Improve performance of sinf and cosf

The second patch improves performance of sinf and cosf using the same
algorithms and polynomials.  The returned values are identical to sincosf
for the same input.  ULP definitions for AArch64 and x64 are updated.

sinf/cosf througput gains on Cortex-A72:
* |x| < 0x1p-12 : 1.2x
* |x| < M_PI_4  : 1.8x
* |x| < 2 * M_PI: 1.7x
* |x| < 120.0   : 2.3x
* |x| < Inf     : 3.0x

	* NEWS: Mention sinf, cosf, sincosf.
	* sysdeps/aarch64/libm-test-ulps: Update ULP for sinf, cosf, sincosf.
	* sysdeps/x86_64/fpu/libm-test-ulps: Update ULP for sinf and cosf.
	* sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c: Add definitions of
	constants rather than including generic sincosf.h.
	* sysdeps/x86_64/fpu/s_sincosf_data.c: Remove.
	* sysdeps/ieee754/flt-32/s_cosf.c (cosf): Rewrite.
	* sysdeps/ieee754/flt-32/s_sincosf.h (reduced_sin): Remove.
	(reduced_cos): Remove.
	(sinf_poly): New function.
	* sysdeps/ieee754/flt-32/s_sinf.c (sinf): Rewrite.

3 files changed, 44 insertions, 2 deletions

diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps
index 912db31..a51faec 100644
--- a/sysdeps/x86_64/fpu/libm-test-ulps
+++ b/sysdeps/x86_64/fpu/libm-test-ulps
@@ -1271,24 +1271,30 @@ ldouble: 1
 
 Function: "cos_downward":
 double: 1
+float: 1
 float128: 3
 idouble: 1
+ifloat: 1
 ifloat128: 3
 ildouble: 3
 ldouble: 3
 
 Function: "cos_towardzero":
 double: 1
+float: 1
 float128: 1
 idouble: 1
+ifloat: 1
 ifloat128: 1
 ildouble: 2
 ldouble: 2
 
 Function: "cos_upward":
 double: 1
+float: 1
 float128: 2
 idouble: 1
+ifloat: 1
 ifloat128: 2
 ildouble: 2
 ldouble: 2
@@ -2539,24 +2545,30 @@ ldouble: 1
 
 Function: "sin_downward":
 double: 1
+float: 1
 float128: 3
 idouble: 1
+ifloat:1
 ifloat128: 3
 ildouble: 3
 ldouble: 3
 
 Function: "sin_towardzero":
 double: 1
+float:1
 float128: 2
 idouble: 1
+ifloat: 1
 ifloat128: 2
 ildouble: 2
 ldouble: 2
 
 Function: "sin_upward":
 double: 1
+float:1
 float128: 3
 idouble: 1
+ifloat: 1
 ifloat128: 3
 ildouble: 3
 ldouble: 3
diff --git a/sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c b/sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c
index 64abe7a..0b80c4f 100644
--- a/sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c
+++ b/sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c
@@ -21,7 +21,6 @@
 #include <math_private.h>
 #include <x86intrin.h>
 #include <libm-alias-float.h>
-#include "s_sincosf.h"
 
 #define SINCOSF __sincosf_fma
 
@@ -31,6 +30,38 @@
 # define SINCOSF_FUNC SINCOSF
 #endif
 
+/* PI/2 with 98 bits of accuracy.  */
+static const double PI_2_hi = 0x1.921fb544p+0;
+static const double PI_2_lo = 0x1.0b4611a626332p-34;
+
+static const double SMALL = 0x1p-50; /* 2^-50.  */
+static const double inv_PI_4 = 0x1.45f306dc9c883p+0; /* 4/PI.  */
+
+#define FLOAT_EXPONENT_SHIFT 23
+#define FLOAT_EXPONENT_BIAS 127
+
+static const double pio2_table[] = {
+  0 * M_PI_2,
+  1 * M_PI_2,
+  2 * M_PI_2,
+  3 * M_PI_2,
+  4 * M_PI_2,
+  5 * M_PI_2
+};
+
+static const double invpio4_table[] = {
+  0x0p+0,
+  0x1.45f306cp+0,
+  0x1.c9c882ap-28,
+  0x1.4fe13a8p-58,
+  0x1.f47d4dp-85,
+  0x1.bb81b6cp-112,
+  0x1.4acc9ep-142,
+  0x1.0e4107cp-169
+};
+
+static const double ones[] = { 1.0, -1.0 };
+
 /* Chebyshev constants for sin and cos, range -PI/4 - PI/4.  */
 static const __v2df V0 = { -0x1.5555555551cd9p-3, -0x1.ffffffffe98aep-2};
 static const __v2df V1 = { 0x1.1111110c2688bp-7, 0x1.55555545c50c7p-5 };
diff --git a/sysdeps/x86_64/fpu/s_sincosf_data.c b/sysdeps/x86_64/fpu/s_sincosf_data.c
deleted file mode 100644
index 1cc8931..0000000
--- a/sysdeps/x86_64/fpu/s_sincosf_data.c
+++ /dev/null
@@ -1 +0,0 @@
-/* Not needed.  */