[PATCH v2 0/4] Enable generic math code for more arches

unofficial mirror of libc-alpha@sourceware.org
 help / color / mirror / Atom feed

* [PATCH v2 0/4] Enable generic math code for more arches
@ 2020-06-02  0:35 Vineet Gupta via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
                   ` (3 more replies)
  0 siblings, 4 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02  0:35 UTC (permalink / raw)
  To: libc-alpha; +Cc: Vineet Gupta, linux-snps-arc

Hi,

This was suggested by Adhemerval Zanella as part of ARC hard float
review.

There was testing snafu last time around (wrong branch on build server).
Although only relevant configs are aarch64 and PPC, the series passes
build-many-glibcs for following:

ARCH="aarch64-linux-gnu arm-linux-gnueabi arm-linux-gnueabihf hppa-linux-gnu x86_64-linux-gnu arm-linux-gnueabihf riscv64-linux-gnu-rv64imac-lp64 riscv64-linux-gnu-rv64imafdc-lp64 powerpc-linux-gnu microblaze-linux-gnu nios2-linux-gnu hppa-linux-gnu"

Thx,
-Vineet

Vineet Gupta (4):
  iee754: provide gcc builtins based generic sqrt functions
  iee754: provide gcc builtins based generic fma functions
  aarch/fpu: use generic sqrt{,f}, fma{,f}, nearbyint{,f} functions
  powerpc/fpu: use generic fma functions

 sysdeps/aarch64/fpu/e_sqrt.c                | 27 --------
 sysdeps/aarch64/fpu/e_sqrtf.c               | 27 --------
 sysdeps/aarch64/fpu/math-use-builtins.h     | 71 +++++++++++++++++++++
 sysdeps/aarch64/fpu/s_floor.c               | 29 ---------
 sysdeps/aarch64/fpu/s_floorf.c              | 29 ---------
 sysdeps/aarch64/fpu/s_fma.c                 | 28 --------
 sysdeps/aarch64/fpu/s_fmaf.c                | 28 --------
 sysdeps/aarch64/fpu/s_nearbyint.c           | 28 --------
 sysdeps/aarch64/fpu/s_nearbyintf.c          | 28 --------
 sysdeps/aarch64/fpu/s_rint.c                | 29 ---------
 sysdeps/aarch64/fpu/s_rintf.c               | 29 ---------
 sysdeps/aarch64/fpu/s_round.c               | 29 ---------
 sysdeps/aarch64/fpu/s_roundf.c              | 29 ---------
 sysdeps/aarch64/fpu/s_trunc.c               | 29 ---------
 sysdeps/aarch64/fpu/s_truncf.c              | 29 ---------
 sysdeps/generic/math-use-builtins.h         |  8 +++
 sysdeps/ieee754/dbl-64/e_sqrt.c             |  6 ++
 sysdeps/ieee754/dbl-64/s_fma.c              |  6 ++
 sysdeps/ieee754/dbl-64/s_fmaf.c             |  6 ++
 sysdeps/ieee754/float128/float128_private.h |  2 +
 sysdeps/ieee754/flt-32/e_sqrtf.c            | 16 +++--
 sysdeps/ieee754/ldbl-128/s_fmal.c           |  5 ++
 sysdeps/powerpc/fpu/math-use-builtins.h     | 69 ++++++++++++++++++++
 sysdeps/powerpc/fpu/s_fma.c                 | 27 --------
 sysdeps/powerpc/fpu/s_fmaf.c                | 27 --------
 25 files changed, 183 insertions(+), 458 deletions(-)
 delete mode 100644 sysdeps/aarch64/fpu/e_sqrt.c
 delete mode 100644 sysdeps/aarch64/fpu/e_sqrtf.c
 create mode 100644 sysdeps/aarch64/fpu/math-use-builtins.h
 delete mode 100644 sysdeps/aarch64/fpu/s_floor.c
 delete mode 100644 sysdeps/aarch64/fpu/s_floorf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_fma.c
 delete mode 100644 sysdeps/aarch64/fpu/s_fmaf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_nearbyint.c
 delete mode 100644 sysdeps/aarch64/fpu/s_nearbyintf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_rint.c
 delete mode 100644 sysdeps/aarch64/fpu/s_rintf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_round.c
 delete mode 100644 sysdeps/aarch64/fpu/s_roundf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_trunc.c
 delete mode 100644 sysdeps/aarch64/fpu/s_truncf.c
 create mode 100644 sysdeps/powerpc/fpu/math-use-builtins.h
 delete mode 100644 sysdeps/powerpc/fpu/s_fma.c
 delete mode 100644 sysdeps/powerpc/fpu/s_fmaf.c

-- 
2.20.1


^ permalink raw reply	[flat|nested] 16+ messages in thread

* [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions
  2020-06-02  0:35 [PATCH v2 0/4] Enable generic math code for more arches Vineet Gupta via Libc-alpha
@ 2020-06-02  0:35 ` Vineet Gupta via Libc-alpha
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
  2020-06-02 17:17   ` Adhemerval Zanella via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions Vineet Gupta via Libc-alpha
                   ` (2 subsequent siblings)
  3 siblings, 2 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02  0:35 UTC (permalink / raw)
  To: libc-alpha; +Cc: Vineet Gupta, linux-snps-arc

Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
---
 sysdeps/generic/math-use-builtins.h |  3 +++
 sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
 sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
 3 files changed, 19 insertions(+), 6 deletions(-)

diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
index 8a39ef58bc95..fc724c824a17 100644
--- a/sysdeps/generic/math-use-builtins.h
+++ b/sysdeps/generic/math-use-builtins.h
@@ -60,4 +60,7 @@
 # define USE_COPYSIGNF128_BUILTIN 0
 #endif
 
+#define USE_SQRT_BUILTIN 0
+#define USE_SQRTF_BUILTIN 0
+
 #endif /* math-use-builtins.h */
diff --git a/sysdeps/ieee754/dbl-64/e_sqrt.c b/sysdeps/ieee754/dbl-64/e_sqrt.c
index d42a1a4eb6e9..518a8ae5cdaf 100644
--- a/sysdeps/ieee754/dbl-64/e_sqrt.c
+++ b/sysdeps/ieee754/dbl-64/e_sqrt.c
@@ -41,6 +41,7 @@
 #include <math_private.h>
 #include <fenv_private.h>
 #include <libm-alias-finite.h>
+#include <math-use-builtins.h>
 
 /*********************************************************************/
 /* An ultimate sqrt routine. Given an IEEE double machine number x   */
@@ -50,6 +51,10 @@
 double
 __ieee754_sqrt (double x)
 {
+#if USE_SQRT_BUILTIN
+  return __builtin_sqrt (x);
+#else
+  /* Use generic implementation.  */
   static const double
     rt0 = 9.99999999859990725855365213134618E-01,
     rt1 = 4.99999999495955425917856814202739E-01,
@@ -138,6 +143,7 @@ __ieee754_sqrt (double x)
 	return (x - x) / (x - x); /* sqrt(-ve)=sNaN */
       return 0x1p-256 * __ieee754_sqrt (x * 0x1p512);
     }
+#endif /* ! USE_SQRT_BUILTIN  */
 }
 #ifndef __ieee754_sqrt
 libm_alias_finite (__ieee754_sqrt, __sqrt)
diff --git a/sysdeps/ieee754/flt-32/e_sqrtf.c b/sysdeps/ieee754/flt-32/e_sqrtf.c
index b339444301aa..d85a04162983 100644
--- a/sysdeps/ieee754/flt-32/e_sqrtf.c
+++ b/sysdeps/ieee754/flt-32/e_sqrtf.c
@@ -16,12 +16,15 @@
 #include <math.h>
 #include <math_private.h>
 #include <libm-alias-finite.h>
-
-static	const float	one	= 1.0, tiny=1.0e-30;
+#include <math-use-builtins.h>
 
 float
 __ieee754_sqrtf(float x)
 {
+#if USE_SQRTF_BUILTIN
+	return __builtin_sqrtf (x);
+#else
+	/* Use generic implementation.  */
 	float z;
 	int32_t sign = (int)0x80000000;
 	int32_t ix,s,q,m,t,i;
@@ -70,10 +73,10 @@ __ieee754_sqrtf(float x)
 
     /* use floating add to find out rounding direction */
 	if(ix!=0) {
-	    z = one-tiny; /* trigger inexact flag */
-	    if (z>=one) {
-		z = one+tiny;
-		if (z>one)
+	    z = 0x1p0 - 0x1.4484cp-100; /* trigger inexact flag */
+	    if (z >= 0x1p0) {
+		z = 0x1p0 + 0x1.4484cp-100;
+		if (z > 0x1p0)
 		    q += 2;
 		else
 		    q += (q&1);
@@ -83,6 +86,7 @@ __ieee754_sqrtf(float x)
 	ix += (m <<23);
 	SET_FLOAT_WORD(z,ix);
 	return z;
+#endif /* ! USE_SQRTF_BUILTIN  */
 }
 #ifndef __ieee754_sqrtf
 libm_alias_finite (__ieee754_sqrtf, __sqrtf)
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 16+ messages in thread

* [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02  0:35 [PATCH v2 0/4] Enable generic math code for more arches Vineet Gupta via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
@ 2020-06-02  0:35 ` Vineet Gupta via Libc-alpha
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
  2020-06-02 17:28   ` Adhemerval Zanella via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions Vineet Gupta via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 4/4] powerpc/fpu: use generic fma functions Vineet Gupta via Libc-alpha
  3 siblings, 2 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02  0:35 UTC (permalink / raw)
  To: libc-alpha; +Cc: Vineet Gupta, linux-snps-arc

