S/390: Hardware iconv modules.

5 files changed, 1601 insertions, 0 deletions

diff --git a/sysdeps/s390/s390-64/Makefile b/sysdeps/s390/s390-64/Makefile
index 0a5051449d..1814f37abd 100644
--- a/sysdeps/s390/s390-64/Makefile
+++ b/sysdeps/s390/s390-64/Makefile
@@ -9,3 +9,70 @@ CFLAGS-rtld.c += -Wno-uninitialized -Wno-unused
 CFLAGS-dl-load.c += -Wno-unused
 CFLAGS-dl-reloc.c += -Wno-unused
 endif
+
+ifeq ($(subdir),iconvdata)
+ISO-8859-1_CP037_Z900-routines := iso-8859-1_cp037_z900
+ISO-8859-1_CP037_Z900-map := gconv.map
+
+UTF8_UTF32_Z9-routines := utf8-utf32-z9
+UTF8_UTF32_Z9-map := gconv.map
+
+UTF16_UTF32_Z9-routines := utf16-utf32-z9
+UTF16_UTF32_Z9-map := gconv.map
+
+UTF8_UTF16_Z9-routines := utf8-utf16-z9
+UTF8_UTF16_Z9-map := gconv.map
+
+s390x-iconv-modules = ISO-8859-1_CP037_Z900 UTF8_UTF16_Z9 UTF16_UTF32_Z9 UTF8_UTF32_Z9
+
+extra-modules-left += $(s390x-iconv-modules)
+include extra-module.mk
+
+extra-objs      += $(addsuffix .so, $(s390x-iconv-modules))
+install-others  += $(patsubst %, $(inst_gconvdir)/%.so, $(s390x-iconv-modules))
+
+distribute += iso-8859-1_cp037_z900.c utf8-utf32-z9.c utf16-utf32-z9.c utf8-utf16-z9.c
+
+$(patsubst %, $(inst_gconvdir)/%.so, $(s390x-iconv-modules)) : \
+$(inst_gconvdir)/%.so: $(objpfx)%.so $(+force)
+	$(do-install-program)
+
+$(objpfx)gconv-modules-s390: gconv-modules $(+force)
+	cp $< $@
+	echo >> $@
+	echo "# S/390 hardware accelerated modules" >> $@
+	echo -n "module	ISO-8859-1//		IBM037//	" >> $@
+	echo "	ISO-8859-1_CP037_Z900	1" >> $@
+	echo -n "module	IBM037//		ISO-8859-1//	" >> $@
+	echo "	ISO-8859-1_CP037_Z900	1" >> $@
+	echo -n "module	ISO-10646/UTF8/		UTF-32//	" >> $@
+	echo "	UTF8_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-32BE//		ISO-10646/UTF8/	" >> $@
+	echo "	UTF8_UTF32_Z9		1" >> $@
+	echo -n "module	ISO-10646/UTF8/		UTF-32BE//	" >> $@
+	echo "	UTF8_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-16BE//		UTF-32//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-32BE//		UTF-16//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	INTERNAL		UTF-16//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-32BE//		UTF-16BE//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	INTERNAL		UTF-16BE//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-16BE//		UTF-32BE//	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-16BE//		INTERNAL	" >> $@
+	echo "	UTF16_UTF32_Z9		1" >> $@
+	echo -n "module	UTF-16BE//		ISO-10646/UTF8/	" >> $@
+	echo "	UTF8_UTF16_Z9		1" >> $@
+	echo -n "module	ISO-10646/UTF8/		UTF-16//	" >> $@
+	echo "	UTF8_UTF16_Z9		1" >> $@
+	echo -n "module	ISO-10646/UTF8/		UTF-16BE//	" >> $@
+	echo "	UTF8_UTF16_Z9		1" >> $@
+
+$(inst_gconvdir)/gconv-modules: $(objpfx)gconv-modules-s390 $(+force)
+	$(do-install)
+
+endif
diff --git a/sysdeps/s390/s390-64/iso-8859-1_cp037_z900.c b/sysdeps/s390/s390-64/iso-8859-1_cp037_z900.c
new file mode 100644
index 0000000000..d4c4931f22
--- /dev/null
+++ b/sysdeps/s390/s390-64/iso-8859-1_cp037_z900.c
@@ -0,0 +1,238 @@
+/* Conversion between ISO 8859-1 and IBM037.
+
+   This module uses the Z900 variant of the Translate One To One
+   instruction.
+   Copyright (C) 1997-2009 Free Software Foundation, Inc.
+
+   Author: Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
+   Based on the work by Ulrich Drepper  <drepper@cygnus.com>, 1997.
+
+   Thanks to Daniel Appich who covered the relevant performance work
+   in his diploma thesis.
+
+   This 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.
+
+   This 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+
+// conversion table from ISO-8859-1 to IBM037
+static const unsigned char table_iso8859_1_to_cp037[256]
+__attribute__ ((aligned (8))) =
+{
+  [0x00] = 0x00, [0x01] = 0x01, [0x02] = 0x02, [0x03] = 0x03,
+  [0x04] = 0x37, [0x05] = 0x2D, [0x06] = 0x2E, [0x07] = 0x2F,
+  [0x08] = 0x16, [0x09] = 0x05, [0x0A] = 0x25, [0x0B] = 0x0B,
+  [0x0C] = 0x0C, [0x0D] = 0x0D, [0x0E] = 0x0E, [0x0F] = 0x0F,
+  [0x10] = 0x10, [0x11] = 0x11, [0x12] = 0x12, [0x13] = 0x13,
+  [0x14] = 0x3C, [0x15] = 0x3D, [0x16] = 0x32, [0x17] = 0x26,
+  [0x18] = 0x18, [0x19] = 0x19, [0x1A] = 0x3F, [0x1B] = 0x27,
+  [0x1C] = 0x1C, [0x1D] = 0x1D, [0x1E] = 0x1E, [0x1F] = 0x1F,
+  [0x20] = 0x40, [0x21] = 0x5A, [0x22] = 0x7F, [0x23] = 0x7B,
+  [0x24] = 0x5B, [0x25] = 0x6C, [0x26] = 0x50, [0x27] = 0x7D,
+  [0x28] = 0x4D, [0x29] = 0x5D, [0x2A] = 0x5C, [0x2B] = 0x4E,
+  [0x2C] = 0x6B, [0x2D] = 0x60, [0x2E] = 0x4B, [0x2F] = 0x61,
+  [0x30] = 0xF0, [0x31] = 0xF1, [0x32] = 0xF2, [0x33] = 0xF3,
+  [0x34] = 0xF4, [0x35] = 0xF5, [0x36] = 0xF6, [0x37] = 0xF7,
+  [0x38] = 0xF8, [0x39] = 0xF9, [0x3A] = 0x7A, [0x3B] = 0x5E,
+  [0x3C] = 0x4C, [0x3D] = 0x7E, [0x3E] = 0x6E, [0x3F] = 0x6F,
+  [0x40] = 0x7C, [0x41] = 0xC1, [0x42] = 0xC2, [0x43] = 0xC3,
+  [0x44] = 0xC4, [0x45] = 0xC5, [0x46] = 0xC6, [0x47] = 0xC7,
+  [0x48] = 0xC8, [0x49] = 0xC9, [0x4A] = 0xD1, [0x4B] = 0xD2,
+  [0x4C] = 0xD3, [0x4D] = 0xD4, [0x4E] = 0xD5, [0x4F] = 0xD6,
+  [0x50] = 0xD7, [0x51] = 0xD8, [0x52] = 0xD9, [0x53] = 0xE2,
+  [0x54] = 0xE3, [0x55] = 0xE4, [0x56] = 0xE5, [0x57] = 0xE6,
