diff options
author | Ulrich Drepper <drepper@redhat.com> | 2004-12-22 20:10:10 +0000 |
---|---|---|
committer | Ulrich Drepper <drepper@redhat.com> | 2004-12-22 20:10:10 +0000 |
commit | a334319f6530564d22e775935d9c91663623a1b4 (patch) | |
tree | b5877475619e4c938e98757d518bb1e9cbead751 /sysdeps/mips/mips64 | |
parent | 0ecb606cb6cf65de1d9fc8a919bceb4be476c602 (diff) | |
download | glibc-a334319f6530564d22e775935d9c91663623a1b4.tar.gz glibc-a334319f6530564d22e775935d9c91663623a1b4.tar.xz glibc-a334319f6530564d22e775935d9c91663623a1b4.zip |
(CFLAGS-tst-align.c): Add -mpreferred-stack-boundary=4.
Diffstat (limited to 'sysdeps/mips/mips64')
-rw-r--r-- | sysdeps/mips/mips64/Dist | 1 | ||||
-rw-r--r-- | sysdeps/mips/mips64/Implies | 3 | ||||
-rw-r--r-- | sysdeps/mips/mips64/__longjmp.c | 98 | ||||
-rw-r--r-- | sysdeps/mips/mips64/add_n.S | 130 | ||||
-rw-r--r-- | sysdeps/mips/mips64/addmul_1.S | 107 | ||||
-rw-r--r-- | sysdeps/mips/mips64/bsd-_setjmp.S | 49 | ||||
-rw-r--r-- | sysdeps/mips/mips64/bsd-setjmp.S | 47 | ||||
-rw-r--r-- | sysdeps/mips/mips64/gmp-mparam.h | 31 | ||||
-rw-r--r-- | sysdeps/mips/mips64/lshift.S | 105 | ||||
-rw-r--r-- | sysdeps/mips/mips64/memcpy.S | 140 | ||||
-rw-r--r-- | sysdeps/mips/mips64/memset.S | 92 | ||||
-rw-r--r-- | sysdeps/mips/mips64/mul_1.S | 96 | ||||
-rw-r--r-- | sysdeps/mips/mips64/n32/Implies | 4 | ||||
-rw-r--r-- | sysdeps/mips/mips64/n32/Makefile | 6 | ||||
-rw-r--r-- | sysdeps/mips/mips64/n64/Implies | 4 | ||||
-rw-r--r-- | sysdeps/mips/mips64/n64/Makefile | 6 | ||||
-rw-r--r-- | sysdeps/mips/mips64/rshift.S | 102 | ||||
-rw-r--r-- | sysdeps/mips/mips64/setjmp.S | 46 | ||||
-rw-r--r-- | sysdeps/mips/mips64/setjmp_aux.c | 78 | ||||
-rw-r--r-- | sysdeps/mips/mips64/soft-fp/Dist | 1 | ||||
-rw-r--r-- | sysdeps/mips/mips64/soft-fp/sfp-machine.h | 47 | ||||
-rw-r--r-- | sysdeps/mips/mips64/sub_n.S | 130 | ||||
-rw-r--r-- | sysdeps/mips/mips64/submul_1.S | 108 |
23 files changed, 1431 insertions, 0 deletions
diff --git a/sysdeps/mips/mips64/Dist b/sysdeps/mips/mips64/Dist new file mode 100644 index 0000000000..ad6ea0313a --- /dev/null +++ b/sysdeps/mips/mips64/Dist @@ -0,0 +1 @@ +setjmp_aux.c diff --git a/sysdeps/mips/mips64/Implies b/sysdeps/mips/mips64/Implies new file mode 100644 index 0000000000..8c18cb3034 --- /dev/null +++ b/sysdeps/mips/mips64/Implies @@ -0,0 +1,3 @@ +# MIPS uses IEEE 754 floating point. +ieee754/flt-32 +ieee754/dbl-64 diff --git a/sysdeps/mips/mips64/__longjmp.c b/sysdeps/mips/mips64/__longjmp.c new file mode 100644 index 0000000000..546493f842 --- /dev/null +++ b/sysdeps/mips/mips64/__longjmp.c @@ -0,0 +1,98 @@ +/* Copyright (C) 1992, 1995, 1997, 2000, 2003, 2004 + Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Brendan Kehoe (brendan@zen.org). + + 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 <setjmp.h> +#include <sgidefs.h> +#include <stdlib.h> + +#undef __longjmp + +#ifndef __GNUC__ + #error This file uses GNU C extensions; you must compile with GCC. +#endif + +void +__longjmp (env, val_arg) + __jmp_buf env; + int val_arg; +{ + /* gcc 1.39.19 miscompiled the longjmp routine (as it did setjmp before + the hack around it); force it to use $a1 for the longjmp value. + Without this it saves $a1 in a register which gets clobbered + along the way. */ + register int val asm ("a1"); + + /* Pull back the floating point callee-saved registers. */ +#if _MIPS_SIM == _ABI64 + asm volatile ("l.d $f24, %0" : : "m" (env[0].__fpregs[0])); + asm volatile ("l.d $f25, %0" : : "m" (env[0].__fpregs[1])); + asm volatile ("l.d $f26, %0" : : "m" (env[0].__fpregs[2])); + asm volatile ("l.d $f27, %0" : : "m" (env[0].__fpregs[3])); + asm volatile ("l.d $f28, %0" : : "m" (env[0].__fpregs[4])); + asm volatile ("l.d $f29, %0" : : "m" (env[0].__fpregs[5])); + asm volatile ("l.d $f30, %0" : : "m" (env[0].__fpregs[6])); + asm volatile ("l.d $f31, %0" : : "m" (env[0].__fpregs[7])); +#else + asm volatile ("l.d $f20, %0" : : "m" (env[0].__fpregs[0])); + asm volatile ("l.d $f22, %0" : : "m" (env[0].