#define synd x5
#define shift x6
#define tmp x7
-#define wtmp w7
#define vrepchr v0
#define qdata q1
#define vdata v1
#define vhas_chr v2
-#define vrepmask v3
-#define vend v4
-#define dend d4
+#define vend v3
+#define dend d3
/*
Core algorithm:
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting leading zeros identifies
+ exactly which byte matched. */
ENTRY (MEMCHR)
PTR_ARG (0)
cbz cntin, L(nomatch)
ld1 {vdata.16b}, [src]
dup vrepchr.16b, chrin
- mov wtmp, 0xf00f
- dup vrepmask.8h, wtmp
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
lsl shift, srcin, 2
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
lsr synd, synd, shift
cbz synd, L(start_loop)
fmov synd, dend
cbz synd, L(loop32)
L(end):
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
add tmp, srcin, cntin
sub cntrem, tmp, src
#define synd x5
#define shift x6
#define tmp x7
-#define wtmp w7
#define end x8
#define endm1 x9
#define qdata q1
#define vdata v1
#define vhas_chr v2
-#define vrepmask v3
-#define vend v4
-#define dend d4
+#define vend v3
+#define dend d3
/*
Core algorithm:
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting leading zeros identifies
+ exactly which byte matched. */
ENTRY (__memrchr)
PTR_ARG (0)
cbz cntin, L(nomatch)
ld1 {vdata.16b}, [src]
dup vrepchr.16b, chrin
- mov wtmp, 0xf00f
- dup vrepmask.8h, wtmp
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
neg shift, end, lsl 2
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
lsl synd, synd, shift
cbz synd, L(start_loop)
fmov synd, dend
cbz synd, L(loop32)
L(end):
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
add tmp, src, 15
#define src x2
#define tmp1 x1
#define tmp2 x3
-#define tmp2w w3
#define vrepchr v0
#define vdata v1
#define qdata q1
#define vhas_nul v2
#define vhas_chr v3
-#define vrepmask v4
-#define vend v5
-#define dend d5
+#define vend v4
+#define dend d4
-/* Core algorithm:
-
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+/*
+ Core algorithm:
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting leading zeros identifies
+ exactly which byte matched. */
ENTRY (__strchrnul)
PTR_ARG (0)
bic src, srcin, 15
dup vrepchr.16b, chrin
ld1 {vdata.16b}, [src]
- mov tmp2w, 0xf00f
- dup vrepmask.8h, tmp2w
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
lsl tmp2, srcin, 2
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov tmp1, dend
lsr tmp1, tmp1, tmp2 /* Mask padding bits. */
cbz tmp1, L(loop)
fmov tmp1, dend
cbz tmp1, L(loop)
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov tmp1, dend
#ifndef __AARCH64EB__
rbit tmp1, tmp1
#define len x4
#define synd x4
#define tmp x5
-#define wtmp w5
#define shift x5
#define data1 x6
#define dataw1 w6
#define dataq q0
#define vdata v0
#define vhas_nul v1
-#define vrepmask v2
-#define vend v3
-#define dend d3
+#define vend v2
+#define dend d2
#define dataq2 q1
#ifdef BUILD_STPCPY
# define IFSTPCPY(X,...)
#endif
-/* Core algorithm:
-
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+/*
+ Core algorithm:
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting leading zeros identifies
+ exactly which byte matched. */
ENTRY (STRCPY)
PTR_ARG (0)
PTR_ARG (1)
bic src, srcin, 15
- mov wtmp, 0xf00f
ld1 {vdata.16b}, [src]
- dup vrepmask.8h, wtmp
cmeq vhas_nul.16b, vdata.16b, 0
lsl shift, srcin, 2
- and vhas_nul.16b, vhas_nul.16b, vrepmask.16b
- addp vend.16b, vhas_nul.16b, vhas_nul.16b
+ shrn vend.8b, vhas_nul.8h, 4 /* 128->64 */
fmov synd, dend
lsr synd, synd, shift
cbnz synd, L(tail)
ldr dataq, [src, 16]!
