diff options
Diffstat (limited to 'sysdeps/powerpc')
| -rw-r--r-- | sysdeps/powerpc/Makefile | 6 | ||||
| -rw-r--r-- | sysdeps/powerpc/bits/mathinline.h | 44 | ||||
| -rw-r--r-- | sysdeps/powerpc/bzero.S | 27 | ||||
| -rw-r--r-- | sysdeps/powerpc/dl-machine.h | 71 | ||||
| -rw-r--r-- | sysdeps/powerpc/machine-gmon.h | 1 | ||||
| -rw-r--r-- | sysdeps/powerpc/stpcpy.S | 100 | ||||
| -rw-r--r-- | sysdeps/powerpc/strcat.c | 30 | ||||
| -rw-r--r-- | sysdeps/powerpc/strcpy.S | 100 |
8 files changed, 338 insertions, 41 deletions
diff --git a/sysdeps/powerpc/Makefile b/sysdeps/powerpc/Makefile index 3e8f22b573..1813c78dd8 100644 --- a/sysdeps/powerpc/Makefile +++ b/sysdeps/powerpc/Makefile @@ -21,7 +21,7 @@ endif # so that's at least 8192 entries. Since libc only uses about 1200 entries, # we want to use -fpic, because this generates fewer relocs. ifeq (yes,$(build-shared)) -CFLAGS-.os = -fpic -fno-common +pic-ccflag = -fpic endif # The initfini generation code doesn't work in the presence of -fPIC, so @@ -29,3 +29,7 @@ endif ifeq ($(subdir),csu) CFLAGS-initfini.s = -g0 -fpic endif + +ifeq ($(subdir),string) +CFLAGS-memcmp.c += -Wno-uninitialized +endif diff --git a/sysdeps/powerpc/bits/mathinline.h b/sysdeps/powerpc/bits/mathinline.h index d7f2a3ac0d..d7cabaf28b 100644 --- a/sysdeps/powerpc/bits/mathinline.h +++ b/sysdeps/powerpc/bits/mathinline.h @@ -32,34 +32,22 @@ __sgn1 (double __x) { return __x >= 0.0 ? 1.0 : -1.0; } +#endif /* __NO_MATH_INLINES && __OPTIMZE__ */ -/* We'd want to use this if it was implemented in hardware, but - how can we tell? */ -#if 0 -__MATH_INLINE double sqrt (double __x); -__MATH_INLINE double -sqrt (double __x) -{ - register double __value; - __asm - ("fsqrt %0,%1" - : "=f" (__value) : "f" (__x)); - - return __value; -} -#endif - -__MATH_INLINE double fabs (double __x); -__MATH_INLINE double -fabs (double __x) -{ - register double __value; - __asm - ("fabs %0,%1" - : "=f" (__value) : "f" (__x)); - - return __value; -} +#if __USE_ISOC9X +# define __unordered_cmp(x, y) \ + (__extension__ \ + ({ __typeof__(x) __x = (x); __typeof__(y) __y = (y); \ + unsigned __r; \ + __asm__("fcmpu 7,%1,%2 ; mfcr %0" : "=r" (__r) : "f" (__x), "f"(__y)); \ + __r; })) + +# define isgreater(x, y) (__unordered_cmp (x, y) >> 2 & 1) +# define isgreaterequal(x, y) ((__unordered_cmp (x, y) & 6) != 0) +# define isless(x, y) (__unordered_cmp (x, y) >> 3 & 1) +# define islessequal(x, y) ((__unordered_cmp (x, y) & 0xA) != 0) +# define islessgreater(x, y) ((__unordered_cmp (x, y) & 0xC) != 0) +# define isunordered(x, y) (__unordered_cmp (x, y) & 1) +#endif /* __USE_ISOC9X */ -#endif /* __NO_MATH_INLINES && __OPTIMZE__ */ #endif /* __GNUC__ */ diff --git a/sysdeps/powerpc/bzero.S b/sysdeps/powerpc/bzero.S new file mode 100644 index 0000000000..783a91fdcb --- /dev/null +++ b/sysdeps/powerpc/bzero.S @@ -0,0 +1,27 @@ +/* Optimized bzero `implementation' for PowerPC. + Copyright (C) 1997 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 