S390: Use convert-to-fixed instruction for lrint functions.

If compiled with z196 zarch support, the convert-to-fixed instruction
is used to implement lrint, lrintf, lrintl.
Otherwise the common-code implementation is used.

3 files changed, 166 insertions, 0 deletions

diff --git a/sysdeps/s390/fpu/s_lrint.c b/sysdeps/s390/fpu/s_lrint.c
new file mode 100644
index 0000000000..7be60665b5
--- /dev/null
+++ b/sysdeps/s390/fpu/s_lrint.c
@@ -0,0 +1,55 @@
+/* lrint() - S390 version.
+   Copyright (C) 2019 Free Software Foundation, Inc.
+
+   This file is part of the GNU C Library.
+
+   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
+   <https://www.gnu.org/licenses/>.  */
+
+#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
+# include <math.h>
+# include <libm-alias-double.h>
+
+/* The sizeof (long int) differs between s390x (8byte) and s390 (4byte).
+   Thus we need different instructions as the target size is encoded there.
+   Note: On s390 this instruction is only used if build with -mzarch.  */
+# ifdef __s390x__
+#  define INSN "cgdbra"
+# else
+#  define INSN "cfdbra"
+# endif
+
+long int
+__lrint (double x)
+{
+  long int y;
+  /* The z196 zarch "convert to fixed" (cgdbra) instruction is rounding
+     according to current rounding mode (M3-field: 0).
+     First convert x with suppressed inexact exception and check if the
+     resulting value is beyond the target limits (indicated by cc=3;
+     Note: a nan is also indicated by cc=3).
+     If the resulting value is within the target limits, redo
+     without suppressing the inexact exception.  */
+  __asm__ (INSN " %0,0,%1,4 \n\t"
+	   "jo 1f \n\t"
+	   INSN " %0,0,%1,0 \n\t"
+	   "1:"
+	   : "=&d" (y) : "f" (x) : "cc");
+  return y;
+}
+libm_alias_double (__lrint, lrint)
+
+#else
+# include <sysdeps/ieee754/dbl-64/s_lrint.c>
+#endif
diff --git a/sysdeps/s390/fpu/s_lrintf.c b/sysdeps/s390/fpu/s_lrintf.c
new file mode 100644
index 0000000000..d6a2a4081a
--- /dev/null
+++ b/sysdeps/s390/fpu/s_lrintf.c
@@ -0,0 +1,55 @@
+/* lrintf() - S390 version.
+   Copyright (C) 2019 Free Software Foundation, Inc.
+
+   This file is part of the GNU C Library.
+
+   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
+   <https://www.gnu.org/licenses/>.  */
+
+#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
+# include <math.h>
+# include <libm-alias-float.h>
+
+/* The sizeof (long int) differs between s390x (8byte) and s390 (4byte).
+   Thus we need different instructions as the target size is encoded there.
+   Note: On s390 this instruction is only used if build with -mzarch.  */
+# ifdef __s390x__
+#  define INSN "cgebra"
+# else
+#  define INSN "cfebra"
+# endif
+
+long int
+__lrintf (float x)
+{
+  long int y;
+  /* The z196 zarch "convert to fixed" (cgebra) instruction is rounding
+     according to current rounding mode (M3-field: 0).
+     First convert x with suppressed inexact exception and check if the
+     resulting value is beyond the target limits (indicated by cc=3;
+     Note: a nan is also indicated by cc=3).
+     If the resulting value is within the target limits, redo
+     without suppressing the inexact exception.  */
+  __asm__ (INSN " %0,0,%1,4 \n\t"
+	   "jo 1f \n\t"
+	   INSN " %0,0,%1,0 \n\t"
+	   "1:"
+	   : "=&d" (y) : "f" (x) : "cc");
+  return y;
+}
+libm_alias_float (__lrint, lrint)
+
+#else
+# include <sysdeps/ieee754/flt-32/s_lrintf.c>
+#endif
diff --git a/sysdeps/s390/fpu/s_lrintl.c b/sysdeps/s390/fpu/s_lrintl.c
new file mode 100644
index 0000000000..2d386ecff9
--- /dev/null
+++ b/sysdeps/s390/fpu/s_lrintl.c
@@ -0,0 +1,56 @@
+/* lrintl() - S390 version.
+   Copyright (C) 2019 Free Software Foundation, Inc.
+
+   This file is part of the GNU C Library.
+
+   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
+   <https://www.gnu.org/licenses/>.  */
+
+#ifdef HAVE_S390_MIN_Z196_ZARCH_ASM_SUPPORT
+# include <math.h>
+# include <math_private.h>
+# include <libm-alias-ldouble.h>
+
+/* The sizeof (long int) differs between s390x (8byte) and s390 (4byte).
+   Thus we need different instructions as the target size is encoded there.
+   Note: On s390 this instruction is only used if build with -mzarch.  */
+# ifdef __s390x__
+#  define INSN "cgxbra"
+# else
+#  define INSN "cfxbra"
+# endif
+
+long int
+__lrintl (_Float128 x)
+{
+  long int y;
+  /* The z196 zarch "convert to fixed" (cgxbra) instruction is rounding
+     according to current rounding mode (M3-field: 0).
+     First convert x with suppressed inexact exception and check if the
+     resulting value is beyond the target limits (indicated by cc=3;
+     Note: a nan is also indicated by cc=3).
+     If the resulting value is within the target limits, redo
+     without suppressing the inexact exception.  */
+  __asm__ (INSN " %0,0,%1,4 \n\t"
+	   "jo 1f \n\t"
+	   INSN " %0,0,%1,0 \n\t"
+	   "1:"
+	   : "=&d" (y) : "f" (x) : "cc");
+  return y;
+}
+libm_alias_ldouble (__lrint, lrint)
+
+#else
+# include <sysdeps/ieee754/ldbl-128/s_lrintl.c>
+#endif