+  [0x58] = 0xE7, [0x59] = 0xE8, [0x5A] = 0xE9, [0x5B] = 0xBA,
+  [0x5C] = 0xE0, [0x5D] = 0xBB, [0x5E] = 0xB0, [0x5F] = 0x6D,
+  [0x60] = 0x79, [0x61] = 0x81, [0x62] = 0x82, [0x63] = 0x83,
+  [0x64] = 0x84, [0x65] = 0x85, [0x66] = 0x86, [0x67] = 0x87,
+  [0x68] = 0x88, [0x69] = 0x89, [0x6A] = 0x91, [0x6B] = 0x92,
+  [0x6C] = 0x93, [0x6D] = 0x94, [0x6E] = 0x95, [0x6F] = 0x96,
+  [0x70] = 0x97, [0x71] = 0x98, [0x72] = 0x99, [0x73] = 0xA2,
+  [0x74] = 0xA3, [0x75] = 0xA4, [0x76] = 0xA5, [0x77] = 0xA6,
+  [0x78] = 0xA7, [0x79] = 0xA8, [0x7A] = 0xA9, [0x7B] = 0xC0,
+  [0x7C] = 0x4F, [0x7D] = 0xD0, [0x7E] = 0xA1, [0x7F] = 0x07,
+  [0x80] = 0x20, [0x81] = 0x21, [0x82] = 0x22, [0x83] = 0x23,
+  [0x84] = 0x24, [0x85] = 0x15, [0x86] = 0x06, [0x87] = 0x17,
+  [0x88] = 0x28, [0x89] = 0x29, [0x8A] = 0x2A, [0x8B] = 0x2B,
+  [0x8C] = 0x2C, [0x8D] = 0x09, [0x8E] = 0x0A, [0x8F] = 0x1B,
+  [0x90] = 0x30, [0x91] = 0x31, [0x92] = 0x1A, [0x93] = 0x33,
+  [0x94] = 0x34, [0x95] = 0x35, [0x96] = 0x36, [0x97] = 0x08,
+  [0x98] = 0x38, [0x99] = 0x39, [0x9A] = 0x3A, [0x9B] = 0x3B,
+  [0x9C] = 0x04, [0x9D] = 0x14, [0x9E] = 0x3E, [0x9F] = 0xFF,
+  [0xA0] = 0x41, [0xA1] = 0xAA, [0xA2] = 0x4A, [0xA3] = 0xB1,
+  [0xA4] = 0x9F, [0xA5] = 0xB2, [0xA6] = 0x6A, [0xA7] = 0xB5,
+  [0xA8] = 0xBD, [0xA9] = 0xB4, [0xAA] = 0x9A, [0xAB] = 0x8A,
+  [0xAC] = 0x5F, [0xAD] = 0xCA, [0xAE] = 0xAF, [0xAF] = 0xBC,
+  [0xB0] = 0x90, [0xB1] = 0x8F, [0xB2] = 0xEA, [0xB3] = 0xFA,
+  [0xB4] = 0xBE, [0xB5] = 0xA0, [0xB6] = 0xB6, [0xB7] = 0xB3,
+  [0xB8] = 0x9D, [0xB9] = 0xDA, [0xBA] = 0x9B, [0xBB] = 0x8B,
+  [0xBC] = 0xB7, [0xBD] = 0xB8, [0xBE] = 0xB9, [0xBF] = 0xAB,
+  [0xC0] = 0x64, [0xC1] = 0x65, [0xC2] = 0x62, [0xC3] = 0x66,
+  [0xC4] = 0x63, [0xC5] = 0x67, [0xC6] = 0x9E, [0xC7] = 0x68,
+  [0xC8] = 0x74, [0xC9] = 0x71, [0xCA] = 0x72, [0xCB] = 0x73,
+  [0xCC] = 0x78, [0xCD] = 0x75, [0xCE] = 0x76, [0xCF] = 0x77,
+  [0xD0] = 0xAC, [0xD1] = 0x69, [0xD2] = 0xED, [0xD3] = 0xEE,
+  [0xD4] = 0xEB, [0xD5] = 0xEF, [0xD6] = 0xEC, [0xD7] = 0xBF,
+  [0xD8] = 0x80, [0xD9] = 0xFD, [0xDA] = 0xFE, [0xDB] = 0xFB,
+  [0xDC] = 0xFC, [0xDD] = 0xAD, [0xDE] = 0xAE, [0xDF] = 0x59,
+  [0xE0] = 0x44, [0xE1] = 0x45, [0xE2] = 0x42, [0xE3] = 0x46,
+  [0xE4] = 0x43, [0xE5] = 0x47, [0xE6] = 0x9C, [0xE7] = 0x48,
+  [0xE8] = 0x54, [0xE9] = 0x51, [0xEA] = 0x52, [0xEB] = 0x53,
+  [0xEC] = 0x58, [0xED] = 0x55, [0xEE] = 0x56, [0xEF] = 0x57,
+  [0xF0] = 0x8C, [0xF1] = 0x49, [0xF2] = 0xCD, [0xF3] = 0xCE,
+  [0xF4] = 0xCB, [0xF5] = 0xCF, [0xF6] = 0xCC, [0xF7] = 0xE1,
+  [0xF8] = 0x70, [0xF9] = 0xDD, [0xFA] = 0xDE, [0xFB] = 0xDB,
+  [0xFC] = 0xDC, [0xFD] = 0x8D, [0xFE] = 0x8E, [0xFF] = 0xDF
+};
+
+// conversion table from IBM037 to ISO-8859-1
+static const unsigned char table_cp037_iso8859_1[256]
+__attribute__ ((aligned (8))) =
+{
+  [0x00] = 0x00, [0x01] = 0x01, [0x02] = 0x02, [0x03] = 0x03,
+  [0x04] = 0x9C, [0x05] = 0x09, [0x06] = 0x86, [0x07] = 0x7F,
+  [0x08] = 0x97, [0x09] = 0x8D, [0x0A] = 0x8E, [0x0B] = 0x0B,
+  [0x0C] = 0x0C, [0x0D] = 0x0D, [0x0E] = 0x0E, [0x0F] = 0x0F,
+  [0x10] = 0x10, [0x11] = 0x11, [0x12] = 0x12, [0x13] = 0x13,
+  [0x14] = 0x9D, [0x15] = 0x85, [0x16] = 0x08, [0x17] = 0x87,
+  [0x18] = 0x18, [0x19] = 0x19, [0x1A] = 0x92, [0x1B] = 0x8F,
+  [0x1C] = 0x1C, [0x1D] = 0x1D, [0x1E] = 0x1E, [0x1F] = 0x1F,
+  [0x20] = 0x80, [0x21] = 0x81, [0x22] = 0x82, [0x23] = 0x83,
+  [0x24] = 0x84, [0x25] = 0x0A, [0x26] = 0x17, [0x27] = 0x1B,
+  [0x28] = 0x88, [0x29] = 0x89, [0x2A] = 0x8A, [0x2B] = 0x8B,
+  [0x2C] = 0x8C, [0x2D] = 0x05, [0x2E] = 0x06, [0x2F] = 0x07,
+  [0x30] = 0x90, [0x31] = 0x91, [0x32] = 0x16, [0x33] = 0x93,
+  [0x34] = 0x94, [0x35] = 0x95, [0x36] = 0x96, [0x37] = 0x04,
+  [0x38] = 0x98, [0x39] = 0x99, [0x3A] = 0x9A, [0x3B] = 0x9B,
+  [0x3C] = 0x14, [0x3D] = 0x15, [0x3E] = 0x9E, [0x3F] = 0x1A,
+  [0x40] = 0x20, [0x41] = 0xA0, [0x42] = 0xE2, [0x43] = 0xE4,
+  [0x44] = 0xE0, [0x45] = 0xE1, [0x46] = 0xE3, [0x47] = 0xE5,
+  [0x48] = 0xE7, [0x49] = 0xF1, [0x4A] = 0xA2, [0x4B] = 0x2E,
+  [0x4C] = 0x3C, [0x4D] = 0x28, [0x4E] = 0x2B, [0x4F] = 0x7C,
+  [0x50] = 0x26, [0x51] = 0xE9, [0x52] = 0xEA, [0x53] = 0xEB,
+  [0x54] = 0xE8, [0x55] = 0xED, [0x56] = 0xEE, [0x57] = 0xEF,
+  [0x58] = 0xEC, [0x59] = 0xDF, [0x5A] = 0x21, [0x5B] = 0x24,
+  [0x5C] = 0x2A, [0x5D] = 0x29, [0x5E] = 0x3B, [0x5F] = 0xAC,
+  [0x60] = 0x2D, [0x61] = 0x2F, [0x62] = 0xC2, [0x63] = 0xC4,
+  [0x64] = 0xC0, [0x65] = 0xC1, [0x66] = 0xC3, [0x67] = 0xC5,
+  [0x68] = 0xC7, [0x69] = 0xD1, [0x6A] = 0xA6, [0x6B] = 0x2C,
+  [0x6C] = 0x25, [0x6D] = 0x5F, [0x6E] = 0x3E, [0x6F] = 0x3F,
+  [0x70] = 0xF8, [0x71] = 0xC9, [0x72] = 0xCA, [0x73] = 0xCB,
+  [0x74] = 0xC8, [0x75] = 0xCD, [0x76] = 0xCE, [0x77] = 0xCF,
+  [0x78] = 0xCC, [0x79] = 0x60, [0x7A] = 0x3A, [0x7B] = 0x23,
+  [0x7C] = 0x40, [0x7D] = 0x27, [0x7E] = 0x3D, [0x7F] = 0x22,
+  [0x80] = 0xD8, [0x81] = 0x61, [0x82] = 0x62, [0x83] = 0x63,
+  [0x84] = 0x64, [0x85] = 0x65, [0x86] = 0x66, [0x87] = 0x67,
+  [0x88] = 0x68, [0x89] = 0x69, [0x8A] = 0xAB, [0x8B] = 0xBB,
+  [0x8C] = 0xF0, [0x8D] = 0xFD, [0x8E] = 0xFE, [0x8F] = 0xB1,
