Skip to contentSkip to conversation

Enable generic math code for more arches

PATCH Superseded Vineet Gupta 3 participants 13 messages Jun 2, 2020 – Jun 3, 2020 3 messages not in the archive Synopsys ARC architecture

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Enable generic math code for more arches

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
Signature · 7 lines
2.20.1


_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc
#1 lore raw

PATCH v2 1/4 iee754: provide gcc builtins based generic sqrt functions

---
 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(-)
sysdeps/generic/math-use-builtins.h +3 -0
index 8a39ef58bc95..fc724c824a17 100644
@@ -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 */
sysdeps/ieee754/dbl-64/e_sqrt.c +6 -0
index d42a1a4eb6e9..518a8ae5cdaf 100644
@@ -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)
sysdeps/ieee754/flt-32/e_sqrtf.c +10 -6
index b339444301aa..d85a04162983 100644
@@ -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)
Signature · 7 lines
2.20.1


_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw

Re: [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions

Quoted 16 lines · Stefan Liebler via Libc-alpha
Quoted 15 lines · Vineet Gupta via Libc-alpha
---
 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(-)
sysdeps/generic/math-use-builtins.h +0 -0
index 8a39ef58bc95..fc724c824a17 100644
@@ -60,4 +60,7 @@
 # define USE_COPYSIGNF128_BUILTIN 0
 #endif
 
Quoted 6 lines

sysdeps/generic/math-use-builtins.h line 66

Quoted 4 lines
sysdeps/generic/math-use-builtins.h +3 -0
+#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

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

Reviewed-by on v2 1/4
#4 lore raw

Re: [PATCH v2 1/4] iee754: provide gcc builtins based generic sqrt functions

In reply to Vineet Gupta
Quoted 1 line · Vineet Gupta

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

Quoted 11 lines
---
 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(-)
sysdeps/generic/math-use-builtins.h +0 -0
index 8a39ef58bc95..fc724c824a17 100644

sysdeps/generic/math-use-builtins.h line 66

Quoted 7 lines
sysdeps/generic/math-use-builtins.h +3 -0
@@ -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.

Quoted 24 lines
sysdeps/ieee754/dbl-64/e_sqrt.c +5 -0
index d42a1a4eb6e9..518a8ae5cdaf 100644
@@ -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,

sysdeps/ieee754/dbl-64/e_sqrt.c __ieee754_sqrt (double x) · line 149

Quoted 7 lines
sysdeps/ieee754/dbl-64/e_sqrt.c +1 -0
@@ -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.

Quoted 23 lines
sysdeps/ieee754/flt-32/e_sqrtf.c +5 -2
index b339444301aa..d85a04162983 100644
@@ -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;

sysdeps/ieee754/flt-32/e_sqrtf.c __ieee754_sqrtf(float x) · line 76

Quoted 8 lines
sysdeps/ieee754/flt-32/e_sqrtf.c +1 -4
@@ -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 '*/'.

Quoted 8 lines
sysdeps/ieee754/flt-32/e_sqrtf.c +3 -0
+	    if (z >= 0x1p0) {
+		z = 0x1p0 + 0x1.4484cp-100;
+		if (z > 0x1p0)
 		    q += 2;
 		else
 		    q += (q&1);

sysdeps/ieee754/flt-32/e_sqrtf.c __ieee754_sqrtf(float x) · line 92

Quoted 7 lines
sysdeps/ieee754/flt-32/e_sqrtf.c +1 -0
@@ -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.

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

#5 lore raw

PATCH v2 2/4 iee754: provide gcc builtins based generic fma functions

In reply to Vineet Gupta
---
 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(+)
sysdeps/generic/math-use-builtins.h +5 -0
index fc724c824a17..cf25ed8a2138 100644
@@ -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 */
sysdeps/ieee754/dbl-64/s_fma.c +6 -0
index 876df6e78bdc..9dc5b132b9ee 100644
@@ -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)
sysdeps/ieee754/dbl-64/s_fmaf.c +6 -0
index 57329d0a87fe..93b8660d5242 100644
@@ -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)
sysdeps/ieee754/float128/float128_private.h +2 -0
index f97463d9dc1b..ab6fc9f3c9cf 100644
@@ -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
sysdeps/ieee754/ldbl-128/s_fmal.c +5 -0
index 7475015bcec6..a610499e47c7 100644
@@ -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)
Signature · 7 lines
2.20.1


_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

A message that is not in the archive (sent off-list or not yet received): [email protected]

#7 lore raw

Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions

Quoted 4 lines · Stefan Liebler via Libc-alpha

#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.

