Optimize sparc {ceil,floor}{,f} using vis2 'siam' instruction.

	* sysdeps/sparc/sparc-ifunc.h (SPARC_ASM_IFUNC2): New macro.
	(SPARC_ASM_VIS2_IFUNC): Likewise.
	(SPARC_ASM_VIS3_VIS2_IFUNC): Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis3.S: Make
	use of 'siam' instruction.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceilf-vis3.S:
	Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis3.S:
	Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floorf-vis3.S:
	Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceil-vis3.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceilf-vis3.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floor-vis3.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floorf-vis3.S: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil-vis2.S: New
	file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceilf-vis2.S: New
	file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis2.S: New
	file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floorf-vis2.S: New
	file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceil-vis2.S: New file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceilf-vis2.S: New file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floor-vis2.S: New file.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floorf-vis2.S: New file.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceil.S: Hook in
	new VIS2 routines.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_ceilf.S: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor.S: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floorf.S:
	Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceil.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_ceilf.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floor.S: Likewise.
	* sysdeps/sparc/sparc64/fpu/multiarch/s_floorf.S: Likewise.
	* sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/Makefile: Add new VIS2
	routines to libm-sysdep_routines.
	* sysdeps/sparc/sparc64/fpu/multiarch/Makefile: Likewise.

1 files changed, 12 insertions, 28 deletions

diff --git a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis3.S b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis3.S
index 8445f1d7a6..41fdfac3b3 100644
--- a/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis3.S
+++ b/sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_floor-vis3.S
@@ -19,27 +19,21 @@
 
 #include <sysdep.h>
 
-	/* Since changing the rounding mode is extremely expensive, we
-	   try to round up using a method that is rounding mode
-	   agnostic.
+	/* 'siam' (Set Interval Arithmetic Mode) is used to quickly override
+	   the rounding mode during this routine.
 
 	   We add then subtract (or subtract than add if the initial
 	   value was negative) 2**23 to the value, then subtract it
 	   back out.
 
-	   This will clear out the fractional portion of the value.
-	   One of two things will happen for non-whole initial values.
-	   Either the rounding mode will round it up, or it will be
-	   rounded down.  If the value started out whole, it will be
-	   equal after the addition and subtraction.  This means we
-	   can accurately detect with one test whether we need to add
-	   another 1.0 to round it up properly.
+	   This will clear out the fractional portion of the value and,
+	   with suitable 'siam' initiated rouding mode settings, round
+	   the final result in the proper direction.
 
-	   VIS instructions are used to facilitate the formation of
-	   easier constants, and the propagation of the sign bit.  */
+	   We also use VIS3 moves to avoid using the stack to transfer
+	   values between float and integer registers.  */
 
 #define TWO_FIFTYTWO	0x43300000		/* 2**52 */
-#define ONE_DOT_ZERO	0x3ff00000		/* 1.0 */
 
 #define ZERO		%f10			/* 0.0 */
 #define SIGN_BIT	%f12			/* -0.0 */
@@ -47,32 +41,22 @@
 ENTRY (__floor_vis3)
 	sethi	%hi(TWO_FIFTYTWO), %o2
 	sllx	%o0, 32, %o0
-	sethi	%hi(ONE_DOT_ZERO), %o3
+	sllx	%o2, 32, %o2
 	or	%o0, %o1, %o0
 	movxtod	%o0, %f0
-	sllx	%o2, 32, %o2
 	fzero	ZERO
-	sllx	%o3, 32, %o3
-
 	fnegd	ZERO, SIGN_BIT
-
 	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
+	siam	(1 << 2) | 3
 	faddd	%f0, %f16, %f18
+	siam	(1 << 2) | 0
 	fsubd	%f18, %f16, %f18
-	fcmpd	%fcc2, %f18, %f0
-	movxtod	%o3, %f20
-
-	fmovdule %fcc2, ZERO, %f20
-	fsubd	%f18, %f20, %f0
-	fabsd	%f0, %f0
+	siam	(0 << 2)
 	retl
-	 for	%f0, SIGN_BIT, %f0
+	 for	%f18, SIGN_BIT, %f0
 END (__floor_vis3)