+  [0x90] = 0xB0, [0x91] = 0x6A, [0x92] = 0x6B, [0x93] = 0x6C,
+  [0x94] = 0x6D, [0x95] = 0x6E, [0x96] = 0x6F, [0x97] = 0x70,
+  [0x98] = 0x71, [0x99] = 0x72, [0x9A] = 0xAA, [0x9B] = 0xBA,
+  [0x9C] = 0xE6, [0x9D] = 0xB8, [0x9E] = 0xC6, [0x9F] = 0xA4,
+  [0xA0] = 0xB5, [0xA1] = 0x7E, [0xA2] = 0x73, [0xA3] = 0x74,
+  [0xA4] = 0x75, [0xA5] = 0x76, [0xA6] = 0x77, [0xA7] = 0x78,
+  [0xA8] = 0x79, [0xA9] = 0x7A, [0xAA] = 0xA1, [0xAB] = 0xBF,
+  [0xAC] = 0xD0, [0xAD] = 0xDD, [0xAE] = 0xDE, [0xAF] = 0xAE,
+  [0xB0] = 0x5E, [0xB1] = 0xA3, [0xB2] = 0xA5, [0xB3] = 0xB7,
+  [0xB4] = 0xA9, [0xB5] = 0xA7, [0xB6] = 0xB6, [0xB7] = 0xBC,
+  [0xB8] = 0xBD, [0xB9] = 0xBE, [0xBA] = 0x5B, [0xBB] = 0x5D,
+  [0xBC] = 0xAF, [0xBD] = 0xA8, [0xBE] = 0xB4, [0xBF] = 0xD7,
+  [0xC0] = 0x7B, [0xC1] = 0x41, [0xC2] = 0x42, [0xC3] = 0x43,
+  [0xC4] = 0x44, [0xC5] = 0x45, [0xC6] = 0x46, [0xC7] = 0x47,
+  [0xC8] = 0x48, [0xC9] = 0x49, [0xCA] = 0xAD, [0xCB] = 0xF4,
+  [0xCC] = 0xF6, [0xCD] = 0xF2, [0xCE] = 0xF3, [0xCF] = 0xF5,
+  [0xD0] = 0x7D, [0xD1] = 0x4A, [0xD2] = 0x4B, [0xD3] = 0x4C,
+  [0xD4] = 0x4D, [0xD5] = 0x4E, [0xD6] = 0x4F, [0xD7] = 0x50,
+  [0xD8] = 0x51, [0xD9] = 0x52, [0xDA] = 0xB9, [0xDB] = 0xFB,
+  [0xDC] = 0xFC, [0xDD] = 0xF9, [0xDE] = 0xFA, [0xDF] = 0xFF,
+  [0xE0] = 0x5C, [0xE1] = 0xF7, [0xE2] = 0x53, [0xE3] = 0x54,
+  [0xE4] = 0x55, [0xE5] = 0x56, [0xE6] = 0x57, [0xE7] = 0x58,
+  [0xE8] = 0x59, [0xE9] = 0x5A, [0xEA] = 0xB2, [0xEB] = 0xD4,
+  [0xEC] = 0xD6, [0xED] = 0xD2, [0xEE] = 0xD3, [0xEF] = 0xD5,
+  [0xF0] = 0x30, [0xF1] = 0x31, [0xF2] = 0x32, [0xF3] = 0x33,
+  [0xF4] = 0x34, [0xF5] = 0x35, [0xF6] = 0x36, [0xF7] = 0x37,
+  [0xF8] = 0x38, [0xF9] = 0x39, [0xFA] = 0xB3, [0xFB] = 0xDB,
+  [0xFC] = 0xDC, [0xFD] = 0xD9, [0xFE] = 0xDA, [0xFF] = 0x9F
+};
+
+/* Definitions used in the body of the `gconv' function.  */
+#define CHARSET_NAME		"ISO-8859-1//"
+#define FROM_LOOP		iso8859_1_to_cp037_z900
+#define TO_LOOP			cp037_to_iso8859_1_z900
+#define DEFINE_INIT		1
+#define DEFINE_FINI		1
+#define MIN_NEEDED_FROM		1
+#define MIN_NEEDED_TO		1
+
+/* The Z900 variant of troo forces us to always specify a test
+   character which ends the translation.  So if we run into the
+   situation where the translation has been interrupted due to the
+   test character we translate the character by hand and jump back
+   into the instruction.  */
+
+#define TROO_LOOP(TABLE)						\
+  {									\
+    register const unsigned char test asm ("0") = 0;			\
+    register const unsigned char *pTable asm ("1") = TABLE;		\
+    register unsigned char *pOutput asm ("2") = outptr;			\
+    register uint64_t length asm ("3");					\
+    const unsigned char* pInput = inptr;				\
+    uint64_t tmp;							\
+									\
+    length = (inend - inptr < outend - outptr				\
+	      ? inend - inptr : outend - outptr);			\
+									\
+    asm volatile ("0:                        \n\t"			\
+                  "  troo    %0,%1           \n\t"			\
+                  "  jz      1f              \n\t"			\
+                  "  jo      0b              \n\t"			\
+                  "  llgc    %3,0(%1)        \n\t"			\
+                  "  la      %3,0(%3,%4)     \n\t"			\
+                  "  mvc     0(1,%0),0(%3)   \n\t"			\
+                  "  aghi    %1,1            \n\t"			\
+                  "  aghi    %0,1            \n\t"			\
+                  "  aghi    %2,-1           \n\t"			\
+                  "  j       0b              \n\t"			\
+                  "1:                        \n"			\
+									\
+     : "+a" (pOutput), "+a" (pInput), "+d" (length), "=&a" (tmp)        \
+     : "a" (pTable), "d" (test)						\
+     : "cc");								\
+									\
+    inptr = pInput;							\
+    outptr = pOutput;							\
+  }
+
+/* First define the conversion function from ISO 8859-1 to CP037.  */
+#define MIN_NEEDED_INPUT	MIN_NEEDED_FROM
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_TO
+#define LOOPFCT			FROM_LOOP
+#define BODY TROO_LOOP (table_iso8859_1_to_cp037)
+
+#include <iconv/loop.c>
+
+
+/* Next, define the conversion function from CP037 to ISO 8859-1.  */
+#define MIN_NEEDED_INPUT	MIN_NEEDED_TO
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_FROM
+#define LOOPFCT			TO_LOOP
+#define BODY TROO_LOOP (table_cp037_iso8859_1);
+
+#include <iconv/loop.c>
+
+
+/* Now define the toplevel functions.  */
+#include <iconv/skeleton.c>
diff --git a/sysdeps/s390/s390-64/utf16-utf32-z9.c b/sysdeps/s390/s390-64/utf16-utf32-z9.c
new file mode 100644
index 0000000000..c9bccc98a7
--- /dev/null
+++ b/sysdeps/s390/s390-64/utf16-utf32-z9.c
@@ -0,0 +1,325 @@
+/* Conversion between UTF-16 and UTF-32 BE/internal.
+
+   This module uses the Z9-109 variants of the Convert Unicode
+   instructions.
+   Copyright (C) 1997-2009 Free Software Foundation, Inc.
+
+   Author: Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
+   Based on the work by Ulrich Drepper  <drepper@cygnus.com>, 1997.
+
+   Thanks to Daniel Appich who covered the relevant performance work
+   in his diploma thesis.
+
+   This 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.
