Optimize nearbyint{,f} on sparc.

	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile: Add vis3
	nearbyint{,f} to libm-sysdep_routes.
	* sysdeps/sparc/sparc64/fpu/multiarch/Makefile: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint-vis3.S:
	New file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint.S: New
	file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf-vis3.S:
	New file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf.S: New
	file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyint.S: New file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyintf.S: New file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_nearbyint-vis3.S: New
	file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_nearbyint.S: New file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_nearbyintf-vis3.S: New
	file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_nearbyintf.S: New file.
	* sysdeps/sparc/sparc64/fpu/s_nearbyint.S: New file.
	* sysdeps/sparc/sparc64/fpu/s_nearbyintf.S: New file.

7 files changed, 295 insertions, 1 deletions

diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile
index 561b0ee4c4..aecd93bbef 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile
@@ -7,7 +7,8 @@ libm-sysdep_routines += m_copysignf-vis3 m_copysign-vis3 s_ceilf-vis3 \
 			s_floorf-vis3 s_llrintf-vis3 s_llrint-vis3 \
 			s_rintf-vis3 s_rint-vis3 w_sqrt-vis3 w_sqrtf-vis3 \
 			s_fminf-vis3 s_fmin-vis3 s_fmaxf-vis3 s_fmax-vis3 \
-			s_fmaf-vis3 s_fma-vis3 s_fdimf-vis3 s_fdim-vis3
+			s_fmaf-vis3 s_fma-vis3 s_fdimf-vis3 s_fdim-vis3 \
+			s_nearbyint-vis3 s_nearbyintf-vis3
 sysdep_routines += s_copysignf-vis3 s_copysign-vis3
 endif
 endif
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint-vis3.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint-vis3.S
new file mode 100644
index 0000000000..b509500ed4
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint-vis3.S
@@ -0,0 +1,65 @@
+/* Round float to int floating-point values without generating
+   an inexact exception, sparc32 v9 vis3 version.
+
+   Copyright (C) 2013 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David S. Miller <davem@davemloft.net>, 2013.
+
+   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
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+	/* We pop constants into the FPU registers using the incoming
+	   argument stack slots, since this avoid having to use any PIC
+	   references.  We also thus avoid having to allocate a register
+	   window.
+
+	   VIS instructions are used to facilitate the formation of
+	   easier constants, and the propagation of the sign bit.  */
+
+#define TWO_FIFTYTWO	0x43300000		/* 2**52 */
+
+#define ZERO		%f10			/* 0.0 */
+#define SIGN_BIT	%f12			/* -0.0 */
+
+ENTRY (__nearbyint_vis3)
+	st	%fsr, [%sp + 88]
+	sethi	%hi(TWO_FIFTYTWO), %o2
+	sethi	%hi(0xf8003e0), %o5
+	ld	[%sp + 88], %o4
+	sllx	%o0, 32, %o0
+	or	%o5, %lo(0xf8003e0), %o5
+	or	%o0, %o1, %o0
+	movxtod	%o0, %f0
+	andn	%o4, %o5, %o4
+	fzero	ZERO
+	st	%o4, [%sp + 80]
+	sllx	%o2, 32, %o2
+	fnegd	ZERO, SIGN_BIT
+	ld	[%sp + 80], %fsr
+	movxtod	%o2, %f16
+	fabsd	%f0, %f14
+	fcmpd	%fcc3, %f14, %f16
+	fmovduge %fcc3, ZERO, %f16
+	fand	%f0, SIGN_BIT, SIGN_BIT
+	for	%f16, SIGN_BIT, %f16
+	faddd	%f0, %f16, %f6
+	fsubd	%f6, %f16, %f0
+	fabsd	%f0, %f0
+	for	%f0, SIGN_BIT, %f0
+	retl
+	 ld	[%sp + 88], %fsr
+END (__nearbyint_vis3)
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint.S
new file mode 100644
index 0000000000..47da9eaafe
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyint.S
@@ -0,0 +1,19 @@
+#include <sparc-ifunc.h>
+#include <math_ldbl_opt.h>
+
+SPARC_ASM_VIS3_IFUNC(nearbyint)
+
+weak_alias (__nearbyint, nearbyint)
+
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __nearbyint, nearbyintl, GLIBC_2_1)
+#endif
+
+# undef weak_alias
+# define weak_alias(a, b)
+# undef compat_symbol
+# define compat_symbol(a, b, c, d)
+
+#define __nearbyint __nearbyint_generic
+
+#include "../s_nearbyint.S"
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf-vis3.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf-vis3.S
new file mode 100644
index 0000000000..336126deef
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf-vis3.S
@@ -0,0 +1,61 @@
+/* Round float to int floating-point values without generating
+   an inexact exception, sparc32 v9 vis3 version.