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., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +#include <sysdep.h> + +ENTRY(__bzero) + mr %r5,%r4 + li %r4,0 + b memset@local +END(__bzero) +weak_alias (__bzero, bzero) diff --git a/sysdeps/powerpc/dl-machine.h b/sysdeps/powerpc/dl-machine.h index 771b711a14..70a3f20b31 100644 --- a/sysdeps/powerpc/dl-machine.h +++ b/sysdeps/powerpc/dl-machine.h @@ -199,6 +199,53 @@ _dl_runtime_resolve: bctr 0: .size _dl_runtime_resolve,0b-_dl_runtime_resolve + + .align 2 + .globl _dl_prof_resolve + .type _dl_prof_resolve,@function +_dl_prof_resolve: + # We need to save the registers used to pass parameters, and register 0, + # which is used by _mcount; the registers are saved in a stack frame. + stwu 1,-48(1) + stw 0,12(1) + stw 3,16(1) + stw 4,20(1) + # The code that calls this has put parameters for `fixup' in r12 and r11. + mr 3,12 + stw 5,24(1) + mr 4,11 + stw 6,28(1) + mflr 5 + # We also need to save some of the condition register fields. + stw 7,32(1) + stw 5,52(1) + stw 8,36(1) + mfcr 0 + stw 9,40(1) + stw 10,44(1) + stw 0,8(1) + bl profile_fixup@local + # 'fixup' returns the address we want to branch to. + mtctr 3 + # Put the registers back... + lwz 0,52(1) + lwz 10,44(1) + lwz 9,40(1) + mtlr 0 + lwz 8,36(1) + lwz 0,8(1) + lwz 7,32(1) + lwz 6,28(1) + mtcrf 0xFF,0 + lwz 5,24(1) + lwz 4,20(1) + lwz 3,16(1) + lwz 0,12(1) + # ...unwind the stack frame, and jump to the PLT entry we updated. + addi 1,1,48 + bctr +0: + .size _dl_prof_resolve,0b-_dl_prof_resolve # Undo '.section text'. .previous "); @@ -409,8 +456,14 @@ elf_machine_runtime_setup (struct link_map *map, int lazy, int profile) Elf32_Word num_plt_entries = (map->l_info[DT_PLTRELSZ]->d_un.d_val / sizeof (Elf32_Rela)); Elf32_Word rel_offset_words = PLT_DATA_START_WORDS (num_plt_entries); - extern void _dl_runtime_resolve (void); Elf32_Word size_modified; + extern void _dl_runtime_resolve (void); + extern void _dl_prof_resolve (void); + Elf32_Word dlrr; + + dlrr = (Elf32_Word)(char *)(profile + ? _dl_prof_resolve + : _dl_runtime_resolve); if (lazy) for (i = 0; i < num_plt_entries; i++) @@ -433,8 +486,7 @@ elf_machine_runtime_setup (struct link_map *map, int lazy, int profile) /* Multiply index of entry by 3 (in r11). */ plt[0] = OPCODE_SLWI (12, 11, 1); plt[1] = OPCODE_ADD (11, 12, 11); - if ((Elf32_Word) (char *) _dl_runtime_resolve <= 0x01fffffc || - (Elf32_Word) (char *) _dl_runtime_resolve >= 0xfe000000) + if (dlrr <= 0x01fffffc || dlrr >= 0xfe000000) { /* Load address of link map in r12. */ plt[2] = OPCODE_LI (12, (Elf32_Word) (char *) map); @@ -442,15 +494,13 @@ elf_machine_runtime_setup (struct link_map *map, int lazy, int profile) + 0x8000) >> 16)); /* Call _dl_runtime_resolve. */ - plt[4] = OPCODE_BA ((Elf32_Word) (char *) _dl_runtime_resolve); + plt[4] = OPCODE_BA (dlrr); } else { /* Get address of _dl_runtime_resolve in CTR. */ - plt[2] = OPCODE_LI (12, (Elf32_Word) (char *) _dl_runtime_resolve); - plt[3] = OPCODE_ADDIS (12, 