---
 sysdeps/generic/math-use-builtins.h         | 5 +++++
 sysdeps/ieee754/dbl-64/s_fma.c              | 6 ++++++
 sysdeps/ieee754/dbl-64/s_fmaf.c             | 6 ++++++
 sysdeps/ieee754/float128/float128_private.h | 2 ++
 sysdeps/ieee754/ldbl-128/s_fmal.c           | 5 +++++
 5 files changed, 24 insertions(+)

diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
index fc724c824a17..cf25ed8a2138 100644
--- a/sysdeps/generic/math-use-builtins.h
+++ b/sysdeps/generic/math-use-builtins.h
@@ -63,4 +63,9 @@
 #define USE_SQRT_BUILTIN 0
 #define USE_SQRTF_BUILTIN 0
 
+#define USE_FMA_BUILTIN 0
+#define USE_FMAF_BUILTIN 0
+#define USE_FMAL_BUILTIN 0
+#define USE_FMAF128_BUILTIN 0
+
 #endif /* math-use-builtins.h */
diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
index 876df6e78bdc..9dc5b132b9ee 100644
--- a/sysdeps/ieee754/dbl-64/s_fma.c
+++ b/sysdeps/ieee754/dbl-64/s_fma.c
@@ -25,6 +25,7 @@
 #include <fenv_private.h>
 #include <libm-alias-double.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
 /* This implementation uses rounding to odd to avoid problems with
    double rounding.  See a paper by Boldo and Melquiond:
@@ -33,6 +34,10 @@
 double
 __fma (double x, double y, double z)
 {
+#if USE_FMA_BUILTIN
+  return __builtin_fma (x, y, z);
+#else
+  /* Use generic implementation.  */
   union ieee754_double u, v, w;
   int adjust = 0;
   u.d = x;
@@ -292,6 +297,7 @@ __fma (double x, double y, double z)
       v.ieee.mantissa1 |= j;
       return v.d * 0x1p-108;
     }
+#endif /* ! USE_FMA_BUILTIN  */
 }
 #ifndef __fma
 libm_alias_double (__fma, fma)
diff --git a/sysdeps/ieee754/dbl-64/s_fmaf.c b/sysdeps/ieee754/dbl-64/s_fmaf.c
index 57329d0a87fe..93b8660d5242 100644
--- a/sysdeps/ieee754/dbl-64/s_fmaf.c
+++ b/sysdeps/ieee754/dbl-64/s_fmaf.c
@@ -23,6 +23,7 @@
 #include <math-barriers.h>
 #include <fenv_private.h>
 #include <libm-alias-float.h>
+#include <math-use-builtins.h>
 
 /* This implementation relies on double being more than twice as
    precise as float and uses rounding to odd in order to avoid problems
@@ -33,6 +34,10 @@
 float
 __fmaf (float x, float y, float z)
 {
+#if USE_FMAF_BUILTIN
+  return __builtin_fmaf (x, y, z);
+#else
+  /* Use generic implementation.  */
   fenv_t env;
 
   /* Multiplication is always exact.  */
@@ -60,6 +65,7 @@ __fmaf (float x, float y, float z)
 
   /* And finally truncation with round to nearest.  */
   return (float) u.d;
+#endif /* ! USE_FMAF_BUILTIN  */
 }
 #ifndef __fmaf
 libm_alias_float (__fma, fma)
diff --git a/sysdeps/ieee754/float128/float128_private.h b/sysdeps/ieee754/float128/float128_private.h
index f97463d9dc1b..ab6fc9f3c9cf 100644
--- a/sysdeps/ieee754/float128/float128_private.h
+++ b/sysdeps/ieee754/float128/float128_private.h
@@ -154,6 +154,8 @@
 #define USE_ROUNDL_BUILTIN USE_ROUNDF128_BUILTIN
 #undef USE_COPYSIGNL_BUILTIN
 #define USE_COPYSIGNL_BUILTIN USE_COPYSIGNF128_BUILTIN
+#undef USE_FMAL_BUILTIN
+#define USE_FMAL_BUILTIN USE_FMAF128_BUILTIN
 
 /* IEEE function renames.  */
 #define __ieee754_acoshl __ieee754_acoshf128
diff --git a/sysdeps/ieee754/ldbl-128/s_fmal.c b/sysdeps/ieee754/ldbl-128/s_fmal.c
index 7475015bcec6..a610499e47c7 100644
--- a/sysdeps/ieee754/ldbl-128/s_fmal.c
+++ b/sysdeps/ieee754/ldbl-128/s_fmal.c
@@ -25,6 +25,7 @@
 #include <math_private.h>
 #include <libm-alias-ldouble.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
 /* This implementation uses rounding to odd to avoid problems with
    double rounding.  See a paper by Boldo and Melquiond:
@@ -33,6 +34,9 @@
 _Float128
 __fmal (_Float128 x, _Float128 y, _Float128 z)
 {
+#if USE_FMAL_BUILTIN
+  return __builtin_fmal (x, y, z);
+#else
   union ieee854_long_double u, v, w;
   int adjust = 0;
   u.d = x;
@@ -296,5 +300,6 @@ __fmal (_Float128 x, _Float128 y, _Float128 z)
       v.ieee.mantissa3 |= j;
       return v.d * L(0x1p-228);
     }
+#endif /* ! USE_FMAL_BUILTIN  */
 }
 libm_alias_ldouble (__fma, fma)
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 16+ messages in thread

* [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions
  2020-06-02  0:35 [PATCH v2 0/4] Enable generic math code for more arches Vineet Gupta via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions Vineet Gupta via Libc-alpha
@ 2020-06-02  0:35 ` Vineet Gupta via Libc-alpha
  2020-06-02 17:31   ` Adhemerval Zanella via Libc-alpha
  2020-06-02  0:35 ` [PATCH v2 4/4] powerpc/fpu: use generic fma functions Vineet Gupta via Libc-alpha
  3 siblings, 1 reply; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02  0:35 UTC (permalink / raw)
  To: libc-alpha; +Cc: Vineet Gupta, linux-snps-arc

introduce sysdep header math-use-builtins.h to replace aarch64
impementations with corresponding generic ones

 - newly inroduced generic sqrt{,f}, fma{,f}
 - existing floor{,f}, nearbyint{,f}, rint{,f}, round{,f}, trunc{,f}
 - Note that generic copysign was already enabled (via generic
   math-use-builtins.h) now thru sysdep header

This is supposed to be a non functional change
Passes build-many for aarch64-linux-gnu
---
 sysdeps/aarch64/fpu/e_sqrt.c            | 27 ----------
 sysdeps/aarch64/fpu/e_sqrtf.c           | 27 ----------
 sysdeps/aarch64/fpu/math-use-builtins.h | 71 +++++++++++++++++++++++++
 sysdeps/aarch64/fpu/s_floor.c           | 29 ----------
 sysdeps/aarch64/fpu/s_floorf.c          | 29 ----------
 sysdeps/aarch64/fpu/s_fma.c             | 28 ----------
 sysdeps/aarch64/fpu/s_fmaf.c            | 28 ----------
 sysdeps/aarch64/fpu/s_nearbyint.c       | 28 ----------
 sysdeps/aarch64/fpu/s_nearbyintf.c      | 28 ----------
 sysdeps/aarch64/fpu/s_rint.c            | 29 ----------
 sysdeps/aarch64/fpu/s_rintf.c           | 29 ----------
 sysdeps/aarch64/fpu/s_round.c           | 29 ----------
 sysdeps/aarch64/fpu/s_roundf.c          | 29 ----------
 sysdeps/aarch64/fpu/s_trunc.c           | 29 ----------
 sysdeps/aarch64/fpu/s_truncf.c          | 29 ----------
 15 files changed, 71 insertions(+), 398 deletions(-)
 delete mode 100644 sysdeps/aarch64/fpu/e_sqrt.c
 delete mode 100644 sysdeps/aarch64/fpu/e_sqrtf.c
 create mode 100644 sysdeps/aarch64/fpu/math-use-builtins.h
 delete mode 100644 sysdeps/aarch64/fpu/s_floor.c
 delete mode 100644 sysdeps/aarch64/fpu/s_floorf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_fma.c
 delete mode 100644 sysdeps/aarch64/fpu/s_fmaf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_nearbyint.c
 delete mode 100644 sysdeps/aarch64/fpu/s_nearbyintf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_rint.c
 delete mode 100644 sysdeps/aarch64/fpu/s_rintf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_round.c
 delete mode 100644 sysdeps/aarch64/fpu/s_roundf.c
 delete mode 100644 sysdeps/aarch64/fpu/s_trunc.c
 delete mode 100644 sysdeps/aarch64/fpu/s_truncf.c

