aboutsummaryrefslogtreecommitdiff
path: root/sysdeps/s390/wcsrchr-vx.S
diff options
context:
space:
mode:
authorStefan Liebler <stli@linux.ibm.com>2018-12-18 13:57:22 +0100
committerStefan Liebler <stli@linux.ibm.com>2018-12-18 13:57:22 +0100
commit4753713aaed400d88cfc598e25d504478120b065 (patch)
tree99af0df98fb82cd59be28de908c79b95d78258b9 /sysdeps/s390/wcsrchr-vx.S
parentc09c1b6f014a45f6c4ba189f5ae94b1fb1a2982a (diff)
downloadglibc-4753713aaed400d88cfc598e25d504478120b065.zip
glibc-4753713aaed400d88cfc598e25d504478120b065.tar.gz
glibc-4753713aaed400d88cfc598e25d504478120b065.tar.bz2
S390: Refactor wcsrchr ifunc handling.
The ifunc handling for wcsrchr is adjusted in order to omit ifunc if the minimum architecture level already supports newer CPUs by default. Unfortunately the c ifunc variant can't be omitted at all as it is used by the z13 ifunc variant as fallback if the pointers are not 4-byte aligned. ChangeLog: * sysdeps/s390/multiarch/Makefile (sysdep_routines): Remove wcsrchr variants. * sysdeps/s390/Makefile (sysdep_routines): Add wcsrchr variants. * sysdeps/s390/multiarch/ifunc-impl-list.c (__libc_ifunc_impl_list): Refactor ifunc handling for wcsrchr. * sysdeps/s390/multiarch/wcsrchr-c.c: Move to ... * sysdeps/s390/wcsrchr-c.c: ... here and adjust ifunc handling. * sysdeps/s390/multiarch/wcsrchr-vx.S: Move to ... * sysdeps/s390/wcsrchr-vx.S: ... here and adjust ifunc handling. * sysdeps/s390/multiarch/wcsrchr.c: Move to ... * sysdeps/s390/wcsrchr.c: ... here and adjust ifunc handling. * sysdeps/s390/ifunc-wcsrchr.h: New file.
Diffstat (limited to 'sysdeps/s390/wcsrchr-vx.S')
-rw-r--r--sysdeps/s390/wcsrchr-vx.S195
1 files changed, 195 insertions, 0 deletions
diff --git a/sysdeps/s390/wcsrchr-vx.S b/sysdeps/s390/wcsrchr-vx.S
new file mode 100644
index 0000000..c7b6585
--- /dev/null
+++ b/sysdeps/s390/wcsrchr-vx.S
@@ -0,0 +1,195 @@
+/* Vector optimized 32/64 bit S/390 version of wcsrchr.
+ Copyright (C) 2015-2018 Free Software Foundation, Inc.
+ This file is part of the GNU C Library.
+
+ The GNU C Library is free software; you can redistribute it and/or
+ modify it under the terms of the GNU Lesser General Public
+ License as published by the Free Software Foundation; either
+ version 2.1 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
+ Lesser General Public License for more details.
+
+ You should have received a copy of the GNU Lesser General Public
+ License along with the GNU C Library; if not, see
+ <http://www.gnu.org/licenses/>. */
+
+#include <ifunc-wcsrchr.h>
+#if HAVE_WCSRCHR_Z13
+
+# include "sysdep.h"
+# include "asm-syntax.h"
+
+ .text
+
+/* wchar_t *wcsrchr (const wchar_t *s, wchar_t c)
+ Locate the last character c in string.
+
+ Register usage:
+ -r0=loaded bytes in first part of s.
+ -r1=pointer to last occurence of c or NULL if not found.
+ -r2=s
+ -r3=c
+ -r4=tmp
+ -r5=current_len
+ -v16=part of s
+ -v17=index of found element
+ -v18=replicated c
+ -v19=part of s with last occurence of c.
+ -v20=permute pattern
+*/
+ENTRY(WCSRCHR_Z13)
+ .machine "z13"
+ .machinemode "zarch_nohighgprs"
+
+ vlbb %v16,0(%r2),6 /* Load s until next 4k-byte boundary. */
+ lcbb %r0,0(%r2),6 /* Get bytes to 4k-byte boundary or 16. */
+
+ tmll %r2,3 /* Test if s is 4-byte aligned? */
+ jne .Lfallback /* And use common-code variant if not. */
+
+ vlvgf %v18,%r3,0 /* Generate vector which elements are all c. */
+ vrepf %v18,%v18,0
+
+ lghi %r1,-1 /* Currently no c found. */
+ lghi %r5,0 /* current_len = 0. */
+
+ vfeezfs %v17,%v16,%v18 /* Find element equal or zero. */
+ vlgvb %r4,%v17,7 /* Load byte index of c/zero or 16. */
+ clrjl %r4,%r0,.Lfound_first_part /* Found c/zero in loaded bytes. */
+.Lalign:
+ /* Align s to 16 byte. */
+ risbgn %r4,%r2,60,128+63,0 /* %r3 = bits 60-63 of %r2 'and' 15. */
+ lghi %r5,16 /* current_len = 16. */
+ slr %r5,%r4 /* Compute bytes to 16bytes boundary. */
+
+.Lloop:
+ vl %v16,0(%r5,%r2) /* Load s. */
+ vfeezfs %v17,%v16,%v18 /* Find element equal with zero search. */
+ jno .Lfound /* Found c/zero (cc=0|1|2). */
+ vl %v16,16(%r5,%r2)
+ vfeezfs %v17,%v16,%v18
+ jno .Lfound16
+ vl %v16,32(%r5,%r2)
+ vfeezfs %v17,%v16,%v18
+ jno .Lfound32
+ vl %v16,48(%r5,%r2)
+ vfeezfs %v17,%v16,%v18
+ jno .Lfound48
+
+ aghi %r5,64
+ j .Lloop /* No character and no zero -> loop. */
+
+.Lfound48:
+ la %r5,16(%r5) /* Use la since aghi would clobber cc. */
+.Lfound32:
+ la %r5,16(%r5)
+.Lfound16:
+ la %r5,16(%r5)
+.Lfound:
+ je .Lzero /* Found zero, but no c before that zero. */
+ /* Save this part of s to check for further matches after reaching
+ the end of the complete string. */
+ vlr %v19,%v16
+ lgr %r1,%r5
+
+ jh .Lzero /* Found a zero after the found c. */
+ aghi %r5,16 /* Start search of next part of s. */
+ j .Lloop
+
+.Lfound_first_part:
+ /* This code is only executed if the found c/zero is whithin loaded
+ bytes. If no c/zero was found (cc==3) the found index = 16, thus
+ this code is not called.
