diff options
Diffstat (limited to 'sysdeps/i386')
| -rw-r--r-- | sysdeps/i386/fpu/e_exp2.S | 34 | ||||
| -rw-r--r-- | sysdeps/i386/fpu/e_exp2f.S | 34 | ||||
| -rw-r--r-- | sysdeps/i386/fpu/e_exp2l.S | 29 |
3 files changed, 94 insertions, 3 deletions
diff --git a/sysdeps/i386/fpu/e_exp2.S b/sysdeps/i386/fpu/e_exp2.S index f802cf8b98..b75a63a0b3 100644 --- a/sysdeps/i386/fpu/e_exp2.S +++ b/sysdeps/i386/fpu/e_exp2.S @@ -6,7 +6,24 @@ #include <machine/asm.h> + .section .rodata.cst8,"aM",@progbits,8 + + .p2align 3 + .type dbl_min,@object +dbl_min: .byte 0, 0, 0, 0, 0, 0, 0x10, 0 + ASM_SIZE_DIRECTIVE(dbl_min) + +#ifdef PIC +# define MO(op) op##@GOTOFF(%ecx) +#else +# define MO(op) op +#endif + + .text ENTRY(__ieee754_exp2) +#ifdef PIC + LOAD_PIC_REG (cx) +#endif fldl 4(%esp) /* I added the following ugly construct because exp(+-Inf) resulted in NaN. The ugliness results from the bright minds at Intel. @@ -27,7 +44,22 @@ ENTRY(__ieee754_exp2) faddp /* 2^(fract(x)) */ fscale /* e^x */ fstp %st(1) - ret + fldl MO(dbl_min) + fld %st(1) + fucompp + fnstsw + sahf + jnc 3f + subl $8, %esp + cfi_adjust_cfa_offset (8) + fld %st(0) + fmul %st(0) + fstpl (%esp) + fstpl (%esp) + fldl (%esp) + addl $8, %esp + cfi_adjust_cfa_offset (-8) +3: ret 1: testl $0x200, %eax /* Test sign. */ jz 2f /* If positive, jump. */ diff --git a/sysdeps/i386/fpu/e_exp2f.S b/sysdeps/i386/fpu/e_exp2f.S index f867d0d472..042c83b1ef 100644 --- a/sysdeps/i386/fpu/e_exp2f.S +++ b/sysdeps/i386/fpu/e_exp2f.S @@ -6,7 +6,24 @@ #include <machine/asm.h> + .section .rodata.cst4,"aM",@progbits,4 + + .p2align 2 + .type flt_min,@object +flt_min: .byte 0, 0, 0x80, 0 + ASM_SIZE_DIRECTIVE(flt_min) + +#ifdef PIC +# define MO(op) op##@GOTOFF(%ecx) +#else +# define MO(op) op +#endif + + .text ENTRY(__ieee754_exp2f) +#ifdef PIC + LOAD_PIC_REG (cx) +#endif flds 4(%esp) /* I added the following ugly construct because exp(+-Inf) resulted in NaN. The ugliness results from the bright minds at Intel. @@ -27,7 +44,22 @@ ENTRY(__ieee754_exp2f) faddp /* 2^(fract(x)) */ fscale /* e^x */ fstp %st(1) - ret + flds MO(flt_min) + fld %st(1) + fucompp + fnstsw + sahf + jnc 3f + subl $4, %esp + cfi_adjust_cfa_offset (4) + fld %st(0) + fmul %st(0) + fstps (%esp) + fstps (%esp) + flds (%esp) + addl $4, %esp + cfi_adjust_cfa_offset (-4) +3: ret 1: testl $0x200, %eax /* Test sign. */ jz 2f /* If positive, jump. */ diff --git a/sysdeps/i386/fpu/e_exp2l.S b/sysdeps/i386/fpu/e_exp2l.S index 2bf9a25727..9a5ff95a6c 100644 --- a/sysdeps/i386/fpu/e_exp2l.S +++ b/sysdeps/i386/fpu/e_exp2l.S @@ -6,7 +6,24 @@ #include <machine/asm.h> + .section .rodata.cst16,"aM",@progbits,16 + .p2align 4 + .type ldbl_min,@object +ldbl_min: .byte 0, 0, 0, 0, 0, 0, 0, 0x80, 0x1, 0 + .byte 0, 0, 0, 0, 0, 0 + ASM_SIZE_DIRECTIVE(ldbl_min) + +#ifdef PIC +# define MO(op) op##@GOTOFF(%ecx) +#else +# define MO(op) op +#endif + + .text ENTRY(__ieee754_exp2l) +#ifdef PIC + LOAD_PIC_REG (cx) +#endif fldt 4(%esp) /* I added the following ugly construct because exp(+-Inf) resulted in NaN. The ugliness results from the bright minds at Intel. @@ -35,7 +52,17 @@ ENTRY(__ieee754_exp2l) faddp /* 2^(fract(x)) */ fscale /* e^x */ fstp %st(1) - ret + /* Ensure underflow for tiny result. */ + fldt MO(ldbl_min) + fld %st(1) + fucompp + fnstsw + sahf + jnc 4f + fld %st + fmul %st + fstp %st +4: ret 1: testl $0x200, %eax /* Test sign. */ jz 2f /* If positive, jump. */ |