__fpregs[1])); + asm volatile ("l.d $f24, %0" : : "m" (env[0].__fpregs[2])); + asm volatile ("l.d $f26, %0" : : "m" (env[0].__fpregs[3])); + asm volatile ("l.d $f28, %0" : : "m" (env[0].__fpregs[4])); + asm volatile ("l.d $f30, %0" : : "m" (env[0].__fpregs[5])); +#endif + + /* Get and reconstruct the floating point csr. */ + asm volatile ("lw $2, %0" : : "m" (env[0].__fpc_csr)); + asm volatile ("ctc1 $2, $31"); + + /* Get the GP. */ + asm volatile ("ld $gp, %0" : : "m" (env[0].__gp)); + + /* Get the callee-saved registers. */ + asm volatile ("ld $16, %0" : : "m" (env[0].__regs[0])); + asm volatile ("ld $17, %0" : : "m" (env[0].__regs[1])); + asm volatile ("ld $18, %0" : : "m" (env[0].__regs[2])); + asm volatile ("ld $19, %0" : : "m" (env[0].__regs[3])); + asm volatile ("ld $20, %0" : : "m" (env[0].__regs[4])); + asm volatile ("ld $21, %0" : : "m" (env[0].__regs[5])); + asm volatile ("ld $22, %0" : : "m" (env[0].__regs[6])); + asm volatile ("ld $23, %0" : : "m" (env[0].__regs[7])); + + /* Get the PC. */ + asm volatile ("ld $31, %0" : : "m" (env[0].__pc)); + + + /* Restore the stack pointer and the FP. They have to be restored + last and in a single asm as gcc, depending on options used, may + use either of them to access env. */ + asm volatile ("ld $29, %0\n\t" + "ld $30, %1\n\t" : : "m" (env[0].__sp), "m" (env[0].__fp)); + +/* Give setjmp 1 if given a 0, or what they gave us if non-zero. */ + if (val == 0) + asm volatile ("dli $2, 1"); + else + asm volatile ("move $2, %0" : : "r" (val)); + + asm volatile ("j $31"); + + /* Avoid `volatile function does return' warnings. */ + for (;;); +} diff --git a/sysdeps/mips/mips64/add_n.S b/sysdeps/mips/mips64/add_n.S new file mode 100644 index 0000000000..072f4f0b73 --- /dev/null +++ b/sysdeps/mips/mips64/add_n.S @@ -0,0 +1,130 @@ +/* MIPS3 __mpn_add_n -- Add two limb vectors of the same length > 0 and + * store sum in a third limb vector. + * + * Copyright (C) 1995, 2000, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* + * INPUT PARAMETERS + * res_ptr $4 + * s1_ptr $5 + * s2_ptr $6 + * size $7 + */ +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 2 + .globl __mpn_add_n + .ent __mpn_add_n +__mpn_add_n: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + ld $10,0($5) + ld $11,0($6) + + daddiu $7,$7,-1 + and $9,$7,4-1 # number of limbs in first loop + beq $9,$0,L(L0) # if multiple of 4 limbs, skip first loop + move $2,$0 + + dsubu $7,$7,$9 + +L(Loop0): daddiu $9,$9,-1 + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + or $2,$2,$8 + + daddiu $5,$5,8 + daddiu $6,$6,8 + move $10,$12 + move $11,$13 + bne $9,$0,L(Loop0) + daddiu $4,$4,8 + +L(L0): beq $7,$0,L(Lend) + nop + +L(Loop): daddiu $7,$7,-4 + + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + or $2,$2,$8 + + ld $10,16($5) + daddu $13,$13,$2 + ld $11,16($6) + sltu $8,$13,$2 + daddu $13,$12,$13 + sltu $2,$13,$12 + sd $13,8($4) + or $2,$2,$8 + + ld $12,24($5) + daddu $11,$11,$2 + ld $13,24($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,16($4) + or $2,$2,$8 + + ld $10,32($5) + daddu $13,$13,$2 + ld $11,32($6) + sltu $8,$13,$2 + daddu $13,$12,$13 + sltu $2,$13,$12 + sd $13,24($4) + or $2,$2,$8 + + daddiu $5,$5,32 + daddiu $6,$6,32 + + bne $7,$0,L(Loop) + daddiu $4,$4,32 + +L(Lend): daddu $11,$11,$2 + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + j $31 + or $2,$2,$8 + + .end __mpn_add_n diff --git a/sysdeps/mips/mips64/addmul_1.S b/sysdeps/mips/mips64/addmul_1.S new file mode 100644 index 0000000000..f5ecd83702 --- /dev/null +++ b/sysdeps/mips/mips64/addmul_1.S @@ -0,0 +1,107 @@ +/* MIPS3 __mpn_addmul_1 -- Multiply a limb vector with a single limb and + * add the product to a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * s1_ptr $5 + * size $6 + * s2_limb $7 + */ + +#ifdef PIC + .option pic2 +#endif + .text + .align 4 + .globl __mpn_addmul_1 + .ent __mpn_addmul_1 +__mpn_addmul_1: +#ifdef PIC + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + # warm up phase 0 + ld $8,0($5) + + # warm up phase 1 + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,L(LC0) + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,L(LC1) + ld $8,0($5) # load new s1 limb as early as possible + +L(Loop): ld $10,0($4) + mflo $3 + mfhi $9 + daddiu $5,$5,8 + daddu $3,$3,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$3,$2 # carry from previous addition -> $2 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + bne $6,$0,L(Loop) + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 1 +L(LC1): ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dmultu $8,$7 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 0 +L(LC0): ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end __mpn_addmul_1 diff --git a/sysdeps/mips/mips64/bsd-_setjmp.S b/sysdeps/mips/mips64/bsd-_setjmp.S new file mode 100644 index 0000000000..7620cf391f --- /dev/null +++ b/sysdeps/mips/mips64/bsd-_setjmp.S @@ -0,0 +1,49 @@ +/* BSD `_setjmp' entry point to `sigsetjmp (..., 0)'. MIPS64 version. + Copyright (C) 1996, 1997, 2000, 2002, 2003, 2004 + 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. */ + +/* This just does a tail-call to `__sigsetjmp (ARG, 0)'. + We cannot do it in C because it must be a tail-call, so frame-unwinding + in setjmp doesn't clobber the state restored by longjmp. */ + +#include <sgidefs.h> +#include <sysdep.h> +#include <sys/asm.h> + +#ifdef __PIC__ + .option pic2 +#endif +ENTRY (_setjmp) +#ifdef __PIC__ + SETUP_GP +#endif + SETUP_GP64 (v0, C_SYMBOL_NAME (_setjmp)) + PTR_LA t9, C_SYMBOL_NAME (__sigsetjmp) +#if _MIPS_SIM == _ABIO32 + nop +#endif + RESTORE_GP64 + move a1, zero /* Pass a second argument of zero. */ +#ifdef __PIC__ + jr t9 +#else + j C_SYMBOL_NAME (__sigsetjmp) +#endif + .end _setjmp +libc_hidden_def (_setjmp) diff --git a/sysdeps/mips/mips64/bsd-setjmp.S b/sysdeps/mips/mips64/bsd-setjmp.S new file mode 100644 index 0000000000..2a1fd9ce71 --- /dev/null +++ b/sysdeps/mips/mips64/bsd-setjmp.S @@ -0,0 +1,47 @@ +/* BSD `setjmp' entry point to `sigsetjmp (..., 1)'. MIPS64 version. + Copyright (C) 1996, 1997, 2002, 2003, 2004 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. */ + +/* This just does a tail-call to `__sigsetjmp (ARG, 1)'. + We cannot do it in C because it must be a tail-call, so frame-unwinding + in setjmp doesn't clobber the state restored by longjmp. */ + +#include <sgidefs.h> +#include <sysdep.h> +#include <sys/asm.h> + +#ifdef PIC + .option pic2 +#endif +ENTRY (setjmp) +#ifdef __PIC__ + SETUP_GP +#endif + SETUP_GP64 (v0, C_SYMBOL_NAME (setjmp)) + PTR_LA t9, C_SYMBOL_NAME (__sigsetjmp) +#if _MIPS_SIM == _ABIO32 + nop +#endif + RESTORE_GP64 + dli a1, 1 /* Pass a second argument of one. */ +#ifdef __PIC__ + jr t9 +#else + j C_SYMBOL_NAME (__sigsetjmp) +#endif + .end setjmp diff --git a/sysdeps/mips/mips64/gmp-mparam.h b/sysdeps/mips/mips64/gmp-mparam.h new file mode 100644 index 0000000000..7666137a21 --- /dev/null +++ b/sysdeps/mips/mips64/gmp-mparam.h @@ -0,0 +1,31 @@ +/* gmp-mparam.h -- Compiler/machine parameter header file. + +Copyright (C) 1991, 1993, 1994, 2002, 2003 Free Software Foundation, Inc. + +This file is part of the GNU MP Library. + +The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to +the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */ + +#if defined __GMP_H__ && ! defined _LONG_LONG_LIMB +#error "Included too late for _LONG_LONG_LIMB to take effect" +#endif + +#define _LONG_LONG_LIMB +#define BITS_PER_MP_LIMB 64 +#define BYTES_PER_MP_LIMB 8 +#define BITS_PER_LONGINT __WORDSIZE +#define BITS_PER_INT 32 +#define BITS_PER_SHORTINT 16 +#define BITS_PER_CHAR 8 diff --git a/sysdeps/mips/mips64/lshift.S b/sysdeps/mips/mips64/lshift.S new file mode 100644 index 0000000000..20f9e3da19 --- /dev/null +++ b/sysdeps/mips/mips64/lshift.S @@ -0,0 +1,105 @@ +/* MIPS3 __mpn_lshift -- + * + * Copyright (C) 1995, 2000, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * src_ptr $5 + * size $6 + * cnt $7 + */ + +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 2 + .globl __mpn_lshift + .ent __mpn_lshift +__mpn_lshift: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + dsll $2,$6,3 + daddu $5,$5,$2 # make r5 point at end of src + ld $10,-8($5) # load first limb + dsubu $13,$0,$7 + daddu $4,$4,$2 # make r4 point at end of res + daddiu $6,$6,-1 + and $9,$6,4-1 # number of limbs in first loop + beq $9,$0,L(L0) # if multiple of 4 limbs, skip first loop + dsrl $2,$10,$13 # compute function result + + dsubu $6,$6,$9 + +L(Loop0): ld $3,-16($5) + daddiu $4,$4,-8 + daddiu $5,$5,-8 + daddiu $9,$9,-1 + dsll $11,$10,$7 + dsrl $12,$3,$13 + move $10,$3 + or $8,$11,$12 + bne $9,$0,L(Loop0) + sd $8,0($4) + +L(L0): beq $6,$0,L(Lend) + nop + +L(Loop): ld $3,-16($5) + daddiu $4,$4,-32 + daddiu $6,$6,-4 + dsll $11,$10,$7 + dsrl $12,$3,$13 + + ld $10,-24($5) + dsll $14,$3,$7 + or $8,$11,$12 + sd $8,24($4) + dsrl $9,$10,$13 + + ld $3,-32($5) + dsll $11,$10,$7 + or $8,$14,$9 + sd $8,16($4) + dsrl $12,$3,$13 + + ld $10,-40($5) + dsll $14,$3,$7 + or $8,$11,$12 + sd $8,8($4) + dsrl $9,$10,$13 + + daddiu $5,$5,-32 + or $8,$14,$9 + bgtz $6,L(Loop) + sd $8,0($4) + +L(Lend): dsll $8,$10,$7 + j $31 + sd $8,-8($4) + .end __mpn_lshift diff --git a/sysdeps/mips/mips64/memcpy.S b/sysdeps/mips/mips64/memcpy.S new file mode 100644 index 0000000000..eab9463f86 --- /dev/null +++ b/sysdeps/mips/mips64/memcpy.S @@ -0,0 +1,140 @@ +/* Copyright (C) 2002, 2003 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Hartvig Ekner <hartvige@mips.com>, 2002. + Ported to mips3 n32/n64 by Alexandre Oliva <aoliva@redhat.com> + + 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 <endian.h> +#include <sys/asm.h> + + +/* void *memcpy(void *s1, const void *s2, size_t n); + + This could probably be optimized further. */ + +#if __BYTE_ORDER == __BIG_ENDIAN +# define LDHI ldl /* high part is left in big-endian */ +# define SDHI sdl /* high part is left in big-endian */ +# define LDLO ldr /* low part is right in big-endian */ +# define SDLO sdr /* low part is right in big-endian */ +#else +# define LDHI ldr /* high part is right in little-endian */ +# define SDHI sdr /* high part is right in little-endian */ +# define LDLO ldl /* low part is left in little-endian */ +# define SDLO sdl /* low part is left in little-endian */ +#endif + +ENTRY (memcpy) + .set noreorder + + slti t0, a2, 16 # Less than 16? + bne t0, zero, L(last16) + move v0, a0 # Setup exit value before too late + + xor t0, a1, a0 # Find a0/a1 displacement + andi t0, 0x7 + bne t0, zero, L(shift) # Go handle the unaligned case + PTR_SUBU t1, zero, a1 + andi t1, 0x7 # a0/a1 are aligned, but are we + beq t1, zero, L(chk8w) # starting in the middle of a word? + PTR_SUBU a2, t1 + LDHI t0, 0(a1) # Yes we are... take care of that + PTR_ADDU a1, t1 + SDHI t0, 0(a0) + PTR_ADDU a0, t1 + +L(chk8w): + andi t0, a2, 0x3f # 64 or more bytes left? + beq t0, a2, L(chk1w) + PTR_SUBU a3, a2, t0 # Yes + PTR_ADDU a3, a1 # a3 = end address of loop + move a2, t0 # a2 = what will be left after loop +L(lop8w): + ld t0, 0(a1) # Loop taking 8 words at a time + ld t1, 8(a1) + ld t2, 16(a1) + ld t3, 24(a1) + ld ta0, 32(a1) + ld ta1, 40(a1) + ld ta2, 48(a1) + ld ta3, 56(a1) + PTR_ADDIU a0, 64 + PTR_ADDIU a1, 64 + sd t0, -64(a0) + sd t1, -56(a0) + sd t2, -48(a0) + sd t3, -40(a0) + sd ta0, -32(a0) + sd ta1, -24(a0) + sd ta2, -16(a0) + bne a1, a3, L(lop8w) + sd ta3, -8(a0) + +L(chk1w): + andi t0, a2, 0x7 # 8 or more bytes left? + beq t0, a2, L(last16) + PTR_SUBU a3, a2, t0 # Yes, handle them one dword at a time + PTR_ADDU a3, a1 # a3 again end address + move a2, t0 +L(lop1w): + ld t0, 0(a1) + PTR_ADDIU a0, 8 + PTR_ADDIU a1, 8 + bne a1, a3, L(lop1w) + sd t0, -8(a0) + +L(last16): + blez a2, L(lst16e) # Handle last 16 bytes, one at a time + PTR_ADDU a3, a2, a1 +L(lst16l): + lb t0, 0(a1) + PTR_ADDIU a0, 1 + PTR_ADDIU a1, 1 + bne a1, a3, L(lst16l) + sb t0, -1(a0) +L(lst16e): + jr ra # Bye, bye + nop + +L(shift): + PTR_SUBU a3, zero, a0 # Src and Dest unaligned + andi a3, 0x7 # (unoptimized case...) + beq a3, zero, L(shft1) + PTR_SUBU a2, a3 # a2 = bytes left + LDHI t0, 0(a1) # Take