From b3f27d8150d4f3c64063a9a257ec1d228de66398 Mon Sep 17 00:00:00 2001 From: Joseph Myers Date: Wed, 22 Sep 2021 21:25:31 +0000 Subject: Add narrowing fma functions This patch adds the narrowing fused multiply-add functions from TS 18661-1 / TS 18661-3 / C2X to glibc's libm: ffma, ffmal, dfmal, f32fmaf64, f32fmaf32x, f32xfmaf64 for all configurations; f32fmaf64x, f32fmaf128, f64fmaf64x, f64fmaf128, f32xfmaf64x, f32xfmaf128, f64xfmaf128 for configurations with _Float64x and _Float128; __f32fmaieee128 and __f64fmaieee128 aliases in the powerpc64le case (for calls to ffmal and dfmal when long double is IEEE binary128). Corresponding tgmath.h macro support is also added. The changes are mostly similar to those for the other narrowing functions previously added, especially that for sqrt, so the description of those generally applies to this patch as well. As with sqrt, I reused the same test inputs in auto-libm-test-in as for non-narrowing fma rather than adding extra or separate inputs for narrowing fma. The tests in libm-test-narrow-fma.inc also follow those for non-narrowing fma. The non-narrowing fma has a known bug (bug 6801) that it does not set errno on errors (overflow, underflow, Inf * 0, Inf - Inf). Rather than fixing this or having narrowing fma check for errors when non-narrowing does not (complicating the cases when narrowing fma can otherwise be an alias for a non-narrowing function), this patch does not attempt to check for errors from narrowing fma and set errno; the CHECK_NARROW_FMA macro is still present, but as a placeholder that does nothing, and this missing errno setting is considered to be covered by the existing bug rather than needing a separate open bug. missing-errno annotations are duly added to many of the auto-libm-test-in test inputs for fma. This completes adding all the new functions from TS 18661-1 to glibc, so will be followed by corresponding stdc-predef.h changes to define __STDC_IEC_60559_BFP__ and __STDC_IEC_60559_COMPLEX__, as the support for TS 18661-1 will be at a similar level to that for C standard floating-point facilities up to C11 (pragmas not implemented, but library functions done). (There are still further changes to be done to implement changes to the types of fromfp functions from N2548.) Tested as followed: natively with the full glibc testsuite for x86_64 (GCC 11, 7, 6) and x86 (GCC 11); with build-many-glibcs.py with GCC 11, 7 and 6; cross testing of math/ tests for powerpc64le, powerpc32 hard float, mips64 (all three ABIs, both hard and soft float). The different GCC versions are to cover the different cases in tgmath.h and tgmath.h tests properly (GCC 6 has _Float* only as typedefs in glibc headers, GCC 7 has proper _Float* support, GCC 8 adds __builtin_tgmath). --- sysdeps/ieee754/soft-fp/s_dfmal.c | 74 +++++++++++++++++++++++++++++++++++++++ sysdeps/ieee754/soft-fp/s_ffma.c | 72 +++++++++++++++++++++++++++++++++++++ sysdeps/ieee754/soft-fp/s_ffmal.c | 70 ++++++++++++++++++++++++++++++++++++ sysdeps/ieee754/soft-fp/s_fma.c | 6 ++++ sysdeps/ieee754/soft-fp/s_fmal.c | 4 +++ 5 files changed, 226 insertions(+) create mode 100644 sysdeps/ieee754/soft-fp/s_dfmal.c create mode 