Quoted 10 lines
Quoted 10 lines
sysdeps/ieee754/dbl-64/s_fma.c +1 -0
index 876df6e78bdc..9dc5b132b9ee 100644
@@ -25,6 +25,7 @@
 #include <fenv_private.h>
 #include <libm-alias-double.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
Quoted 13 lines

sysdeps/ieee754/dbl-64/s_fma.c line 38

Quoted 8 lines
sysdeps/ieee754/dbl-64/s_fma.c +2 -0
 /* 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
_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

A message that is not in the archive (sent off-list or not yet received): [email protected]

#9 lore raw

Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions

Quoted 8 lines · Adhemerval Zanella via Libc-alpha
Quoted 4 lines

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:

Quoted 6 lines

/* 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 !

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

#10 lore raw

Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions

In reply to Vineet Gupta
Quoted 42 lines · Vineet Gupta
Quoted 4 lines · Stefan Liebler via Libc-alpha

#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.

Quoted 10 lines
Quoted 10 lines
sysdeps/ieee754/dbl-64/s_fma.c +1 -0
index 876df6e78bdc..9dc5b132b9ee 100644
@@ -25,6 +25,7 @@
 #include <fenv_private.h>
 #include <libm-alias-double.h>
 #include <tininess.h>
+#include <math-use-builtins.h>
 
Quoted 13 lines

sysdeps/ieee754/dbl-64/s_fma.c line 38

Quoted 8 lines
sysdeps/ieee754/dbl-64/s_fma.c +2 -0
 /* 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.

Quoted 8 lines

#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

Quoted 2 lines

-Vineet

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

Reviewed-by on v2 2/4
#11 lore raw

Re: [PATCH v2 2/4] iee754: provide gcc builtins based generic fma functions

In reply to Vineet Gupta
Quoted 7 lines · Vineet Gupta
---
 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.

Quoted 114 lines
sysdeps/generic/math-use-builtins.h +5 -0
index fc724c824a17..cf25ed8a2138 100644
@@ -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 */
sysdeps/ieee754/dbl-64/s_fma.c +6 -0
index 876df6e78bdc..9dc5b132b9ee 100644
@@ -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)
sysdeps/ieee754/dbl-64/s_fmaf.c +6 -0
index 57329d0a87fe..93b8660d5242 100644
@@ -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)
sysdeps/ieee754/float128/float128_private.h +2 -0
index f97463d9dc1b..ab6fc9f3c9cf 100644
@@ -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
sysdeps/ieee754/ldbl-128/s_fmal.c +4 -0
index 7475015bcec6..a610499e47c7 100644
@@ -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;

sysdeps/ieee754/ldbl-128/s_fmal.c __fmal (_Float128 x, _Float128 y, _Float128 z) · line 305

Quoted 6 lines
sysdeps/ieee754/ldbl-128/s_fmal.c +1 -0
@@ -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)

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

#12 lore raw

PATCH v2 3/4 aarch/fpu: use generic builtins based math functions

In reply to Vineet Gupta

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
sysdeps/aarch64/fpu/e_sqrt.c deleted +0 -27
deleted file mode 100644
index abb67ef7b061..000000000000
@@ -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)
sysdeps/aarch64/fpu/e_sqrtf.c deleted +0 -27
deleted file mode 100644
index 13008a4f45d6..000000000000
@@ -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)
sysdeps/aarch64/fpu/math-use-builtins.h new file +71 -0