care of first odd part + LDLO t0, 7(a1) + PTR_ADDU a1, a3 + SDHI t0, 0(a0) + PTR_ADDU a0, a3 +L(shft1): + andi t0, a2, 0x7 + PTR_SUBU a3, a2, t0 + PTR_ADDU a3, a1 +L(shfth): + LDHI t1, 0(a1) # Limp through, dword by dword + LDLO t1, 7(a1) + PTR_ADDIU a0, 8 + PTR_ADDIU a1, 8 + bne a1, a3, L(shfth) + sd t1, -8(a0) + b L(last16) # Handle anything which may be left + move a2, t0 + + .set reorder +END (memcpy) +libc_hidden_builtin_def (memcpy) diff --git a/sysdeps/mips/mips64/memset.S b/sysdeps/mips/mips64/memset.S new file mode 100644 index 0000000000..b50aaba816 --- /dev/null +++ b/sysdeps/mips/mips64/memset.S @@ -0,0 +1,92 @@ +/* Copyright (C) 2002, 2003 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Hartvig Ekner <hartvige@mips.com>, 2002. + Ported to mips3 n32/n64 by Alexandre Oliva <aoliva@redhat.com> + + 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 <endian.h> +#include <sys/asm.h> + + +/* void *memset(void *s, int c, size_t n); + + This could probably be optimized further. */ + +#if __BYTE_ORDER == __BIG_ENDIAN +# define SDHI sdl /* high part is left in big-endian */ +#else +# define SDHI sdr /* high part is right in little-endian */ +#endif + +ENTRY (memset) + .set noreorder + + slti ta1, a2, 16 # Less than 16? + bne ta1, zero, L(last16) + move v0, a0 # Setup exit value before too late + + beq a1, zero, L(ueven) # If zero pattern, no need to extend + andi a1, 0xff # Avoid problems with bogus arguments + dsll ta0, a1, 8 + or a1, ta0 + dsll ta0, a1, 16 + or a1, ta0 # a1 is now pattern in full word + dsll ta0, a1, 32 + or a1, ta0 # a1 is now pattern in double word + +L(ueven): + PTR_SUBU ta0, zero, a0 # Unaligned address? + andi ta0, 0x7 + beq ta0, zero, L(chkw) + PTR_SUBU a2, ta0 + SDHI a1, 0(a0) # Yes, handle first unaligned part + PTR_ADDU a0, ta0 # Now both a0 and a2 are updated + +L(chkw): + andi ta0, a2, 0xf # Enough left for one loop iteration? + beq ta0, a2, L(chkl) + PTR_SUBU a3, a2, ta0 + PTR_ADDU a3, a0 # a3 is last loop address +1 + move a2, ta0 # a2 is now # of bytes left after loop +L(loopw): + PTR_ADDIU a0, 16 # Handle 2 dwords pr. iteration + sd a1, -16(a0) + bne a0, a3, L(loopw) + sd a1, -8(a0) + +L(chkl): + andi ta0, a2, 0x8 # Check if there is at least a double + beq ta0, zero, L(last16) # word remaining after the loop + PTR_SUBU a2, ta0 + sd a1, 0(a0) # Yes... + PTR_ADDIU a0, 8 + +L(last16): + blez a2, L(exit) # Handle last 16 bytes (if cnt>0) + PTR_ADDU a3, a2, a0 # a3 is last address +1 +L(lst16l): + PTR_ADDIU a0, 1 + bne a0, a3, L(lst16l) + sb a1, -1(a0) +L(exit): + j ra # Bye, bye + nop + + .set reorder +END (memset) +libc_hidden_builtin_def (memset) diff --git a/sysdeps/mips/mips64/mul_1.S b/sysdeps/mips/mips64/mul_1.S new file mode 100644 index 0000000000..c711783001 --- /dev/null +++ b/sysdeps/mips/mips64/mul_1.S @@ -0,0 +1,96 @@ +/* MIPS3 __mpn_mul_1 -- Multiply a limb vector with a single limb and + * store the product in a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 2000, 2002, 2003 + * Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * s1_ptr $5 + * size $6 + * s2_limb $7 + */ + +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 4 + .globl __mpn_mul_1 + .ent __mpn_mul_1 +__mpn_mul_1: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + # warm up phase 0 + ld $8,0($5) + + # warm up phase 1 + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,L(LC0) + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,L(LC1) + ld $8,0($5) # load new s1 limb as early as possible + +L(Loop): mflo $10 + mfhi $9 + daddiu $5,$5,8 + daddu $10,$10,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$10,$2 # carry from previous addition -> $2 + sd $10,0($4) + daddiu $4,$4,8 + bne $6,$0,L(Loop) + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 1 +L(LC1): mflo $10 + mfhi $9 + daddu $10,$10,$2 + sltu $2,$10,$2 + dmultu $8,$7 + sd $10,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 0 +L(LC0): mflo $10 + mfhi $9 + daddu $10,$10,$2 + sltu $2,$10,$2 + sd $10,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end __mpn_mul_1 diff --git a/sysdeps/mips/mips64/n32/Implies b/sysdeps/mips/mips64/n32/Implies new file mode 100644 index 0000000000..a7cb280ec4 --- /dev/null +++ b/sysdeps/mips/mips64/n32/Implies @@ -0,0 +1,4 @@ +ieee754/ldbl-128 +mips/mips64 +mips +wordsize-32 diff --git a/sysdeps/mips/mips64/n32/Makefile b/sysdeps/mips/mips64/n32/Makefile new file mode 100644 index 0000000000..a84d2a51bd --- /dev/null +++ b/sysdeps/mips/mips64/n32/Makefile @@ -0,0 +1,6 @@ +# `long double' is a distinct type we support. +long-double-fcts = yes + +ifeq ($(filter -mabi=n32,$(CC)),) +CC += -mabi=n32 +endif diff --git a/sysdeps/mips/mips64/n64/Implies b/sysdeps/mips/mips64/n64/Implies new file mode 100644 index 0000000000..e507786789 --- /dev/null +++ b/sysdeps/mips/mips64/n64/Implies @@ -0,0 +1,4 @@ +ieee754/ldbl-128 +mips/mips64 +mips +wordsize-64 diff --git a/sysdeps/mips/mips64/n64/Makefile b/sysdeps/mips/mips64/n64/Makefile new file mode 100644 index 0000000000..a823f32b53 --- /dev/null +++ b/sysdeps/mips/mips64/n64/Makefile @@ -0,0 +1,6 @@ +# `long double' is a distinct type we support. +long-double-fcts = yes + +ifeq ($(filter -mabi=64,$(CC)),) +CC += -mabi=64 +endif diff --git a/sysdeps/mips/mips64/rshift.S b/sysdeps/mips/mips64/rshift.S new file mode 100644 index 0000000000..e6a8a06d3d --- /dev/null +++ b/sysdeps/mips/mips64/rshift.S @@ -0,0 +1,102 @@ +/* MIPS3 __mpn_rshift -- + * + * Copyright (C) 1995, 2000, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * src_ptr $5 + * size $6 + * cnt $7 + */ + +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 2 + .globl __mpn_rshift + .ent __mpn_rshift +__mpn_rshift: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + ld $10,0($5) # load first limb + dsubu $13,$0,$7 + daddiu $6,$6,-1 + and $9,$6,4-1 # number of limbs in first loop + beq $9,$0,L(L0) # if multiple of 4 limbs, skip first loop + dsll $2,$10,$13 # compute function result + + dsubu $6,$6,$9 + +L(Loop0): ld $3,8($5) + daddiu $4,$4,8 + daddiu $5,$5,8 + daddiu $9,$9,-1 + dsrl $11,$10,$7 + dsll $12,$3,$13 + move $10,$3 + or $8,$11,$12 + bne $9,$0,L(Loop0) + sd $8,-8($4) + +L(L0): beq $6,$0,L(Lend) + nop + +L(Loop): ld $3,8($5) + daddiu $4,$4,32 + daddiu $6,$6,-4 + dsrl $11,$10,$7 + dsll $12,$3,$13 + + ld $10,16($5) + dsrl $14,$3,$7 + or $8,$11,$12 + sd $8,-32($4) + dsll $9,$10,$13 + + ld $3,24($5) + dsrl $11,$10,$7 + or $8,$14,$9 + sd $8,-24($4) + dsll $12,$3,$13 + + ld $10,32($5) + dsrl $14,$3,$7 + or $8,$11,$12 + sd $8,-16($4) + dsll $9,$10,$13 + + daddiu $5,$5,32 + or $8,$14,$9 + bgtz $6,L(Loop) + sd $8,-8($4) + +L(Lend): dsrl $8,$10,$7 + j $31 + sd $8,0($4) + .end __mpn_rshift diff --git a/sysdeps/mips/mips64/setjmp.S b/sysdeps/mips/mips64/setjmp.S new file mode 100644 index 0000000000..bdfd9cd51c --- /dev/null +++ b/sysdeps/mips/mips64/setjmp.S @@ -0,0 +1,46 @@ +/* Copyright (C) 1996, 1997, 2000, 2002, 2003, 2004 + 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 <sgidefs.h> +#include <sysdep.h> +#include <sys/asm.h> + +/* The function __sigsetjmp_aux saves all the registers, but it can't + reliably access the stack or frame pointers, so we pass them in as + extra arguments. */ +#ifdef __PIC__ + .option pic2 +#endif +ENTRY (__sigsetjmp) +#ifdef __PIC__ + SETUP_GP +#endif + SETUP_GP64 (v0, C_SYMBOL_NAME (__sigsetjmp)) + move a2, sp + move a3, fp + PTR_LA t9, __sigsetjmp_aux +#if _MIPS_SIM == _ABIO32 + nop +#endif + RESTORE_GP64 +#if _MIPS_SIM != _ABIO32 + move a4, gp +#endif + jr t9 + .end __sigsetjmp diff --git a/sysdeps/mips/mips64/setjmp_aux.c b/sysdeps/mips/mips64/setjmp_aux.c new file mode 100644 index 0000000000..26b4739c32 --- /dev/null +++ b/sysdeps/mips/mips64/setjmp_aux.c @@ -0,0 +1,78 @@ +/* Copyright (C) 1996, 1997, 2003, 2004 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Contributed by Brendan Kehoe (brendan@zen.org). + + 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 <setjmp.h> +#include <sgidefs.h> + +/* This function is only called via the assembly language routine + __sigsetjmp, which arranges