diff --git a/sysdeps/aarch64/fpu/e_sqrt.c b/sysdeps/aarch64/fpu/e_sqrt.c
deleted file mode 100644
index abb67ef7b061..000000000000
--- a/sysdeps/aarch64/fpu/e_sqrt.c
+++ /dev/null
@@ -1,27 +0,0 @@
-/* Square root of floating point number.
-   Copyright (C) 2015-2020 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/>.  */
-
-#include <math_private.h>
-#include <libm-alias-finite.h>
-
-double
-__ieee754_sqrt (double d)
-{
-  return __builtin_sqrt (d);
-}
-libm_alias_finite (__ieee754_sqrt, __sqrt)
diff --git a/sysdeps/aarch64/fpu/e_sqrtf.c b/sysdeps/aarch64/fpu/e_sqrtf.c
deleted file mode 100644
index 13008a4f45d6..000000000000
--- a/sysdeps/aarch64/fpu/e_sqrtf.c
+++ /dev/null
@@ -1,27 +0,0 @@
-/* Single-precision floating point square root.
-   Copyright (C) 2015-2020 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/>.  */
-
-#include <math_private.h>
-#include <libm-alias-finite.h>
-
-float
-__ieee754_sqrtf (float s)
-{
-  return __builtin_sqrtf (s);
-}
-libm_alias_finite (__ieee754_sqrtf, __sqrtf)
diff --git a/sysdeps/aarch64/fpu/math-use-builtins.h b/sysdeps/aarch64/fpu/math-use-builtins.h
new file mode 100644
index 000000000000..50a1ba0db5ea
--- /dev/null
+++ b/sysdeps/aarch64/fpu/math-use-builtins.h
@@ -0,0 +1,71 @@
+/* Using math gcc builtins instead of generic implementation.  aarch64 version.
+   Copyright (C) 2020 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/>.  */
+
+#ifndef MATH_USE_BUILTINS_H
+#define MATH_USE_BUILTINS_H	1
+
+#include <features.h> /* For __GNUC_PREREQ.  */
+
+/* Define these macros to 1 to use __builtin_xyz instead of the
+   generic implementation.  */
+#define USE_NEARBYINT_BUILTIN 1
+#define USE_NEARBYINTF_BUILTIN 1
+#define USE_NEARBYINTL_BUILTIN 0
+#define USE_NEARBYINTF128_BUILTIN 0
+
+#define USE_RINT_BUILTIN 1
+#define USE_RINTF_BUILTIN 1
+#define USE_RINTL_BUILTIN 0
+#define USE_RINTF128_BUILTIN 0
+
+#define USE_FLOOR_BUILTIN 1
+#define USE_FLOORF_BUILTIN 1
+#define USE_FLOORL_BUILTIN 0
+#define USE_FLOORF128_BUILTIN 0
+
+#define USE_CEIL_BUILTIN 1
+#define USE_CEILF_BUILTIN 1
+#define USE_CEILL_BUILTIN 0
+#define USE_CEILF128_BUILTIN 0
+
+#define USE_TRUNC_BUILTIN 1
+#define USE_TRUNCF_BUILTIN 1
+#define USE_TRUNCL_BUILTIN 0
+#define USE_TRUNCF128_BUILTIN 0
+
+#define USE_ROUND_BUILTIN 1
+#define USE_ROUNDF_BUILTIN 1
+#define USE_ROUNDL_BUILTIN 0
+#define USE_ROUNDF128_BUILTIN 0
+
+#define USE_COPYSIGNL_BUILTIN 1
+#if __GNUC_PREREQ (7, 0)
+# define USE_COPYSIGNF128_BUILTIN 1
+#else
+# define USE_COPYSIGNF128_BUILTIN 0
+#endif
+
+#define USE_SQRT_BUILTIN 1
+#define USE_SQRTF_BUILTIN 1
+
+#define USE_FMA_BUILTIN 1
+#define USE_FMAF_BUILTIN 1
+#define USE_FMAL_BUILTIN 0
+#define USE_FMAF128_BUILTIN 0
+
+#endif
diff --git a/sysdeps/aarch64/fpu/s_floor.c b/sysdeps/aarch64/fpu/s_floor.c
deleted file mode 100644
index d7d3e9336a6e..000000000000
--- a/sysdeps/aarch64/fpu/s_floor.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__floor (double x)
-{
-  return __builtin_floor (x);
-}
-
-libm_alias_double (__floor, floor)
diff --git a/sysdeps/aarch64/fpu/s_floorf.c b/sysdeps/aarch64/fpu/s_floorf.c
deleted file mode 100644
index 625ff4375702..000000000000
--- a/sysdeps/aarch64/fpu/s_floorf.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__floorf (float x)
-{
-  return __builtin_floorf (x);
-}
-
-libm_alias_float (__floor, floor)
diff --git a/sysdeps/aarch64/fpu/s_fma.c b/sysdeps/aarch64/fpu/s_fma.c
deleted file mode 100644
index 48bb40482dc9..000000000000
--- a/sysdeps/aarch64/fpu/s_fma.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Copyright (C) 1996-2020 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/>.  */
-
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__fma (double x, double y, double z)
-{
-  return __builtin_fma (x, y, z);
-}
-
-libm_alias_double (__fma, fma)
diff --git a/sysdeps/aarch64/fpu/s_fmaf.c b/sysdeps/aarch64/fpu/s_fmaf.c
deleted file mode 100644
index 544f32e27aec..000000000000
--- a/sysdeps/aarch64/fpu/s_fmaf.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Copyright (C) 2011-2020 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/>.  */
-
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__fmaf (float x, float y, float z)
-{
-  return __builtin_fmaf (x, y, z);
-}
-
-libm_alias_float (__fma, fma)
diff --git a/sysdeps/aarch64/fpu/s_nearbyint.c b/sysdeps/aarch64/fpu/s_nearbyint.c
deleted file mode 100644
index 59a57dd39ba1..000000000000
--- a/sysdeps/aarch64/fpu/s_nearbyint.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Copyright (C) 2011-2020 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/>.  */
-
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__nearbyint (double x)
-{
-  return __builtin_nearbyint (x);
-}
-
-libm_alias_double (__nearbyint, nearbyint)
diff --git a/sysdeps/aarch64/fpu/s_nearbyintf.c b/sysdeps/aarch64/fpu/s_nearbyintf.c
deleted file mode 100644
index f52968ae2bbd..000000000000
--- a/sysdeps/aarch64/fpu/s_nearbyintf.c
+++ /dev/null
@@ -1,28 +0,0 @@
-/* Copyright (C) 2011-2020 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/>.  */
-
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__nearbyintf (float x)
-{
-  return __builtin_nearbyintf (x);
-}
-
-libm_alias_float (__nearbyint, nearbyint)
diff --git a/sysdeps/aarch64/fpu/s_rint.c b/sysdeps/aarch64/fpu/s_rint.c
deleted file mode 100644
index bdd757167e0c..000000000000
--- a/sysdeps/aarch64/fpu/s_rint.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__rint (double x)
-{
-  return __builtin_rint (x);
-}
-
-libm_alias_double (__rint, rint)
diff --git a/sysdeps/aarch64/fpu/s_rintf.c b/sysdeps/aarch64/fpu/s_rintf.c
deleted file mode 100644
index 8cf7c63d6dda..000000000000
--- a/sysdeps/aarch64/fpu/s_rintf.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__rintf (float x)
-{
-  return __builtin_rintf (x);
-}
-
-libm_alias_float (__rint, rint)
diff --git a/sysdeps/aarch64/fpu/s_round.c b/sysdeps/aarch64/fpu/s_round.c
deleted file mode 100644
index d2fd5f34117a..000000000000
--- a/sysdeps/aarch64/fpu/s_round.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__round (double x)
-{
-  return __builtin_round (x);
-}
-
-libm_alias_double (__round, round)
diff --git a/sysdeps/aarch64/fpu/s_roundf.c b/sysdeps/aarch64/fpu/s_roundf.c
deleted file mode 100644
index e51214e86cd2..000000000000
--- a/sysdeps/aarch64/fpu/s_roundf.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__roundf (float x)
-{
-  return __builtin_roundf (x);
-}
-
-libm_alias_float (__round, round)
diff --git a/sysdeps/aarch64/fpu/s_trunc.c b/sysdeps/aarch64/fpu/s_trunc.c
deleted file mode 100644
index bac23a5e269d..000000000000
--- a/sysdeps/aarch64/fpu/s_trunc.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__trunc (double x)
-{
-  return __builtin_trunc (x);
-}
-
-libm_alias_double (__trunc, trunc)
diff --git a/sysdeps/aarch64/fpu/s_truncf.c b/sysdeps/aarch64/fpu/s_truncf.c
deleted file mode 100644
index 6a025c084a5a..000000000000
--- a/sysdeps/aarch64/fpu/s_truncf.c
+++ /dev/null
@@ -1,29 +0,0 @@
-/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__truncf (float x)
-{
-  return __builtin_truncf (x);
-}
-
-libm_alias_float (__trunc, trunc)
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 16+ messages in thread

* [PATCH v2 4/4] powerpc/fpu: use generic fma functions
  2020-06-02  0:35 [PATCH v2 0/4] Enable generic math code for more arches Vineet Gupta via Libc-alpha
                   ` (2 preceding siblings ...)
  2020-06-02  0:35 ` [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions Vineet Gupta via Libc-alpha
@ 2020-06-02  0:35 ` Vineet Gupta via Libc-alpha
  3 siblings, 0 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02  0:35 UTC (permalink / raw)
  To: libc-alpha; +Cc: Vineet Gupta, linux-snps-arc

This is supposed to be a non functional change
Passes build-many for powerpc-linux-gnu

Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
---
 sysdeps/powerpc/fpu/math-use-builtins.h | 69 +++++++++++++++++++++++++
 sysdeps/powerpc/fpu/s_fma.c             | 27 ----------
 sysdeps/powerpc/fpu/s_fmaf.c            | 27 ----------
 3 files changed, 69 insertions(+), 54 deletions(-)
 create mode 100644 sysdeps/powerpc/fpu/math-use-builtins.h
 delete mode 100644 sysdeps/powerpc/fpu/s_fma.c
 delete mode 100644 sysdeps/powerpc/fpu/s_fmaf.c

diff --git a/sysdeps/powerpc/fpu/math-use-builtins.h b/sysdeps/powerpc/fpu/math-use-builtins.h
new file mode 100644
index 000000000000..4780934379ba
--- /dev/null
+++ b/sysdeps/powerpc/fpu/math-use-builtins.h
@@ -0,0 +1,69 @@
+/* Using math gcc builtins instead of generic implementation.  PowerPC version.
+   Copyright (C) 2019-2020 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/>.  */
+
+#ifndef MATH_USE_BUILTINS_H
+#define MATH_USE_BUILTINS_H	1
+
+#include <features.h> /* For __GNUC_PREREQ.  */
+
+/* Define these macros to 1 to use __builtin_xyz instead of the
+   generic implementation.  */
+#define USE_NEARBYINT_BUILTIN 0
+#define USE_NEARBYINTF_BUILTIN 0
+#define USE_NEARBYINTL_BUILTIN 0
+#define USE_NEARBYINTF128_BUILTIN 0
+
+#define USE_RINT_BUILTIN 0
+#define USE_RINTF_BUILTIN 0
+#define USE_RINTL_BUILTIN 0
+#define USE_RINTF128_BUILTIN 0
+
+#define USE_FLOOR_BUILTIN 0
+#define USE_FLOORF_BUILTIN 0
+#define USE_FLOORL_BUILTIN 0
+#define USE_FLOORF128_BUILTIN 0
+
+#define USE_CEIL_BUILTIN 0
+#define USE_CEILF_BUILTIN 0
+#define USE_CEILL_BUILTIN 0
+#define USE_CEILF128_BUILTIN 0
+
+#define USE_TRUNC_BUILTIN 0
+#define USE_TRUNCF_BUILTIN 0
+#define USE_TRUNCL_BUILTIN 0
+#define USE_TRUNCF128_BUILTIN 0
+
+#define USE_ROUND_BUILTIN 0
+#define USE_ROUNDF_BUILTIN 0
+#define USE_ROUNDL_BUILTIN 0
+#define USE_ROUNDF128_BUILTIN 0
+
+#define USE_COPYSIGNL_BUILTIN 1
+#if __GNUC_PREREQ (7, 0)
+# define USE_COPYSIGNF128_BUILTIN 1
+#else
+# define USE_COPYSIGNF128_BUILTIN 0
+#endif
+
+#define USE_SQRT_BUILTIN 0
+#define USE_SQRTF_BUILTIN 0
+
+#define USE_FMA_BUILTIN 1
+#define USE_FMAF_BUILTIN 1
+
+#endif
diff --git a/sysdeps/powerpc/fpu/s_fma.c b/sysdeps/powerpc/fpu/s_fma.c
deleted file mode 100644
index 9ddd13253485..000000000000
--- a/sysdeps/powerpc/fpu/s_fma.c
+++ /dev/null
@@ -1,27 +0,0 @@
-/* Compute x * y + z as ternary operation.  PowerPC version.
-   Copyright (C) 2019-2020 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/>.  */
-
-#include <sysdep.h>
-#include <libm-alias-double.h>
-
-double
-__fma (double x, double y, double z)
-{
-  return __builtin_fma (x, y, z);
-}
-libm_alias_double (__fma, fma)
diff --git a/sysdeps/powerpc/fpu/s_fmaf.c b/sysdeps/powerpc/fpu/s_fmaf.c
deleted file mode 100644
index d132c20125bb..000000000000
--- a/sysdeps/powerpc/fpu/s_fmaf.c
+++ /dev/null
@@ -1,27 +0,0 @@
-/* Compute x * y + z as ternary operation.  PowerPC version.
-   Copyright (C) 2019-2020 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/>.  */
-
-#include <sysdep.h>
-#include <libm-alias-float.h>
-
-float
-__fmaf (float x, float y, float z)
-{
-  return __builtin_fmaf (x, y, z);
-}
-libm_alias_float (__fma, fma)
-- 
2.20.1


^ permalink raw reply related	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02  0:35 ` [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions Vineet Gupta via Libc-alpha
@ 2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
  2020-06-02 17:13     ` Vineet Gupta via Libc-alpha
  2020-06-02 17:28   ` Adhemerval Zanella via Libc-alpha
  1 sibling, 1 reply; 16+ messages in thread
From: Stefan Liebler via Libc-alpha @ 2020-06-02 12:51 UTC (permalink / raw)
  To: libc-alpha

On 6/2/20 2:35 AM, Vineet Gupta via Libc-alpha wrote:
> ---
>  sysdeps/generic/math-use-builtins.h         | 5 +++++
>  sysdeps/ieee754/dbl-64/s_fma.c              | 6 ++++++
>  sysdeps/ieee754/dbl-64/s_fmaf.c             | 6 ++++++
>  sysdeps/ieee754/float128/float128_private.h | 2 ++
>  sysdeps/ieee754/ldbl-128/s_fmal.c           | 5 +++++
>  5 files changed, 24 insertions(+)
> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index fc724c824a17..cf25ed8a2138 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -63,4 +63,9 @@
>  #define USE_SQRT_BUILTIN 0
>  #define USE_SQRTF_BUILTIN 0
> 
> +#define USE_FMA_BUILTIN 0
> +#define USE_FMAF_BUILTIN 0
> +#define USE_FMAL_BUILTIN 0
> +#define USE_FMAF128_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */
Please also update the current architecture specific math-use-builtins.h
file: sysdeps/s390/fpu/math-use-builtins.h
Otherwise it will break build on s390x.

> diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
> index 876df6e78bdc..9dc5b132b9ee 100644
> --- a/sysdeps/ieee754/dbl-64/s_fma.c
> +++ b/sysdeps/ieee754/dbl-64/s_fma.c
> @@ -25,6 +25,7 @@
>  #include <fenv_private.h>
>  #include <libm-alias-double.h>
>  #include <tininess.h>
> +#include <math-use-builtins.h>
> 
>  /* This implementation uses rounding to odd to avoid problems with
>     double rounding.  See a paper by Boldo and Melquiond:
> @@ -33,6 +34,10 @@
>  double
>  __fma (double x, double y, double z)
>  {
> +#if USE_FMA_BUILTIN
> +  return __builtin_fma (x, y, z);
Architectures which have support for ldbl-128 will use the file
sysdeps/ieee754/ldbl-128/s_fma.c instead of
sysdeps/ieee754/dbl-64/s_fma.c. Should this file also be adjusted in
order to use the builtin if USE_FMA_BUILTIN is set to one?
> +#else
> +  /* Use generic implementation.  */
>    union ieee754_double u, v, w;
>    int adjust = 0;
>    u.d = x;
> @@ -292,6 +297,7 @@ __fma (double x, double y, double z)
>        v.ieee.mantissa1 |= j;
>        return v.d * 0x1p-108;
>      }
> +#endif /* ! USE_FMA_BUILTIN  */
>  }
>  #ifndef __fma
>  libm_alias_double (__fma, fma)
> diff --git a/sysdeps/ieee754/dbl-64/s_fmaf.c b/sysdeps/ieee754/dbl-64/s_fmaf.c
> index 57329d0a87fe..93b8660d5242 100644
> --- a/sysdeps/ieee754/dbl-64/s_fmaf.c
> +++ b/sysdeps/ieee754/dbl-64/s_fmaf.c
> @@ -23,6 +23,7 @@
>  #include <math-barriers.h>
>  #include <fenv_private.h>
>  #include <libm-alias-float.h>
> +#include <math-use-builtins.h>
> 
>  /* This implementation relies on double being more than twice as
>     precise as float and uses rounding to odd in order to avoid problems
> @@ -33,6 +34,10 @@
>  float
>  __fmaf (float x, float y, float z)
>  {
> +#if USE_FMAF_BUILTIN
> +  return __builtin_fmaf (x, y, z);
> +#else
> +  /* Use generic implementation.  */
>    fenv_t env;
> 
>    /* Multiplication is always exact.  */
> @@ -60,6 +65,7 @@ __fmaf (float x, float y, float z)
> 
>    /* And finally truncation with round to nearest.  */
>    return (float) u.d;
> +#endif /* ! USE_FMAF_BUILTIN  */
>  }
>  #ifndef __fmaf
>  libm_alias_float (__fma, fma)
> diff --git a/sysdeps/ieee754/float128/float128_private.h b/sysdeps/ieee754/float128/float128_private.h
> index f97463d9dc1b..ab6fc9f3c9cf 100644
> --- a/sysdeps/ieee754/float128/float128_private.h
> +++ b/sysdeps/ieee754/float128/float128_private.h
> @@ -154,6 +154,8 @@
>  #define USE_ROUNDL_BUILTIN USE_ROUNDF128_BUILTIN
>  #undef USE_COPYSIGNL_BUILTIN
>  #define USE_COPYSIGNL_BUILTIN USE_COPYSIGNF128_BUILTIN
> +#undef USE_FMAL_BUILTIN
> +#define USE_FMAL_BUILTIN USE_FMAF128_BUILTIN
> 
>  /* IEEE function renames.  */
>  #define __ieee754_acoshl __ieee754_acoshf128
> diff --git a/sysdeps/ieee754/ldbl-128/s_fmal.c b/sysdeps/ieee754/ldbl-128/s_fmal.c
> index 7475015bcec6..a610499e47c7 100644
> --- a/sysdeps/ieee754/ldbl-128/s_fmal.c
> +++ b/sysdeps/ieee754/ldbl-128/s_fmal.c
> @@ -25,6 +25,7 @@
>  #include <math_private.h>
>  #include <libm-alias-ldouble.h>
>  #include <tininess.h>
> +#include <math-use-builtins.h>
> 
>  /* This implementation uses rounding to odd to avoid problems with
>     double rounding.  See a paper by Boldo and Melquiond:
> @@ -33,6 +34,9 @@
>  _Float128
>  __fmal (_Float128 x, _Float128 y, _Float128 z)
>  {
> +#if USE_FMAL_BUILTIN
> +  return __builtin_fmal (x, y, z);
> +#else
>    union ieee854_long_double u, v, w;
>    int adjust = 0;
>    u.d = x;
> @@ -296,5 +300,6 @@ __fmal (_Float128 x, _Float128 y, _Float128 z)
>        v.ieee.mantissa3 |= j;
>        return v.d * L(0x1p-228);
>      }
> +#endif /* ! USE_FMAL_BUILTIN  */
>  }
>  libm_alias_ldouble (__fma, fma)
> 


