diff options
author | Roland McGrath <roland@gnu.org> | 2002-09-17 23:50:03 +0000 |
---|---|---|
committer | Roland McGrath <roland@gnu.org> | 2002-09-17 23:50:03 +0000 |
commit | cfc91acd418d2ea62a4520b42406a16ef07257c3 (patch) | |
tree | c7d291182781c1534a03528b33d417983b8d2e54 /sysdeps/powerpc/powerpc64/strlen.S | |
parent | 81cb0d8214c425e356c3f79b4b024caab36540b8 (diff) | |
download | glibc-cfc91acd418d2ea62a4520b42406a16ef07257c3.tar glibc-cfc91acd418d2ea62a4520b42406a16ef07257c3.tar.gz glibc-cfc91acd418d2ea62a4520b42406a16ef07257c3.tar.bz2 glibc-cfc91acd418d2ea62a4520b42406a16ef07257c3.zip |
2002-09-17 Steven Munroe <sjmunroe@us.ibm.com>
Ported to PowerPC64 running Linux.
* sysdeps/powerpc/powerpc64/Dist: New file.
* sysdeps/powerpc/powerpc64/Implies: New file.
* sysdeps/powerpc/powerpc64/Makefile: New file.
* sysdeps/powerpc/powerpc64/__longjmp.S: New file.
* sysdeps/powerpc/powerpc64/atomicity.h: New file.
* sysdeps/powerpc/powerpc64/backtrace.c: New file.
* sysdeps/powerpc/powerpc64/bp-asm.h: New file.
* sysdeps/powerpc/powerpc64/bsd-_setjmp.S: New file.
* sysdeps/powerpc/powerpc64/bsd-setjmp.S: New file.
* sysdeps/powerpc/powerpc64/dl-dtprocnum.h: New file.
* sysdeps/powerpc/powerpc64/dl-lookupcfg.h: New file.
* sysdeps/powerpc/powerpc64/dl-machine.c: New file.
* sysdeps/powerpc/powerpc64/dl-machine.h: New file.
* sysdeps/powerpc/powerpc64/memset.S: New file.
* sysdeps/powerpc/powerpc64/ppc-mcount.S: New file.
* sysdeps/powerpc/powerpc64/register-dump.h: New file.
* sysdeps/powerpc/powerpc64/setjmp.S: New file.
* sysdeps/powerpc/powerpc64/stpcpy.S: New file.
* sysdeps/powerpc/powerpc64/strchr.S: New file.
* sysdeps/powerpc/powerpc64/strcmp.S: New file.
* sysdeps/powerpc/powerpc64/strcpy.S: New file.
* sysdeps/powerpc/powerpc64/strlen.S: New file.
* sysdeps/powerpc/powerpc64/elf/bzero.S: New file.
* sysdeps/powerpc/powerpc64/elf/start.S: New file.
* sysdeps/powerpc/powerpc64/fpu/s_copysign.S: New file.
* sysdeps/powerpc/powerpc64/fpu/s_copysignf.S: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/Versions: New File.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/brk.S: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/clone.S: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/glob64.c: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/socket.S: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/syscalls.list: New file.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/sysdep.h: New file.
Diffstat (limited to 'sysdeps/powerpc/powerpc64/strlen.S')
-rw-r--r-- | sysdeps/powerpc/powerpc64/strlen.S | 163 |
1 files changed, 163 insertions, 0 deletions
diff --git a/sysdeps/powerpc/powerpc64/strlen.S b/sysdeps/powerpc/powerpc64/strlen.S new file mode 100644 index 0000000000..7907382002 --- /dev/null +++ b/sysdeps/powerpc/powerpc64/strlen.S @@ -0,0 +1,163 @@ +/* Optimized strlen implementation for PowerPC64. + Copyright (C) 1997, 1999, 2000, 2002 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, write to the Free + Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + 02111-1307 USA. */ + +#include <sysdep.h> +#include <bp-sym.h> +#include <bp-asm.h> + +/* The algorithm here uses the following techniques: + + 1) Given a word 'x', we can test to see if it contains any 0 bytes + by subtracting 0x01010101, and seeing if any of the high bits of each + byte changed from 0 to 1. This works because the least significant + 0 byte must have had no incoming carry (otherwise it's not the least + significant), so it is 0x00 - 0x01 == 0xff. For all other + byte values, either they have the high bit set initially, or when + 1 is subtracted you get a value in the range 0x00-0x7f, none of which + have their high bit set. The expression here is + (x + 0xfefefeff) & ~(x | 0x7f7f7f7f), which gives 0x00000000 when + there were no 0x00 bytes in the word. + + 2) Given a word 'x', we can test to see _which_ byte was zero by + calculating ~(((x & 0x7f7f7f7f) + 0x7f7f7f7f) | x | 0x7f7f7f7f). + This produces 0x80 in each byte that was zero, and 0x00 in all + the other bytes. The '| 0x7f7f7f7f' clears the low 7 bits in each + byte, and the '| x' part ensures that bytes with the high bit set + produce 