From 37825319ee64d51bf8b1774f4964e9743eaf5ef6 Mon Sep 17 00:00:00 2001 From: Joseph Myers Date: Thu, 24 Sep 2015 21:41:00 +0000 Subject: Refactor i386 libm code forcing underflow exceptions. This patch refactors code in sysdeps/i386/fpu that forces underflow exceptions to use common macros for that purpose as far as possible. (Although some of the macros end up used in only one place, I think it's cleanest to define all these macros together so that all the code forcing underflow uses such macros. Some more uses of such macros will also be introduced when fixing remaining bugs about missing underflow exceptions, and it would be possible to do further refactoring of the macros in i386-math-asm.h to share more code by using other macros internally. Places that test for underflow by examining the representation of the argument with integer operations, rather that using floating-point comparisons on the argument or result, are unchanged by this patch.) Most of this code uses a macro MO to abstract away the differences between PIC and non-PIC memory references to constants. log1p functions, however, hardcoded PIC conditionals for this. Because the common macros rely on the use of MO, I changed the log1p functions to use the normal style here, and, for consistency, also made that change to log1pl which is otherwise unaffected by this patch. Tested for x86. * sysdeps/i386/fpu/i386-math-asm.h (DEFINE_LDBL_MIN): New macro. (FLT_CHECK_FORCE_UFLOW): Likewise. (DBL_CHECK_FORCE_UFLOW): Likewise. (FLT_CHECK_FORCE_UFLOW_NARROW): Likewise. (DBL_CHECK_FORCE_UFLOW_NARROW): Likewise. (LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN): Likewise. (FLT_CHECK_FORCE_UFLOW_NONNAN): Likewise. (DBL_CHECK_FORCE_UFLOW_NONNAN): Likewise. (FLT_CHECK_FORCE_UFLOW_NONNEG): Likewise. (DBL_CHECK_FORCE_UFLOW_NONNEG): Likewise. (LDBL_CHECK_FORCE_UFLOW_NONNEG): Likewise. * sysdeps/i386/fpu/e_asin.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (__ieee754_asin): Use DBL_CHECK_FORCE_UFLOW. * sysdeps/i386/fpu/e_asinf.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (__ieee754_asinf): Use FLT_CHECK_FORCE_UFLOW. * sysdeps/i386/fpu/e_atan2.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (__ieee754_atan2): Use DBL_CHECK_FORCE_UFLOW_NARROW. * sysdeps/i386/fpu/e_atan2f.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (__ieee754_atan2f): Use FLT_CHECK_FORCE_UFLOW_NARROW. * sysdeps/i386/fpu/e_atanh.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (__ieee754_atanh): Use DBL_CHECK_FORCE_UFLOW_NONNEG. * sysdeps/i386/fpu/e_atanhf.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (__ieee754_atanhf): Use FLT_CHECK_FORCE_UFLOW_NONNEG. * sysdeps/i386/fpu/e_exp2l.S: Include . (ldbl_min): Replace with use of DEFINE_LDBL_MIN. (__ieee754_exp2l): Use LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN. * sysdeps/i386/fpu/e_expl.S: Include . [!USE_AS_EXPM1L] (cmin): Replace with use of DEFINE_LDBL_MIN. (IEEE754_EXPL): Use LDBL_CHECK_FORCE_UFLOW_NONNEG. * sysdeps/i386/fpu/s_atan.