+
+   Copyright (C) 2013 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David S. Miller <davem@davemloft.net>, 2013.
+
+   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
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+
+	/* We pop constants into the FPU registers using the incoming
+	   argument stack slots, since this avoid having to use any PIC
+	   references.  We also thus avoid having to allocate a register
+	   window.
+
+	   VIS instructions are used to facilitate the formation of
+	   easier constants, and the propagation of the sign bit.  */
+
+#define TWO_TWENTYTHREE	0x4b000000		/* 2**23 */
+
+#define ZERO		%f10			/* 0.0 */
+#define SIGN_BIT	%f12			/* -0.0 */
+
+ENTRY (__nearbyintf_vis3)
+	st	%fsr, [%sp + 88]
+	movwtos	%o0, %f1
+	sethi	%hi(TWO_TWENTYTHREE), %o2
+	sethi	%hi(0xf8003e0), %o5
+	ld	[%sp + 88], %o4
+	fzeros	ZERO
+	or	%o5, %lo(0xf8003e0), %o5
+	fnegs	ZERO, SIGN_BIT
+	andn	%o4, %o5, %o4
+	st	%o4, [%sp + 80]
+	ld	[%sp + 80], %fsr
+	movwtos	%o2, %f16
+	fabss	%f1, %f14
+	fcmps	%fcc3, %f14, %f16
+	fmovsuge %fcc3, ZERO, %f16
+	fands	%f1, SIGN_BIT, SIGN_BIT
+	fors	%f16, SIGN_BIT, %f16
+	fadds	%f1, %f16, %f5
+	fsubs	%f5, %f16, %f0
+	fabss	%f0, %f0
+	fors	%f0, SIGN_BIT, %f0
+	retl
+	 ld	[%sp + 88], %fsr
+END (__nearbyintf_vis3)
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf.S
new file mode 100644
index 0000000000..95100c1bfc
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_nearbyintf.S
@@ -0,0 +1,12 @@
+#include <sparc-ifunc.h>
+
+SPARC_ASM_VIS3_IFUNC(nearbyintf)
+
+weak_alias (__nearbyintf, nearbyintf)
+
+# undef weak_alias
+# define weak_alias(a, b)
+
+#define __nearbyintf __nearbyintf_generic
+
+#include "../s_nearbyintf.S"
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyint.S b/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyint.S
new file mode 100644
index 0000000000..ee6a575e19
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyint.S
@@ -0,0 +1,72 @@
+/* Round float to int floating-point values without generating
+   an inexact exception, sparc32 v9 version.
+
+   Copyright (C) 2013 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David S. Miller <davem@davemloft.net>, 2013.
+
+   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
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+#include <math_ldbl_opt.h>
+
+	/* We pop constants into the FPU registers using the incoming
+	   argument stack slots, since this avoid having to use any PIC
+	   references.  We also thus avoid having to allocate a register
+	   window.