100644 sysdeps/ieee754/soft-fp/s_ffma.c create mode 100644 sysdeps/ieee754/soft-fp/s_ffmal.c (limited to 'sysdeps/ieee754/soft-fp') diff --git a/sysdeps/ieee754/soft-fp/s_dfmal.c b/sysdeps/ieee754/soft-fp/s_dfmal.c new file mode 100644 index 0000000000..6e8b081a9e --- /dev/null +++ b/sysdeps/ieee754/soft-fp/s_dfmal.c @@ -0,0 +1,74 @@ +/* Fused multiply-add of long double (ldbl-128) values, narrowing the result to + double, using soft-fp. + Copyright (C) 2021 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, see + . */ + +#define f32xfmaf64x __hide_f32xfmaf64x +#define f32xfmaf128 __hide_f32xfmaf128 +#define f64fmaf64x __hide_f64fmaf64x +#define f64fmaf128 __hide_f64fmaf128 +#include +#undef f32xfmaf64x +#undef f32xfmaf128 +#undef f64fmaf64x +#undef f64fmaf128 + +#include +#include + +/* R_e is not set in cases where it is not used in packing, but the + compiler does not see that it is set in all cases where it is + used, resulting in warnings that it may be used uninitialized. + The location of the warning differs in different versions of GCC, + it may be where R is defined using a macro or it may be where the + macro is defined. */ +DIAG_PUSH_NEEDS_COMMENT; +DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wmaybe-uninitialized"); + +#include +#include +#include + +double +__dfmal (_Float128 x, _Float128 y, _Float128 z) +{ + FP_DECL_EX; + FP_DECL_Q (X); + FP_DECL_Q (Y); + FP_DECL_Q (Z); + FP_DECL_Q (R); + FP_DECL_D (RN); + double ret; + + FP_INIT_ROUNDMODE; + FP_UNPACK_Q (X, x); + FP_UNPACK_Q (Y, y); + FP_UNPACK_Q (Z, z); + FP_FMA_Q (R, X, Y, Z); +#if _FP_W_TYPE_SIZE < 64 + FP_TRUNC_COOKED (D, Q, 2, 4, RN, R); +#else + FP_TRUNC_COOKED (D, Q, 1, 2, RN, R); +#endif + FP_PACK_D (ret, RN); + FP_HANDLE_EXCEPTIONS; + CHECK_NARROW_FMA (ret, x, y, z); + return ret; +} +DIAG_POP_NEEDS_COMMENT; + +libm_alias_double_ldouble (fma) diff --git a/sysdeps/ieee754/soft-fp/s_ffma.c b/sysdeps/ieee754/soft-fp/s_ffma.c new file mode 100644 index 0000000000..3ab5844b0b --- /dev/null +++ b/sysdeps/ieee754/soft-fp/s_ffma.c @@ -0,0 +1,72 @@ +/* Fused multiply-add of double values, narrowing the result to float, + using soft-fp. + Copyright (C) 2021 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, see + . */ + +#define f32fmaf64 __hide_f32fmaf64 +#define f32fmaf32x __hide_f32fmaf32x +#define ffmal __hide_ffmal +#include +#undef f32fmaf64 +#undef f32fmaf32x +#undef ffmal + +#include +#include + +/* R_e is not set in cases where it is not used in packing, but the + compiler does not see that it is set in all cases where it is + used, resulting in warnings that it may be used uninitialized. + The location of the warning differs in different versions of GCC, + it may be where R is defined using a macro or it may be where the + macro is defined. */ +DIAG_PUSH_NEEDS_COMMENT; +DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wmaybe-uninitialized"); + +#include +#include +#include + +float +__ffma (double x, double y, double z) +{ + FP_DECL_EX; + FP_DECL_D (X); + FP_DECL_D (Y); + FP_DECL_D (Z); + FP_DECL_D (R); + FP_DECL_S (RN); + float ret; + + FP_INIT_ROUNDMODE; + FP_UNPACK_D (X, x); + FP_UNPACK_D (Y, y); + FP_UNPACK_D (Z, z); + FP_FMA_D (R, X, Y, Z); +#if _FP_W_TYPE_SIZE < _FP_FRACBITS_D + FP_TRUNC_COOKED (S, D, 1, 2, RN, R); +#else + FP_TRUNC_COOKED (S, D, 1, 1, RN, R); +#endif + FP_PACK_S (ret, RN); + FP_HANDLE_EXCEPTIONS; + CHECK_NARROW_FMA (ret, x, y, z); + return ret; +} +DIAG_POP_NEEDS_COMMENT; + +libm_alias_float_double (fma) diff --git a/sysdeps/ieee754/soft-fp/s_ffmal.c b/sysdeps/ieee754/soft-fp/s_ffmal.c new file mode 100644 index 0000000000..15e537e978 --- /dev/null +++ b/sysdeps/ieee754/soft-fp/s_ffmal.c @@ -0,0 +1,70 @@ +/* Fused multiply-add of long double (ldbl-128) values, narrowing the result to + float, using soft-fp. + Copyright (C) 2021 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, see + . */ + +#define f32fmaf64x __hide_f32fmaf64x +#define f32fmaf128 __hide_f32fmaf128 +#include +#undef f32fmaf64x +#undef f32fmaf128 + +#include +#include + +/* R_e is not set in cases where it is not used in packing, but the + compiler does not see that it is set in all cases where it is + used, resulting in warnings that it may be used uninitialized. + The location of the warning differs in different versions of GCC, + it may be where R is defined using a macro or it may be where the + macro is defined. */ +DIAG_PUSH_NEEDS_COMMENT; +DIAG_IGNORE_NEEDS_COMMENT (4.9, "-Wmaybe-uninitialized"); + +#include +#include +#include + +float +__ffmal (_Float128 x, _Float128 y, _Float128 z) +{ + FP_DECL_EX; + FP_DECL_Q (X); + FP_DECL_Q (Y); + FP_DECL_Q (Z); + FP_DECL_Q (R); + FP_DECL_S (RN); + float ret; + + FP_INIT_ROUNDMODE; + FP_UNPACK_Q (X, x); + FP_UNPACK_Q (Y, y); + FP_UNPACK_Q (Z, z); + FP_FMA_Q (R, X, Y, Z); +#if _FP_W_TYPE_SIZE < 64 + FP_TRUNC_COOKED (S, Q, 1, 4, RN, R); +#else + FP_TRUNC_COOKED (S, Q, 1, 2, RN, R); +#endif + FP_PACK_S (ret, RN); + FP_HANDLE_EXCEPTIONS; + CHECK_NARROW_FMA (ret, x, y, z); + return ret; +} +DIAG_POP_NEEDS_COMMENT; + +libm_alias_float_ldouble (fma) diff --git a/sysdeps/ieee754/soft-fp/s_fma.c b/sysdeps/ieee754/soft-fp/s_fma.c index 4b0d6b50aa..d3870452dd 100644 --- a/sysdeps/ieee754/soft-fp/s_fma.c +++ b/sysdeps/ieee754/soft-fp/s_fma.c @@ -26,9 +26,14 @@ . */ #define NO_MATH_REDIRECT +#define dfmal __hide_dfmal +#define f32xfmaf64 __hide_f32xfmaf64 #include +#undef dfmal +#undef f32xfmaf64 #include #include +#include /* R_e is not set in cases where it is not used in packing, but the compiler does not see that it is set in all cases where it is @@ -66,4 +71,5 @@ DIAG_POP_NEEDS_COMMENT; #ifndef __fma libm_alias_double (__fma, fma) +libm_alias_double_narrow (__fma, fma) #endif diff --git a/sysdeps/ieee754/soft-fp/s_fmal.c b/sysdeps/ieee754/soft-fp/s_fmal.c index aecec13923..7be917a46e 100644 --- a/sysdeps/ieee754/soft-fp/s_fmal.c +++ b/sysdeps/ieee754/soft-fp/s_fmal.c @@ -26,9 +26,12 @@ . */ #define NO_MATH_REDIRECT +#define f64xfmaf128 __hide_f64xfmaf128 #include +#undef f64xfmaf128 #include #include +#include /* R_e is not set in cases where it is not used in packing, but the compiler does not see that it is set in all cases where it is @@ -65,3 +68,4 @@ __fmal (long double a, long double b, long double c) DIAG_POP_NEEDS_COMMENT; libm_alias_ldouble (__fma, fma) +libm_alias_ldouble_narrow (__fma, fma) -- cgit 1.4.1