| /* Optimized strcmp implementation for POWER7/PowerPC32. |
| Copyright (C) 2010-2014 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> |
| |
| /* See strlen.s for comments on how the end-of-string testing works. */ |
| |
| /* int [r3] strncmp (const char *s1 [r3], |
| const char *s2 [r4], |
| size_t size [r5]) */ |
| |
| EALIGN (strncmp,5,0) |
| |
| #define rTMP2 r0 |
| #define rRTN r3 |
| #define rSTR1 r3 /* first string arg */ |
| #define rSTR2 r4 /* second string arg */ |
| #define rN r5 /* max string length */ |
| #define rWORD1 r6 /* current word in s1 */ |
| #define rWORD2 r7 /* current word in s2 */ |
| #define rWORD3 r10 |
| #define rWORD4 r11 |
| #define rFEFE r8 /* constant 0xfefefeff (-0x01010101) */ |
| #define r7F7F r9 /* constant 0x7f7f7f7f */ |
| #define rNEG r10 /* ~(word in s1 | 0x7f7f7f7f) */ |
| #define rBITDIF r11 /* bits that differ in s1 & s2 words */ |
| #define rTMP r12 |
| |
| dcbt 0,rSTR1 |
| nop |
| or rTMP,rSTR2,rSTR1 |
| lis r7F7F,0x7f7f |
| dcbt 0,rSTR2 |
| nop |
| clrlwi. rTMP,rTMP,30 |
| cmplwi cr1,rN,0 |
| lis rFEFE,-0x101 |
| bne L(unaligned) |
| /* We are word aligned so set up for two loops. first a word |
| loop, then fall into the byte loop if any residual. */ |
| srwi. rTMP,rN,2 |
| clrlwi rN,rN,30 |
| addi rFEFE,rFEFE,-0x101 |
| addi r7F7F,r7F7F,0x7f7f |
| cmplwi cr1,rN,0 |
| beq L(unaligned) |
| |
| mtctr rTMP |
| lwz rWORD1,0(rSTR1) |
| lwz rWORD2,0(rSTR2) |
| b L(g1) |
| |
| L(g0): |
| lwzu rWORD1,4(rSTR1) |
| bne cr1,L(different) |
| lwzu rWORD2,4(rSTR2) |
| L(g1): add rTMP,rFEFE,rWORD1 |
| nor rNEG,r7F7F,rWORD1 |
| bdz L(tail) |
| and. rTMP,rTMP,rNEG |
| cmpw cr1,rWORD1,rWORD2 |
| beq L(g0) |
| |
| /* OK. We've hit the end of the string. We need to be careful that |
| we don't compare two strings as different because of gunk beyond |
| the end of the strings... */ |
| #ifdef __LITTLE_ENDIAN__ |
| L(endstring): |
| slwi rTMP, rTMP, 1 |
| addi rTMP2, rTMP, -1 |
| andc rTMP2, rTMP2, rTMP |
| and rWORD2, rWORD2, rTMP2 /* Mask off gunk. */ |
| and rWORD1, rWORD1, rTMP2 |
| rlwinm rTMP2, rWORD2, 8, 0xffffffff /* Byte reverse word. */ |
| rlwinm rTMP, rWORD1, 8, 0xffffffff |
| rldimi rTMP2, rWORD2, 24, 32 |
| rldimi rTMP, rWORD1, 24, 32 |
| rlwimi rTMP2, rWORD2, 24, 16, 23 |
| rlwimi rTMP, rWORD1, 24, 16, 23 |
| xor. rBITDIF, rTMP, rTMP2 |
| sub rRTN, rTMP, rTMP2 |
| bgelr |
| ori rRTN, rTMP2, 1 |
| blr |
| |
| L(different): |
| lwz rWORD1, -4(rSTR1) |
| rlwinm rTMP2, rWORD2, 8, 0xffffffff /* Byte reverse word. */ |
| rlwinm rTMP, rWORD1, 8, 0xffffffff |
| rldimi rTMP2, rWORD2, 24, 32 |
| rldimi rTMP, rWORD1, 24, 32 |
| rlwimi rTMP2, rWORD2, 24, 16, 23 |
| rlwimi rTMP, rWORD1, 24, 16, 23 |
| xor. rBITDIF, rTMP, rTMP2 |
| sub rRTN, rTMP, rTMP2 |
| bgelr |
| ori rRTN, rTMP2, 1 |
| blr |
| |
| #else |
| L(endstring): |
| and rTMP,r7F7F,rWORD1 |
| beq cr1,L(equal) |
| add rTMP,rTMP,r7F7F |
| xor. rBITDIF,rWORD1,rWORD2 |
| andc rNEG,rNEG,rTMP |
| blt L(highbit) |
| cntlzw rBITDIF,rBITDIF |
| cntlzw rNEG,rNEG |
| addi rNEG,rNEG,7 |
| cmpw cr1,rNEG,rBITDIF |
| sub rRTN,rWORD1,rWORD2 |
| bgelr cr1 |
| L(equal): |
| li rRTN,0 |
| blr |
| |
| L(different): |
| lwz rWORD1,-4(rSTR1) |
| xor. rBITDIF,rWORD1,rWORD2 |
| sub rRTN,rWORD1,rWORD2 |
| bgelr |
| L(highbit): |
| ori rRTN, rWORD2, 1 |
| blr |
| #endif |
| |
| /* Oh well. In this case, we just do a byte-by-byte comparison. */ |
| .align 4 |
| L(tail): |
| and. rTMP,rTMP,rNEG |
| cmpw cr1,rWORD1,rWORD2 |
| bne L(endstring) |
| addi rSTR1,rSTR1,4 |
| bne cr1,L(different) |
| addi rSTR2,rSTR2,4 |
| cmplwi cr1,rN,0 |
| L(unaligned): |
| mtctr rN |
| ble cr1,L(ux) |
| L(uz): |
| lbz rWORD1,0(rSTR1) |
| lbz rWORD2,0(rSTR2) |
| .align 4 |
| L(u1): |
| cmpwi cr1,rWORD1,0 |
| bdz L(u4) |
| cmpw rWORD1,rWORD2 |
| beq cr1,L(u4) |
| bne L(u4) |
| lbzu rWORD3,1(rSTR1) |
| lbzu rWORD4,1(rSTR2) |
| cmpwi cr1,rWORD3,0 |
| bdz L(u3) |
| cmpw rWORD3,rWORD4 |
| beq cr1,L(u3) |
| bne L(u3) |
| lbzu rWORD1,1(rSTR1) |
| lbzu rWORD2,1(rSTR2) |
| cmpwi cr1,rWORD1,0 |
| bdz L(u4) |
| cmpw rWORD1,rWORD2 |
| beq cr1,L(u4) |
| bne L(u4) |
| lbzu rWORD3,1(rSTR1) |
| lbzu rWORD4,1(rSTR2) |
| cmpwi cr1,rWORD3,0 |
| bdz L(u3) |
| cmpw rWORD3,rWORD4 |
| beq cr1,L(u3) |
| bne L(u3) |
| lbzu rWORD1,1(rSTR1) |
| lbzu rWORD2,1(rSTR2) |
| b L(u1) |
| |
| L(u3): sub rRTN,rWORD3,rWORD4 |
| blr |
| L(u4): sub rRTN,rWORD1,rWORD2 |
| blr |
| L(ux): |
| li rRTN,0 |
| blr |
| END (strncmp) |
| libc_hidden_builtin_def (strncmp) |