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (__atan): Use DBL_CHECK_FORCE_UFLOW. * sysdeps/i386/fpu/s_atanf.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (__atanf): Use FLT_CHECK_FORCE_UFLOW. * sysdeps/i386/fpu/s_expm1.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (__expm1): Use DBL_CHECK_FORCE_UFLOW. Move underflow check after main computation. * sysdeps/i386/fpu/s_expm1f.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (__expm1f): Use FLT_CHECK_FORCE_UFLOW. Move underflow check after main computation. * sysdeps/i386/fpu/s_log1p.S: Include . (dbl_min): Replace with use of DEFINE_DBL_MIN. (MO): New macro. (__log1p): Use MO. Use DBL_CHECK_FORCE_UFLOW_NONNAN. * sysdeps/i386/fpu/s_log1pf.S: Include . (flt_min): Replace with use of DEFINE_FLT_MIN. (MO): New macro. (__log1pf): Use MO. Use FLT_CHECK_FORCE_UFLOW_NONNAN. * sysdeps/i386/fpu/s_log1pl.S (MO): New macro. (__log1pl): Use MO. --- ChangeLog | 60 ++++++++++++++ sysdeps/i386/fpu/e_asin.S | 25 +----- sysdeps/i386/fpu/e_asinf.S | 23 +----- sysdeps/i386/fpu/e_atan2.S | 27 +------ sysdeps/i386/fpu/e_atan2f.S | 27 +------ sysdeps/i386/fpu/e_atanh.S | 22 +---- sysdeps/i386/fpu/e_atanhf.S | 22 +---- sysdeps/i386/fpu/e_exp2l.S | 21 +---- sysdeps/i386/fpu/e_expl.S | 19 +---- sysdeps/i386/fpu/i386-math-asm.h | 168 +++++++++++++++++++++++++++++++++++++++ sysdeps/i386/fpu/s_atan.S | 25 +----- sysdeps/i386/fpu/s_atanf.S | 25 +----- sysdeps/i386/fpu/s_expm1.S | 24 +----- sysdeps/i386/fpu/s_expm1f.S | 24 +----- sysdeps/i386/fpu/s_log1p.S | 45 +++-------- sysdeps/i386/fpu/s_log1pf.S | 45 +++-------- sysdeps/i386/fpu/s_log1pl.S | 18 ++--- 17 files changed, 304 insertions(+), 316 deletions(-) diff --git a/ChangeLog b/ChangeLog index df079ca..dbe8c77 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,5 +1,65 @@ 2015-09-24 Joseph Myers + * sysdeps/i386/fpu/i386-math-asm.h (DEFINE_LDBL_MIN): New macro. + (FLT_CHECK_FORCE_UFLOW): Likewise. + (DBL_CHECK_FORCE_UFLOW): Likewise. + (FLT_CHECK_FORCE_UFLOW_NARROW): Likewise. + (DBL_CHECK_FORCE_UFLOW_NARROW): Likewise. + (LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN): Likewise. + (FLT_CHECK_FORCE_UFLOW_NONNAN): Likewise. + (DBL_CHECK_FORCE_UFLOW_NONNAN): Likewise. + (FLT_CHECK_FORCE_UFLOW_NONNEG): Likewise. + (DBL_CHECK_FORCE_UFLOW_NONNEG): Likewise. + (LDBL_CHECK_FORCE_UFLOW_NONNEG): Likewise. + * sysdeps/i386/fpu/e_asin.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (__ieee754_asin): Use DBL_CHECK_FORCE_UFLOW. + * sysdeps/i386/fpu/e_asinf.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (__ieee754_asinf): Use FLT_CHECK_FORCE_UFLOW. + * sysdeps/i386/fpu/e_atan2.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (__ieee754_atan2): Use DBL_CHECK_FORCE_UFLOW_NARROW. + * sysdeps/i386/fpu/e_atan2f.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (__ieee754_atan2f): Use FLT_CHECK_FORCE_UFLOW_NARROW. + * sysdeps/i386/fpu/e_atanh.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (__ieee754_atanh): Use DBL_CHECK_FORCE_UFLOW_NONNEG. + * sysdeps/i386/fpu/e_atanhf.