diff options
Diffstat (limited to 'sysdeps/alpha')
-rw-r--r-- | sysdeps/alpha/Dist | 2 | ||||
-rw-r--r-- | sysdeps/alpha/strchr.c | 84 | ||||
-rw-r--r-- | sysdeps/alpha/stxcpy.S | 33 |
3 files changed, 19 insertions, 100 deletions
diff --git a/sysdeps/alpha/Dist b/sysdeps/alpha/Dist index c9419e6..3cc9595 100644 --- a/sysdeps/alpha/Dist +++ b/sysdeps/alpha/Dist @@ -3,3 +3,5 @@ DEFS.h divrem.h divl.S divlu.S divq.S divqu.S reml.S remlu.S remq.S remqu.S _mcount.S +stxcpy.S +stxncpy.S diff --git a/sysdeps/alpha/strchr.c b/sysdeps/alpha/strchr.c deleted file mode 100644 index 69afa4b..0000000 --- a/sysdeps/alpha/strchr.c +++ /dev/null @@ -1,84 +0,0 @@ -/* Copyright (C) 1992, 1993, 1994, 1995 Free Software Foundation, Inc. - -The GNU C Library is free software; you can redistribute it and/or -modify it under the terms of the GNU Library General Public License as -published by the Free Software Foundation; either version 2 of the -License, or (at your option) any later version. - -The GNU C Library is distributed in the hope that it will be useful, -but WITHOUT ANY WARRANTY; without even the implied warranty of -MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU -Library General Public License for more details. - -You should have received a copy of the GNU Library General Public -License along with the GNU C Library; see the file COPYING.LIB. If -not, write to the Free Software Foundation, Inc., 675 Mass Ave, -Cambridge, MA 02139, USA. */ - -#include <string.h> - -/* Return the length of the null-terminated string STR. Scan for - the null terminator quickly by testing eight bytes at a time. */ - -char * -strchr (const char *str, int c) -{ - const char *char_ptr; - const unsigned long int *longword_ptr; - unsigned long int charmask; - - c = (unsigned char) c; - - /* Handle the first few characters by reading one character at a time. - Do this until STR is aligned on a 8-byte border. */ - for (char_ptr = str; ((unsigned long int) char_ptr & 7) != 0; ++char_ptr) - if (*char_ptr == c) - return (char *) char_ptr; - else if (*char_ptr == '\0') - return NULL; - - longword_ptr = (unsigned long int *) char_ptr; - - /* Set up a longword, each of whose bytes is C. */ - charmask = c | (c << 8); - charmask |= charmask << 16; - charmask |= charmask << 32; - - for (;;) - { - const unsigned long int longword = *longword_ptr++; - int ge, le, zero; - - /* Set bits in ZERO if bytes in LONGWORD are zero. */ - asm ("cmpbge $31, %1, %0" : "=r" (zero) : "r" (longword)); - - /* Set bits in GE if bytes in CHARMASK are >= bytes in LONGWORD. */ - asm ("cmpbge %1, %2, %0" : "=r" (ge) : "r" (charmask), "r" (longword)); - - /* Set bits in LE if bytes in CHARMASK are <= bytes in LONGWORD. */ - asm ("cmpbge %2, %1, %0" : "=r" (le) : "r" (charmask), "r" (longword)); - - /* Bytes that are both <= and >= are == to C. */ - if (zero || (ge & le)) - { - /* Which of the bytes was the C? */ - - char *cp = (char *) (longword_ptr - 1); - int i; - - for (i = 0; i < 8; i++) - { - if (cp[i] == c) - return &cp[i]; - if (cp[i] == 0) - return NULL; - } - return NULL; - } - } -} - -#ifdef weak_alias -#undef index -weak_alias (strchr, index) -#endif diff --git a/sysdeps/alpha/stxcpy.S b/sysdeps/alpha/stxcpy.S index e381b70..49dd8e5 100644 --- a/sysdeps/alpha/stxcpy.S +++ b/sysdeps/alpha/stxcpy.S @@ -36,7 +36,7 @@ Cambridge, MA 02139, USA. */ Furthermore, v0, a3-a5, t11, and t12 are untouched. */ -/* This is generally scheduled for the EV5, but should