to pass in the stack pointer and the frame + pointer. We do things this way because it's difficult to reliably + access them in C. */ + +int +__sigsetjmp_aux (jmp_buf env, int savemask, long long sp, long long fp, + long long gp) +{ + /* Store the floating point callee-saved registers... */ +#if _MIPS_SIM == _ABI64 + asm volatile ("s.d $f24, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[0])); + asm volatile ("s.d $f25, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[1])); + asm volatile ("s.d $f26, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[2])); + asm volatile ("s.d $f27, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[3])); + asm volatile ("s.d $f28, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[4])); + asm volatile ("s.d $f29, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[5])); + asm volatile ("s.d $f30, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[6])); + asm volatile ("s.d $f31, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[7])); +#else + asm volatile ("s.d $f20, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[0])); + asm volatile ("s.d $f22, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[1])); + asm volatile ("s.d $f24, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[2])); + asm volatile ("s.d $f26, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[3])); + asm volatile ("s.d $f28, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[4])); + asm volatile ("s.d $f30, %0" : : "m" (env[0].__jmpbuf[0].__fpregs[5])); +#endif + + /* .. and the PC; */ + asm volatile ("sd $31, %0" : : "m" (env[0].__jmpbuf[0].__pc)); + + /* .. and the stack pointer; */ + env[0].__jmpbuf[0].__sp = sp; + + /* .. and the FP; it'll be in s8. */ + env[0].__jmpbuf[0].__fp = fp; + + /* .. and the GP; */ + env[0].__jmpbuf[0].__gp = gp; + + /* .. and the callee-saved registers; */ + asm volatile ("sd $16, %0" : : "m" (env[0].__jmpbuf[0].__regs[0])); + asm volatile ("sd $17, %0" : : "m" (env[0].__jmpbuf[0].__regs[1])); + asm volatile ("sd $18, %0" : : "m" (env[0].__jmpbuf[0].__regs[2])); + asm volatile ("sd $19, %0" : : "m" (env[0].__jmpbuf[0].__regs[3])); + asm volatile ("sd $20, %0" : : "m" (env[0].__jmpbuf[0].__regs[4])); + asm volatile ("sd $21, %0" : : "m" (env[0].__jmpbuf[0].__regs[5])); + asm volatile ("sd $22, %0" : : "m" (env[0].__jmpbuf[0].__regs[6])); + asm volatile ("sd $23, %0" : : "m" (env[0].__jmpbuf[0].__regs[7])); + + /* .. and finally get and reconstruct the floating point csr. */ + asm ("cfc1 %0, $31" : "=r" (env[0].__jmpbuf[0].__fpc_csr)); + + /* Save the signal mask if requested. */ + return __sigjmp_save (env, savemask); +} diff --git a/sysdeps/mips/mips64/soft-fp/Dist b/sysdeps/mips/mips64/soft-fp/Dist new file mode 100644 index 0000000000..7e9914fe58 --- /dev/null +++ b/sysdeps/mips/mips64/soft-fp/Dist @@ -0,0 +1 @@ +sfp-machine.h diff --git a/sysdeps/mips/mips64/soft-fp/sfp-machine.h b/sysdeps/mips/mips64/soft-fp/sfp-machine.h new file mode 100644 index 0000000000..309a14a5f8 --- /dev/null +++ b/sysdeps/mips/mips64/soft-fp/sfp-machine.h @@ -0,0 +1,47 @@ +#define _FP_W_TYPE_SIZE 64 +#define _FP_W_TYPE unsigned long long +#define _FP_WS_TYPE signed long long +#define _FP_I_TYPE long long + +#define _FP_MUL_MEAT_S(R,X,Y) \ + _FP_MUL_MEAT_1_imm(_FP_WFRACBITS_S,R,X,Y) +#define _FP_MUL_MEAT_D(R,X,Y) \ + _FP_MUL_MEAT_1_wide(_FP_WFRACBITS_D,R,X,Y,umul_ppmm) +#define _FP_MUL_MEAT_Q(R,X,Y) \ + _FP_MUL_MEAT_2_wide_3mul(_FP_WFRACBITS_Q,R,X,Y,umul_ppmm) + +#define _FP_DIV_MEAT_S(R,X,Y) _FP_DIV_MEAT_1_imm(S,R,X,Y,_FP_DIV_HELP_imm) +#define _FP_DIV_MEAT_D(R,X,Y) _FP_DIV_MEAT_1_udiv_norm(D,R,X,Y) +#define _FP_DIV_MEAT_Q(R,X,Y) _FP_DIV_MEAT_2_udiv(Q,R,X,Y) + +#define _FP_NANFRAC_S ((_FP_QNANBIT_S << 1) - 1) +#define _FP_NANFRAC_D ((_FP_QNANBIT_D << 1) - 1) +#define _FP_NANFRAC_Q ((_FP_QNANBIT_Q << 1) - 1), -1 +#define _FP_NANSIGN_S 0 +#define _FP_NANSIGN_D 0 +#define _FP_NANSIGN_Q 0 + +#define _FP_KEEPNANFRACP 1 +/* From my experiments it seems X is chosen unless one of the + NaNs is