0x00. The addition will carry into the high bit of each byte + iff that byte had one of its low 7 bits set. We can then just see + which was the most significant bit set and divide by 8 to find how + many to add to the index. + This is from the book 'The PowerPC Compiler Writer's Guide', + by Steve Hoxey, Faraydon Karim, Bill Hay and Hank Warren. + + We deal with strings not aligned to a word boundary by taking the + first word and ensuring that bytes not part of the string + are treated as nonzero. To allow for memory latency, we unroll the + loop a few times, being careful to ensure that we do not read ahead + across cache line boundaries. + + Questions to answer: + 1) How long are strings passed to strlen? If they're often really long, + we should probably use cache management instructions and/or unroll the + loop more. If they're often quite short, it might be better to use + fact (2) in the inner loop than have to recalculate it. + 2) How popular are bytes with the high bit set? If they are very rare, + on some processors it might be useful to use the simpler expression + ~((x - 0x01010101) | 0x7f7f7f7f) (that is, on processors with only one + ALU), but this fails when any character has its high bit set. */ + +/* Some notes on register usage: Under the SVR4 ABI, we can use registers + 0 and 3 through 12 (so long as we don't call any procedures) without + saving them. We can also use registers 14 through 31 if we save them. + We can't use r1 (it's the stack pointer), r2 nor r13 because the user + program may expect them to hold their usual value if we get sent + a signal. Integer parameters are passed in r3 through r10. + We can use condition registers cr0, cr1, cr5, cr6, and cr7 without saving + them, the others we must save. */ + +/* int [r3] strlen (char *s [r3]) */ + +ENTRY (BP_SYM (strlen)) + +#define rTMP1 r0 +#define rRTN r3 /* incoming STR arg, outgoing result */ +#define rSTR r4 /* current string position */ +#define rPADN r5 /* number of padding bits we prepend to the + string to make it start at a word boundary */ +#define rFEFE r6 /* constant 0xfefefeff (-0x01010101) */ +#define r7F7F r7 /* constant 0x7f7f7f7f */ +#define rWORD1 r8 /* current string word */ +#define rWORD2 r9 /* next string word */ +#define rMASK r9 /* mask for first string word */ +#define rTMP2 r10 +#define rTMP3 r11 +#define rTMP4 r12 + + CHECK_BOUNDS_LOW (rRTN, rTMP1, rTMP2) + + clrrdi rSTR, rRTN, 2 + lis r7F7F, 0x7f7f + rlwinm rPADN, rRTN, 3, 27, 28 + lwz rWORD1, 0(rSTR) + li rMASK, -1 + addi r7F7F, r7F7F, 0x7f7f +/* That's the setup done, now do the first pair of words. + We make an exception and use method (2) on the first two words, to reduce + overhead. */ + srw rMASK, rMASK, rPADN + and rTMP1, r7F7F, rWORD1 + or rTMP2, r7F7F, rWORD1 + add rTMP1, rTMP1, r7F7F + nor rTMP1, rTMP2, rTMP1 + and. rWORD1, rTMP1, rMASK + mtcrf 0x01, rRTN + bne L(done0) + lis rFEFE, -0x101 + addi rFEFE, rFEFE, -0x101 + clrldi rFEFE,rFEFE,32 /* clear upper 32 */ +/* Are we now aligned to a doubleword boundary? */ + bt 29, L(loop) + +/* Handle second word of pair. */ + lwzu rWORD1, 4(rSTR) + and rTMP1, r7F7F, rWORD1 + or rTMP2, r7F7F, rWORD1 + add rTMP1, rTMP1, r7F7F + nor. rWORD1, rTMP2, rTMP1 + clrldi. rWORD1,rWORD1,32 /* clear upper 32 */ + bne L(done0) + +/* The loop. */ + +L(loop): + lwz rWORD1, 4(rSTR) + lwzu rWORD2, 8(rSTR) + add rTMP1, rFEFE, rWORD1 + nor rTMP2, r7F7F, rWORD1 + and. rTMP1, rTMP1, rTMP2 + clrldi. rTMP1,rTMP1,32 /* clear upper 32 */ + add rTMP3, rFEFE, rWORD2 + nor rTMP4, r7F7F, rWORD2 + bne L(done1) + and. rTMP1, rTMP3, rTMP4 + clrldi. rTMP1,rTMP1,32 /* clear upper 32 */ + beq L(loop) + + and rTMP1, r7F7F, rWORD2 + add rTMP1, rTMP1, r7F7F + andc rWORD1, rTMP4, rTMP1 + b L(done0) + +L(done1): + and rTMP1, r7F7F, rWORD1 + subi rSTR, rSTR, 4 + add rTMP1, rTMP1, r7F7F + andc rWORD1, rTMP2, rTMP1 + +/* When we get to here, rSTR points to the first word in the string that + contains a zero byte, and the most significant set bit in rWORD1 is in that + byte. */ +L(done0): + cntlzw rTMP3, rWORD1 + subf rTMP1, rRTN, rSTR + srwi rTMP3, rTMP3, 3 + add rRTN, rTMP1, rTMP3 + /* GKM FIXME: check high bound. */ + blr +END (BP_SYM (strlen)) |