S390: Use cu21 instruction for converting from utf16 to utf8.

This patch adds an ifunc variant to use the cu instruction on arch12 CPUs.
This new ifunc variant can be built if binutils support z13 vector
instructions.  At runtime, HWCAP_S390_VXE decides if we can use the
cu21 instruction.

ChangeLog:

	* sysdeps/s390/utf8-utf16-z9.c (__to_utf8_loop_vx_cu):
	Use vector and cu21 instruction.
	* sysdeps/s390/multiarch/utf8-utf16-z9.c:
	Add __to_utf8_loop_vx_cu in ifunc resolver.

3 files changed, 129 insertions, 3 deletions

diff --git a/ChangeLog b/ChangeLog
index 62aacea90f..8ef682af42 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,5 +1,12 @@
 2017-06-27  Stefan Liebler  <stli@linux.vnet.ibm.com>
 
+	* sysdeps/s390/utf8-utf16-z9.c (__to_utf8_loop_vx_cu):
+	Use vector and cu21 instruction.
+	* sysdeps/s390/multiarch/utf8-utf16-z9.c:
+	Add __to_utf8_loop_vx_cu in ifunc resolver.
+
+2017-06-27  Stefan Liebler  <stli@linux.vnet.ibm.com>
+
 	* sysdeps/s390/utf16-utf32-z9.c (__from_utf16_loop_vx_cu):
 	Use vector and cu24 instruction.
 
diff --git a/sysdeps/s390/multiarch/utf8-utf16-z9.c b/sysdeps/s390/multiarch/utf8-utf16-z9.c
index b55ef1aaec..1252281e4b 100644
--- a/sysdeps/s390/multiarch/utf8-utf16-z9.c
+++ b/sysdeps/s390/multiarch/utf8-utf16-z9.c
@@ -41,8 +41,10 @@ s390_libc_ifunc_expr (FROM_LOOP_DEFAULT, FROM_LOOP,
 			: FROM_LOOP_DEFAULT);
 
 s390_libc_ifunc_expr (TO_LOOP_DEFAULT, TO_LOOP,
-		      (HAVE_TO_VX && (hwcap & HWCAP_S390_VX))
-		      ? TO_LOOP_VX
-		      : TO_LOOP_DEFAULT);
+		      (HAVE_TO_VX_CU && (hwcap & HWCAP_S390_VXE))
+		      ? TO_LOOP_VX_CU
+		      : (HAVE_TO_VX && (hwcap & HWCAP_S390_VX))
+			? TO_LOOP_VX
+			: TO_LOOP_DEFAULT);
 
 #include <iconv/skeleton.c>
diff --git a/sysdeps/s390/utf8-utf16-z9.c b/sysdeps/s390/utf8-utf16-z9.c
index 42163e7a19..84866820ba 100644
--- a/sysdeps/s390/utf8-utf16-z9.c
+++ b/sysdeps/s390/utf8-utf16-z9.c
@@ -52,9 +52,11 @@
 #if defined HAVE_S390_VX_ASM_SUPPORT && defined USE_MULTIARCH
 # define HAVE_FROM_VX		1
 # define HAVE_TO_VX		1
+# define HAVE_TO_VX_CU		1
 #else
 # define HAVE_FROM_VX		0
 # define HAVE_TO_VX		0
+# define HAVE_TO_VX_CU		0
 #endif
 
 #if defined HAVE_S390_VX_GCC_SUPPORT
@@ -817,6 +819,121 @@ gconv_end (struct __gconv_step *data)
 # define TO_LOOP_VX		NULL
 #endif /* HAVE_TO_VX != 1  */
 