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions
  2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
@ 2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
  2020-06-02 16:54     ` Vineet Gupta via Libc-alpha
  2020-06-02 17:17   ` Adhemerval Zanella via Libc-alpha
  1 sibling, 1 reply; 16+ messages in thread
From: Stefan Liebler via Libc-alpha @ 2020-06-02 12:51 UTC (permalink / raw)
  To: libc-alpha

On 6/2/20 2:35 AM, Vineet Gupta via Libc-alpha wrote:
> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
> ---
>  sysdeps/generic/math-use-builtins.h |  3 +++
>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>  3 files changed, 19 insertions(+), 6 deletions(-)
> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index 8a39ef58bc95..fc724c824a17 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -60,4 +60,7 @@
>  # define USE_COPYSIGNF128_BUILTIN 0
>  #endif
> 
> +#define USE_SQRT_BUILTIN 0
> +#define USE_SQRTF_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */
Please also update the current architecture specific math-use-builtins.h
file: sysdeps/s390/fpu/math-use-builtins.h

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
@ 2020-06-02 16:54     ` Vineet Gupta via Libc-alpha
  0 siblings, 0 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02 16:54 UTC (permalink / raw)
  To: Stefan Liebler, libc-alpha@sourceware.org; +Cc: arcml

On 6/2/20 5:51 AM, Stefan Liebler via Libc-alpha wrote:
> On 6/2/20 2:35 AM, Vineet Gupta via Libc-alpha wrote:
>> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>
>> ---
>>  sysdeps/generic/math-use-builtins.h |  3 +++
>>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>>  3 files changed, 19 insertions(+), 6 deletions(-)
>>
>> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
>> index 8a39ef58bc95..fc724c824a17 100644
>> --- a/sysdeps/generic/math-use-builtins.h
>> +++ b/sysdeps/generic/math-use-builtins.h
>> @@ -60,4 +60,7 @@
>>  # define USE_COPYSIGNF128_BUILTIN 0
>>  #endif
>>
>> +#define USE_SQRT_BUILTIN 0
>> +#define USE_SQRTF_BUILTIN 0
>> +
>>  #endif /* math-use-builtins.h */
> Please also update the current architecture specific math-use-builtins.h
> file: sysdeps/s390/fpu/math-use-builtins.h

Fixed. I should have added s390 to my list of arches to test.

Thx for taking a look.

-Vineet



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
@ 2020-06-02 17:13     ` Vineet Gupta via Libc-alpha
  2020-06-02 17:27       ` Adhemerval Zanella via Libc-alpha
  2020-06-03  6:29       ` Stefan Liebler via Libc-alpha
  0 siblings, 2 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02 17:13 UTC (permalink / raw)
  To: Stefan Liebler, libc-alpha@sourceware.org; +Cc: arcml

On 6/2/20 5:51 AM, Stefan Liebler via Libc-alpha wrote:
>  #endif /* math-use-builtins.h */
> Please also update the current architecture specific math-use-builtins.h
> file: sysdeps/s390/fpu/math-use-builtins.h
> Otherwise it will break build on s390x.

Done.

>> diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
>> index 876df6e78bdc..9dc5b132b9ee 100644
>> --- a/sysdeps/ieee754/dbl-64/s_fma.c
>> +++ b/sysdeps/ieee754/dbl-64/s_fma.c
>> @@ -25,6 +25,7 @@
>>  #include <fenv_private.h>
>>  #include <libm-alias-double.h>
>>  #include <tininess.h>
>> +#include <math-use-builtins.h>
>>
>>  /* This implementation uses rounding to odd to avoid problems with
>>     double rounding.  See a paper by Boldo and Melquiond:
>> @@ -33,6 +34,10 @@
>>  double
>>  __fma (double x, double y, double z)
>>  {
>> +#if USE_FMA_BUILTIN
>> +  return __builtin_fma (x, y, z);
>
> Architectures which have support for ldbl-128 will use the file
> sysdeps/ieee754/ldbl-128/s_fma.c instead of
> sysdeps/ieee754/dbl-64/s_fma.c. Should this file also be adjusted in
> order to use the builtin if USE_FMA_BUILTIN is set to one?

Right.

I used commit f82996f815 "Use GCC builtins for round functions if desired" as
starting point for my change. And seems it was not an ideal reference :-) as round
has far fewer instances than fma. Indeed fma is present in ldbl-128 and dbl-64 so
needs updating in both.

But just to be sure s390 is currently not using the newly introduced builtins so
I'll keep them as follows.

#define USE_SQRT_BUILTIN 0
#define USE_SQRTF_BUILTIN 0

#define USE_FMA_BUILTIN 0
#define USE_FMAF_BUILTIN 0
#define USE_FMAL_BUILTIN 0
#define USE_FMAF128_BUILTIN 0

-Vineet

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions
  2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
@ 2020-06-02 17:17   ` Adhemerval Zanella via Libc-alpha
  1 sibling, 0 replies; 16+ messages in thread
From: Adhemerval Zanella via Libc-alpha @ 2020-06-02 17:17 UTC (permalink / raw)
  To: Vineet Gupta, libc-alpha; +Cc: linux-snps-arc



On 01/06/2020 21:35, Vineet Gupta wrote:
> Reviewed-by: Adhemerval Zanella  <adhemerval.zanella@linaro.org>

LGTM with the small nit below and the s390 fix pointed by Stefan.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>

> ---
>  sysdeps/generic/math-use-builtins.h |  3 +++
>  sysdeps/ieee754/dbl-64/e_sqrt.c     |  6 ++++++
>  sysdeps/ieee754/flt-32/e_sqrtf.c    | 16 ++++++++++------
>  3 files changed, 19 insertions(+), 6 deletions(-)
> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index 8a39ef58bc95..fc724c824a17 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -60,4 +60,7 @@
>  # define USE_COPYSIGNF128_BUILTIN 0
>  #endif
>  
> +#define USE_SQRT_BUILTIN 0
> +#define USE_SQRTF_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */

Ok.

> diff --git a/sysdeps/ieee754/dbl-64/e_sqrt.c b/sysdeps/ieee754/dbl-64/e_sqrt.c
> index d42a1a4eb6e9..518a8ae5cdaf 100644
> --- a/sysdeps/ieee754/dbl-64/e_sqrt.c
> +++ b/sysdeps/ieee754/dbl-64/e_sqrt.c
> @@ -41,6 +41,7 @@
>  #include <math_private.h>
>  #include <fenv_private.h>
>  #include <libm-alias-finite.h>
> +#include <math-use-builtins.h>
>  
>  /*********************************************************************/
>  /* An ultimate sqrt routine. Given an IEEE double machine number x   */
> @@ -50,6 +51,10 @@
>  double
>  __ieee754_sqrt (double x)
>  {
> +#if USE_SQRT_BUILTIN
> +  return __builtin_sqrt (x);
> +#else
> +  /* Use generic implementation.  */
>    static const double
>      rt0 = 9.99999999859990725855365213134618E-01,
>      rt1 = 4.99999999495955425917856814202739E-01,
> @@ -138,6 +143,7 @@ __ieee754_sqrt (double x)
>  	return (x - x) / (x - x); /* sqrt(-ve)=sNaN */
>        return 0x1p-256 * __ieee754_sqrt (x * 0x1p512);
>      }
> +#endif /* ! USE_SQRT_BUILTIN  */
>  }
>  #ifndef __ieee754_sqrt
>  libm_alias_finite (__ieee754_sqrt, __sqrt)

Ok.

> diff --git a/sysdeps/ieee754/flt-32/e_sqrtf.c b/sysdeps/ieee754/flt-32/e_sqrtf.c
> index b339444301aa..d85a04162983 100644
> --- a/sysdeps/ieee754/flt-32/e_sqrtf.c
> +++ b/sysdeps/ieee754/flt-32/e_sqrtf.c
> @@ -16,12 +16,15 @@
>  #include <math.h>
>  #include <math_private.h>
>  #include <libm-alias-finite.h>
> -
> -static	const float	one	= 1.0, tiny=1.0e-30;
> +#include <math-use-builtins.h>
>  
>  float
>  __ieee754_sqrtf(float x)
>  {
> +#if USE_SQRTF_BUILTIN
> +	return __builtin_sqrtf (x);
> +#else
> +	/* Use generic implementation.  */
>  	float z;
>  	int32_t sign = (int)0x80000000;
>  	int32_t ix,s,q,m,t,i;
> @@ -70,10 +73,10 @@ __ieee754_sqrtf(float x)
>  
>      /* use floating add to find out rounding direction */
>  	if(ix!=0) {
> -	    z = one-tiny; /* trigger inexact flag */
> -	    if (z>=one) {
> -		z = one+tiny;
> -		if (z>one)
> +	    z = 0x1p0 - 0x1.4484cp-100; /* trigger inexact flag */

Period and double space before '*/'.

> +	    if (z >= 0x1p0) {
> +		z = 0x1p0 + 0x1.4484cp-100;
> +		if (z > 0x1p0)
>  		    q += 2;
>  		else
>  		    q += (q&1);
> @@ -83,6 +86,7 @@ __ieee754_sqrtf(float x)
>  	ix += (m <<23);
>  	SET_FLOAT_WORD(z,ix);
>  	return z;
> +#endif /* ! USE_SQRTF_BUILTIN  */
>  }
>  #ifndef __ieee754_sqrtf
>  libm_alias_finite (__ieee754_sqrtf, __sqrtf)
> 

Ok.