sNaN, in which case the result is NANSIGN/NANFRAC. */ +#define _FP_CHOOSENAN(fs, wc, R, X, Y, OP) \ + do { \ + if ((_FP_FRAC_HIGH_RAW_##fs(X) | \ + _FP_FRAC_HIGH_RAW_##fs(Y)) & _FP_QNANBIT_##fs) \ + { \ + R##_s = _FP_NANSIGN_##fs; \ + _FP_FRAC_SET_##wc(R,_FP_NANFRAC_##fs); \ + } \ + else \ + { \ + R##_s = X##_s; \ + _FP_FRAC_COPY_##wc(R,X); \ + } \ + R##_c = FP_CLS_NAN; \ + } while (0) + +#define FP_EX_INVALID (1 << 4) +#define FP_EX_DIVZERO (1 << 3) +#define FP_EX_OVERFLOW (1 << 2) +#define FP_EX_UNDERFLOW (1 << 1) +#define FP_EX_INEXACT (1 << 0) diff --git a/sysdeps/mips/mips64/sub_n.S b/sysdeps/mips/mips64/sub_n.S new file mode 100644 index 0000000000..aa8b0dcf9a --- /dev/null +++ b/sysdeps/mips/mips64/sub_n.S @@ -0,0 +1,130 @@ +/* MIPS3 __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and + * store difference in a third limb vector. + * + * Copyright (C) 1995, 2000, 2002, 2003 Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * s1_ptr $5 + * s2_ptr $6 + * size $7 + */ + +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 2 + .globl __mpn_sub_n + .ent __mpn_sub_n +__mpn_sub_n: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + ld $10,0($5) + ld $11,0($6) + + daddiu $7,$7,-1 + and $9,$7,4-1 # number of limbs in first loop + beq $9,$0,L(L0) # if multiple of 4 limbs, skip first loop + move $2,$0 + + dsubu $7,$7,$9 + +L(Loop0): daddiu $9,$9,-1 + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + or $2,$2,$8 + + daddiu $5,$5,8 + daddiu $6,$6,8 + move $10,$12 + move $11,$13 + bne $9,$0,L(Loop0) + daddiu $4,$4,8 + +L(L0): beq $7,$0,L(Lend) + nop + +L(Loop): daddiu $7,$7,-4 + + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + or $2,$2,$8 + + ld $10,16($5) + daddu $13,$13,$2 + ld $11,16($6) + sltu $8,$13,$2 + dsubu $13,$12,$13 + sltu $2,$12,$13 + sd $13,8($4) + or $2,$2,$8 + + ld $12,24($5) + daddu $11,$11,$2 + ld $13,24($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,16($4) + or $2,$2,$8 + + ld $10,32($5) + daddu $13,$13,$2 + ld $11,32($6) + sltu $8,$13,$2 + dsubu $13,$12,$13 + sltu $2,$12,$13 + sd $13,24($4) + or $2,$2,$8 + + daddiu $5,$5,32 + daddiu $6,$6,32 + + bne $7,$0,L(Loop) + daddiu $4,$4,32 + +L(Lend): daddu $11,$11,$2 + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + j $31 + or $2,$2,$8 + + .end __mpn_sub_n diff --git a/sysdeps/mips/mips64/submul_1.S b/sysdeps/mips/mips64/submul_1.S new file mode 100644 index 0000000000..4971b992a1 --- /dev/null +++ b/sysdeps/mips/mips64/submul_1.S @@ -0,0 +1,108 @@ +/* MIPS3 __mpn_submul_1 -- Multiply a limb vector with a single limb and + * subtract the product from a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 2000, 2002, 2003 + * Free Software Foundation, Inc. + * + * This file is part of the GNU MP Library. + * + * The GNU MP 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 MP 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 MP 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> +#include <sys/asm.h> + +/* INPUT PARAMETERS + * res_ptr $4 + * s1_ptr $5 + * size $6 + * s2_limb $7 + */ + +#ifdef __PIC__ + .option pic2 +#endif + .text + .align 4 + .globl __mpn_submul_1 + .ent __mpn_submul_1 +__mpn_submul_1: +#ifdef __PIC__ + SETUP_GP /* ??? unused */ +#endif + .set noreorder + .set nomacro + + # warm up phase 0 + ld $8,0($5) + + # warm up phase 1 + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,L(LC0) + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,L(LC1) + ld $8,0($5) # load new s1 limb as early as possible + +L(Loop): ld $10,0($4) + mflo $3 + mfhi $9 + daddiu $5,$5,8 + daddu $3,$3,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$3,$2 # carry from previous addition -> $2 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + bne $6,$0,L(Loop) + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 1 +L(LC1): ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dmultu $8,$7 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + + # cool down phase 0 +L(LC0): ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end __mpn_submul_1 |