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (__ieee754_atanhf): Use FLT_CHECK_FORCE_UFLOW_NONNEG. + * sysdeps/i386/fpu/e_exp2l.S: Include . + (ldbl_min): Replace with use of DEFINE_LDBL_MIN. + (__ieee754_exp2l): Use LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN. + * sysdeps/i386/fpu/e_expl.S: Include . + [!USE_AS_EXPM1L] (cmin): Replace with use of DEFINE_LDBL_MIN. + (IEEE754_EXPL): Use LDBL_CHECK_FORCE_UFLOW_NONNEG. + * sysdeps/i386/fpu/s_atan.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (__atan): Use DBL_CHECK_FORCE_UFLOW. + * sysdeps/i386/fpu/s_atanf.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (__atanf): Use FLT_CHECK_FORCE_UFLOW. + * sysdeps/i386/fpu/s_expm1.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (__expm1): Use DBL_CHECK_FORCE_UFLOW. Move underflow check after + main computation. + * sysdeps/i386/fpu/s_expm1f.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (__expm1f): Use FLT_CHECK_FORCE_UFLOW. Move underflow check after + main computation. + * sysdeps/i386/fpu/s_log1p.S: Include . + (dbl_min): Replace with use of DEFINE_DBL_MIN. + (MO): New macro. + (__log1p): Use MO. Use DBL_CHECK_FORCE_UFLOW_NONNAN. + * sysdeps/i386/fpu/s_log1pf.S: Include . + (flt_min): Replace with use of DEFINE_FLT_MIN. + (MO): New macro. + (__log1pf): Use MO. Use FLT_CHECK_FORCE_UFLOW_NONNAN. + * sysdeps/i386/fpu/s_log1pl.S (MO): New macro. + (__log1pl): Use MO. + [BZ #19003] * sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-e_pow-fma4.c): Add $(config-cflags-nofma). diff --git a/sysdeps/i386/fpu/e_asin.S b/sysdeps/i386/fpu/e_asin.S index ea8cb6f..39c8b47 100644 --- a/sysdeps/i386/fpu/e_asin.S +++ b/sysdeps/i386/fpu/e_asin.S @@ -4,15 +4,11 @@ */ #include +#include RCSID("$NetBSD: e_asin.S,v 1.4 1995/05/08 23:45:40 jtc Exp $") - .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) +DEFINE_DBL_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -36,20 +32,7 @@ ENTRY(__ieee754_asin) fmulp /* 1 - x^2 */ fsqrt /* sqrt (1 - x^2) */ fpatan - fldl MO(dbl_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) -1: ret + DBL_CHECK_FORCE_UFLOW + ret END (__ieee754_asin) strong_alias (__ieee754_asin, __asin_finite) diff --git a/sysdeps/i386/fpu/e_asinf.S b/sysdeps/i386/fpu/e_asinf.S index f7bda77..1102bde 100644 --- a/sysdeps/i386/fpu/e_asinf.S +++ b/sysdeps/i386/fpu/e_asinf.S @@ -5,15 +5,13 @@ */ #include +#include RCSID("$NetBSD: $") .section .rodata.cst4,"aM",@progbits,4 - .p2align 2 - .type flt_min,@object -flt_min: .byte 0, 0, 0x80, 0 - ASM_SIZE_DIRECTIVE(flt_min) +DEFINE_FLT_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -35,20 +33,7 @@ ENTRY(__ieee754_asinf) fsubp /* 1 - x^2 */ fsqrt /* sqrt (1 - x^2) */ fpatan - flds MO(flt_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $4, %esp - cfi_adjust_cfa_offset (4) - fld %st(0) - fmul %st(0) - fstps (%esp) - addl $4, %esp - cfi_adjust_cfa_offset (-4) -1: ret + FLT_CHECK_FORCE_UFLOW + ret END (__ieee754_asinf) strong_alias (__ieee754_asinf, __asinf_finite) diff --git a/sysdeps/i386/fpu/e_atan2.S b/sysdeps/i386/fpu/e_atan2.S index 6f5481f..25f43bb 100644 --- a/sysdeps/i386/fpu/e_atan2.S +++ b/sysdeps/i386/fpu/e_atan2.S @@ -4,15 +4,11 @@ */ #include +#include RCSID("$NetBSD: e_atan2.S,v 1.4 1995/05/08 23:46:28 jtc Exp $") - .