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02 17:13     ` Vineet Gupta via Libc-alpha
@ 2020-06-02 17:27       ` Adhemerval Zanella via Libc-alpha
  2020-06-03  0:20         ` Vineet Gupta via Libc-alpha
  2020-06-03  6:29       ` Stefan Liebler via Libc-alpha
  1 sibling, 1 reply; 16+ messages in thread
From: Adhemerval Zanella via Libc-alpha @ 2020-06-02 17:27 UTC (permalink / raw)
  To: libc-alpha



On 02/06/2020 14:13, Vineet Gupta via Libc-alpha wrote:
> On 6/2/20 5:51 AM, Stefan Liebler via Libc-alpha wrote:
>>  #endif /* math-use-builtins.h */
>> Please also update the current architecture specific math-use-builtins.h
>> file: sysdeps/s390/fpu/math-use-builtins.h
>> Otherwise it will break build on s390x.
> 
> Done.
> 
>>> diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
>>> index 876df6e78bdc..9dc5b132b9ee 100644
>>> --- a/sysdeps/ieee754/dbl-64/s_fma.c
>>> +++ b/sysdeps/ieee754/dbl-64/s_fma.c
>>> @@ -25,6 +25,7 @@
>>>  #include <fenv_private.h>
>>>  #include <libm-alias-double.h>
>>>  #include <tininess.h>
>>> +#include <math-use-builtins.h>
>>>
>>>  /* This implementation uses rounding to odd to avoid problems with
>>>     double rounding.  See a paper by Boldo and Melquiond:
>>> @@ -33,6 +34,10 @@
>>>  double
>>>  __fma (double x, double y, double z)
>>>  {
>>> +#if USE_FMA_BUILTIN
>>> +  return __builtin_fma (x, y, z);
>>
>> Architectures which have support for ldbl-128 will use the file
>> sysdeps/ieee754/ldbl-128/s_fma.c instead of
>> sysdeps/ieee754/dbl-64/s_fma.c. Should this file also be adjusted in
>> order to use the builtin if USE_FMA_BUILTIN is set to one?
> 
> Right.
> 
> I used commit f82996f815 "Use GCC builtins for round functions if desired" as
> starting point for my change. And seems it was not an ideal reference :-) as round
> has far fewer instances than fma. Indeed fma is present in ldbl-128 and dbl-64 so
> needs updating in both.

I think after this set is upstream I will refactor to make each symbol and its
variant (i.g sqrt, sqrtf, etc.) to be define on its own file.  The default
math-use-builtins will then include each file:

  /* math-use-builtins.h  */
  #inclde <math-use-builtins-sqrt.h>
  [...]

With default USE_* begin 0. This would allow to add new builtin usage without 
require to sync with every architecture (and without breaking its build as well).

> 
> But just to be sure s390 is currently not using the newly introduced builtins so
> I'll keep them as follows.
> 
> #define USE_SQRT_BUILTIN 0
> #define USE_SQRTF_BUILTIN 0
> 
> #define USE_FMA_BUILTIN 0
> #define USE_FMAF_BUILTIN 0
> #define USE_FMAL_BUILTIN 0
> #define USE_FMAF128_BUILTIN 0
> 
> -Vineet

It is ok, we can optimize s390 in a subsequent patch (since it does provide fma
and sqrtf builtins).


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02  0:35 ` [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions Vineet Gupta via Libc-alpha
  2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
@ 2020-06-02 17:28   ` Adhemerval Zanella via Libc-alpha
  1 sibling, 0 replies; 16+ messages in thread
From: Adhemerval Zanella via Libc-alpha @ 2020-06-02 17:28 UTC (permalink / raw)
  To: Vineet Gupta, libc-alpha; +Cc: linux-snps-arc



On 01/06/2020 21:35, Vineet Gupta wrote:
> ---
>  sysdeps/generic/math-use-builtins.h         | 5 +++++
>  sysdeps/ieee754/dbl-64/s_fma.c              | 6 ++++++
>  sysdeps/ieee754/dbl-64/s_fmaf.c             | 6 ++++++
>  sysdeps/ieee754/float128/float128_private.h | 2 ++
>  sysdeps/ieee754/ldbl-128/s_fmal.c           | 5 +++++
>  5 files changed, 24 insertions(+)


LGTM with s390 fix pointed by Stefan.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>

> 
> diff --git a/sysdeps/generic/math-use-builtins.h b/sysdeps/generic/math-use-builtins.h
> index fc724c824a17..cf25ed8a2138 100644
> --- a/sysdeps/generic/math-use-builtins.h
> +++ b/sysdeps/generic/math-use-builtins.h
> @@ -63,4 +63,9 @@
>  #define USE_SQRT_BUILTIN 0
>  #define USE_SQRTF_BUILTIN 0
>  
> +#define USE_FMA_BUILTIN 0
> +#define USE_FMAF_BUILTIN 0
> +#define USE_FMAL_BUILTIN 0
> +#define USE_FMAF128_BUILTIN 0
> +
>  #endif /* math-use-builtins.h */
> diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
> index 876df6e78bdc..9dc5b132b9ee 100644
> --- a/sysdeps/ieee754/dbl-64/s_fma.c
> +++ b/sysdeps/ieee754/dbl-64/s_fma.c
> @@ -25,6 +25,7 @@
>  #include <fenv_private.h>
>  #include <libm-alias-double.h>
>  #include <tininess.h>
> +#include <math-use-builtins.h>
>  
>  /* This implementation uses rounding to odd to avoid problems with
>     double rounding.  See a paper by Boldo and Melquiond:
> @@ -33,6 +34,10 @@
>  double
>  __fma (double x, double y, double z)
>  {
> +#if USE_FMA_BUILTIN
> +  return __builtin_fma (x, y, z);
> +#else
> +  /* Use generic implementation.  */
>    union ieee754_double u, v, w;
>    int adjust = 0;
>    u.d = x;
> @@ -292,6 +297,7 @@ __fma (double x, double y, double z)
>        v.ieee.mantissa1 |= j;
>        return v.d * 0x1p-108;
>      }
> +#endif /* ! USE_FMA_BUILTIN  */
>  }
>  #ifndef __fma
>  libm_alias_double (__fma, fma)
> diff --git a/sysdeps/ieee754/dbl-64/s_fmaf.c b/sysdeps/ieee754/dbl-64/s_fmaf.c
> index 57329d0a87fe..93b8660d5242 100644
> --- a/sysdeps/ieee754/dbl-64/s_fmaf.c
> +++ b/sysdeps/ieee754/dbl-64/s_fmaf.c
> @@ -23,6 +23,7 @@
>  #include <math-barriers.h>
>  #include <fenv_private.h>
>  #include <libm-alias-float.h>
> +#include <math-use-builtins.h>
>  
>  /* This implementation relies on double being more than twice as
>     precise as float and uses rounding to odd in order to avoid problems
> @@ -33,6 +34,10 @@
>  float
>  __fmaf (float x, float y, float z)
>  {
> +#if USE_FMAF_BUILTIN
> +  return __builtin_fmaf (x, y, z);
> +#else
> +  /* Use generic implementation.  */
>    fenv_t env;
>  
>    /* Multiplication is always exact.  */
> @@ -60,6 +65,7 @@ __fmaf (float x, float y, float z)
>  
>    /* And finally truncation with round to nearest.  */
>    return (float) u.d;
> +#endif /* ! USE_FMAF_BUILTIN  */
>  }
>  #ifndef __fmaf
>  libm_alias_float (__fma, fma)
> diff --git a/sysdeps/ieee754/float128/float128_private.h b/sysdeps/ieee754/float128/float128_private.h
> index f97463d9dc1b..ab6fc9f3c9cf 100644
> --- a/sysdeps/ieee754/float128/float128_private.h
> +++ b/sysdeps/ieee754/float128/float128_private.h
> @@ -154,6 +154,8 @@
>  #define USE_ROUNDL_BUILTIN USE_ROUNDF128_BUILTIN
>  #undef USE_COPYSIGNL_BUILTIN
>  #define USE_COPYSIGNL_BUILTIN USE_COPYSIGNF128_BUILTIN
> +#undef USE_FMAL_BUILTIN
> +#define USE_FMAL_BUILTIN USE_FMAF128_BUILTIN
>  
>  /* IEEE function renames.  */
>  #define __ieee754_acoshl __ieee754_acoshf128
> diff --git a/sysdeps/ieee754/ldbl-128/s_fmal.c b/sysdeps/ieee754/ldbl-128/s_fmal.c
> index 7475015bcec6..a610499e47c7 100644
> --- a/sysdeps/ieee754/ldbl-128/s_fmal.c
> +++ b/sysdeps/ieee754/ldbl-128/s_fmal.c
> @@ -25,6 +25,7 @@
>  #include <math_private.h>
>  #include <libm-alias-ldouble.h>
>  #include <tininess.h>
> +#include <math-use-builtins.h>
>  
>  /* This implementation uses rounding to odd to avoid problems with
>     double rounding.  See a paper by Boldo and Melquiond:
> @@ -33,6 +34,9 @@
>  _Float128
>  __fmal (_Float128 x, _Float128 y, _Float128 z)
>  {
> +#if USE_FMAL_BUILTIN
> +  return __builtin_fmal (x, y, z);
> +#else
>    union ieee854_long_double u, v, w;
>    int adjust = 0;
>    u.d = x;
> @@ -296,5 +300,6 @@ __fmal (_Float128 x, _Float128 y, _Float128 z)
>        v.ieee.mantissa3 |= j;
>        return v.d * L(0x1p-228);
>      }
> +#endif /* ! USE_FMAL_BUILTIN  */
>  }
>  libm_alias_ldouble (__fma, fma)
> 

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions
  2020-06-02  0:35 ` [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions Vineet Gupta via Libc-alpha
@ 2020-06-02 17:31   ` Adhemerval Zanella via Libc-alpha
  2020-06-02 19:14     ` Vineet Gupta via Libc-alpha
  0 siblings, 1 reply; 16+ messages in thread
From: Adhemerval Zanella via Libc-alpha @ 2020-06-02 17:31 UTC (permalink / raw)
  To: Vineet Gupta, libc-alpha; +Cc: linux-snps-arc



On 01/06/2020 21:35, Vineet Gupta wrote:
> introduce sysdep header math-use-builtins.h to replace aarch64
> impementations with corresponding generic ones

s/impementations/implementations and missing ':'.