+
+   This 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <dl-procinfo.h>
+#include <gconv.h>
+
+/* UTF-32 big endian byte order mark.  */
+#define BOM_UTF32               0x0000feffu
+
+/* UTF-16 big endian byte order mark.  */
+#define BOM_UTF16	        0xfeff
+
+#define DEFINE_INIT		0
+#define DEFINE_FINI		0
+#define MIN_NEEDED_FROM		2
+#define MAX_NEEDED_FROM		4
+#define MIN_NEEDED_TO		4
+#define FROM_LOOP		from_utf16_loop
+#define TO_LOOP			to_utf16_loop
+#define FROM_DIRECTION		(dir == from_utf16)
+#define PREPARE_LOOP							\
+  enum direction dir = ((struct utf16_data *) step->__data)->dir;	\
+  int emit_bom = ((struct utf16_data *) step->__data)->emit_bom;	\
+									\
+  if (emit_bom && !data->__internal_use					\
+      && data->__invocation_counter == 0)				\
+    {									\
+      if (dir == to_utf16)						\
+	{								\
+          /* Emit the UTF-16 Byte Order Mark.  */			\
+          if (__builtin_expect (outbuf + 2 > outend, 0))		\
+	    return __GCONV_FULL_OUTPUT;					\
+									\
+	  put16u (outbuf, BOM_UTF16);					\
+	  outbuf += 2;							\
+	}								\
+      else								\
+	{								\
+          /* Emit the UTF-32 Byte Order Mark.  */			\
+	  if (__builtin_expect (outbuf + 4 > outend, 0))		\
+	    return __GCONV_FULL_OUTPUT;					\
+									\
+	  put32u (outbuf, BOM_UTF32);					\
+	  outbuf += 4;							\
+	}								\
+    }
+
+/* Direction of the transformation.  */
+enum direction
+{
+  illegal_dir,
+  to_utf16,
+  from_utf16
+};
+
+struct utf16_data
+{
+  enum direction dir;
+  int emit_bom;
+};
+
+
+extern int gconv_init (struct __gconv_step *step);
+int
+gconv_init (struct __gconv_step *step)
+{
+  /* Determine which direction.  */
+  struct utf16_data *new_data;
+  enum direction dir = illegal_dir;
+  int emit_bom;
+  int result;
+
+  emit_bom = (__strcasecmp (step->__to_name, "UTF-32//") == 0
+	      || __strcasecmp (step->__to_name, "UTF-16//") == 0);
+
+  if (__strcasecmp (step->__from_name, "UTF-16BE//") == 0
+      && (__strcasecmp (step->__to_name, "UTF-32//") == 0
+	  || __strcasecmp (step->__to_name, "UTF-32BE//") == 0
+      	  || __strcasecmp (step->__to_name, "INTERNAL") == 0))
+    {
+      dir = from_utf16;
+    }
+  else if ((__strcasecmp (step->__to_name, "UTF-16//") == 0
+	    || __strcasecmp (step->__to_name, "UTF-16BE//") == 0)
+	   && (__strcasecmp (step->__from_name, "UTF-32BE//") == 0
+	       || __strcasecmp (step->__from_name, "INTERNAL") == 0))
+    {
+      dir = to_utf16;
+    }
+
+  result = __GCONV_NOCONV;
+  if (dir != illegal_dir)
+    {
+      new_data = (struct utf16_data *) malloc (sizeof (struct utf16_data));
+
+      result = __GCONV_NOMEM;
+      if (new_data != NULL)
+	{
+	  new_data->dir = dir;
+	  new_data->emit_bom = emit_bom;
+	  step->__data = new_data;
+
+	  if (dir == from_utf16)
+	    {
+	      step->__min_needed_from = MIN_NEEDED_FROM;
+	      step->__max_needed_from = MIN_NEEDED_FROM;
+	      step->__min_needed_to = MIN_NEEDED_TO;
+	      step->__max_needed_to = MIN_NEEDED_TO;
+	    }
+	  else
+	    {
+	      step->__min_needed_from = MIN_NEEDED_TO;
+	      step->__max_needed_from = MIN_NEEDED_TO;
+	      step->__min_needed_to = MIN_NEEDED_FROM;
+	      step->__max_needed_to = MIN_NEEDED_FROM;
+	    }
+
+	  step->__stateful = 0;
+
+	  result = __GCONV_OK;
+	}
+    }
+
+  return result;
+}
+
+
+extern void gconv_end (struct __gconv_step *data);
+void
+gconv_end (struct __gconv_step *data)
+{
+  free (data->__data);
+}
+
+/* The macro for the hardware loop.  This is used for both
+   directions.  */
+#define HARDWARE_CONVERT(INSTRUCTION)					\
+  {									\
+    register const unsigned char* pInput asm ("8") = inptr;		\
+    register unsigned long long inlen asm ("9") = inend - inptr;	\
+    register unsigned char* pOutput asm ("10") = outptr;		\
+    register unsigned long long outlen asm("11") = outend - outptr;	\
+    uint64_t cc = 0;							\
+									\
+    asm volatile ("0: " INSTRUCTION "  \n\t"				\
+                  "   jo     0b        \n\t"				\
+		  "   ipm    %2        \n"			        \
+		  : "+a" (pOutput), "+a" (pInput), "+d" (cc),		\
+		    "+d" (outlen), "+d" (inlen)				\
+		  :							\
+		  : "cc", "memory");					\
+									\
+    inptr = pInput;							\
+    outptr = pOutput;							\
+    cc >>= 28;          						\
+									\
+    if (cc == 1)							\
+      {									\
+	result = __GCONV_FULL_OUTPUT;					\
+	break;								\
+      }									\
+    else if (cc == 2)							\
+      {									\
+	result = __GCONV_ILLEGAL_INPUT;					\
+	break;								\
+      }									\
+  }
+
+/* Conversion function from UTF-16 to UTF-32 internal/BE.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_INPUT	MAX_NEEDED_FROM
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_TO
+#define LOOPFCT			FROM_LOOP
+/* The software routine is copied from utf-16.c (minus bytes
+   swapping).  */
+#define BODY								\
+  {									\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)				\
+      {									\
+	HARDWARE_CONVERT ("cu24 %0, %1, 1");				\
+	if (inptr != inend)						\
+	  {								\
+	    /* Check if the third byte is				\
+	       a valid start of a UTF-16 surrogate.  */			\
+	    if (inend - inptr == 3 && (inptr[3] & 0xfc) != 0xdc)	\
+	      STANDARD_FROM_LOOP_ERR_HANDLER (3);			\
+									\
+	    result = __GCONV_INCOMPLETE_INPUT;				\
+	    break;							\
+	  }								\
+	continue;							\
+      }									\
+									\
+    uint16_t u1 = get16 (inptr);					\
+									\
+    if (__builtin_expect (u1 < 0xd800, 1) || u1 > 0xdfff)		\
+      {									\
+	/* No surrogate.  */						\
+	put32 (outptr, u1);						\
+	inptr += 2;							\
+      }									\
+    else								\
+      {									\
+	/* It's a surrogate character.  At least the first word says	\
+	   it is.  */							\
+	if (__builtin_expect (inptr + 4 > inend, 0))			\
+	  {								\
+	    /* We don't have enough input for another complete input	\
+	       character.  */						\
+	    result = __GCONV_INCOMPLETE_INPUT;				\
+	    break;							\
+	  }								\
+									\
+	inptr += 2;							\
+	uint16_t u2 = get16 (inptr);					\
+	if (__builtin_expect (u2 < 0xdc00, 0)				\
+	    || __builtin_expect (u2 > 0xdfff, 0))			\
+	  {								\
+	    /* This is no valid second word for a surrogate.  */	\
+	    inptr -= 2;							\
+	    STANDARD_FROM_LOOP_ERR_HANDLER (2);				\
+	  }								\
+									\
+	put32 (outptr, ((u1 - 0xd7c0) << 10) + (u2 - 0xdc00));		\
+	inptr += 2;							\
+      }									\
+    outptr += 4;							\
+  }
+#define LOOP_NEED_FLAGS
+#include <iconv/loop.c>
+
+/* Conversion from UTF-32 internal/BE to UTF-16.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_TO
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_OUTPUT	MAX_NEEDED_FROM
+#define LOOPFCT			TO_LOOP
+/* The software routine is copied from utf-16.c (minus bytes
+   swapping).  */
+#define BODY								\
+  {									\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)				\
+      {									\
+	HARDWARE_CONVERT ("cu42 %0, %1");				\
+									\
+	if (inptr != inend)						\
+	  {								\
+	    result = __GCONV_INCOMPLETE_INPUT;				\
+	    break;							\
+	  }								\
+	continue;							\
+      }									\
+									\
+    uint32_t c = get32 (inptr);						\
+									\
+    if (__builtin_expect (c <= 0xd7ff, 1)				\
+	|| (c >=0xdc00 && c <= 0xffff))					\
+      {									\
+        /* Two UTF-16 chars.  */					\
+        put16 (outptr, c);						\
+      }									\
+    else if (__builtin_expect (c >= 0x10000, 1)				\
+	     && __builtin_expect (c <= 0x10ffff, 1))			\
+      {									\
+	/* Four UTF-16 chars.  */					\
+        uint16_t zabcd = ((c & 0x1f0000) >> 16) - 1;			\
+	uint16_t out;							\
+									\
+	/* Generate a surrogate character.  */				\
+	if (__builtin_expect (outptr + 4 > outend, 0))			\
+	  {								\
+	    /* Overflow in the output buffer.  */			\
+	    result = __GCONV_FULL_OUTPUT;				\
+	    break;							\
+	  }								\
+									\
+	out = 0xd800;							\
+	out |= (zabcd & 0xff) << 6;					\
+	out |= (c >> 10) & 0x3f;					\
+	put16 (outptr, out);						\
+	outptr += 2;							\
+									\
+	out = 0xdc00;							\
+	out |= c & 0x3ff;						\
+	put16 (outptr, out);						\
+      }									\
+    else								\
+      {									\
+        STANDARD_TO_LOOP_ERR_HANDLER (4);				\
+      }									\
+    outptr += 2;							\
+    inptr += 4;								\
+  }
+#define LOOP_NEED_FLAGS
+#include <iconv/loop.c>
+
+#include <iconv/skeleton.c>
diff --git a/sysdeps/s390/s390-64/utf8-utf16-z9.c b/sysdeps/s390/s390-64/utf8-utf16-z9.c
new file mode 100644
index 0000000000..3fe3652b35
--- /dev/null
+++ b/sysdeps/s390/s390-64/utf8-utf16-z9.c
@@ -0,0 +1,463 @@
+/* Conversion between UTF-16 and UTF-32 BE/internal.
