diff options
Diffstat (limited to 'sysdeps/i386')
-rw-r--r-- | sysdeps/i386/dl-machine.h | 17 | ||||
-rw-r--r-- | sysdeps/i386/fpu/bits/mathinline.h | 157 | ||||
-rw-r--r-- | sysdeps/i386/fpu/fraiseexcpt.c | 23 |
3 files changed, 103 insertions, 94 deletions
diff --git a/sysdeps/i386/dl-machine.h b/sysdeps/i386/dl-machine.h index e517cff0ff..213060ba84 100644 --- a/sysdeps/i386/dl-machine.h +++ b/sysdeps/i386/dl-machine.h @@ -57,20 +57,12 @@ static inline Elf32_Addr __attribute__ ((unused)) elf_machine_load_address (void) { Elf32_Addr addr; - asm (" call .Lhere\n" - ".Lhere: popl %0\n" - " subl $.Lhere, %0" + asm (" call 1f\n" + "1: popl %0\n" + " subl 1b@GOT(%%ebx), %0" : "=r" (addr)); return addr; } -/* The `subl' insn above will contain an R_386_32 relocation entry - intended to insert the run-time address of the label `.Lhere'. - This will be the first relocation in the text of the dynamic linker; - we skip it to avoid trying to modify read-only text in this early stage. */ -#define ELF_MACHINE_BEFORE_RTLD_RELOC(dynamic_info) \ - ++(const Elf32_Rel *) (dynamic_info)[DT_REL]->d_un.d_ptr; \ - (dynamic_info)[DT_RELSZ]->d_un.d_val -= sizeof (Elf32_Rel); - #ifndef PROF /* We add a declaration of this function here so that in dl-runtime.c @@ -132,6 +124,7 @@ elf_machine_runtime_setup (struct link_map *l, int lazy, int profile) # define ELF_MACHINE_RUNTIME_TRAMPOLINE asm ("\ .globl _dl_runtime_resolve .type _dl_runtime_resolve, @function + .align 16 _dl_runtime_resolve: pushl %eax # Preserve registers otherwise clobbered. pushl %ecx @@ -147,6 +140,7 @@ _dl_runtime_resolve: .globl _dl_runtime_profile .type _dl_runtime_profile, @function + .align 16 _dl_runtime_profile: pushl %eax # Preserve registers otherwise clobbered. pushl %ecx @@ -167,6 +161,7 @@ _dl_runtime_profile: .globl _dl_runtime_profile .type _dl_runtime_resolve, @function .type _dl_runtime_profile, @function + .align 16 _dl_runtime_resolve: _dl_runtime_profile: pushl %eax # Preserve registers otherwise clobbered. diff --git a/sysdeps/i386/fpu/bits/mathinline.h b/sysdeps/i386/fpu/bits/mathinline.h index c6b9331f9a..17f62a080f 100644 --- a/sysdeps/i386/fpu/bits/mathinline.h +++ b/sysdeps/i386/fpu/bits/mathinline.h @@ -73,6 +73,10 @@ #ifdef __GNUC__ #if !defined __NO_MATH_INLINES && defined __OPTIMIZE__ +/* The gcc, version 2.7 or below, has problems with all this inlining + code. So disable it for this version of the compiler. */ +#if __GNUC__ > 2 || (__GNUC__ == 2 && __GNUC_MINOR__ > 7) + #ifdef __cplusplus # define __MATH_INLINE __inline #else @@ -154,6 +158,80 @@ } +/* Miscellaneous functions */ + +__inline_mathcode (__sgn, __x, \ + return __x == 0.0 ? 0.0 : (__x > 0.0 ? 1.0 : -1.0)) + +__inline_mathcode (__pow2, __x, \ + register long double __value; \ + register long double __exponent; \ + long int __p = (long int) __x; \ + if (__x == (long double) __p) \ + { \ + __asm __volatile__ \ + ("fscale" \ + : "=t" (__value) : "0" (1.0), "u" (__x)); \ + return __value; \ + } \ + __asm __volatile__ \ + ("fldl %%st(0)\n\t" \ + "frndint # int(x)\n\t" \ + "fxch\n\t" \ + "fsub %%st(1) # fract(x)\n\t" \ + "f2xm1 # 2^(fract(x)) - 1\n\t" \ + : "=t" (__value), "=u" (__exponent) : "0" (__x)); \ + __value += 1.0; \ + __asm __volatile__ \ + ("fscale" \ + : "=t" (__value) : "0" (__value), "u" (__exponent)); \ + return __value) + +#define __sincos_code \ + register long double __cosr; \ + register long double __sinr; \ + __asm __volatile__ \ + ("fsincos\n\t" \ + "fnstsw %%ax\n\t" \ + "testl $0x400, %%eax\n\t" \ + "jz 1f\n\t" \ + "fldpi\n\t" \ + "fadd %%st(0)\n\t" \ + "fxch %%st(1)\n\t" \ + "2: fprem1\n\t" \ + "fnstsw %%ax\n\t" \ + "testl $0x400, %%eax\n\t" \ + "jnz 2b\n\t" \ + "fstp %%st(1)\n\t" \ + "fsincos\n\t" \ + "1:" \ + : "=t" (__cosr), "=u" (__sinr) : "0" (__x)); \ + *__sinx = __sinr; \ + *__cosx = __cosr + +__MATH_INLINE void __sincos (double __x, double *__sinx, double *__cosx); +__MATH_INLINE void +__sincos (double __x, double *__sinx, double *__cosx) +{ + __sincos_code; +} + +__MATH_INLINE void __sincosf (float __x, float *__sinx, float *__cosx); +__MATH_INLINE void +__sincosf (float __x, float *__sinx, float *__cosx) +{ + __sincos_code; +} + +__MATH_INLINE void __sincosl (long double __x, long double *__sinx, + long double *__cosx); +__MATH_INLINE void +__sincosl (long double __x, long double *__sinx, long double *__cosx) +{ + __sincos_code; +} + + /* Optimized inline implementation, sometimes with reduced precision and/or argument range. */ @@ -278,10 +356,10 @@ __inline_mathop (sqrt, "fsqrt") __inline_mathop_ (long double, __sqrtl, "fsqrt") #if defined __GNUC__ && (__GNUC__ > 2 || __GNUC__ == 2 && __GNUC_MINOR__ >= 8) -__inline_mathcode_ (fabs, __x, return __builtin_fabs (__x)) -__inline_mathcode_ (fabsf, __x, return __builtin_fabsf (__x)) -__inline_mathcode_ (fabsl, __x, return __builtin_fabsl (__x)) -__inline_mathcode_ (__fabsl, __x, return __builtin_fabsl (__x)) +__inline_mathcode_ (double, fabs, __x, return __builtin_fabs (__x)) +__inline_mathcode_ (float, fabsf, __x, return __builtin_fabsf (__x)) +__inline_mathcode_ (long double, fabsl, __x, return __builtin_fabsl (__x)) +__inline_mathcode_ (long double, __fabsl, __x, return __builtin_fabsl (__x)) #else __inline_mathop (fabs, "fabs") __inline_mathop_ (long double, __fabsl, "fabs") @@ -356,7 +434,7 @@ ldexp (double __x, int __y) /* Optimized versions for some non-standardized functions. */ #if defined __USE_ISOC9X || defined __USE_MISC -__inline_mathop_decl (log2, "fyl2x", "u" (1.0), "0" (__x) : "st(1)") +__inline_mathop(log2, "fld1; fxch; fyl2x") __inline_mathcode (expm1, __x, __expm1_code) @@ -443,15 +521,12 @@ __finite (double __x) ("orl $0x800fffff, %0\n\t" "incl %0\n\t" "shrl $31, %0" - : "=q" (__result) : "0" (((int *) &__x)[1])); + : "=q" (__result) : "0" (((int *) &__x)[1]) : "cc"); return __result; } /* Miscellaneous functions */ -__inline_mathcode (__sgn, __x, \ - return __x == 0.0 ? 0.0 : (__x > 0.0 ? 1.0 : -1.0)) - __inline_mathcode (__coshm1, __x, \ register long double __exm1 = __expm1l (__fabsl (__x)); \ return 0.5 * (__exm1 / (__exm1 + 1.0)) * __exm1) @@ -459,69 +534,6 @@ __inline_mathcode (__coshm1, __x, \ __inline_mathcode (__acosh1p, __x, \ return log1pl (__x + __sqrtl (__x) * __sqrtl (__x + 2.0))) -__inline_mathcode (__pow2, __x, \ - register long double __value; \ - register long double __exponent; \ - long int __p = (long int) __x; \ - if (__x == (long double) __p) \ - return ldexpl (1.0, __p); \ - __asm __volatile__ \ - ("fldl %%st(0)\n\t" \ - "frndint # int(x)\n\t" \ - "fxch\n\t" \ - "fsub %%st(1) # fract(x)\n\t" \ - "f2xm1 # 2^(fract(x)) - 1\n\t" \ - : "=t" (__value), "=u" (__exponent) : "0" (__x)); \ - __value += 1.0; \ - __asm __volatile__ \ - ("fscale" \ - : "=t" (__value) : "0" (__value), "u" (__exponent)); \ - return __value) - -#define __sincos_code \ - register long double __cosr; \ - register long double __sinr; \ - __asm __volatile__ \ - ("fsincos\n\t" \ - "fnstsw %%ax\n\t" \ - "testl $0x400, %%eax\n\t" \ - "jz 1f\n\t" \ - "fldpi\n\t" \ - "fadd %%st(0)\n\t" \ - "fxch %%st(1)\n\t" \ - "2: fprem1\n\t" \ - "fnstsw %%ax\n\t" \ - "testl $0x400, %%eax\n\t" \ - "jnz 2b\n\t" \ - "fstp %%st(1)\n\t" \ - "fsincos\n\t" \ - "1:" \ - : "=t" (__cosr), "=u" (__sinr) : "0" (__x)); \ - *__sinx = __sinr; \ - *__cosx = __cosr - -__MATH_INLINE void __sincos (double __x, double *__sinx, double *__cosx); -__MATH_INLINE void -__sincos (double __x, double *__sinx, double *__cosx) -{ - __sincos_code; -} - -__MATH_INLINE void __sincosf (float __x, float *__sinx, float *__cosx); -__MATH_INLINE void -__sincosf (float __x, float *__sinx, float *__cosx) -{ - __sincos_code; -} - -__MATH_INLINE void __sincosl (long double __x, long double *__sinx, - long double *__cosx); -__MATH_INLINE void -__sincosl (long double __x, long double *__sinx, long double *__cosx) -{ - __sincos_code; -} - #endif /* __USE_MISC */ /* Undefine some of the large macros which are not used anymore. */ @@ -530,6 +542,7 @@ __sincosl (long double __x, long double *__sinx, long double *__cosx) #undef __atan2_code #undef __sincos_code +#endif /* Not gcc <= 2.7. */ #endif /* __NO_MATH_INLINES */ #endif /* __GNUC__ */ diff --git a/sysdeps/i386/fpu/fraiseexcpt.c b/sysdeps/i386/fpu/fraiseexcpt.c index 0af8c71ddc..d468449873 100644 --- a/sysdeps/i386/fpu/fraiseexcpt.c +++ b/sysdeps/i386/fpu/fraiseexcpt.c @@ -34,7 +34,7 @@ feraiseexcept (int excepts) { /* One example of a invalid operation is 0.0 / 0.0. */ double d; - __asm__ ("fldz; fdiv %%st, %%st(0); fwait" : "=t" (d)); + __asm__ __volatile__ ("fldz; fdiv %%st, %%st(0); fwait" : "=t" (d)); (void) &d; } @@ -42,7 +42,8 @@ feraiseexcept (int excepts) if ((FE_DIVBYZERO & excepts) != 0) { double d; - __asm__ ("fldz; fld1; fdivp %%st, %%st(1); fwait" : "=t" (d)); + __asm__ __volatile__ ("fldz; fld1; fdivp %%st, %%st(1); fwait" + : "=t" (d)); (void) &d; } @@ -55,16 +56,16 @@ feraiseexcept (int excepts) /* Bah, we have to clear selected exceptions. Since there is no `fldsw' instruction we have to do it the hard way. */ - __asm__ ("fnstenv %0" : "=m" (*&temp)); + __asm__ __volatile__ ("fnstenv %0" : "=m" (*&temp)); /* Set the relevant bits. */ temp.status_word |= FE_OVERFLOW; /* Put the new data in effect. */ - __asm__ ("fldenv %0" : : "m" (*&temp)); + __asm__ __volatile__ ("fldenv %0" : : "m" (*&temp)); /* And raise the exception. */ - __asm__ ("fwait"); + __asm__ __volatile__ ("fwait"); } /* Next: underflow. */ @@ -76,16 +77,16 @@ feraiseexcept (int excepts) /* Bah, we have to clear selected exceptions. Since there is no `fldsw' instruction we have to do it the hard way. */ - __asm__ ("fnstenv %0" : "=m" (*&temp)); + __asm__ __volatile__ ("fnstenv %0" : "=m" (*&temp)); /* Set the relevant bits. */ temp.status_word |= FE_UNDERFLOW; /* Put the new data in effect. */ - __asm__ ("fldenv %0" : : "m" (*&temp)); + __asm__ __volatile__ ("fldenv %0" : : "m" (*&temp)); /* And raise the exception. */ - __asm__ ("fwait"); + __asm__ __volatile__ ("fwait"); } /* Last: inexact. */ @@ -97,15 +98,15 @@ feraiseexcept (int excepts) /* Bah, we have to clear selected exceptions. Since there is no `fldsw' instruction we have to do it the hard way. */ - __asm__ ("fnstenv %0" : "=m" (*&temp)); + __asm__ __volatile__ ("fnstenv %0" : "=m" (*&temp)); /* Set the relevant bits. */ temp.status_word |= FE_INEXACT; /* Put the new data in effect. */ - __asm__ ("fldenv %0" : : "m" (*&temp)); + __asm__ __volatile__ ("fldenv %0" : : "m" (*&temp)); /* And raise the exception. */ - __asm__ ("fwait"); + __asm__ __volatile__ ("fwait"); } } |