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) +DEFINE_DBL_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -28,22 +24,7 @@ ENTRY(__ieee754_atan2) fldl 4(%esp) fldl 12(%esp) fpatan - fldl MO(dbl_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - fstpl (%esp) - fldl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) -1: ret + DBL_CHECK_FORCE_UFLOW_NARROW + ret END (__ieee754_atan2) strong_alias (__ieee754_atan2, __atan2_finite) diff --git a/sysdeps/i386/fpu/e_atan2f.S b/sysdeps/i386/fpu/e_atan2f.S index ec0eb3f..2bc909a 100644 --- a/sysdeps/i386/fpu/e_atan2f.S +++ b/sysdeps/i386/fpu/e_atan2f.S @@ -4,15 +4,11 @@ */ #include +#include RCSID("$NetBSD: e_atan2f.S,v 1.1 1995/05/08 23:35:10 jtc Exp $") - .section .rodata.cst4,"aM",@progbits,4 - - .p2align 2 - .type flt_min,@object -flt_min: .byte 0, 0, 0x80, 0 - ASM_SIZE_DIRECTIVE(flt_min) +DEFINE_FLT_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -28,22 +24,7 @@ ENTRY(__ieee754_atan2f) flds 4(%esp) flds 8(%esp) fpatan - flds MO(flt_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - 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) -1: ret + FLT_CHECK_FORCE_UFLOW_NARROW + ret END (__ieee754_atan2f) strong_alias (__ieee754_atan2f, __atan2f_finite) diff --git a/sysdeps/i386/fpu/e_atanh.S b/sysdeps/i386/fpu/e_atanh.S index 90d19bc..5772bbf 100644 --- a/sysdeps/i386/fpu/e_atanh.S +++ b/sysdeps/i386/fpu/e_atanh.S @@ -18,6 +18,7 @@ . */ #include +#include .section .rodata @@ -35,12 +36,7 @@ limit: .double 0.29 ln2_2: .tfloat 0.3465735902799726547086160 ASM_SIZE_DIRECTIVE(ln2_2) - .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) +DEFINE_DBL_MIN #ifdef PIC #define MO(op) op##@GOTOFF(%edx) @@ -88,18 +84,8 @@ ENTRY(__ieee754_atanh) sahf jae 4f fyl2xp1 // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|)) - fcoml MO(dbl_min) - fnstsw - sahf - jae 8f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) -8: jecxz 3f + DBL_CHECK_FORCE_UFLOW_NONNEG + jecxz 3f fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x)) 3: ret diff --git a/sysdeps/i386/fpu/e_atanhf.S b/sysdeps/i386/fpu/e_atanhf.S index 1c8969e..6e41849 100644 --- a/sysdeps/i386/fpu/e_atanhf.S +++ b/sysdeps/i386/fpu/e_atanhf.S @@ -18,6 +18,7 @@ . */ #include +#include .section .rodata @@ -36,12 +37,7 @@ limit: .double 0.29 ln2_2: .tfloat 0.3465735902799726547086160 ASM_SIZE_DIRECTIVE(ln2_2) - .section .rodata.cst4,"aM",@progbits,4 - - .p2align 2 - .type flt_min,@object -flt_min: .byte 0, 0, 0x80, 0 - ASM_SIZE_DIRECTIVE(flt_min) +DEFINE_FLT_MIN #ifdef PIC #define MO(op) op##@GOTOFF(%edx) @@ -84,18 +80,8 @@ ENTRY(__ieee754_atanhf) sahf jae 4f fyl2xp1 // 0.5*ln2*ld(1+2*|x|+(2*|x|^2)/(1-|x|)) - fcoms MO(flt_min) - fnstsw - sahf - jae 6f - subl $4, %esp - cfi_adjust_cfa_offset (4) - fld %st(0) - fmul %st(0) - fstps (%esp) - addl $4, %esp - cfi_adjust_cfa_offset (-4) -6: jecxz 3f + FLT_CHECK_FORCE_UFLOW_NONNEG + jecxz 3f fchs // 0.5*ln2*ld(1+2*x+(2*x^2)/(1-x)) 3: ret diff --git a/sysdeps/i386/fpu/e_exp2l.S b/sysdeps/i386/fpu/e_exp2l.S index 9a5ff95..c4cb73d 100644 --- a/sysdeps/i386/fpu/e_exp2l.S +++ b/sysdeps/i386/fpu/e_exp2l.S @@ -5,13 +5,9 @@ */ #include +#include - .