+ Resulting condition code of vector find element equal:
+ cc==0: no c, found zero
+ cc==1: c found, no zero
+ cc==2: c found, found zero after c
+ cc==3: no c, no zero (this case can be ignored). */
+ je .Lzero /* Found zero, but no c before that zero. */
+
+ locgrne %r1,%r5 /* Mark c as found in first part of s. */
+ vlr %v19,%v16
+
+ jl .Lalign /* No zero (e.g. if vr was fully loaded)
+ -> Align and loop afterwards. */
+
+ /* Found a zero in vr. If vr was not fully loaded due to block
+ boundary, the remaining bytes are filled with zero and we can't
+ rely on zero indication of condition code here! */
+
+ vfenezf %v17,%v16,%v16
+ vlgvb %r4,%v17,7 /* Load byte index of zero or 16. */
+ clrjl %r4,%r0,.Lzero /* Zero within loaded bytes -> end. */
+ j .Lalign /* Align and loop afterwards. */
+
+.Lend_searched_zero:
+ vlgvb %r4,%v17,7 /* Load byte index of zero. */
+ algr %r5,%r4
+ la %r2,0(%r5,%r2) /* Return pointer to zero. */
+ br %r14
+
+.Lzero:
+ /* Reached end of string. Check if one c was found before. */
+ clije %r3,0,.Lend_searched_zero /* Found zero and c is zero. */
+
+ cgfi %r1,-1 /* No c found -> return NULL. */
+ locghie %r2,0
+ ber %r14
+
+ larl %r3,.Lpermute_mask /* Load permute mask. */
+ vl %v20,0(%r3)
+
+ /* c was found and is part of v19. */
+ vfenezf %v17,%v19,%v19 /* Find zero. */
+ vlgvb %r4,%v17,7 /* Load byte index of zero or 16. */
+ ahi %r4,3 /* Found zero index is first byte,
+ thus highest byte index is last byte of
+ wchar_t zero. */
+
+ clgfi %r5,0 /* Loaded byte count in v19 is 16, ... */
+ lochine %r0,16 /* ... if v19 is not the first part of s. */
+ ahi %r0,-1 /* Convert byte count to highest index. */
+
+ clr %r0,%r4
+ locrl %r4,%r0 /* r4 = min (zero-index, highest-index). */
+
+ /* Right-shift of v19 to mask bytes after zero. */
+ clije %r4,15,.Lzero_permute /* No shift is needed if highest index
+ in vr is 15. */
+ lhi %r0,15
+ slr %r0,%r4 /* Compute byte count for vector shift left. */
+ sll %r0,3 /* Convert to bit count. */
+ vlvgb %v17,%r0,7
+ vsrlb %v19,%v19,%v17 /* Vector shift right by byte by number of bytes
+ specified in bits 1-4 of byte 7 in v17. */
+
+ /* Reverse bytes in v19. */
+.Lzero_permute:
+ vperm %v19,%v19,%v19,%v20 /* Permute v19 to reversed order. */
+
+ /* Find c in reversed v19. */
+ vfeef %v19,%v19,%v18 /* Find c. */
+ la %r2,0(%r1,%r2)
+ vlgvb %r3,%v19,7 /* Load byte index of c. */
+
+ /* Compute index in real s and return. */
+ slgr %r4,%r3
+ lay %r2,-3(%r4,%r2) /* Return pointer to zero. -3 is needed,
+ because the found byte index is reversed in
+ vector-register. Thus point to first byte of
+ wchar_t. */
+ br %r14
+.Lpermute_mask:
+ .byte 0x0C,0x0D,0x0E,0x0F,0x08,0x09,0x0A,0x0B
+ .byte 0x04,0x05,0x06,0x07,0x00,0x01,0x02,0x03
+.Lfallback:
+ jg WCSRCHR_C
+END(WCSRCHR_Z13)
+
+# if ! HAVE_WCSRCHR_IFUNC
+strong_alias (WCSRCHR_Z13, wcsrchr)
+# endif
+#endif