+
+	   VIS instructions are used to facilitate the formation of
+	   easier constants, and the propagation of the sign bit.  */
+
+#define TWO_FIFTYTWO	0x43300000		/* 2**52 */
+
+#define ZERO		%f10			/* 0.0 */
+#define SIGN_BIT	%f12			/* -0.0 */
+
+ENTRY (__nearbyint)
+	st	%fsr, [%sp + 88]
+	sethi	%hi(TWO_FIFTYTWO), %o2
+	sethi	%hi(0xf8003e0), %o5
+	ld	[%sp + 88], %o4
+	sllx	%o0, 32, %o0
+	or	%o5, %lo(0xf8003e0), %o5
+	or	%o0, %o1, %o0
+	andn	%o4, %o5, %o4
+	fzero	ZERO
+	st	%o4, [%sp + 80]
+	stx	%o0, [%sp + 72]
+	sllx	%o2, 32, %o2
+	fnegd	ZERO, SIGN_BIT
+	ldd	[%sp + 72], %f0
+	ld	[%sp + 80], %fsr
+	stx	%o2, [%sp + 72]
+	fabsd	%f0, %f14
+	ldd	[%sp + 72], %f16
+	fcmpd	%fcc3, %f14, %f16
+	fmovduge %fcc3, ZERO, %f16
+	fand	%f0, SIGN_BIT, SIGN_BIT
+	for	%f16, SIGN_BIT, %f16
+	faddd	%f0, %f16, %f6
+	fsubd	%f6, %f16, %f0
+	fabsd	%f0, %f0
+	for	%f0, SIGN_BIT, %f0
+	retl
+	 ld	[%sp + 88], %fsr
+END (__nearbyint)
+weak_alias (__nearbyint, nearbyint)
+
+#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
+compat_symbol (libm, __nearbyint, nearbyintl, GLIBC_2_1)
+#endif
diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyintf.S b/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyintf.S
new file mode 100644
index 0000000000..4225b54497
--- /dev/null
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/s_nearbyintf.S
@@ -0,0 +1,64 @@
+/* Round float to int floating-point values without generating
+   an inexact exception, sparc32 v9 version.
+
+   Copyright (C) 2013 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by David S. Miller <davem@davemloft.net>, 2013.
+
+   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
+   <http://www.gnu.org/licenses/>.  */
+
+#include <sysdep.h>
+
+	/* We pop constants into the FPU registers using the incoming
+	   argument stack slots, since this avoid having to use any PIC
+	   references.  We also thus avoid having to allocate a register
+	   window.
+
+	   VIS instructions are used to facilitate the formation of
+	   easier constants, and the propagation of the sign bit.  */
+
+#define TWO_TWENTYTHREE	0x4b000000		/* 2**23 */
+
+#define ZERO		%f10			/* 0.0 */
+#define SIGN_BIT	%f12			/* -0.0 */
+
+ENTRY (__nearbyintf)
+	st	%fsr, [%sp + 88]
+	st	%o0, [%sp + 68]
+	sethi	%hi(TWO_TWENTYTHREE), %o2
+	sethi	%hi(0xf8003e0), %o5
+	ld	[%sp + 88], %o4
+	fzeros	ZERO
+	or	%o5, %lo(0xf8003e0), %o5
+	fnegs	ZERO, SIGN_BIT
+	andn	%o4, %o5, %o4
+	st	%o4, [%sp + 80]
+	ld	[%sp + 68], %f1
+	ld	[%sp + 80], %fsr
+	st	%o2, [%sp + 68]
+	fabss	%f1, %f14
+	ld	[%sp + 68], %f16
+	fcmps	%fcc3, %f14, %f16
+	fmovsuge %fcc3, ZERO, %f16
+	fands	%f1, SIGN_BIT, SIGN_BIT
+	fors	%f16, SIGN_BIT, %f16
+	fadds	%f1, %f16, %f5
+	fsubs	%f5, %f16, %f0
+	fabss	%f0, %f0
+	fors	%f0, SIGN_BIT, %f0
+	retl
+	 ld	[%sp + 88], %fsr
+END (__nearbyintf)
+weak_alias (__nearbyintf, nearbyintf)