Update.

1999-12-29  Ulrich Drepper  <drepper@cygnus.com>

	* soft-fp/*: Tons of new files to implement floating-point arithmetic
	in software.
	Contributed by  Richard Henderson, Jakub Jelinek and others.

1 files changed, 87 insertions, 0 deletions

diff --git a/soft-fp/sysdeps/i386/sfp-machine.h b/soft-fp/sysdeps/i386/sfp-machine.h
new file mode 100644
index 0000000000..047452b905
--- /dev/null
+++ b/soft-fp/sysdeps/i386/sfp-machine.h
@@ -0,0 +1,87 @@
+#define _FP_W_TYPE_SIZE		32
+#define _FP_W_TYPE		unsigned long
+#define _FP_WS_TYPE		signed long
+#define _FP_I_TYPE		long
+
+#define __FP_FRAC_ADD_2(rh, rl, xh, xl, yh, yl)				\
+  __asm__("addl %5,%1; adcl %3,%0"					\
+	  : "=r"(rh), "=r"(rl)						\
+	  : "%0"(xh), "g"(yh), "%1"(xl), "g"(yl)			\
+	  : "cc")
+
+#define __FP_FRAC_ADD_4(r3,r2,r1,r0,x3,x2,x1,x0,y3,y2,y1,y0)		\
+  do { 									\
+    __asm__ volatile("addl %5,%1; adcl %3,%0"				\
+		     : "=r"(r1), "=r"(r0)				\
+		     : "%0"(x1), "g"(y1), "%1"(x0), "g"(y0)		\
+		     : "cc");						\
+    __asm__ volatile("adcl %5,%1; adcl %3,%0"				\
+		     : "=r"(r3), "=r"(r2)				\
+		     : "%0"(x3), "g"(y3), "%1"(x2), "g"(y2)		\
+		     : "cc");						\
+  } while (0)
+
+#define __FP_FRAC_SUB_2(rh, rl, xh, xl, yh, yl)				\
+  __asm__("subl %5,%1; sbbl %4,%0"					\
+	  : "=r"(rh), "=r"(rl)						\
+	  : "0"(xh), "1"(xl), "g"(yh), "g"(yl)				\
+	  : "cc")
+
+#define __FP_CLZ(r, x)							\
+  do {									\
+    __asm__("bsrl %1,%0" : "=r"(r) : "g"(x) : "cc");			\
+    r ^= 31;								\
+  } while (0)
+
+#define _i386_mul_32_64(rh, rl, x, y)					\
+  __asm__("mull %2" : "=d"(rh), "=a"(rl) : "%g"(x), "1"(y) : "cc")
+
+#define _i386_div_64_32(q, r, nh, nl, d)				\
+  __asm__ ("divl %4" : "=a"(q), "=d"(r) : "0"(nl), "1"(nh), "g"(d) : "cc")
+
+
+#define _FP_MUL_MEAT_S(R,X,Y)					\
+  _FP_MUL_MEAT_1_wide(_FP_WFRACBITS_S,R,X,Y,_i386_mul_32_64)
+#define _FP_MUL_MEAT_D(R,X,Y)					\
+  _FP_MUL_MEAT_2_wide(_FP_WFRACBITS_D,R,X,Y,_i386_mul_32_64)
+
+#define _FP_DIV_MEAT_S(R,X,Y)	_FP_DIV_MEAT_1_udiv(S,R,X,Y)
+#define _FP_DIV_MEAT_D(R,X,Y)	_FP_DIV_MEAT_2_udiv(D,R,X,Y)
+
+#define _FP_NANFRAC_S		_FP_QNANBIT_S
+#define _FP_NANFRAC_D		_FP_QNANBIT_D, 0
+#define _FP_NANSIGN_S		1
+#define _FP_NANSIGN_D		1
+
+#define _FP_KEEPNANFRACP 1
+/* Here is something Intel misdesigned: the specs don't define
+   the case where we have two NaNs with same mantissas, but
+   different sign. Different operations pick up different NaNs.
+ */
+#define _FP_CHOOSENAN(fs, wc, R, X, Y, OP)			\
+  do {								\
+    if (_FP_FRAC_GT_##wc(X, Y)					\
+	|| (_FP_FRAC_EQ_##wc(X,Y) && (OP == '+' || OP == '*')))	\
+      {								\
+	R##_s = X##_s;						\
+        _FP_FRAC_COPY_##wc(R,X);				\
+      }								\
+    else							\
+      {								\
+	R##_s = Y##_s;						\
+        _FP_FRAC_COPY_##wc(R,Y);				\
+      }								\
+    R##_c = FP_CLS_NAN;						\
+  } while (0)
+
+#define FP_EX_INVALID           (1 << 0)
+#define FP_EX_DENORM		(1 << 1)
+#define FP_EX_DIVZERO           (1 << 2)
+#define FP_EX_OVERFLOW          (1 << 3)
+#define FP_EX_UNDERFLOW         (1 << 4)
+#define FP_EX_INEXACT           (1 << 5)
+
+#define FP_RND_NEAREST		0
+#define FP_RND_ZERO		3
+#define FP_RND_PINF		2
+#define FP_RND_MINF		1