+
+   This module uses the Z9-109 variants of the Convert Unicode
+   instructions.
+   Copyright (C) 1997-2009 Free Software Foundation, Inc.
+
+   Author: Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
+   Based on the work by Ulrich Drepper  <drepper@cygnus.com>, 1997.
+
+   Thanks to Daniel Appich who covered the relevant performance work
+   in his diploma thesis.
+
+   This 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.
+
+   This 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <dl-procinfo.h>
+#include <gconv.h>
+
+/* UTF-16 big endian byte order mark.  */
+#define BOM_UTF16	0xfeff
+
+#define DEFINE_INIT		0
+#define DEFINE_FINI		0
+#define MIN_NEEDED_FROM		1
+#define MAX_NEEDED_FROM		4
+#define MIN_NEEDED_TO		2
+#define MAX_NEEDED_TO		4
+#define FROM_LOOP		from_utf8_loop
+#define TO_LOOP			to_utf8_loop
+#define FROM_DIRECTION		(dir == from_utf8)
+#define PREPARE_LOOP							\
+  enum direction dir = ((struct utf8_data *) step->__data)->dir;	\
+  int emit_bom = ((struct utf8_data *) step->__data)->emit_bom;		\
+									\
+  if (emit_bom && !data->__internal_use					\
+      && data->__invocation_counter == 0)				\
+    {									\
+      /* Emit the UTF-16 Byte Order Mark.  */				\
+      if (__builtin_expect (outbuf + 2 > outend, 0))			\
+	return __GCONV_FULL_OUTPUT;					\
+									\
+      put16u (outbuf, BOM_UTF16);					\
+      outbuf += 2;							\
+    }
+
+/* Direction of the transformation.  */
+enum direction
+{
+  illegal_dir,
+  to_utf8,
+  from_utf8
+};
+
+struct utf8_data
+{
+  enum direction dir;
+  int emit_bom;
+};
+
+
+extern int gconv_init (struct __gconv_step *step);
+int
+gconv_init (struct __gconv_step *step)
+{
+  /* Determine which direction.  */
+  struct utf8_data *new_data;
+  enum direction dir = illegal_dir;
+  int emit_bom;
+  int result;
+
+  emit_bom = (__strcasecmp (step->__to_name, "UTF-16//") == 0);
+
+  if (__strcasecmp (step->__from_name, "ISO-10646/UTF8/") == 0
+      && (__strcasecmp (step->__to_name, "UTF-16//") == 0
+	  || __strcasecmp (step->__to_name, "UTF-16BE//") == 0))
+    {
+      dir = from_utf8;
+    }
+  else if (__strcasecmp (step->__from_name, "UTF-16BE//") == 0
+	   && __strcasecmp (step->__to_name, "ISO-10646/UTF8/") == 0)
+    {
+      dir = to_utf8;
+    }
+
+  result = __GCONV_NOCONV;
+  if (dir != illegal_dir)
+    {
+      new_data = (struct utf8_data *) malloc (sizeof (struct utf8_data));
+
+      result = __GCONV_NOMEM;
+      if (new_data != NULL)
+	{
+	  new_data->dir = dir;
+	  new_data->emit_bom = emit_bom;
+	  step->__data = new_data;
+
+	  if (dir == from_utf8)
+	    {
+	      step->__min_needed_from = MIN_NEEDED_FROM;
+	      step->__max_needed_from = MIN_NEEDED_FROM;
+	      step->__min_needed_to = MIN_NEEDED_TO;
+	      step->__max_needed_to = MIN_NEEDED_TO;
+	    }
+	  else
+	    {
+	      step->__min_needed_from = MIN_NEEDED_TO;
+	      step->__max_needed_from = MIN_NEEDED_TO;
+	      step->__min_needed_to = MIN_NEEDED_FROM;
+	      step->__max_needed_to = MIN_NEEDED_FROM;
+	    }
+
+	  step->__stateful = 0;
+
+	  result = __GCONV_OK;
+	}
+    }
+
+  return result;
+}
+
+
+extern void gconv_end (struct __gconv_step *data);
+void
+gconv_end (struct __gconv_step *data)
+{
+  free (data->__data);
+}
+
+/* The macro for the hardware loop.  This is used for both
+   directions.  */
+#define HARDWARE_CONVERT(INSTRUCTION)					\
+  {									\
+    register const unsigned char* pInput asm ("8") = inptr;		\
+    register unsigned long long inlen asm ("9") = inend - inptr;	\
+    register unsigned char* pOutput asm ("10") = outptr;		\
+    register unsigned long long outlen asm("11") = outend - outptr;	\
+    uint64_t cc = 0;							\
+									\
+    asm volatile ("0: " INSTRUCTION "  \n\t"				\
+                  "   jo     0b        \n\t"				\
+		  "   ipm    %2        \n"			        \
+		  : "+a" (pOutput), "+a" (pInput), "+d" (cc),		\
+		    "+d" (outlen), "+d" (inlen)				\
+		  :							\
+		  : "cc", "memory");					\
+									\
+    inptr = pInput;							\
+    outptr = pOutput;							\
+    cc >>= 28;								\
+									\
+    if (cc == 1)							\
+      {									\
+	result = __GCONV_FULL_OUTPUT;					\
+	break;								\
+      }									\
+    else if (cc == 2)							\
+      {									\
+	result = __GCONV_ILLEGAL_INPUT;					\
+	break;								\
+      }									\
+  }
+
+/* Conversion function from UTF-8 to UTF-16.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_INPUT	MAX_NEEDED_FROM
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_TO
+#define LOOPFCT			FROM_LOOP
+/* The software implementation is based on the code in gconv_simple.c.  */
+#define BODY								\
+  {									\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)				\
+      {									\
+	HARDWARE_CONVERT ("cu12 %0, %1, 1");				\
+	  								\
+	if (inptr != inend)						\
+	  {								\
+	    int i;							\
+	    for (i = 1; inptr + i < inend; ++i)				\
+	      if ((inptr[i] & 0xc0) != 0x80)				\
+		break;							\
+	    								\
+	    if (__builtin_expect (inptr + i == inend, 1))		\
+	      {								\
+		result = __GCONV_INCOMPLETE_INPUT;			\
+		break;							\
+	      }								\
+	    STANDARD_FROM_LOOP_ERR_HANDLER (i);				\
+	  }								\
+	continue;							\
+    }									\
+									\
+    /* Next input byte.  */						\
+    uint16_t ch = *inptr;						\
+									\
+    if (__builtin_expect (ch < 0x80, 1))				\
+      {									\
+	/* One byte sequence.  */					\
+	++inptr;							\
+      }									\
+    else								\
+      {									\
+	uint_fast32_t cnt;						\
+	uint_fast32_t i;						\
+									\
+	if (ch >= 0xc2 && ch < 0xe0)					\
+	  {								\
+	    /* We expect two bytes.  The first byte cannot be 0xc0	\
+	       or 0xc1, otherwise the wide character could have been	\
+	       represented using a single byte.  */			\
+	    cnt = 2;							\
+	    ch &= 0x1f;							\
+	  }								\
+        else if (__builtin_expect ((ch & 0xf0) == 0xe0, 1))		\
+	  {								\
+	    /* We expect three bytes.  */				\
+	    cnt = 3;							\
+	    ch &= 0x0f;							\
+	  }								\
+	else if (__builtin_expect ((ch & 0xf8) == 0xf0, 1))		\
+	  {								\
+	    /* We expect four bytes.  */				\
+	    cnt = 4;							\
+	    ch &= 0x07;							\
+	  }								\
+	else								\
+	  {								\
+	    /* Search the end of this ill-formed UTF-8 character.  This	\
+	       is the next byte with (x & 0xc0) != 0x80.  */		\
+	    i = 0;							\
+	    do								\
+	      ++i;							\
+	    while (inptr + i < inend					\
+		   && (*(inptr + i) & 0xc0) == 0x80			\
+		   && i < 5);						\
+									\
+	  errout:							\
+	    STANDARD_FROM_LOOP_ERR_HANDLER (i);				\
+	  }								\
+									\