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) +DEFINE_LDBL_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -52,17 +48,8 @@ ENTRY(__ieee754_exp2l) faddp /* 2^(fract(x)) */ fscale /* e^x */ fstp %st(1) - /* 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 + LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN + ret 1: testl $0x200, %eax /* Test sign. */ jz 2f /* If positive, jump. */ diff --git a/sysdeps/i386/fpu/e_expl.S b/sysdeps/i386/fpu/e_expl.S index 711c86a..64e0d96 100644 --- a/sysdeps/i386/fpu/e_expl.S +++ b/sysdeps/i386/fpu/e_expl.S @@ -23,6 +23,7 @@ */ #include +#include #ifdef USE_AS_EXP10L # define IEEE754_EXPL __ieee754_exp10l @@ -65,10 +66,7 @@ c1: .byte 0x20, 0xfa, 0xee, 0xc2, 0x5f, 0x70, 0xa5, 0xec, 0xed, 0x3f csat: .byte 0, 0, 0, 0, 0, 0, 0, 0x80, 0x0e, 0x40 .byte 0, 0, 0, 0, 0, 0 ASM_SIZE_DIRECTIVE(csat) - .type cmin,@object -cmin: .byte 0, 0, 0, 0, 0, 0, 0, 0x80, 0x1, 0 - .byte 0, 0, 0, 0, 0, 0 - ASM_SIZE_DIRECTIVE(cmin) +DEFINE_LDBL_MIN #endif #ifdef PIC @@ -199,18 +197,9 @@ ENTRY(IEEE754_EXPL) fstp %st(1) /* 2 */ fscale /* 2 scale factor is st(1); base^x */ fstp %st(1) /* 1 */ - /* Ensure underflow for tiny result. */ - fldt MO(cmin) /* 2 cmin */ - fld %st(1) /* 3 */ - fcompp /* 1 */ - fnstsw - sahf - jnc 6f - fld %st - fmul %st - fstp %st + LDBL_CHECK_FORCE_UFLOW_NONNEG #endif -6: fstp %st(1) /* 0 */ + fstp %st(1) /* 0 */ jmp 2f 1: #ifdef USE_AS_EXPM1L diff --git a/sysdeps/i386/fpu/i386-math-asm.h b/sysdeps/i386/fpu/i386-math-asm.h index c15029d..5d8d847 100644 --- a/sysdeps/i386/fpu/i386-math-asm.h +++ b/sysdeps/i386/fpu/i386-math-asm.h @@ -52,6 +52,14 @@ flt_min: \ dbl_min: \ .byte 0, 0, 0, 0, 0, 0, 0x10, 0; \ .size dbl_min, .-dbl_min; +#define DEFINE_LDBL_MIN \ + .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; \ + .size ldbl_min, .-ldbl_min; /* Remove excess range and precision by storing a value on the stack and loading it back. The value is given to be nonnegative or NaN; @@ -123,4 +131,164 @@ dbl_min: \ addl $8, %esp; \ cfi_adjust_cfa_offset (-8); +/* Force an underflow exception if the given value is subnormal. The + relevant constant for the minimum of the type must have been + defined, the MO macro must have been defined for access to memory + operands, and, if PIC, the PIC register must have been loaded. */ +#define FLT_CHECK_FORCE_UFLOW \ + flds MO(flt_min); \ + fld %st(1); \ + fabs; \ + fucompp; \ + fnstsw; \ + sahf; \ + jnc 6424f; \ + subl $4, %esp; \ + cfi_adjust_cfa_offset (4); \ + fld %st(0); \ + fmul %st(0); \ + fstps (%esp); \ + addl $4, %esp; \ + cfi_adjust_cfa_offset (-4); \ +6424: +#define DBL_CHECK_FORCE_UFLOW \ + fldl MO(dbl_min); \ + fld %st(1); \ + fabs; \ + fucompp; \ + fnstsw; \ + sahf; \ + jnc 6453f; \ + subl $8, %esp; \ + cfi_adjust_cfa_offset (8); \ + fld %st(0); \ + fmul %st(0); \ + fstpl (%esp); \ + addl $8, %esp; \ + cfi_adjust_cfa_offset (-8); \ +6453: + +/* Likewise, but also remove excess range and precision if the value + is subnormal. */ +#define FLT_CHECK_FORCE_UFLOW_NARROW \ + flds MO(flt_min); \ + fld %st(1); \ + fabs; \ + fucompp; \ + fnstsw; \ + sahf; \ + jnc 6424f; \ + 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); \ +6424: +#define DBL_CHECK_FORCE_UFLOW_NARROW \ + fldl MO(dbl_min); \ + fld %st(1); \ + fabs; \ + fucompp; \ + fnstsw; \ + sahf; \ + jnc 6453f; \ + 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); \ +6453: + +/* Likewise, but the argument is nonnegative or NaN. */ +#define LDBL_CHECK_FORCE_UFLOW_NONNEG_NAN \ + fldt MO(ldbl_min); \ + fld %st(1); \ + fucompp; \ + fnstsw; \ + sahf; \ + jnc 6464f; \ + fld %st(0); \ + fmul %st(0); \ + fstp %st(0); \ +6464: + +/* Likewise, but the argument is not a NaN. */ +#define FLT_CHECK_FORCE_UFLOW_NONNAN \ + fld %st(0); \ + fabs; \ + fcomps MO(flt_min); \ + fnstsw; \ + sahf; \ + jnc 6424f; \ + subl $4, %esp; \ + cfi_adjust_cfa_offset (4); \ + fld %st(0); \ + fmul %st(0); \ + fstps (%esp); \ + addl $4, %esp; \ + cfi_adjust_cfa_offset (-4); \ +6424: +#define DBL_CHECK_FORCE_UFLOW_NONNAN \ + fld %st(0); \ + fabs; \ + fcompl MO(dbl_min); \ + fnstsw; \ + sahf; \ + jnc 6453f; \ + subl $8, %esp; \ + cfi_adjust_cfa_offset (8); \ + fld %st(0); \ + fmul %st(0); \ + fstpl (%esp); \ + addl $8, %esp; \ + cfi_adjust_cfa_offset (-8); \ +6453: + +/* Likewise, but the argument is nonnegative and not a NaN. */ +#define FLT_CHECK_FORCE_UFLOW_NONNEG \ + fcoms MO(flt_min); \ + fnstsw; \ + sahf; \ + jnc 6424f; \ + subl $4, %esp; \ + cfi_adjust_cfa_offset (4); \ + fld %st(0); \ + fmul %st(0); \ + fstps (%esp); \ + addl $4, %esp; \ + cfi_adjust_cfa_offset (-4); \ +6424: +#define DBL_CHECK_FORCE_UFLOW_NONNEG \ + fcoml MO(dbl_min); \ + fnstsw; \ + sahf; \ + jnc 6453f; \ + subl $8, %esp; \ + cfi_adjust_cfa_offset (8); \ + fld %st(0); \ + fmul %st(0); \ + fstpl (%esp); \ + addl $8, %esp; \ + cfi_adjust_cfa_offset (-8); \ +6453: +#define LDBL_CHECK_FORCE_UFLOW_NONNEG \ + fldt MO(ldbl_min); \ + fld %st(1); \ + fcompp; \ + fnstsw; \ + sahf; \ + jnc 6464f; \ + fld %st(0); \ + fmul %st(0); \ + fstp %st(0); \ +6464: + #endif /* i386-math-asm.h. */ diff --git a/sysdeps/i386/fpu/s_atan.S b/sysdeps/i386/fpu/s_atan.S index c28b73ce..644de78 100644 --- a/sysdeps/i386/fpu/s_atan.S +++ b/sysdeps/i386/fpu/s_atan.S @@ -4,15 +4,11 @@ */ #include +#include RCSID("$NetBSD: s_atan.S,v 1.4 1995/05/08 23:50:41 jtc Exp $") - .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) +DEFINE_DBL_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -28,20 +24,7 @@ ENTRY(__atan) fldl 4(%esp) fld1 fpatan - fldl MO(dbl_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) -1: ret + DBL_CHECK_FORCE_UFLOW + ret END (__atan) weak_alias (__atan, atan) diff --git a/sysdeps/i386/fpu/s_atanf.S b/sysdeps/i386/fpu/s_atanf.S index da3c2a6..0589c11 100644 --- a/sysdeps/i386/fpu/s_atanf.S +++ b/sysdeps/i386/fpu/s_atanf.S @@ -4,15 +4,11 @@ */ #include +#include RCSID("$NetBSD: s_atanf.S,v 1.3 1995/05/08 23:51:33 jtc Exp $") - .section .rodata.cst4,"aM",@progbits,4 - - .p2align 2 - .type flt_min,@object -flt_min: .byte 0, 0, 0x80, 0 - ASM_SIZE_DIRECTIVE(flt_min) +DEFINE_FLT_MIN #ifdef PIC # define MO(op) op##@GOTOFF(%ecx) @@ -28,20 +24,7 @@ ENTRY(__atanf) flds 4(%esp) fld1 fpatan - flds MO(flt_min) - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $4, %esp - cfi_adjust_cfa_offset (4) - fld %st(0) - fmul %st(0) - fstps (%esp) - addl $4, %esp - cfi_adjust_cfa_offset (-4) -1: ret + FLT_CHECK_FORCE_UFLOW + ret END (__atanf) weak_alias (__atanf, atanf) diff --git a/sysdeps/i386/fpu/s_expm1.S b/sysdeps/i386/fpu/s_expm1.S index 05e5285..b4694c8 100644 --- a/sysdeps/i386/fpu/s_expm1.S +++ b/sysdeps/i386/fpu/s_expm1.S @@ -23,6 +23,7 @@ #include #include +#include .section .rodata @@ -37,12 +38,7 @@ one: .double 1.0 l2e: .tfloat 1.442695040888963407359924681002 ASM_SIZE_DIRECTIVE(l2e) - .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) +DEFINE_DBL_MIN #ifdef PIC #define MO(op) op##@GOTOFF(%edx) @@ -81,21 +77,6 @@ ENTRY(__expm1) #ifdef PIC LOAD_PIC_REG (dx) #endif - fld %st - fabs - fcoml MO(dbl_min) - fstp %st - fnstsw - sahf - jae 5f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) - ret 5: fldt MO(l2e) // log2(e) : x fmulp // log2(e)*x @@ -122,6 +103,7 @@ ENTRY(__expm1) fsubrl MO(one) // 1-2^int(log2(e)*x) : int(log2(e)*x) : 2^(log2(e)*x)-2^int(log2(e)*x) fstp %st(1) // 1-2^int(log2(e)*x) : 2^(log2(e)*x)-2^int(log2(e)*x) fsubrp %st, %st(1) // 2^(log2(e)*x) + DBL_CHECK_FORCE_UFLOW ret 2: fstp %st diff --git a/sysdeps/i386/fpu/s_expm1f.S b/sysdeps/i386/fpu/s_expm1f.S index a83e435..5522902 100644 --- a/sysdeps/i386/fpu/s_expm1f.S +++ b/sysdeps/i386/fpu/s_expm1f.S @@ -23,6 +23,7 @@ #include #include +#include .section .rodata @@ -37,12 +38,7 @@ one: .double 1.0 l2e: .tfloat 1.442695040888963407359924681002 ASM_SIZE_DIRECTIVE(l2e) - .section .rodata.cst4,"aM",@progbits,4 - - .p2align 2 - .type flt_min,@object -flt_min: .byte 0, 0, 0x80, 0 - ASM_SIZE_DIRECTIVE(flt_min) +DEFINE_FLT_MIN #ifdef PIC #define MO(op) op##@GOTOFF(%edx) @@ -81,21 +77,6 @@ ENTRY(__expm1f) #ifdef PIC LOAD_PIC_REG (dx) #endif - fld %st - fabs - fcoms MO(flt_min) - fstp %st - fnstsw - sahf - jae 5f - subl $4, %esp - cfi_adjust_cfa_offset (4) - fld %st(0) - fmul %st(0) - fstps (%esp) - addl $4, %esp - cfi_adjust_cfa_offset (-4) - ret 5: fldt MO(l2e) // log2(e) : x fmulp // log2(e)*x @@ -122,6 +103,7 @@ ENTRY(__expm1f) fsubrl MO(one) // 1-2^int(log2(e)*x) : int(log2(e)*x) : 2^(log2(e)*x)-2^int(log2(e)*x) fstp %st(1) // 1-2^int(log2(e)*x) : 2^(log2(e)*x)-2^int(log2(e)*x) fsubrp %st, %st(1) // 2^(log2(e)*x) + FLT_CHECK_FORCE_UFLOW ret 2: fstp %st diff --git a/sysdeps/i386/fpu/s_log1p.S b/sysdeps/i386/fpu/s_log1p.S index c2559a3..7978e76 100644 --- a/sysdeps/i386/fpu/s_log1p.S +++ b/sysdeps/i386/fpu/s_log1p.S @@ -4,6 +4,7 @@ */ #include +#include RCSID("$NetBSD: s_log1p.S,v 1.7 1995/05/09 00:10:58 jtc Exp $") @@ -17,12 +18,13 @@ RCSID("$NetBSD: s_log1p.S,v 1.7 1995/05/09 00:10:58 jtc Exp $") limit: .double 0.29 one: .double 1.0 - .section .rodata.cst8,"aM",@progbits,8 +DEFINE_DBL_MIN - .