http://gcc.gnu.org/bugzilla/show_bug.cgi?id=50110
--- Comment #4 from David Meggy <david.meggy at icron dot com> 2011-08-17 23:15:26 UTC --- >From rtl.def /* Reference to a signed bit-field of specified size and position. Operand 0 is the memory unit (usually SImode or QImode) which contains the field's first bit. Operand 1 is the width, in bits. Operand 2 is the number of bits in the memory unit before the first bit of this field. If BITS_BIG_ENDIAN is defined, the first bit is the msb and operand 2 counts from the msb of the memory unit. Otherwise, the first bit is the lsb and operand 2 counts from the lsb of the memory unit. This kind of expression can not appear as an lvalue in RTL. */ DEF_RTL_EXPR(SIGN_EXTRACT, "sign_extract", "eee", RTX_BITFIELD_OPS) /* Similar for unsigned bit-field. But note! This kind of expression _can_ appear as an lvalue. */ DEF_RTL_EXPR(ZERO_EXTRACT, "zero_extract", "eee", RTX_BITFIELD_OPS) It appears that this behaviour is intended. To me it seems logical that the extract instruction is really just a "right shift", followed by an "and", where the "and" uses a mask based on the width of the bitfield. To count from the MSB seems counter-intuitive to me. If my understanding is correct this code comment should appear in http://gcc.gnu.org/onlinedocs/gccint/Standard-Names.html#Standard-Names under the extzv and extv headings. I can fix up my define_insn to do the conversion in our implementation.