still be pretty +/* This is generally scheduled for the EV5, but should still be pretty good for the EV4 too. */ #include <sysdep.h> @@ -65,7 +65,7 @@ stxcpy_aligned: lda t2, -1 # e1 : build a mask against false zero mskqh t2, a1, t2 # e0 : detection in the src word mskqh t1, a1, t3 # e0 : - ornot t1, t2, t2 # .. e1 : + ornot t1, t2, t2 # .. e1 : mskql t0, a1, t0 # e0 : assemble the first output word cmpbge zero, t2, t7 # .. e1 : bits set iff null found or t0, t3, t1 # e0 : @@ -99,9 +99,9 @@ $a_eos: /* We're doing a partial word store and so need to combine our source and original destination words. */ ldq_u t0, 0(a0) # e0 : - subq t8, 1, t6 # .. e1 : + subq t8, 1, t6 # .. e1 : zapnot t1, t6, t1 # e0 : clear src bytes >= null - or t8, t6, t7 # .. e1 : + or t8, t6, t7 # .. e1 : zap t0, t7, t0 # e0 : clear dst bytes <= null or t0, t1, t1 # e1 : @@ -156,13 +156,13 @@ $u_head: or t1, t4, t1 # .. e1 : mskqh t1, a0, t1 # e0 : or t0, t1, t1 # e1 : - + or t1, t6, t6 # e0 : cmpbge zero, t6, t7 # .. e1 : lda t6, -1 # e0 : for masking just below bne t7, $u_final # .. e1 : - mskql t6, a1, t6 # e0 : mask out the bits we have + mskql t6, a1, t6 # e0 : mask out the bits we have or t6, t2, t2 # e1 : already extracted before cmpbge zero, t2, t7 # e0 : testing eos bne t7, $u_late_head_exit # .. e1 (zdb) @@ -181,7 +181,7 @@ $u_head: /* Unaligned copy main loop. In order to avoid reading too much, the loop is structured to detect zeros in aligned source words. - This has, unfortunately, effectively pulled half of a loop + This has, unfortunately, effectively pulled half of a loop iteration out into the head and half into the tail, but it does prevent nastiness from accumulating in the very thing we want to run as fast as possible. @@ -207,7 +207,7 @@ $u_loop: /* We've found a zero somewhere in the source word we just read. If it resides in the lower half, we have one (probably partial) - word to write out, and if it resides in the upper half, we + word to write out, and if it resides in the upper half, we have one full and one partial word left to write out. On entry to this basic block: @@ -234,7 +234,7 @@ $u_final: negq t7, t6 # e0 : isolate low bit set and t6, t7, t8 # e1 : - and t8, 0x80, t6 # e0 : avoid dest word load if we can + and t8, 0x80, t6 # e0 : avoid dest word load if we can bne t6, 1f # .. e1 (zdb) ldq_u t0, 0(a0) # e0 : @@ -256,7 +256,7 @@ $unaligned: and a0, 7, t4 # .. e1 : find dest misalignment and a1, 7, t5 # e0 : find src misalignment - /* Conditionally load the first destination word and a bytemask + /* Conditionally load the first destination word and a bytemask with 0xff indicating that the destination byte is sacrosanct. */ mov zero, t0 # .. e1 : @@ -290,18 +290,19 @@ $unaligned: negq t7, t6 # .. e1 : build bitmask of bytes <= zero and t6, t7, t8 # e0 : - nop # .. e1 : + and a1, 7, t5 # .. e1 : subq t8, 1, t6 # e0 : or t6, t8, t7 # e1 : + srl t8, t5, t8 # e0 : adjust final null return value - zapnot t2, t7, t2 # e0 : prepare source word; mirror changes + zapnot t2, t7, t2 # .. e1 : prepare source word; mirror changes and t1, t2, t1 # e1 : to source validity mask - extql t2, a1, t2 # e0 : + extql t2, a1, t2 # .. e0 : extql t1, a1, t1 # e0 : - andnot t0, t2, t0 # e0 : zero place for source to reside + andnot t0, t2, t0 # .. e1 : zero place for source to reside or t0, t1, t1 # e1 : and put it there - stq_u t1, 0(a0) # e0 : - ret (t9) # .. e1 : + stq_u t1, 0(a0) # .. e0 : + ret (t9) .end __stxcpy |