powerpc64: add single-instruction math functions

while the official elfv2 abi for "powerpc64le" sets power8 as the
baseline isa, we use it for both little and big endian powerpc64
targets and need to maintain compatibility with pre-power8 models. the
instructions for sqrt, fabs, and fma are in the baseline isa; support
for the rest is conditional via predefined isa-level macros.

patch by David Edelsohn.

src/math/powerpc64/ceil.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+double ceil(double x)
+{
+	__asm__ ("frip %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}
+
+#else
+
+#include "../ceil.c"
+
+#endif

src/math/powerpc64/ceilf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+float ceilf(float x)
+{
+	__asm__ ("frip %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}
+
+#else
+
+#include "../ceilf.c"
+
+#endif

src/math/powerpc64/fabs.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+double fabs(double x)
+{
+	__asm__ ("fabs %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}

src/math/powerpc64/fabsf.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+float fabsf(float x)
+{
+	__asm__ ("fabs %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}

src/math/powerpc64/floor.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+double floor(double x)
+{
+	__asm__ ("frim %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}
+
+#else
+
+#include "../floor.c"
+
+#endif

src/math/powerpc64/floorf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+float floorf(float x)
+{
+	__asm__ ("frim %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}
+
+#else
+
+#include "../floorf.c"
+
+#endif

src/math/powerpc64/fma.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+double fma(double x, double y, double z)
+{
+	__asm__ ("fmadd %0, %1, %2, %3" : "=d"(x) : "d"(x), "d"(y), "d"(z));
+	return x;
+}

src/math/powerpc64/fmaf.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+float fmaf(float x, float y, float z)
+{
+	__asm__ ("fmadds %0, %1, %2, %3" : "=f"(x) : "f"(x), "f"(y), "f"(z));
+	return x;
+}

src/math/powerpc64/fmax.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+double fmax(double x, double y)
+{
+	__asm__ ("xsmaxdp %x0, %x1, %x2" : "=ws"(x) : "ws"(x), "ws"(y));
+	return x;
+}
+
+#else
+
+#include "../fmax.c"
+
+#endif

src/math/powerpc64/fmaxf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+float fmaxf(float x, float y)
+{
+	__asm__ ("xsmaxdp %x0, %x1, %x2" : "=ww"(x) : "ww"(x), "ww"(y));
+	return x;
+}
+
+#else
+
+#include "../fmaxf.c"
+
+#endif

src/math/powerpc64/fmin.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+double fmin(double x, double y)
+{
+	__asm__ ("xsmindp %x0, %x1, %x2" : "=ws"(x) : "ws"(x), "ws"(y));
+	return x;
+}
+
+#else
+
+#include "../fmin.c"
+
+#endif

src/math/powerpc64/fminf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+float fminf(float x, float y)
+{
+	__asm__ ("xsmindp %x0, %x1, %x2" : "=ww"(x) : "ww"(x), "ww"(y));
+	return x;
+}
+
+#else
+
+#include "../fminf.c"
+
+#endif

src/math/powerpc64/lrint.c

@@ -0,0 +1,16 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+long lrint(double x)
+{
+	long n;
+	__asm__ ("fctid %0, %1" : "=d"(n) : "d"(x));
+	return n;
+}
+
+#else
+
+#include "../lrint.c"
+
+#endif

src/math/powerpc64/lrintf.c

@@ -0,0 +1,16 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+long lrintf(float x)
+{
+	long n;
+	__asm__ ("fctid %0, %1" : "=d"(n) : "f"(x));
+	return n;
+}
+
+#else
+
+#include "../lrintf.c"
+
+#endif

src/math/powerpc64/lround.c

@@ -0,0 +1,18 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+long lround(double x)
+{
+	long n;
+	__asm__ (
+		"xsrdpi %1, %1\n"
+		"fctid %0, %1\n" : "=d"(n), "+d"(x));
+	return n;
+}
+
+#else
+
+#include "../lround.c"
+
+#endif

src/math/powerpc64/lroundf.c

@@ -0,0 +1,18 @@
+#include <math.h>
+
+#ifdef __VSX__
+
+long lroundf(float x)
+{
+	long n;
+	__asm__ (
+		"xsrdpi %1, %1\n"
+		"fctid %0, %1\n" : "=d"(n), "+f"(x));
+	return n;
+}
+
+#else
+
+#include "../lroundf.c"
+
+#endif

src/math/powerpc64/round.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+double round(double x)
+{
+	__asm__ ("frin %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}
+
+#else
+
+#include "../round.c"
+
+#endif

src/math/powerpc64/roundf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+float roundf(float x)
+{
+	__asm__ ("frin %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}
+
+#else
+
+#include "../roundf.c"
+
+#endif

src/math/powerpc64/sqrt.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+double sqrt(double x)
+{
+	__asm__ ("fsqrt %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}

src/math/powerpc64/sqrtf.c

@@ -0,0 +1,7 @@
+#include <math.h>
+
+float sqrtf(float x)
+{
+	__asm__ ("fsqrts %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}

src/math/powerpc64/trunc.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+double trunc(double x)
+{
+	__asm__ ("friz %0, %1" : "=d"(x) : "d"(x));
+	return x;
+}
+
+#else
+
+#include "../trunc.c"
+
+#endif

src/math/powerpc64/truncf.c

@@ -0,0 +1,15 @@
+#include <math.h>
+
+#ifdef _ARCH_PWR5X
+
+float truncf(float x)
+{
+	__asm__ ("friz %0, %1" : "=f"(x) : "f"(x));
+	return x;
+}
+
+#else
+
+#include "../truncf.c"
+
+#endif