> 
>  - newly inroduced generic sqrt{,f}, fma{,f}
>  - existing floor{,f}, nearbyint{,f}, rint{,f}, round{,f}, trunc{,f}
>  - Note that generic copysign was already enabled (via generic
>    math-use-builtins.h) now thru sysdep header
> 
> This is supposed to be a non functional change
> Passes build-many for aarch64-linux-gnu

LGTM (just double check if the objects generated by the removed 
implementations did not changed).

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>

> ---
>  sysdeps/aarch64/fpu/e_sqrt.c            | 27 ----------
>  sysdeps/aarch64/fpu/e_sqrtf.c           | 27 ----------
>  sysdeps/aarch64/fpu/math-use-builtins.h | 71 +++++++++++++++++++++++++
>  sysdeps/aarch64/fpu/s_floor.c           | 29 ----------
>  sysdeps/aarch64/fpu/s_floorf.c          | 29 ----------
>  sysdeps/aarch64/fpu/s_fma.c             | 28 ----------
>  sysdeps/aarch64/fpu/s_fmaf.c            | 28 ----------
>  sysdeps/aarch64/fpu/s_nearbyint.c       | 28 ----------
>  sysdeps/aarch64/fpu/s_nearbyintf.c      | 28 ----------
>  sysdeps/aarch64/fpu/s_rint.c            | 29 ----------
>  sysdeps/aarch64/fpu/s_rintf.c           | 29 ----------
>  sysdeps/aarch64/fpu/s_round.c           | 29 ----------
>  sysdeps/aarch64/fpu/s_roundf.c          | 29 ----------
>  sysdeps/aarch64/fpu/s_trunc.c           | 29 ----------
>  sysdeps/aarch64/fpu/s_truncf.c          | 29 ----------
>  15 files changed, 71 insertions(+), 398 deletions(-)
>  delete mode 100644 sysdeps/aarch64/fpu/e_sqrt.c
>  delete mode 100644 sysdeps/aarch64/fpu/e_sqrtf.c
>  create mode 100644 sysdeps/aarch64/fpu/math-use-builtins.h
>  delete mode 100644 sysdeps/aarch64/fpu/s_floor.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_floorf.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_fma.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_fmaf.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_nearbyint.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_nearbyintf.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_rint.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_rintf.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_round.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_roundf.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_trunc.c
>  delete mode 100644 sysdeps/aarch64/fpu/s_truncf.c
> 
> diff --git a/sysdeps/aarch64/fpu/e_sqrt.c b/sysdeps/aarch64/fpu/e_sqrt.c
> deleted file mode 100644
> index abb67ef7b061..000000000000
> --- a/sysdeps/aarch64/fpu/e_sqrt.c
> +++ /dev/null
> @@ -1,27 +0,0 @@
> -/* Square root of floating point number.
> -   Copyright (C) 2015-2020 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/>.  */
> -
> -#include <math_private.h>
> -#include <libm-alias-finite.h>
> -
> -double
> -__ieee754_sqrt (double d)
> -{
> -  return __builtin_sqrt (d);
> -}
> -libm_alias_finite (__ieee754_sqrt, __sqrt)

Ok.

> diff --git a/sysdeps/aarch64/fpu/e_sqrtf.c b/sysdeps/aarch64/fpu/e_sqrtf.c
> deleted file mode 100644
> index 13008a4f45d6..000000000000
> --- a/sysdeps/aarch64/fpu/e_sqrtf.c
> +++ /dev/null
> @@ -1,27 +0,0 @@
> -/* Single-precision floating point square root.
> -   Copyright (C) 2015-2020 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/>.  */
> -
> -#include <math_private.h>
> -#include <libm-alias-finite.h>
> -
> -float
> -__ieee754_sqrtf (float s)
> -{
> -  return __builtin_sqrtf (s);
> -}
> -libm_alias_finite (__ieee754_sqrtf, __sqrtf)

Ok.

> diff --git a/sysdeps/aarch64/fpu/math-use-builtins.h b/sysdeps/aarch64/fpu/math-use-builtins.h
> new file mode 100644
> index 000000000000..50a1ba0db5ea
> --- /dev/null
> +++ b/sysdeps/aarch64/fpu/math-use-builtins.h
> @@ -0,0 +1,71 @@
> +/* Using math gcc builtins instead of generic implementation.  aarch64 version.
> +   Copyright (C) 2020 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/>.  */
> +
> +#ifndef MATH_USE_BUILTINS_H
> +#define MATH_USE_BUILTINS_H	1
> +
> +#include <features.h> /* For __GNUC_PREREQ.  */
> +
> +/* Define these macros to 1 to use __builtin_xyz instead of the
> +   generic implementation.  */
> +#define USE_NEARBYINT_BUILTIN 1
> +#define USE_NEARBYINTF_BUILTIN 1
> +#define USE_NEARBYINTL_BUILTIN 0
> +#define USE_NEARBYINTF128_BUILTIN 0
> +
> +#define USE_RINT_BUILTIN 1
> +#define USE_RINTF_BUILTIN 1
> +#define USE_RINTL_BUILTIN 0
> +#define USE_RINTF128_BUILTIN 0
> +
> +#define USE_FLOOR_BUILTIN 1
> +#define USE_FLOORF_BUILTIN 1
> +#define USE_FLOORL_BUILTIN 0
> +#define USE_FLOORF128_BUILTIN 0
> +
> +#define USE_CEIL_BUILTIN 1
> +#define USE_CEILF_BUILTIN 1
> +#define USE_CEILL_BUILTIN 0
> +#define USE_CEILF128_BUILTIN 0
> +
> +#define USE_TRUNC_BUILTIN 1
> +#define USE_TRUNCF_BUILTIN 1
> +#define USE_TRUNCL_BUILTIN 0
> +#define USE_TRUNCF128_BUILTIN 0
> +
> +#define USE_ROUND_BUILTIN 1
> +#define USE_ROUNDF_BUILTIN 1
> +#define USE_ROUNDL_BUILTIN 0
> +#define USE_ROUNDF128_BUILTIN 0
> +
> +#define USE_COPYSIGNL_BUILTIN 1
> +#if __GNUC_PREREQ (7, 0)
> +# define USE_COPYSIGNF128_BUILTIN 1
> +#else
> +# define USE_COPYSIGNF128_BUILTIN 0
> +#endif
> +
> +#define USE_SQRT_BUILTIN 1
> +#define USE_SQRTF_BUILTIN 1
> +
> +#define USE_FMA_BUILTIN 1
> +#define USE_FMAF_BUILTIN 1
> +#define USE_FMAL_BUILTIN 0
> +#define USE_FMAF128_BUILTIN 0
> +
> +#endif

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_floor.c b/sysdeps/aarch64/fpu/s_floor.c
> deleted file mode 100644
> index d7d3e9336a6e..000000000000
> --- a/sysdeps/aarch64/fpu/s_floor.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__floor (double x)
> -{
> -  return __builtin_floor (x);
> -}
> -
> -libm_alias_double (__floor, floor)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_floorf.c b/sysdeps/aarch64/fpu/s_floorf.c
> deleted file mode 100644
> index 625ff4375702..000000000000
> --- a/sysdeps/aarch64/fpu/s_floorf.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__floorf (float x)
> -{
> -  return __builtin_floorf (x);
> -}
> -
> -libm_alias_float (__floor, floor)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_fma.c b/sysdeps/aarch64/fpu/s_fma.c
> deleted file mode 100644
> index 48bb40482dc9..000000000000
> --- a/sysdeps/aarch64/fpu/s_fma.c
> +++ /dev/null
> @@ -1,28 +0,0 @@
> -/* Copyright (C) 1996-2020 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/>.  */
> -
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__fma (double x, double y, double z)
> -{
> -  return __builtin_fma (x, y, z);
> -}
> -
> -libm_alias_double (__fma, fma)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_fmaf.c b/sysdeps/aarch64/fpu/s_fmaf.c
> deleted file mode 100644
> index 544f32e27aec..000000000000
> --- a/sysdeps/aarch64/fpu/s_fmaf.c
> +++ /dev/null
> @@ -1,28 +0,0 @@
> -/* Copyright (C) 2011-2020 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/>.  */
> -
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__fmaf (float x, float y, float z)
> -{
> -  return __builtin_fmaf (x, y, z);
> -}
> -
> -libm_alias_float (__fma, fma)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_nearbyint.c b/sysdeps/aarch64/fpu/s_nearbyint.c
> deleted file mode 100644
> index 59a57dd39ba1..000000000000
> --- a/sysdeps/aarch64/fpu/s_nearbyint.c
> +++ /dev/null
> @@ -1,28 +0,0 @@
> -/* Copyright (C) 2011-2020 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/>.  */
> -
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__nearbyint (double x)
> -{
> -  return __builtin_nearbyint (x);
> -}
> -
> -libm_alias_double (__nearbyint, nearbyint)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_nearbyintf.c b/sysdeps/aarch64/fpu/s_nearbyintf.c
> deleted file mode 100644
> index f52968ae2bbd..000000000000
> --- a/sysdeps/aarch64/fpu/s_nearbyintf.c
> +++ /dev/null
> @@ -1,28 +0,0 @@
> -/* Copyright (C) 2011-2020 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/>.  */
> -
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__nearbyintf (float x)
> -{
> -  return __builtin_nearbyintf (x);
> -}
> -
> -libm_alias_float (__nearbyint, nearbyint)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_rint.c b/sysdeps/aarch64/fpu/s_rint.c
> deleted file mode 100644
> index bdd757167e0c..000000000000
> --- a/sysdeps/aarch64/fpu/s_rint.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__rint (double x)
> -{
> -  return __builtin_rint (x);
> -}
> -
> -libm_alias_double (__rint, rint)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_rintf.c b/sysdeps/aarch64/fpu/s_rintf.c
> deleted file mode 100644
> index 8cf7c63d6dda..000000000000
> --- a/sysdeps/aarch64/fpu/s_rintf.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__rintf (float x)
> -{
> -  return __builtin_rintf (x);
> -}
> -
> -libm_alias_float (__rint, rint)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_round.c b/sysdeps/aarch64/fpu/s_round.c
> deleted file mode 100644
> index d2fd5f34117a..000000000000
> --- a/sysdeps/aarch64/fpu/s_round.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__round (double x)
> -{
> -  return __builtin_round (x);
> -}
> -
> -libm_alias_double (__round, round)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_roundf.c b/sysdeps/aarch64/fpu/s_roundf.c
> deleted file mode 100644
> index e51214e86cd2..000000000000
> --- a/sysdeps/aarch64/fpu/s_roundf.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__roundf (float x)
> -{
> -  return __builtin_roundf (x);
> -}
> -
> -libm_alias_float (__round, round)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_trunc.c b/sysdeps/aarch64/fpu/s_trunc.c
> deleted file mode 100644
> index bac23a5e269d..000000000000
> --- a/sysdeps/aarch64/fpu/s_trunc.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-double.h>
> -
> -double
> -__trunc (double x)
> -{
> -  return __builtin_trunc (x);
> -}
> -
> -libm_alias_double (__trunc, trunc)