71 lines not on this page · raw message

sysdeps/aarch64/fpu/s_floor.c deleted +0 -29
deleted file mode 100644
index d7d3e9336a6e..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_floorf.c deleted +0 -29
deleted file mode 100644
index 625ff4375702..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_fma.c deleted +0 -28
deleted file mode 100644
index 48bb40482dc9..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_fmaf.c deleted +0 -28
deleted file mode 100644
index 544f32e27aec..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_nearbyint.c deleted +0 -28
deleted file mode 100644
index 59a57dd39ba1..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_nearbyintf.c deleted +0 -28
deleted file mode 100644
index f52968ae2bbd..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_rint.c deleted +0 -29
deleted file mode 100644
index bdd757167e0c..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_rintf.c deleted +0 -29
deleted file mode 100644
index 8cf7c63d6dda..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_round.c deleted +0 -29
deleted file mode 100644
index d2fd5f34117a..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_roundf.c deleted +0 -29
deleted file mode 100644
index e51214e86cd2..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_trunc.c deleted +0 -29
deleted file mode 100644
index bac23a5e269d..000000000000
@@ -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)
sysdeps/aarch64/fpu/s_truncf.c deleted +0 -29
deleted file mode 100644
index 6a025c084a5a..000000000000
@@ -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)
Signature · 7 lines
2.20.1


_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc
Reviewed-by on v2 3/4
#13 lore raw

Re: [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions

Quoted 2 lines · Vineet Gupta

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

s/impementations/implementations and missing ':'.

Quoted 8 lines

- 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).

Quoted 56 lines
---
 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
sysdeps/aarch64/fpu/e_sqrt.c deleted +0 -17
deleted file mode 100644
index abb67ef7b061..000000000000
@@ -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/>.  */

sysdeps/aarch64/fpu/e_sqrt.c line 27

Quoted 10 lines
sysdeps/aarch64/fpu/e_sqrt.c deleted +0 -10
-
-#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.

Quoted 23 lines
sysdeps/aarch64/fpu/e_sqrtf.c deleted +0 -17
deleted file mode 100644
index 13008a4f45d6..000000000000
@@ -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/>.  */

sysdeps/aarch64/fpu/e_sqrtf.c line 27

Quoted 10 lines
sysdeps/aarch64/fpu/e_sqrtf.c deleted +0 -10
-
-#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.

Quoted 67 lines
sysdeps/aarch64/fpu/math-use-builtins.h new file +61 -0
new file mode 100644
index 000000000000..50a1ba0db5ea
@@ -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

sysdeps/aarch64/fpu/math-use-builtins.h line 71

Quoted 10 lines
sysdeps/aarch64/fpu/math-use-builtins.h new file +10 -0
+
+#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.

Quoted 25 lines
sysdeps/aarch64/fpu/s_floor.c deleted +0 -19
deleted file mode 100644
index d7d3e9336a6e..000000000000
@@ -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

sysdeps/aarch64/fpu/s_floor.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_floor.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__floor (double x)
-{
-  return __builtin_floor (x);
-}
-
-libm_alias_double (__floor, floor)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_floorf.c deleted +0 -19
deleted file mode 100644
index 625ff4375702..000000000000
@@ -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

sysdeps/aarch64/fpu/s_floorf.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_floorf.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__floorf (float x)
-{
-  return __builtin_floorf (x);
-}
-
-libm_alias_float (__floor, floor)

Ok.

Quoted 24 lines
sysdeps/aarch64/fpu/s_fma.c deleted +0 -18
deleted file mode 100644
index 48bb40482dc9..000000000000
@@ -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/>.  */
-

sysdeps/aarch64/fpu/s_fma.c line 28

Quoted 10 lines
sysdeps/aarch64/fpu/s_fma.c deleted +0 -10
-#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.

Quoted 24 lines
sysdeps/aarch64/fpu/s_fmaf.c deleted +0 -18
deleted file mode 100644
index 544f32e27aec..000000000000
@@ -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/>.  */
-

sysdeps/aarch64/fpu/s_fmaf.c line 28

Quoted 10 lines
sysdeps/aarch64/fpu/s_fmaf.c deleted +0 -10
-#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.

