diff options
author | Adhemerval Zanella <azanella@linux.vnet.ibm.com> | 2014-11-19 15:24:55 -0500 |
---|---|---|
committer | Adhemerval Zanella <azanella@linux.vnet.ibm.com> | 2014-12-02 07:16:24 -0500 |
commit | bb2542e0ae6f82197e804e7e9b78fed461553fc0 (patch) | |
tree | 2e62b412d0e0991663d888dd31b214c0c69b5f5c /sysdeps/powerpc/powerpc64/power7 | |
parent | 2e8a2de2dafa3238b5b58eecb407af6825a780cd (diff) | |
download | glibc-bb2542e0ae6f82197e804e7e9b78fed461553fc0.tar glibc-bb2542e0ae6f82197e804e7e9b78fed461553fc0.tar.gz glibc-bb2542e0ae6f82197e804e7e9b78fed461553fc0.tar.bz2 glibc-bb2542e0ae6f82197e804e7e9b78fed461553fc0.zip |
powerpc: Add powerpc64 strcspn optimization
This patch makes the POWER7 optimized strcspn generic by using
default doubleword stores to zero the hash, instead of VSX
instructions. Performance on POWER7/POWER8 does not change.
Diffstat (limited to 'sysdeps/powerpc/powerpc64/power7')
-rw-r--r-- | sysdeps/powerpc/powerpc64/power7/strcspn.S | 139 |
1 files changed, 0 insertions, 139 deletions
diff --git a/sysdeps/powerpc/powerpc64/power7/strcspn.S b/sysdeps/powerpc/powerpc64/power7/strcspn.S deleted file mode 100644 index 3f6aa0a506..0000000000 --- a/sysdeps/powerpc/powerpc64/power7/strcspn.S +++ /dev/null @@ -1,139 +0,0 @@ -/* Optimized strcspn implementation for PowerPC64. - Copyright (C) 2014 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 <sysdep.h> - -/* size_t [r3] strcspn (const char [r4] *s, const char [r5] *reject) */ - - .machine power7 -EALIGN (strcspn, 4, 0) - CALL_MCOUNT 3 - - /* The idea to speed up the algorithm is to create a lookup table - for fast check if input character should be considered. For ASCII - or ISO-8859-X character sets it has 256 positions. */ - lbz r10,0(r4) - - /* First the table should be cleared and to avoid unaligned accesses - when using the VSX stores the table address is aligned to 16 - bytes. */ - xxlxor v0,v0,v0 - - /* PPC64 ELF ABI stack is aligned to 16 bytes. */ - addi r9,r1,-256 - - li r8,48 - li r5,16 - li r6,32 - cmpdi cr7,r10,0 /* reject[0] == '\0' ? */ - addi r12,r9,64 - /* Clear the table with 0 values */ - stxvw4x v0,r0,r9 - addi r11,r9,128 - addi r7,r9,192 - stxvw4x v0,r9,r5 - stxvw4x v0,r9,r6 - stxvw4x v0,r9,r8 - stxvw4x v0,r0,r12 - stxvw4x v0,r12,r5 - stxvw4x v0,r12,r6 - stxvw4x v0,r12,r8 - stxvw4x v0,r0,r11 - stxvw4x v0,r11,r5 - stxvw4x v0,r11,r6 - stxvw4x v0,r11,r8 - stxvw4x v0,r0,r7 - stxvw4x v0,r7,r5 - stxvw4x v0,r7,r6 - stxvw4x v0,r7,r8 - li r8,1 - beq cr7,L(finish_table) /* If reject[0] == '\0' skip */ - - /* Initialize the table as: - for (i=0; reject[i]; i++ - table[reject[i]]] = 1 */ - .p2align 4,,15 -L(init_table): - stbx r8,r9,r10 - lbzu r10,1(r4) - cmpdi cr7,r10,0 /* If reject[0] == '\0' finish */ - bne cr7,L(init_table) -L(finish_table): - /* set table[0] = 1 */ - li r10,1 - stb r10,0(r9) - li r10,0 - b L(mainloop) - - /* Unrool the loop 4 times and check using the table as: - i = 0; - while (1) - { - if (table[input[i++]] == 1) - return i - 1; - if (table[input[i++]] == 1) - return i - 1; - if (table[input[i++]] == 1) - return i - 1; - if (table[input[i++]] == 1) - return i - 1; - } */ - .p2align 4,,15 -L(unroll): - lbz r8,1(r3) - addi r10,r10,4 - lbzx r8,r9,r8 - cmpwi r7,r8,1 - beq cr7,L(end) - lbz r8,2(r3) - addi r3,r3,4 - lbzx r8,r9,r8 - cmpwi cr7,r8,1 - beq cr7,L(end2) - lbz r8,3(r7) - lbzx r8,r9,r8 - cmpwi cr7,r8,1 - beq cr7,L(end3) -L(mainloop): - lbz r8,0(r3) - mr r7,r3 - addi r6,r10,1 - addi r4,r10,2 - addi r5,r10,3 - lbzx r8,r9,8 - cmpwi cr7,r8,1 - bne cr7,L(unroll) - mr r3,r10 - blr - - .p2align 4,,15 -L(end): - mr r3,r6 - blr - - .p2align 4,,15 -L(end2): - mr r3,r4 - blr - - .p2align 4,,15 -L(end3): - mr r3,r5 - blr -END (strcspn) -libc_hidden_builtin_def (strcspn) |