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(%edx) +#else +# define MO(op) op +#endif /* * Use the fyl2xp1 function when the argument is in the range -0.29 to 0.29, @@ -44,43 +46,18 @@ ENTRY(__log1p) sahf jc 3f // in case x is NaN or ħInf 4: fabs -#ifdef PIC - fcompl limit@GOTOFF(%edx) -#else - fcompl limit -#endif + fcompl MO(limit) fnstsw sahf jc 2f -#ifdef PIC - faddl one@GOTOFF(%edx) -#else - faddl one -#endif + faddl MO(one) fyl2x ret 2: fyl2xp1 -#ifdef PIC - fldl dbl_min@GOTOFF(%edx) -#else - fldl dbl_min -#endif - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $8, %esp - cfi_adjust_cfa_offset (8) - fld %st(0) - fmul %st(0) - fstpl (%esp) - addl $8, %esp - cfi_adjust_cfa_offset (-8) -1: ret + DBL_CHECK_FORCE_UFLOW_NONNAN + ret 3: jp 4b // in case x is ħInf fstp %st(1) diff --git a/sysdeps/i386/fpu/s_log1pf.S b/sysdeps/i386/fpu/s_log1pf.S index 8fca22e..acaa299 100644 --- a/sysdeps/i386/fpu/s_log1pf.S +++ b/sysdeps/i386/fpu/s_log1pf.S @@ -4,6 +4,7 @@ */ #include +#include RCSID("$NetBSD: s_log1pf.S,v 1.4 1995/05/09 00:13:05 jtc Exp $") @@ -17,12 +18,13 @@ RCSID("$NetBSD: s_log1pf.S,v 1.4 1995/05/09 00:13:05 jtc Exp $") limit: .float 0.29 one: .float 1.0 - .section .rodata.cst4,"aM",@progbits,4 +DEFINE_FLT_MIN - .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(%edx) +#else +# define MO(op) op +#endif /* * Use the fyl2xp1 function when the argument is in the range -0.29 to 0.29, @@ -44,43 +46,18 @@ ENTRY(__log1pf) sahf jc 3f // in case x is NaN or ħInf 4: fabs -#ifdef PIC - fcomps limit@GOTOFF(%edx) -#else - fcomps limit -#endif + fcomps MO(limit) fnstsw sahf jc 2f -#ifdef PIC - fadds one@GOTOFF(%edx) -#else - fadds one -#endif + fadds MO(one) fyl2x ret 2: fyl2xp1 -#ifdef PIC - flds flt_min@GOTOFF(%edx) -#else - flds flt_min -#endif - fld %st(1) - fabs - fucompp - fnstsw - sahf - jnc 1f - subl $4, %esp - cfi_adjust_cfa_offset (4) - fld %st(0) - fmul %st(0) - fstps (%esp) - addl $4, %esp - cfi_adjust_cfa_offset (-4) -1: ret + FLT_CHECK_FORCE_UFLOW_NONNAN + ret 3: jp 4b // in case x is ħInf fstp %st(1) diff --git a/sysdeps/i386/fpu/s_log1pl.S b/sysdeps/i386/fpu/s_log1pl.S index 8f87cf6..98965fa 100644 --- a/sysdeps/i386/fpu/s_log1pl.S +++ b/sysdeps/i386/fpu/s_log1pl.S @@ -22,6 +22,12 @@ limit: .tfloat 0.29 but it helps to optimize the code. */ one: .double 1.0 +#ifdef PIC +# define MO(op) op##@GOTOFF(%edx) +#else +# define MO(op) op +#endif + /* * Use the fyl2xp1 function when the argument is in the range -0.29 to 0.29, * otherwise fyl2x with the needed extra computation. @@ -43,11 +49,7 @@ ENTRY(__log1pl) jc 3f // in case x is NaN or ħInf 4: fabs -#ifdef PIC - fldt limit@GOTOFF(%edx) -#else - fldt limit -#endif + fldt MO(limit) fcompp fnstsw sahf @@ -58,11 +60,7 @@ ENTRY(__log1pl) cmpl $0xc040, %eax jae 5f -#ifdef PIC - faddl one@GOTOFF(%edx) -#else - faddl one -#endif + faddl MO(one) 5: fyl2x ret -- cgit v1.1