src	// correct position when 3 ahead
143 (p9)	adds src=-24,src	// correct position when 4 ahead
146 	sub tmp=8,val1		// which byte in word
150 	mov ar.pfs=saved_pfs	// because of ar.ec, restore no matter what
167 	//	- today we restart from the beginning of the string instead
171 	ld8 val=[base],8	// will fail if unrecoverable fault
180 (p6)	ld8 val=[base],8	// will fail if unrecoverable fault
184 	cmp.eq p6,p0=8,val1	// val1==8 ?
188 	sub tmp=8,val1
191 	sub ret0=ret0,tmp	// length=now - back -1
192 	mov ar.pfs=saved_pfs	// because of ar.ec, restore no matter what