Quoted 24 lines
sysdeps/aarch64/fpu/s_nearbyint.c deleted +0 -18
deleted file mode 100644
index 59a57dd39ba1..000000000000
@@ -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/>.  */
-

sysdeps/aarch64/fpu/s_nearbyint.c line 28

Quoted 10 lines
sysdeps/aarch64/fpu/s_nearbyint.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__nearbyint (double x)
-{
-  return __builtin_nearbyint (x);
-}
-
-libm_alias_double (__nearbyint, nearbyint)

Ok.

Quoted 24 lines
sysdeps/aarch64/fpu/s_nearbyintf.c deleted +0 -18
deleted file mode 100644
index f52968ae2bbd..000000000000
@@ -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/>.  */
-

sysdeps/aarch64/fpu/s_nearbyintf.c line 28

Quoted 10 lines
sysdeps/aarch64/fpu/s_nearbyintf.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__nearbyintf (float x)
-{
-  return __builtin_nearbyintf (x);
-}
-
-libm_alias_float (__nearbyint, nearbyint)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_rint.c deleted +0 -19
deleted file mode 100644
index bdd757167e0c..000000000000
@@ -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

sysdeps/aarch64/fpu/s_rint.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_rint.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__rint (double x)
-{
-  return __builtin_rint (x);
-}
-
-libm_alias_double (__rint, rint)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_rintf.c deleted +0 -19
deleted file mode 100644
index 8cf7c63d6dda..000000000000
@@ -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

sysdeps/aarch64/fpu/s_rintf.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_rintf.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__rintf (float x)
-{
-  return __builtin_rintf (x);
-}
-
-libm_alias_float (__rint, rint)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_round.c deleted +0 -19
deleted file mode 100644
index d2fd5f34117a..000000000000
@@ -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

sysdeps/aarch64/fpu/s_round.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_round.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__round (double x)
-{
-  return __builtin_round (x);
-}
-
-libm_alias_double (__round, round)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_roundf.c deleted +0 -19
deleted file mode 100644
index e51214e86cd2..000000000000
@@ -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

sysdeps/aarch64/fpu/s_roundf.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_roundf.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__roundf (float x)
-{
-  return __builtin_roundf (x);
-}
-
-libm_alias_float (__round, round)

Ok.

Quoted 25 lines
sysdeps/aarch64/fpu/s_trunc.c deleted +0 -19
deleted file mode 100644
index bac23a5e269d..000000000000
@@ -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

sysdeps/aarch64/fpu/s_trunc.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_trunc.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-double.h>
-
-double
-__trunc (double x)
-{
-  return __builtin_trunc (x);
-}
-
-libm_alias_double (__trunc, trunc)

Ok.

Quoted 26 lines
sysdeps/aarch64/fpu/s_truncf.c deleted +0 -19
deleted file mode 100644
index 6a025c084a5a..000000000000
@@ -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

sysdeps/aarch64/fpu/s_truncf.c line 29

Quoted 10 lines
sysdeps/aarch64/fpu/s_truncf.c deleted +0 -10
-#include <math.h>
-#include <libm-alias-float.h>
-
-float
-__truncf (float x)
-{
-  return __builtin_truncf (x);
-}
-
-libm_alias_float (__trunc, trunc)

Ok.

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

#14 lore raw

Re: [PATCH v2 3/4] aarch/fpu: use generic builtins based math functions

Quoted 7 lines · Adhemerval Zanella via Libc-alpha
Quoted 2 lines · Vineet Gupta

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

s/impementations/implementations and missing ':'.

Fixed.

Quoted 10 lines
Quoted 7 lines

- 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.

_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc

#15 lore raw

PATCH v2 4/4 powerpc/fpu: use generic fma functions

In reply to Vineet Gupta

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

---
 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
sysdeps/powerpc/fpu/math-use-builtins.h new file +69 -0
new file mode 100644
index 000000000000..4780934379ba
@@ -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
sysdeps/powerpc/fpu/s_fma.c deleted +0 -27
deleted file mode 100644
index 9ddd13253485..000000000000
@@ -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)
sysdeps/powerpc/fpu/s_fmaf.c deleted +0 -27
deleted file mode 100644
index d132c20125bb..000000000000
@@ -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)
Signature · 7 lines
2.20.1


_______________________________________________
linux-snps-arc mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-snps-arc