diff options
50 files changed, 4447 insertions, 1 deletions
diff --git a/bits/libm-simd-decl-stubs.h b/bits/libm-simd-decl-stubs.h index 73252615ca..845246fab9 100644 --- a/bits/libm-simd-decl-stubs.h +++ b/bits/libm-simd-decl-stubs.h @@ -241,4 +241,15 @@ #define __DECL_SIMD_log2f32x #define __DECL_SIMD_log2f64x #define __DECL_SIMD_log2f128x + +#define __DECL_SIMD_log1p +#define __DECL_SIMD_log1pf +#define __DECL_SIMD_log1pl +#define __DECL_SIMD_log1pf16 +#define __DECL_SIMD_log1pf32 +#define __DECL_SIMD_log1pf64 +#define __DECL_SIMD_log1pf128 +#define __DECL_SIMD_log1pf32x +#define __DECL_SIMD_log1pf64x +#define __DECL_SIMD_log1pf128x #endif diff --git a/math/bits/mathcalls.h b/math/bits/mathcalls.h index bfe52a4666..aa4bc61aa4 100644 --- a/math/bits/mathcalls.h +++ b/math/bits/mathcalls.h @@ -119,7 +119,7 @@ __MATHCALL_VEC (exp10,, (_Mdouble_ __x)); __MATHCALL_VEC (expm1,, (_Mdouble_ __x)); /* Return log(1 + X). */ -__MATHCALL (log1p,, (_Mdouble_ __x)); +__MATHCALL_VEC (log1p,, (_Mdouble_ __x)); /* Return the base 2 signed integral exponent of X. */ __MATHCALL (logb,, (_Mdouble_ __x)); diff --git a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist index fa8b016c5d..68b940606a 100644 --- a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist +++ b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist @@ -55,6 +55,7 @@ GLIBC_2.35 _ZGVbN2v_exp10 F GLIBC_2.35 _ZGVbN2v_exp2 F GLIBC_2.35 _ZGVbN2v_expm1 F GLIBC_2.35 _ZGVbN2v_log10 F +GLIBC_2.35 _ZGVbN2v_log1p F GLIBC_2.35 _ZGVbN2v_log2 F GLIBC_2.35 _ZGVbN2v_sinh F GLIBC_2.35 _ZGVbN2vv_atan2 F @@ -68,6 +69,7 @@ GLIBC_2.35 _ZGVbN4v_exp10f F GLIBC_2.35 _ZGVbN4v_exp2f F GLIBC_2.35 _ZGVbN4v_expm1f F GLIBC_2.35 _ZGVbN4v_log10f F +GLIBC_2.35 _ZGVbN4v_log1pf F GLIBC_2.35 _ZGVbN4v_log2f F GLIBC_2.35 _ZGVbN4v_sinhf F GLIBC_2.35 _ZGVbN4vv_atan2f F @@ -81,6 +83,7 @@ GLIBC_2.35 _ZGVcN4v_exp10 F GLIBC_2.35 _ZGVcN4v_exp2 F GLIBC_2.35 _ZGVcN4v_expm1 F GLIBC_2.35 _ZGVcN4v_log10 F +GLIBC_2.35 _ZGVcN4v_log1p F GLIBC_2.35 _ZGVcN4v_log2 F GLIBC_2.35 _ZGVcN4v_sinh F GLIBC_2.35 _ZGVcN4vv_atan2 F @@ -94,6 +97,7 @@ GLIBC_2.35 _ZGVcN8v_exp10f F GLIBC_2.35 _ZGVcN8v_exp2f F GLIBC_2.35 _ZGVcN8v_expm1f F GLIBC_2.35 _ZGVcN8v_log10f F +GLIBC_2.35 _ZGVcN8v_log1pf F GLIBC_2.35 _ZGVcN8v_log2f F GLIBC_2.35 _ZGVcN8v_sinhf F GLIBC_2.35 _ZGVcN8vv_atan2f F @@ -107,6 +111,7 @@ GLIBC_2.35 _ZGVdN4v_exp10 F GLIBC_2.35 _ZGVdN4v_exp2 F GLIBC_2.35 _ZGVdN4v_expm1 F GLIBC_2.35 _ZGVdN4v_log10 F +GLIBC_2.35 _ZGVdN4v_log1p F GLIBC_2.35 _ZGVdN4v_log2 F GLIBC_2.35 _ZGVdN4v_sinh F GLIBC_2.35 _ZGVdN4vv_atan2 F @@ -120,6 +125,7 @@ GLIBC_2.35 _ZGVdN8v_exp10f F GLIBC_2.35 _ZGVdN8v_exp2f F GLIBC_2.35 _ZGVdN8v_expm1f F GLIBC_2.35 _ZGVdN8v_log10f F +GLIBC_2.35 _ZGVdN8v_log1pf F GLIBC_2.35 _ZGVdN8v_log2f F GLIBC_2.35 _ZGVdN8v_sinhf