cmeq vhas_nul.16b, vdata.16b, 0
- and vhas_nul.16b, vhas_nul.16b, vrepmask.16b
- addp vend.16b, vhas_nul.16b, vhas_nul.16b
+ shrn vend.8b, vhas_nul.8h, 4 /* 128->64 */
fmov synd, dend
cbz synd, L(start_loop)
fmov synd, dend
cbz synd, L(loop)
- and vhas_nul.16b, vhas_nul.16b, vrepmask.16b
- addp vend.16b, vhas_nul.16b, vhas_nul.16b /* 128->64 */
+ shrn vend.8b, vhas_nul.8h, 4 /* 128->64 */
fmov synd, dend
#ifndef __AARCH64EB__
rbit synd, synd
#define src x1
#define synd x2
#define tmp x3
-#define wtmp w3
#define shift x4
#define data q0
#define vdata v0
#define vhas_nul v1
-#define vrepmask v2
-#define vend v3
-#define dend d3
+#define vend v2
+#define dend d2
/* Core algorithm:
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting trailing zeros identifies
+ exactly which byte matched. */
ENTRY (STRLEN)
PTR_ARG (0)
bic src, srcin, 15
- mov wtmp, 0xf00f
ld1 {vdata.16b}, [src]
- dup vrepmask.8h, wtmp
cmeq vhas_nul.16b, vdata.16b, 0
lsl shift, srcin, 2
- and vhas_nul.16b, vhas_nul.16b, vrepmask.16b
- addp vend.16b, vhas_nul.16b, vhas_nul.16b /* 128->64 */
+ shrn vend.8b, vhas_nul.8h, 4 /* 128->64 */
fmov synd, dend
lsr synd, synd, shift
cbz synd, L(loop)
fmov synd, dend
cbz synd, L(loop)
- and vhas_nul.16b, vhas_nul.16b, vrepmask.16b
- addp vend.16b, vhas_nul.16b, vhas_nul.16b /* 128->64 */
+ shrn vend.8b, vhas_nul.8h, 4 /* 128->64 */
sub result, src, srcin
fmov synd, dend
#ifndef __AARCH64EB__
#define src x2
#define synd x3
#define shift x4
-#define wtmp w4
#define tmp x4
#define cntrem x5
#define qdata q0
#define vdata v0
#define vhas_chr v1
-#define vrepmask v2
-#define vend v3
-#define dend d3
+#define vend v2
+#define dend d2
/*
Core algorithm:
- For each 16-byte chunk we calculate a 64-bit syndrome value with four bits
- per byte. For even bytes, bits 0-3 are set if the relevant byte matched the
- requested character or the byte is NUL. Bits 4-7 must be zero. Bits 4-7 are
- set likewise for odd bytes so that adjacent bytes can be merged. Since the
- bits in the syndrome reflect the order in which things occur in the original
- string, counting trailing zeros identifies exactly which byte matched. */
+ For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
+ per byte. We take 4 bits of every comparison byte with shift right and narrow
+ by 4 instruction. Since the bits in the nibble mask reflect the order in
+ which things occur in the original string, counting trailing zeros identifies
+ exactly which byte matched. */
ENTRY (__strnlen)
PTR_ARG (0)
SIZE_ARG (1)
bic src, srcin, 15
- mov wtmp, 0xf00f
cbz cntin, L(nomatch)
ld1 {vdata.16b}, [src], 16
- dup vrepmask.8h, wtmp
cmeq vhas_chr.16b, vdata.16b, 0
lsl shift, srcin, 2
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
lsr synd, synd, shift
cbz synd, L(start_loop)
cbz synd, L(loop32)
L(end):
- and vhas_chr.16b, vhas_chr.16b, vrepmask.16b
- addp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
+ shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
sub src, src, 16
mov synd, vend.d[0]
sub result, src, srcin
projects / thirdparty / glibc.git / commitdiff