+	if (__builtin_expect (inptr + cnt > inend, 0))			\
+	  {								\
+	    /* We don't have enough input.  But before we report	\
+	       that check that all the bytes are correct.  */		\
+	    for (i = 1; inptr + i < inend; ++i)				\
+	      if ((inptr[i] & 0xc0) != 0x80)				\
+		break;							\
+									\
+	    if (__builtin_expect (inptr + i == inend, 1))		\
+	      {								\
+		result = __GCONV_INCOMPLETE_INPUT;			\
+		break;							\
+	      }								\
+									\
+	    goto errout;						\
+	  }								\
+									\
+	if (cnt == 4)							\
+	  {								\
+	    /* For 4 byte UTF-8 chars two UTF-16 chars (high and	\
+	       low) are needed.  */					\
+	    uint16_t zabcd, high, low;					\
+	    								\
+	    if (__builtin_expect (outptr + 4 > outend, 0))		\
+	      {								\
+		/* Overflow in the output buffer.  */			\
+		result = __GCONV_FULL_OUTPUT;				\
+		break;							\
+	      }								\
+									\
+	    /* See Principles of Operations cu12.  */			\
+	    zabcd = (((inptr[0] & 0x7) << 2) |				\
+                     ((inptr[1] & 0x30) >> 4)) - 1;			\
+									\
+	    /* z-bit must be zero after subtracting 1.  */		\
+	    if (zabcd & 0x10)						\
+	      STANDARD_FROM_LOOP_ERR_HANDLER (4)			\
+									\
+	    high = (uint16_t)(0xd8 << 8);       /* high surrogate id */ \
+	    high |= zabcd << 6;	                        /* abcd bits */	\
+	    high |= (inptr[1] & 0xf) << 2;              /* efgh bits */	\
+	    high |= (inptr[2] & 0x30) >> 4;               /* ij bits */	\
+									\
+	    low = (uint16_t)(0xdc << 8);         /* low surrogate id */ \
+	    low |= ((uint16_t)inptr[2] & 0xc) << 6;       /* kl bits */	\
+	    low |= (inptr[2] & 0x3) << 6;                 /* mn bits */	\
+	    low |= inptr[3] & 0x3f;                   /* opqrst bits */	\
+	    								\
+	    put16 (outptr, high);					\
+	    outptr += 2;						\
+	    put16 (outptr, low);					\
+	    outptr += 2;						\
+	    inptr += 4;							\
+	    continue;							\
+	  }								\
+	else								\
+	  {								\
+	    /* Read the possible remaining bytes.  */			\
+	    for (i = 1; i < cnt; ++i)					\
+	      {								\
+		uint16_t byte = inptr[i];				\
+									\
+		if ((byte & 0xc0) != 0x80)				\
+		  /* This is an illegal encoding.  */			\
+		  break;						\
+									\
+		ch <<= 6;						\
+		ch |= byte & 0x3f;					\
+	      }								\
+	    inptr += cnt;						\
+									\
+	  }								\
+      }									\
+    /* Now adjust the pointers and store the result.  */		\
+    *((uint16_t *) outptr) = ch;					\
+    outptr += sizeof (uint16_t);					\
+  }
+
+#define LOOP_NEED_FLAGS
+#include <iconv/loop.c>
+
+/* Conversion from UTF-16 to UTF-8.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_TO
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_OUTPUT	MAX_NEEDED_FROM
+#define LOOPFCT			TO_LOOP
+/* The software routine is based on the functionality of the S/390
+   hardware instruction (cu21) as described in the Principles of
+   Operation.  */
+#define BODY								\
+  {									\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)				\
+      {									\
+	HARDWARE_CONVERT ("cu21 %0, %1");				\
+	if (inptr != inend)						\
+	  {								\
+	    /* Check if the third byte is				\
+	       a valid start of a UTF-16 surrogate.  */			\
+	    if (inend - inptr == 3 && (inptr[3] & 0xfc) != 0xdc)	\
+	      STANDARD_TO_LOOP_ERR_HANDLER (3);				\
+									\
+	    result = __GCONV_INCOMPLETE_INPUT;				\
+	    break;							\
+	  }								\
+	continue;							\
+      }									\
+									\
+    uint16_t c = get16 (inptr);						\
+									\
+    if (__builtin_expect (c <= 0x007f, 1))				\
+      {									\
+	/* Single byte UTF-8 char.  */					\
+	*outptr = c & 0xff;						\
+	outptr++;							\
+      }									\
+    else if (c >= 0x0080 && c <= 0x07ff)				\
+      {									\
+        /* Two byte UTF-8 char.  */					\
+									\
+	if (__builtin_expect (outptr + 2 > outend, 0))			\
+	  {								\
+	    /* Overflow in the output buffer.  */			\
+	    result = __GCONV_FULL_OUTPUT;				\
+	    break;							\
+	  }								\
+									\
+        outptr[0] = 0xc0;						\
+        outptr[0] |= c >> 6;						\
+      									\
+        outptr[1] = 0x80;						\
+        outptr[1] |= c & 0x3f;						\
+									\
+	outptr += 2;							\
+      }									\
+    else if (c >= 0x0800 && c <= 0xd7ff)				\
+      {									\
+	/* Three byte UTF-8 char.  */					\
+									\
+	if (__builtin_expect (outptr + 3 > outend, 0))			\
+	  {								\
+	    /* Overflow in the output buffer.  */			\
+	    result = __GCONV_FULL_OUTPUT;				\
+	    break;							\
+	  }								\
+	outptr[0] = 0xe0;						\
+	outptr[0] |= c >> 12;						\
+									\
+	outptr[1] = 0x80;						\
+	outptr[1] |= (c >> 6) & 0x3f;					\
+									\
+	outptr[2] = 0x80;						\
+	outptr[2] |= c & 0x3f;						\
+									\
+	outptr += 3;							\
+      }									\
+    else if (c >= 0xd800 && c <= 0xdbff)				\
+      {									\
+        /* Four byte UTF-8 char.  */					\
+	uint16_t low, uvwxy;						\
+									\
+	if (__builtin_expect (outptr + 4 > outend, 0))			\
+	  {								\
+	    /* Overflow in the output buffer.  */			\
+	    result = __GCONV_FULL_OUTPUT;				\
+	    break;							\
+	  }								\
+	inptr += 2;							\
+	if (__builtin_expect (inptr + 2 > inend, 0))			\
+	  {								\
+	    result = __GCONV_INCOMPLETE_INPUT;				\
+	    break;							\
+	  }								\
+									\
+	low = get16 (inptr);						\
+									\
+	if ((low & 0xfc00) != 0xdc00)					\
+	  {								\
+	    inptr -= 2;							\
+	    STANDARD_TO_LOOP_ERR_HANDLER (2);				\
+	  }								\
+	uvwxy = ((c >> 6) & 0xf) + 1;					\
+	outptr[0] = 0xf0;						\
+	outptr[0] |= uvwxy >> 2;					\
+									\
+	outptr[1] = 0x80;						\
+	outptr[1] |= (uvwxy << 4) & 0x30;				\
+	outptr[1] |= (c >> 2) & 0x0f;					\
+									\
+	outptr[2] = 0x80;						\
+	outptr[2] |= (c & 0x03) << 4;					\
+	outptr[2] |= (low >> 6) & 0x0f;					\
+									\
+	outptr[3] = 0x80;						\
+	outptr[3] |= low & 0x3f;					\
+									\
+	outptr += 4;							\
+      }									\
+    else								\
+      {									\
+        STANDARD_TO_LOOP_ERR_HANDLER (2);				\
+      }									\
+    inptr += 2;								\
+  }
+#define LOOP_NEED_FLAGS
+#include <iconv/loop.c>
+
+#include <iconv/skeleton.c>
diff --git a/sysdeps/s390/s390-64/utf8-utf32-z9.c b/sysdeps/s390/s390-64/utf8-utf32-z9.c
new file mode 100644
index 0000000000..14847e2403
--- /dev/null
+++ b/sysdeps/s390/s390-64/utf8-utf32-z9.c
@@ -0,0 +1,508 @@
+/* Conversion between UTF-8 and UTF-32 BE/internal.
+
+   This module uses the Z9-109 variants of the Convert Unicode
+   instructions.
+   Copyright (C) 1997-2009 Free Software Foundation, Inc.