F GLIBC_2.35 _ZGVdN8vv_atan2f F @@ -133,6 +139,7 @@ GLIBC_2.35 _ZGVeN16v_exp10f F GLIBC_2.35 _ZGVeN16v_exp2f F GLIBC_2.35 _ZGVeN16v_expm1f F GLIBC_2.35 _ZGVeN16v_log10f F +GLIBC_2.35 _ZGVeN16v_log1pf F GLIBC_2.35 _ZGVeN16v_log2f F GLIBC_2.35 _ZGVeN16v_sinhf F GLIBC_2.35 _ZGVeN16vv_atan2f F @@ -146,6 +153,7 @@ GLIBC_2.35 _ZGVeN8v_exp10 F GLIBC_2.35 _ZGVeN8v_exp2 F GLIBC_2.35 _ZGVeN8v_expm1 F GLIBC_2.35 _ZGVeN8v_log10 F +GLIBC_2.35 _ZGVeN8v_log1p F GLIBC_2.35 _ZGVeN8v_log2 F GLIBC_2.35 _ZGVeN8v_sinh F GLIBC_2.35 _ZGVeN8vv_atan2 F diff --git a/sysdeps/x86/fpu/bits/math-vector.h b/sysdeps/x86/fpu/bits/math-vector.h index 59d284a10a..14c9db3bb3 100644 --- a/sysdeps/x86/fpu/bits/math-vector.h +++ b/sysdeps/x86/fpu/bits/math-vector.h @@ -110,6 +110,10 @@ # define __DECL_SIMD_log2 __DECL_SIMD_x86_64 # undef __DECL_SIMD_log2f # define __DECL_SIMD_log2f __DECL_SIMD_x86_64 +# undef __DECL_SIMD_log1p +# define __DECL_SIMD_log1p __DECL_SIMD_x86_64 +# undef __DECL_SIMD_log1pf +# define __DECL_SIMD_log1pf __DECL_SIMD_x86_64 # endif #endif diff --git a/sysdeps/x86/fpu/finclude/math-vector-fortran.h b/sysdeps/x86/fpu/finclude/math-vector-fortran.h index a2ca9a203f..3dca196432 100644 --- a/sysdeps/x86/fpu/finclude/math-vector-fortran.h +++ b/sysdeps/x86/fpu/finclude/math-vector-fortran.h @@ -54,6 +54,8 @@ !GCC$ builtin (log10f) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (log2) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (log2f) attributes simd (notinbranch) if('x86_64') +!GCC$ builtin (log1p) attributes simd (notinbranch) if('x86_64') +!GCC$ builtin (log1pf) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (cos) attributes simd (notinbranch) if('x32') !GCC$ builtin (cosf) attributes simd (notinbranch) if('x32') @@ -93,3 +95,5 @@ !GCC$ builtin (log10f) attributes simd (notinbranch) if('x32') !GCC$ builtin (log2) attributes simd (notinbranch) if('x32') !GCC$ builtin (log2f) attributes simd (notinbranch) if('x32') +!GCC$ builtin (log1p) attributes simd (notinbranch) if('x32') +!GCC$ builtin (log1pf) attributes simd (notinbranch) if('x32') diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig index 8d6d0915af..378cb06d37 100644 --- a/sysdeps/x86_64/fpu/Makeconfig +++ b/sysdeps/x86_64/fpu/Makeconfig @@ -36,6 +36,7 @@ libmvec-funcs = \ hypot \ log \ log10 \ + log1p \ log2 \ pow \ sin \ diff --git a/sysdeps/x86_64/fpu/Versions b/sysdeps/x86_64/fpu/Versions index 1b48c2d642..155fb115f3 100644 --- a/sysdeps/x86_64/fpu/Versions +++ b/sysdeps/x86_64/fpu/Versions @@ -23,6 +23,7 @@ libmvec { _ZGVbN2v_exp2; _ZGVcN4v_exp2; _ZGVdN4v_exp2; _ZGVeN8v_exp2; _ZGVbN2v_expm1; _ZGVcN4v_expm1; _ZGVdN4v_expm1; _ZGVeN8v_expm1; _ZGVbN2v_log10; _ZGVcN4v_log10; _ZGVdN4v_log10; _ZGVeN8v_log10; + _ZGVbN2v_log1p; _ZGVcN4v_log1p; _ZGVdN4v_log1p; _ZGVeN8v_log1p; _ZGVbN2v_log2; _ZGVcN4v_log2; _ZGVdN4v_log2; _ZGVeN8v_log2; _ZGVbN2v_sinh; _ZGVcN4v_sinh; _ZGVdN4v_sinh; _ZGVeN8v_sinh; _ZGVbN2vv_atan2; _ZGVcN4vv_atan2; _ZGVdN4vv_atan2; _ZGVeN8vv_atan2; @@ -36,6 +37,7 @@ libmvec { _ZGVbN4v_exp2f; _ZGVcN8v_exp2f; _ZGVdN8v_exp2f; _ZGVeN16v_exp2f; _ZGVbN4v_expm1f; _ZGVcN8v_expm1f; _ZGVdN8v_expm1f; _ZGVeN16v_expm1f; _ZGVbN4v_log10f; _ZGVcN8v_log10f; _ZGVdN8v_log10f; _ZGVeN16v_log10f; + _ZGVbN4v_log1pf; _ZGVcN8v_log1pf; _ZGVdN8v_log1pf; _ZGVeN16v_log1pf; _ZGVbN4v_log2f; _ZGVcN8v_log2f; _ZGVdN8v_log2f; _ZGVeN16v_log2f; _ZGVbN4v_sinhf; _ZGVcN8v_sinhf; _ZGVdN8v_sinhf; _ZGVeN16v_sinhf; _ZGVbN4vv_atan2f; _ZGVcN8vv_atan2f; _ZGVdN8vv_atan2f; _ZGVeN16vv_atan2f; diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps index 3b7f3cee6f..a2b15a795b 100644 --- a/sysdeps/x86_64/fpu/libm-test-ulps +++ b/sysdeps/x86_64/fpu/libm-test-ulps @@ -1685,6 +1685,26 @@ float: 2 float128: 2 ldouble: 3 +Function: "log1p_vlen16": +float: 2 + +Function: "log1p_vlen2": +double: 1 + +Function: "log1p_vlen4": +double: 1 +float: 2 + +Function: "log1p_vlen4_avx2": +double: 1 + +Function: "log1p_vlen8": +double: 1 +float: 2 + +Function: "log1p_vlen8_avx2": +float: 2 + Function: "log2": double: 2 float: 1 diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core-sse2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core-sse2.S new file mode 100644 index 0000000000..8004088346 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core-sse2.S @@ -0,0 +1,20 @@ +/* SSE2 version of vectorized log1p, vector length is 2. + 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 + <https://www.gnu.org/licenses/>. */ + +#define _ZGVbN2v_log1p _ZGVbN2v_log1p_sse2 +#include "../svml_d_log1p2_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core.c new file mode 100644 index 0000000000..35ca620aba --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core.c @@ -0,0 +1,27 @@ +/* Multiple versions of vectorized log1p, vector length is 2. + 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 + <https://www.gnu.org/licenses/>. */ + +#define SYMBOL_NAME _ZGVbN2v_log1p +#include "ifunc-mathvec-sse4_1.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVbN2v_log1p, __GI__ZGVbN2v_log1p, __redirect__ZGVbN2v_log1p) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core_sse4.S new file mode 100644 index 0000000000..1f487189fa --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_log1p2_core_sse4.S @@ -0,0 +1,1398 @@ +/* Function log1p vectorized with SSE4. + 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 + https://www.gnu.org/licenses/. */ + +/* + * ALGORITHM DESCRIPTION: + * + * 1+x = 2^k*(xh + xl) is computed in high-low parts; xh in [1,2) + * Get short reciprocal approximation Rcp ~ 1/xh + * R = (Rcp*xh - 1.0) + |