Ok.

> diff --git a/sysdeps/aarch64/fpu/s_truncf.c b/sysdeps/aarch64/fpu/s_truncf.c
> deleted file mode 100644
> index 6a025c084a5a..000000000000
> --- a/sysdeps/aarch64/fpu/s_truncf.c
> +++ /dev/null
> @@ -1,29 +0,0 @@
> -/* Copyright (C) 2011-2020 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 NO_MATH_REDIRECT
> -#include <math.h>
> -#include <libm-alias-float.h>
> -
> -float
> -__truncf (float x)
> -{
> -  return __builtin_truncf (x);
> -}
> -
> -libm_alias_float (__trunc, trunc)
> 


Ok.

^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions
  2020-06-02 17:31   ` Adhemerval Zanella via Libc-alpha
@ 2020-06-02 19:14     ` Vineet Gupta via Libc-alpha
  0 siblings, 0 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-02 19:14 UTC (permalink / raw)
  To: Adhemerval Zanella, libc-alpha@sourceware.org
  Cc: linux-snps-arc@lists.infradead.org

On 6/2/20 10:31 AM, Adhemerval Zanella via Libc-alpha wrote:
> 
> 
> On 01/06/2020 21:35, Vineet Gupta wrote:
>> introduce sysdep header math-use-builtins.h to replace aarch64
>> impementations with corresponding generic ones
> 
> s/impementations/implementations and missing ':'.

Fixed.

>>  - newly inroduced generic sqrt{,f}, fma{,f}
>>  - existing floor{,f}, nearbyint{,f}, rint{,f}, round{,f}, trunc{,f}
>>  - Note that generic copysign was already enabled (via generic
>>    math-use-builtins.h) now thru sysdep header
>>
>> This is supposed to be a non functional change
>> Passes build-many for aarch64-linux-gnu
> 
> LGTM (just double check if the objects generated by the removed 
> implementations did not changed).

I will add the diff output to v3.


^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02 17:27       ` Adhemerval Zanella via Libc-alpha
@ 2020-06-03  0:20         ` Vineet Gupta via Libc-alpha
  0 siblings, 0 replies; 16+ messages in thread
From: Vineet Gupta via Libc-alpha @ 2020-06-03  0:20 UTC (permalink / raw)
  To: Adhemerval Zanella, libc-alpha@sourceware.org; +Cc: arcml

On 6/2/20 10:27 AM, Adhemerval Zanella via Libc-alpha wrote:
>> I used commit f82996f815 "Use GCC builtins for round functions if desired" as
>> starting point for my change. And seems it was not an ideal reference :-) as round
>> has far fewer instances than fma. Indeed fma is present in ldbl-128 and dbl-64 so
>> needs updating in both.
> I think after this set is upstream I will refactor to make each symbol and its
> variant (i.g sqrt, sqrtf, etc.) to be define on its own file.  The default
> math-use-builtins will then include each file:
> 
>   /* math-use-builtins.h  */
>   #inclde <math-use-builtins-sqrt.h>
>   [...]
> 
> With default USE_* begin 0. This would allow to add new builtin usage without 
> require to sync with every architecture (and without breaking its build as well).

OK !



^ permalink raw reply	[flat|nested] 16+ messages in thread

* Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions
  2020-06-02 17:13     ` Vineet Gupta via Libc-alpha
  2020-06-02 17:27       ` Adhemerval Zanella via Libc-alpha
@ 2020-06-03  6:29       ` Stefan Liebler via Libc-alpha
  1 sibling, 0 replies; 16+ messages in thread
From: Stefan Liebler via Libc-alpha @ 2020-06-03  6:29 UTC (permalink / raw)
  To: Vineet Gupta, libc-alpha@sourceware.org; +Cc: arcml

On 6/2/20 7:13 PM, Vineet Gupta wrote:
> On 6/2/20 5:51 AM, Stefan Liebler via Libc-alpha wrote:
>>  #endif /* math-use-builtins.h */
>> Please also update the current architecture specific math-use-builtins.h
>> file: sysdeps/s390/fpu/math-use-builtins.h
>> Otherwise it will break build on s390x.
> 
> Done.
> 
>>> diff --git a/sysdeps/ieee754/dbl-64/s_fma.c b/sysdeps/ieee754/dbl-64/s_fma.c
>>> index 876df6e78bdc..9dc5b132b9ee 100644
>>> --- a/sysdeps/ieee754/dbl-64/s_fma.c
>>> +++ b/sysdeps/ieee754/dbl-64/s_fma.c
>>> @@ -25,6 +25,7 @@
>>>  #include <fenv_private.h>
>>>  #include <libm-alias-double.h>
>>>  #include <tininess.h>
>>> +#include <math-use-builtins.h>
>>>
>>>  /* This implementation uses rounding to odd to avoid problems with
>>>     double rounding.  See a paper by Boldo and Melquiond:
>>> @@ -33,6 +34,10 @@
>>>  double
>>>  __fma (double x, double y, double z)
>>>  {
>>> +#if USE_FMA_BUILTIN
>>> +  return __builtin_fma (x, y, z);
>>
>> Architectures which have support for ldbl-128 will use the file
>> sysdeps/ieee754/ldbl-128/s_fma.c instead of
>> sysdeps/ieee754/dbl-64/s_fma.c. Should this file also be adjusted in
>> order to use the builtin if USE_FMA_BUILTIN is set to one?
> 
> Right.
> 
> I used commit f82996f815 "Use GCC builtins for round functions if desired" as
> starting point for my change. And seems it was not an ideal reference :-) as round
> has far fewer instances than fma. Indeed fma is present in ldbl-128 and dbl-64 so
> needs updating in both.
> 
> But just to be sure s390 is currently not using the newly introduced builtins so
> I'll keep them as follows.
> 
> #define USE_SQRT_BUILTIN 0
> #define USE_SQRTF_BUILTIN 0
> 
> #define USE_FMA_BUILTIN 0
> #define USE_FMAF_BUILTIN 0
> #define USE_FMAL_BUILTIN 0
> #define USE_FMAF128_BUILTIN 0
> 
Yes. Those should be disabled for now in order to not break anything. As
soon as you've committed your patches, I'll have a look if I can
activate some of them and remove s390 specific implementations. But I
first have to check with the gcc guys, if the builtins are always
expanded correctly in various gcc versions.

Thanks.
Stefan
> -Vineet
> 


^ permalink raw reply	[flat|nested] 16+ messages in thread

end of thread, other threads:[~2020-06-03  6:29 UTC | newest]

Thread overview: 16+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-06-02  0:35 [PATCH v2 0/4] Enable generic math code for more arches Vineet Gupta via Libc-alpha
2020-06-02  0:35 ` [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions Vineet Gupta via Libc-alpha
2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
2020-06-02 16:54     ` Vineet Gupta via Libc-alpha
2020-06-02 17:17   ` Adhemerval Zanella via Libc-alpha
2020-06-02  0:35 ` [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions Vineet Gupta via Libc-alpha
2020-06-02 12:51   ` Stefan Liebler via Libc-alpha
2020-06-02 17:13     ` Vineet Gupta via Libc-alpha
2020-06-02 17:27       ` Adhemerval Zanella via Libc-alpha
2020-06-03  0:20         ` Vineet Gupta via Libc-alpha
2020-06-03  6:29       ` Stefan Liebler via Libc-alpha
2020-06-02 17:28   ` Adhemerval Zanella via Libc-alpha
2020-06-02  0:35 ` [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions Vineet Gupta via Libc-alpha
2020-06-02 17:31   ` Adhemerval Zanella via Libc-alpha
2020-06-02 19:14     ` Vineet Gupta via Libc-alpha
2020-06-02  0:35 ` [PATCH v2 4/4] powerpc/fpu: use generic fma functions Vineet Gupta via Libc-alpha

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).