+
+   Author: Andreas Krebbel  <Andreas.Krebbel@de.ibm.com>
+   Based on the work by Ulrich Drepper  <drepper@cygnus.com>, 1997.
+
+   Thanks to Daniel Appich who covered the relevant performance work
+   in his diploma thesis.
+
+   This 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.
+
+   This 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include <dlfcn.h>
+#include <stdint.h>
+#include <unistd.h>
+#include <dl-procinfo.h>
+#include <gconv.h>
+
+/* UTF-32 big endian byte order mark.  */
+#define BOM	                0x0000feffu
+
+#define DEFINE_INIT		0
+#define DEFINE_FINI		0
+/* These definitions apply to the UTF-8 to UTF-32 direction.  The
+   software implementation for UTF-8 still supports multibyte
+   characters up to 6 bytes whereas the hardware variant does not.  */
+#define MIN_NEEDED_FROM		1
+#define MAX_NEEDED_FROM		6
+#define MIN_NEEDED_TO		4
+#define FROM_LOOP		from_utf8_loop
+#define TO_LOOP			to_utf8_loop
+#define FROM_DIRECTION		(dir == from_utf8)
+#define PREPARE_LOOP							\
+  enum direction dir = ((struct utf8_data *) step->__data)->dir;	\
+  int emit_bom = ((struct utf8_data *) step->__data)->emit_bom;		\
+									\
+  if (emit_bom && !data->__internal_use					\
+      && data->__invocation_counter == 0)				\
+    {									\
+      /* Emit the Byte Order Mark.  */					\
+      if (__builtin_expect (outbuf + 4 > outend, 0))			\
+	return __GCONV_FULL_OUTPUT;					\
+      									\
+      put32u (outbuf, BOM);						\
+      outbuf += 4;							\
+    }
+
+/* Direction of the transformation.  */
+enum direction
+{
+  illegal_dir,
+  to_utf8,
+  from_utf8
+};
+
+struct utf8_data
+{
+  enum direction dir;
+  int emit_bom;
+};
+
+
+extern int gconv_init (struct __gconv_step *step);
+int
+gconv_init (struct __gconv_step *step)
+{
+  /* Determine which direction.  */
+  struct utf8_data *new_data;
+  enum direction dir = illegal_dir;
+  int emit_bom;
+  int result;
+
+  emit_bom = (__strcasecmp (step->__to_name, "UTF-32//") == 0);
+
+  if (__strcasecmp (step->__from_name, "ISO-10646/UTF8/") == 0
+      && (__strcasecmp (step->__to_name, "UTF-32//") == 0
+	  || __strcasecmp (step->__to_name, "UTF-32BE//") == 0
+      	  || __strcasecmp (step->__to_name, "INTERNAL") == 0))
+    {
+      dir = from_utf8;
+    }
+  else if (__strcasecmp (step->__to_name, "ISO-10646/UTF8/") == 0
+	   && (__strcasecmp (step->__from_name, "UTF-32BE//") == 0
+	       || __strcasecmp (step->__from_name, "INTERNAL") == 0))
+    {
+      dir = to_utf8;
+    }
+
+  result = __GCONV_NOCONV;
+  if (dir != illegal_dir)
+    {
+      new_data = (struct utf8_data *) malloc (sizeof (struct utf8_data));
+
+      result = __GCONV_NOMEM;
+      if (new_data != NULL)
+	{
+	  new_data->dir = dir;
+	  new_data->emit_bom = emit_bom;
+	  step->__data = new_data;
+
+	  if (dir == from_utf8)
+	    {
+	      step->__min_needed_from = MIN_NEEDED_FROM;
+	      step->__max_needed_from = MIN_NEEDED_FROM;
+	      step->__min_needed_to = MIN_NEEDED_TO;
+	      step->__max_needed_to = MIN_NEEDED_TO;
+	    }
+	  else
+	    {
+	      step->__min_needed_from = MIN_NEEDED_TO;
+	      step->__max_needed_from = MIN_NEEDED_TO;
+	      step->__min_needed_to = MIN_NEEDED_FROM;
+	      step->__max_needed_to = MIN_NEEDED_FROM;
+	    }
+
+	  step->__stateful = 0;
+
+	  result = __GCONV_OK;
+	}
+    }
+
+  return result;
+}
+
+
+extern void gconv_end (struct __gconv_step *data);
+void
+gconv_end (struct __gconv_step *data)
+{
+  free (data->__data);
+}
+
+/* The macro for the hardware loop.  This is used for both
+   directions.  */
+#define HARDWARE_CONVERT(INSTRUCTION)					\
+  {									\
+    register const unsigned char* pInput asm ("8") = inptr;		\
+    register unsigned long long inlen asm ("9") = inend - inptr;	\
+    register unsigned char* pOutput asm ("10") = outptr;		\
+    register unsigned long long outlen asm("11") = outend - outptr;	\
+    uint64_t cc = 0;							\
+									\
+    asm volatile ("0: " INSTRUCTION "  \n\t"				\
+                  "   jo     0b        \n\t"				\
+		  "   ipm    %2        \n"				\
+		  : "+a" (pOutput), "+a" (pInput), "+d" (cc),		\
+		    "+d" (outlen), "+d" (inlen)				\
+		  :							\
+		  : "cc", "memory");					\
+									\
+    inptr = pInput;							\
+    outptr = pOutput;							\
+    cc >>= 28;          						\
+									\
+    if (cc == 1)							\
+      {									\
+	result = __GCONV_FULL_OUTPUT;					\
+	break;								\
+      }									\
+    else if (cc == 2)							\
+      {									\
+	result = __GCONV_ILLEGAL_INPUT;					\
+	break;								\
+      }									\
+  }
+
+/* Conversion function from UTF-8 to UTF-32 internal/BE.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_INPUT	MAX_NEEDED_FROM
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_TO
+#define LOOPFCT			FROM_LOOP
+/* The software routine is copied from gconv_simple.c.  */
+#define BODY             						\
+  {									\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)				\
+      {									\
+	HARDWARE_CONVERT ("cu14 %0, %1, 1");				\
+									\
+	if (inptr != inend)						\
+	  {								\
+	    int i;							\
+	    for (i = 1; inptr + i < inend; ++i)				\
+	      if ((inptr[i] & 0xc0) != 0x80)				\
+		break;							\
+									\
+	    if (__builtin_expect (inptr + i == inend, 1))		\
+	      {								\
+		result = __GCONV_INCOMPLETE_INPUT;			\
+		break;							\
+	      }								\
+	    STANDARD_FROM_LOOP_ERR_HANDLER (i);				\
+	  }								\
+	continue;							\
+      }									\
+									\
+    /* Next input byte.  */						\
+    uint32_t ch = *inptr;						\
+									\
+    if (__builtin_expect (ch < 0x80, 1))				\
+      {									\
+	/* One byte sequence.  */					\
+	++inptr;							\
+      }									\
+    else								\
+      {									\
+	uint_fast32_t cnt;						\
+	uint_fast32_t i;						\
+									\
+	if (ch >= 0xc2 && ch < 0xe0)					\
+	  {								\
+	    /* We expect two bytes.  The first byte cannot be 0xc0 or	\
+	       0xc1, otherwise the wide character could have been	\
+	       represented using a single byte.  */			\
+	    cnt = 2;							\
+	    ch &= 0x1f;							\
+	  }								\
+        else if (__builtin_expect ((ch & 0xf0) == 0xe0, 1))		\
+	  {								\
+	    /* We expect three bytes.  */				\
+	    cnt = 3;							\
+	    ch &= 0x0f;							\
+	  }								\
+	else if (__builtin_expect ((ch & 0xf8) == 0xf0, 1))		\
+	  {								\
+	    /* We expect four bytes.  */				\
+	    cnt = 4;							\
+	    ch &= 0x07;							\
+	  }								\
+	else if (__builtin_expect ((ch & 0xfc) == 0xf8, 1))		\
+	  {								\
+	    /* We expect five bytes.  */				\
+	    cnt = 5;							\
+	    ch &= 0x03;							\
+	  }								\
+	else if (__builtin_expect ((ch & 0xfe) == 0xfc, 1))		\
+	  {								\
+	    /* We expect six bytes.  */					\
+	    cnt = 6;							\
+	    ch &= 0x01;							\
+	  }								\
+	else								\
+	  {								\
+	    /* Search the end of this ill-formed UTF-8 character.  This	\
+	       is the next byte with (x & 0xc0) != 0x80.  */		\
+	    i = 0;							\
+	    do								\
+	      ++i;							\
+	    while (inptr + i < inend					\
+		   && (*(inptr + i) & 0xc0) == 0x80			\
+		   && i < 5);						\
+									\
+	  errout:							\
+	    STANDARD_FROM_LOOP_ERR_HANDLER (i);				\