+#if HAVE_TO_VX_CU == 1
+#define BODY_TO_VX_CU							\
+  {									\
+    register const unsigned char* pInput asm ("8") = inptr;		\
+    register size_t inlen asm ("9") = inend - inptr;			\
+    register unsigned char* pOutput asm ("10") = outptr;		\
+    register size_t outlen asm ("11") = outend - outptr;		\
+    unsigned long tmp, tmp2, tmp3;					\
+    asm volatile (".machine push\n\t"					\
+		  ".machine \"z13\"\n\t"				\
+		  ".machinemode \"zarch_nohighgprs\"\n\t"		\
+		  /* Setup to check for values <= 0x7f.  */		\
+		  "    larl %[R_TMP],9f\n\t"				\
+		  "    vlm %%v30,%%v31,0(%[R_TMP])\n\t"			\
+		  CONVERT_32BIT_SIZE_T ([R_INLEN])			\
+		  CONVERT_32BIT_SIZE_T ([R_OUTLEN])			\
+		  /* Loop which handles UTF-16 chars <=0x7f.  */	\
+		  "0:  clgijl %[R_INLEN],32,20f\n\t"			\
+		  "    clgijl %[R_OUTLEN],16,20f\n\t"			\
+		  "1:  vlm %%v16,%%v17,0(%[R_IN])\n\t"			\
+		  "    lghi %[R_TMP2],0\n\t"				\
+		  /* Check for > 1byte UTF-8 chars.  */			\
+		  "    vstrchs %%v19,%%v16,%%v30,%%v31\n\t"		\
+		  "    jno 10f\n\t" /* Jump away if not all bytes are 1byte \
+				       UTF8 chars.  */			\
+		  "    vstrchs %%v19,%%v17,%%v30,%%v31\n\t"		\
+		  "    jno 11f\n\t" /* Jump away if not all bytes are 1byte \
+				       UTF8 chars.  */			\
+		  /* Shorten to UTF-8.  */				\
+		  "    vpkh %%v18,%%v16,%%v17\n\t"			\
+		  "    la %[R_IN],32(%[R_IN])\n\t"			\
+		  "    aghi %[R_INLEN],-32\n\t"				\
+		  /* Store 16 bytes to buf_out.  */			\
+		  "    vst %%v18,0(%[R_OUT])\n\t"			\
+		  "    aghi %[R_OUTLEN],-16\n\t"			\
+		  "    la %[R_OUT],16(%[R_OUT])\n\t"			\
+		  "    clgijl %[R_INLEN],32,20f\n\t"			\
+		  "    clgijl %[R_OUTLEN],16,20f\n\t"			\
+		  "    j 1b\n\t"					\
+		  /* Setup to check for ch > 0x7f. (v30, v31)  */	\
+		  "9:  .short 0x7f,0x7f,0x0,0x0,0x0,0x0,0x0,0x0\n\t"	\
+		  "    .short 0x2000,0x2000,0x0,0x0,0x0,0x0,0x0,0x0\n\t" \
+		  /* At least one byte is > 0x7f.			\
+		     Store the preceding 1-byte chars.  */		\
+		  "11: lghi %[R_TMP2],16\n\t" /* match was found in v17.  */ \
+		  "10: vlgvb %[R_TMP],%%v19,7\n\t"			\
+		  /* Shorten to UTF-8.  */				\
+		  "    vpkh %%v18,%%v16,%%v17\n\t"			\
+		  "    ar %[R_TMP],%[R_TMP2]\n\t" /* Number of in bytes.  */ \
+		  "    srlg %[R_TMP3],%[R_TMP],1\n\t" /* Number of out bytes.  */ \
+		  "    ahik %[R_TMP2],%[R_TMP3],-1\n\t" /* Highest index to store.  */ \
+		  "    jl 20f\n\t"					\
+		  "    vstl %%v18,%[R_TMP2],0(%[R_OUT])\n\t"		\
+		  /* Update pointers.  */				\
+		  "    la %[R_IN],0(%[R_TMP],%[R_IN])\n\t"		\
+		  "    slgr %[R_INLEN],%[R_TMP]\n\t"			\
+		  "    la %[R_OUT],0(%[R_TMP3],%[R_OUT])\n\t"		\
+		  "    slgr %[R_OUTLEN],%[R_TMP3]\n\t"			\
+		  /* Handles UTF16 surrogates with convert instruction.  */ \
+		  "20: cu21 %[R_OUT],%[R_IN],1\n\t"			\
+		  "    jo 0b\n\t" /* Try vector implemenation again.  */ \
+		  "    lochil %[R_RES],%[RES_OUT_FULL]\n\t" /* cc == 1.  */ \
+		  "    lochih %[R_RES],%[RES_IN_ILL]\n\t" /* cc == 2.  */ \
+		  ".machine pop"					\
+		  : /* outputs */ [R_IN] "+a" (pInput)			\
+		    , [R_INLEN] "+d" (inlen), [R_OUT] "+a" (pOutput)	\
+		    , [R_OUTLEN] "+d" (outlen), [R_TMP] "=a" (tmp)	\
+		    , [R_TMP2] "=d" (tmp2), [R_TMP3] "=a" (tmp3)	\
+		    , [R_RES] "+d" (result)				\
+		  : /* inputs */					\
+		    [RES_OUT_FULL] "i" (__GCONV_FULL_OUTPUT)		\
+		    , [RES_IN_ILL] "i" (__GCONV_ILLEGAL_INPUT)		\
+		  : /* clobber list */ "memory", "cc"			\
+		    ASM_CLOBBER_VR ("v16") ASM_CLOBBER_VR ("v17")	\
+		    ASM_CLOBBER_VR ("v18") ASM_CLOBBER_VR ("v19")	\
+		    ASM_CLOBBER_VR ("v30") ASM_CLOBBER_VR ("v31")	\
+		  );							\
+    inptr = pInput;							\
+    outptr = pOutput;							\
+									\
+    if (__glibc_likely (inlen == 0)					\
+	|| result == __GCONV_FULL_OUTPUT)				\
+      break;								\
+    if (inlen == 1)							\
+      {									\
+	/* Input does not contain a complete utf16 character.  */	\
+	result = __GCONV_INCOMPLETE_INPUT;				\
+	break;								\
+      }									\
+    else if (result != __GCONV_ILLEGAL_INPUT)				\
+      {									\
+	/* Input is >= 2 and < 4 bytes (as cu21 would have processed	\
+	   a possible next utf16 character) and not illegal.		\
+	   => we have a single high surrogate at end of input.  */	\
+	result = __GCONV_INCOMPLETE_INPUT;				\
+	break;								\
+      }									\
+									\
+    STANDARD_TO_LOOP_ERR_HANDLER (2);					\
+  }
+
+/* Generate loop-function with vector and utf-convert instructions.  */
+# define MIN_NEEDED_INPUT	MIN_NEEDED_TO
+# define MAX_NEEDED_INPUT	MAX_NEEDED_TO
+# define MIN_NEEDED_OUTPUT	MIN_NEEDED_FROM
+# define MAX_NEEDED_OUTPUT	MAX_NEEDED_FROM
+# define TO_LOOP_VX_CU		__to_utf8_loop_vx_cu
+# define LOOPFCT		TO_LOOP_VX_CU
+# define BODY                   BODY_TO_VX_CU
+# define LOOP_NEED_FLAGS
+# include <iconv/loop.c>
+#else
+# define TO_LOOP_VX_CU		NULL
+#endif /* HAVE_TO_VX_CU != 1  */
+
 /* This file also exists in sysdeps/s390/multiarch/ which
    generates ifunc resolvers for FROM/TO_LOOP functions
    and includes iconv/skeleton.c afterwards.  */