12, ((((Elf32_Word) (char *) - _dl_runtime_resolve) - + 0x8000) >> 16)); + plt[2] = OPCODE_LI (12, dlrr); + plt[3] = OPCODE_ADDIS (12, 12, (dlrr + 0x8000) >> 16); plt[4] = OPCODE_MTCTR (12); /* Load address of link map in r12. */ @@ -501,7 +551,6 @@ elf_machine_runtime_setup (struct link_map *map, int lazy, int profile) static inline void elf_machine_lazy_rel (struct link_map *map, const Elf32_Rela *reloc) { - assert (ELF32_R_TYPE (reloc->r_info) == R_PPC_JMP_SLOT); /* elf_machine_runtime_setup handles this. */ } @@ -513,7 +562,7 @@ elf_machine_lazy_rel (struct link_map *map, const Elf32_Rela *reloc) LOADADDR is the load address of the object; INFO is an array indexed by DT_* of the .dynamic section info. */ -static inline void +static void elf_machine_rela (struct link_map *map, const Elf32_Rela *reloc, const Elf32_Sym *sym, const struct r_found_version *version, Elf32_Addr *const reloc_addr) @@ -709,4 +758,4 @@ elf_machine_rela (struct link_map *map, const Elf32_Rela *reloc, #define ELF_MACHINE_NO_REL 1 -#endif +#endif /* RESOLVE */ diff --git a/sysdeps/powerpc/machine-gmon.h b/sysdeps/powerpc/machine-gmon.h index ba53807308..b890dd5c53 100644 --- a/sysdeps/powerpc/machine-gmon.h +++ b/sysdeps/powerpc/machine-gmon.h @@ -1,7 +1,6 @@ /* PowerPC-specific implementation of profiling support. Copyright (C) 1997 Free Software Foundation, Inc. This file is part of the GNU C Library. - Contributed by Ulrich Drepper <drepper@cygnus.com>, 1997. 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 diff --git a/sysdeps/powerpc/stpcpy.S b/sysdeps/powerpc/stpcpy.S new file mode 100644 index 0000000000..58ad5b12b6 --- /dev/null +++ b/sysdeps/powerpc/stpcpy.S @@ -0,0 +1,100 @@ +/* Optimized stpcpy implementation for PowerPC. + Copyright (C) 1997 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 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., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +#include <sysdep.h> + +/* See strlen.s for comments on how the end-of-string testing works. */ + +EALIGN(__stpcpy,4,0) +/* char * [r3] stpcpy (char *dest [r3], const char *src [r4]) */ + +/* General register assignments: + r0: temporary + r3: pointer to previous word in dest + r4: pointer to previous word in src + r6: current word from src + r7: 0xfefefeff + r8: 0x7f7f7f7f + r9: ~(word in src | 0x7f7f7f7f) + r10: alternate word from src. */ + + or %r0,%r4,%r3 + clrlwi. %r0,%r0,30 + addi %r3,%r3,-4 + bne L(unaligned) + + lis %r7,0xfeff + lis %r8,0x7f7f + lwz %r6,0(%r4) + addi %r7,%r7,-0x101 + addi %r8,%r8,0x7f7f + b 2f + +0: lwzu %r10,4(%r4) + stwu %r6,4(%r3) + add %r0,%r7,%r10 + nor %r9,%r8,%r10 + and. %r0,%r0,%r9 + bne- 1f + lwzu %r6,4(%r4) + stwu %r10,4(%r3) +2: add %r0,%r7,%r6 + nor %r9,%r8,%r6 + and. %r0,%r0,%r9 + beq+ 0b + + mr %r10,%r6 +/* We've hit the end of the string. Do the rest byte-by-byte. */ +1: rlwinm. %r0,%r10,8,24,31 + stbu %r0,4(%r3) + beqlr- + rlwinm. %r0,%r10,16,24,31 + stbu %r0,1(%r3) + beqlr- + rlwinm. %r0,%r10,24,24,31 + stbu %r0,1(%r3) + beqlr- + stbu %r10,1(%r3) + blr + +/* Oh well. In this case, we just do a byte-by-byte copy. */ + .align 4 + nop +L(unaligned): + lbz %r6,0(%r4) + addi %r3,%r3,3 + cmpwi %r6,0 + beq- 2f + +0: lbzu %r10,1(%r4) + stbu %r6,1(%r3) + cmpwi %r10,0 + beq- 1f + nop /* Let 601 load start of loop. */ + lbzu %r6,1(%r4) + stbu %r10,1(%r3) + cmpwi %r6,0 + bne+ 0b +2: stbu %r6,1(%r3) + blr +1: stbu %r10,1(%r3) + blr +END(__stpcpy) + +weak_alias (__stpcpy, stpcpy) diff --git a/sysdeps/powerpc/strcat.c b/sysdeps/powerpc/strcat.c new file mode 100644 index 0000000000..9d9ab6549a --- /dev/null +++ b/sysdeps/powerpc/strcat.c @@ -0,0 +1,30 @@ +/* strcat version that uses fast strcpy/strlen. + Copyright (C) 1997 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 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., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +#include <string.h> + +#undef strcat + +/* Append SRC on the end of DEST. */ +char * +strcat (char *dest, const char *src) +{ + strcpy (dest + strlen (dest), src); + return dest; +} diff --git a/sysdeps/powerpc/strcpy.S b/sysdeps/powerpc/strcpy.S new file mode 100644 index 0000000000..3c0cce7844 --- /dev/null +++ b/sysdeps/powerpc/strcpy.S @@ -0,0 +1,100 @@ +/* Optimized strcpy implementation for PowerPC. + Copyright (C) 1997 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 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., 59 Temple Place - Suite 330, + Boston, MA 02111-1307, USA. */ + +#include <sysdep.h> + +/* See strlen.s for comments on how the end-of-string testing works. */ + +EALIGN(strcpy,4,0) +/* char * [r3] strcpy (char *dest [r3], const char *src [r4]) */ + +/* General register assignments: + r0: temporary + r3: saved `dest' + r4: pointer to previous word in src + r5: pointer to previous word in dest + r6: current word from src + r7: 0xfefefeff + r8: 0x7f7f7f7f + r9: ~(word in src | 0x7f7f7f7f) + r10: alternate word from src. */ + + or %r0,%r4,%r3 + clrlwi. %r0,%r0,30 + addi %r5,%r3,-4 + bne L(unaligned) + + lis %r7,0xfeff + lis %r8,0x7f7f + lwz %r6,0(%r4) + addi %r7,%r7,-0x101 + addi %r8,%r8,0x7f7f + b 2f + +0: lwzu %r10,4(%r4) + stwu %r6,4(%r5) + add %r0,%r7,%r10 + nor %r9,%r8,%r10 + and. %r0,%r0,%r9 + bne- 1f + lwzu %r6,4(%r4) + stwu %r10,4(%r5) +2: add %r0,%r7,%r6 + nor %r9,%r8,%r6 + and. %r0,%r0,%r9 + beq+ 0b + + mr %r10,%r6 +/* We've hit the end of the string. Do the rest byte-by-byte. */ +1: rlwinm. %r0,%r10,8,24,31 + stb %r0,4(%r5) + beqlr- + rlwinm. %r0,%r10,16,24,31 + stb %r0,5(%r5) + beqlr- + rlwinm. %r0,%r10,24,24,31 + stb %r0,6(%r5) + beqlr- + stb %r10,7(%r5) + blr + +/* Oh well. In this case, we just do a byte-by-byte copy. */ + .align 4 + nop +L(unaligned): + lbz %r6,0(%r4) + addi %r5,%r3,-1 + cmpwi %r6,0 + beq- 2f + +0: lbzu %r10,1(%r4) + stbu %r6,1(%r5) + cmpwi %r10,0 + beq- 1f + nop /* Let 601 load start of loop. */ + lbzu %r6,1(%r4) + stbu %r10,1(%r5) + cmpwi %r6,0 + bne+ 0b +2: stb %r6,1(%r5) + blr +1: stb %r10,1(%r5) + blr + +END(strcpy) |