Hello! Following the patch that allows unaligned operands in pcmpestri [1], we can substitute x86 asm in lex.c with equivalent builtin functions.
2011-06-18 Uros Bizjak <ubiz...@gmail.com> * lex.c (search_line_sse42): Use __builtin_ia32_loaddqu and __builtin_ia32_pcmpestri128 instead of asm. Bootstrapped and regression tested on x86_64-pc-linux-gnu SSE4.2 target. Also, I have checked that the same code is generated for changed function. OK for mainline? [1] http://gcc.gnu.org/ml/gcc-patches/2012-06/msg01189.html Uros.
Index: lex.c =================================================================== --- lex.c (revision 188750) +++ lex.c (working copy) @@ -420,6 +420,7 @@ search_line_sse42 (const uchar *s, const uchar *en { typedef char v16qi __attribute__ ((__vector_size__ (16))); static const v16qi search = { '\n', '\r', '?', '\\' }; + v16qi sv; uintptr_t si = (uintptr_t)s; uintptr_t index; @@ -439,8 +440,9 @@ search_line_sse42 (const uchar *s, const uchar *en /* ??? The builtin doesn't understand that the PCMPESTRI read from memory need not be aligned. */ - __asm ("%vpcmpestri $0, (%1), %2" - : "=c"(index) : "r"(s), "x"(search), "a"(4), "d"(16)); + sv = __builtin_ia32_loaddqu ((const char *) s); + index = __builtin_ia32_pcmpestri128 (search, 4, sv, 16, 0); + if (__builtin_expect (index < 16, 0)) goto found;