+	  }								\
+									\
+	if (__builtin_expect (inptr + cnt > inend, 0))			\
+	  {								\
+	    /* We don't have enough input.  But before we report	\
+	       that check that all the bytes are correct.  */		\
+	    for (i = 1; inptr + i < inend; ++i)				\
+	      if ((inptr[i] & 0xc0) != 0x80)				\
+		break;							\
+									\
+	    if (__builtin_expect (inptr + i == inend, 1))		\
+	      {								\
+		result = __GCONV_INCOMPLETE_INPUT;			\
+		break;							\
+	      }								\
+									\
+	    goto errout;						\
+	  }								\
+									\
+	/* Read the possible remaining bytes.  */			\
+	for (i = 1; i < cnt; ++i)					\
+	  {								\
+	    uint32_t byte = inptr[i];					\
+									\
+	    if ((byte & 0xc0) != 0x80)					\
+	      /* This is an illegal encoding.  */			\
+	      break;							\
+									\
+	    ch <<= 6;							\
+	    ch |= byte & 0x3f;						\
+	  }								\
+									\
+	/* If i < cnt, some trail byte was not >= 0x80, < 0xc0.		\
+	   If cnt > 2 and ch < 2^(5*cnt-4), the wide character ch could	\
+	   have been represented with fewer than cnt bytes.  */		\
+	if (i < cnt || (cnt > 2 && (ch >> (5 * cnt - 4)) == 0))		\
+	  {								\
+	    /* This is an illegal encoding.  */				\
+	    goto errout;						\
+	  }								\
+									\
+	inptr += cnt;							\
+      }									\
+									\
+    /* Now adjust the pointers and store the result.  */		\
+    *((uint32_t *) outptr) = ch;					\
+    outptr += sizeof (uint32_t);					\
+  }
+#define LOOP_NEED_FLAGS
+
+#define STORE_REST							\
+  {									      \
+    /* We store the remaining bytes while converting them into the UCS4	      \
+       format.  We can assume that the first byte in the buffer is	      \
+       correct and that it requires a larger number of bytes than there	      \
+       are in the input buffer.  */					      \
+    wint_t ch = **inptrp;						      \
+    size_t cnt, r;							      \
+									      \
+    state->__count = inend - *inptrp;					      \
+									      \
+    if (ch >= 0xc2 && ch < 0xe0)					      \
+      {									      \
+	/* We expect two bytes.  The first byte cannot be 0xc0 or	      \
+	   0xc1, otherwise the wide character could have been		      \
+	   represented using a single byte.  */				      \
+	cnt = 2;							      \
+	ch &= 0x1f;							      \
+      }									      \
+    else if (__builtin_expect ((ch & 0xf0) == 0xe0, 1))			      \
+      {									      \
+	/* We expect three bytes.  */					      \
+	cnt = 3;							      \
+	ch &= 0x0f;							      \
+      }									      \
+    else if (__builtin_expect ((ch & 0xf8) == 0xf0, 1))			      \
+      {									      \
+	/* We expect four bytes.  */					      \
+	cnt = 4;							      \
+	ch &= 0x07;							      \
+      }									      \
+    else if (__builtin_expect ((ch & 0xfc) == 0xf8, 1))			      \
+      {									      \
+	/* We expect five bytes.  */					      \
+	cnt = 5;							      \
+	ch &= 0x03;							      \
+      }									      \
+    else								      \
+      {									      \
+	/* We expect six bytes.  */					      \
+	cnt = 6;							      \
+	ch &= 0x01;							      \
+      }									      \
+									      \
+    /* The first byte is already consumed.  */				      \
+    r = cnt - 1;							      \
+    while (++(*inptrp) < inend)						      \
+      {									      \
+	ch <<= 6;							      \
+	ch |= **inptrp & 0x3f;						      \
+	--r;								      \
+      }									      \
+									      \
+    /* Shift for the so far missing bytes.  */				      \
+    ch <<= r * 6;							      \
+									      \
+    /* Store the number of bytes expected for the entire sequence.  */	      \
+    state->__count |= cnt << 8;						      \
+									      \
+    /* Store the value.  */						      \
+    state->__value.__wch = ch;						      \
+  }
+
+#define UNPACK_BYTES \
+  {									      \
+    static const unsigned char inmask[5] = { 0xc0, 0xe0, 0xf0, 0xf8, 0xfc };  \
+    wint_t wch = state->__value.__wch;					      \
+    size_t ntotal = state->__count >> 8;				      \
+									      \
+    inlen = state->__count & 255;					      \
+									      \
+    bytebuf[0] = inmask[ntotal - 2];					      \
+									      \
+    do									      \
+      {									      \
+	if (--ntotal < inlen)						      \
+	  bytebuf[ntotal] = 0x80 | (wch & 0x3f);			      \
+	wch >>= 6;							      \
+      }									      \
+    while (ntotal > 1);							      \
+									      \
+    bytebuf[0] |= wch;							      \
+  }
+
+#define CLEAR_STATE \
+  state->__count = 0
+
+#include <iconv/loop.c>
+
+/* Conversion from UTF-32 internal/BE to UTF-8.  */
+
+#define MIN_NEEDED_INPUT	MIN_NEEDED_TO
+#define MIN_NEEDED_OUTPUT	MIN_NEEDED_FROM
+#define MAX_NEEDED_OUTPUT	MAX_NEEDED_FROM
+#define LOOPFCT			TO_LOOP
+/* The software routine mimics the S/390 cu41 instruction.  */
+#define BODY							\
+  {								\
+    if (GLRO (dl_hwcap) & HWCAP_S390_ETF3EH)			\
+      {								\
+	HARDWARE_CONVERT ("cu41 %0, %1");			\
+								\
+	if (inptr != inend)					\
+	  {							\
+	    result = __GCONV_INCOMPLETE_INPUT;			\
+	    break;						\
+	  }							\
+	continue;						\
+      }								\
+								\
+    uint32_t wc = *((const uint32_t *) inptr);			\
+								\
+    if (__builtin_expect (wc <= 0x7f, 1))			\
+      {								\
+        /* Single UTF-8 char.  */				\
+        *outptr = (uint8_t)wc;					\
+	outptr++;						\
+      }								\
+    else if (wc <= 0x7ff)					\
+      {								\
+        /* Two UTF-8 chars.  */					\
+        if (__builtin_expect (outptr + 2 > outend, 0))		\
+	  {							\
+	    /* Overflow in the output buffer.  */		\
+	    result = __GCONV_FULL_OUTPUT;			\
+	    break;						\
+	  }							\
+								\
+        outptr[0] = 0xc0;					\
+	outptr[0] |= wc >> 6;					\
+								\
+	outptr[1] = 0x80;					\
+	outptr[1] |= wc & 0x3f;					\
+								\
+	outptr += 2;						\
+      }								\
+    else if (wc <= 0xffff)					\
+      {								\
+	/* Three UTF-8 chars.  */				\
+	if (__builtin_expect (outptr + 3 > outend, 0))		\
+	  {							\
+	    /* Overflow in the output buffer.  */		\
+	    result = __GCONV_FULL_OUTPUT;			\
+	    break;						\
+	  }							\
+	outptr[0] = 0xe0;					\
+	outptr[0] |= wc >> 12;					\
+								\
+	outptr[1] = 0x80;					\
+	outptr[1] |= (wc >> 6) & 0x3f;				\
+								\
+	outptr[2] = 0x80;					\
+	outptr[2] |= wc & 0x3f;					\
+								\
+	outptr += 3;						\
+      }								\
+      else if (wc <= 0x10ffff)					\
+	{							\
+	  /* Four UTF-8 chars.  */				\
+	  if (__builtin_expect (outptr + 4 > outend, 0))	\
+	    {							\
+	      /* Overflow in the output buffer.  */		\
+	      result = __GCONV_FULL_OUTPUT;			\
+	      break;						\
+	    }							\
+	  outptr[0] = 0xf0;					\
+	  outptr[0] |= wc >> 18;				\
+								\
+	  outptr[1] = 0x80;					\
+	  outptr[1] |= (wc >> 12) & 0x3f;			\
+								\
+	  outptr[2] = 0x80;					\
+	  outptr[2] |= (wc >> 6) & 0x3f;			\
+								\
+	  outptr[3] = 0x80;					\
+	  outptr[3] |= wc & 0x3f;				\
+								\
+	  outptr += 4;						\
+	}							\
+      else							\
+	{							\
+	  STANDARD_TO_LOOP_ERR_HANDLER (4);			\
+	}							\
+    inptr += 4;							\
+  }
+#define LOOP_NEED_FLAGS
+#include <iconv/loop.c>
+
+#include <iconv/skeleton.c>