Re: [RFC] patch to fix an ICE involving sign-extract of mmx expression

2005-09-26 Thread Fariborz Jahanian
On Sep 23, 2005, at 12:41 PM, Richard Henderson wrote: On Thu, Sep 22, 2005 at 01:21:06PM -0700, Fariborz Jahanian wrote: /* Avoid creating invalid subregs, for example when simplifying (x>>32)&255. */ ! if (final_word >= GET_MODE_SIZE (

Re: [RFC] patch to fix an ICE involving sign-extract of mmx expression

2005-09-23 Thread Fariborz Jahanian
On Sep 23, 2005, at 12:41 PM, Richard Henderson wrote: On Thu, Sep 22, 2005 at 01:21:06PM -0700, Fariborz Jahanian wrote: /* Avoid creating invalid subregs, for example when simplifying (x>>32)&255. */ ! if (final_word >= GET_MODE_SIZE (

Re: [RFC] patch to fix an ICE involving sign-extract of mmx expression

2005-09-23 Thread Richard Henderson
On Thu, Sep 22, 2005 at 01:21:06PM -0700, Fariborz Jahanian wrote: > /* Avoid creating invalid subregs, for example when > simplifying (x>>32)&255. */ > ! if (final_word >= GET_MODE_SIZE (inner_mode) > ! || (final_word % GE

Re: [RFC] patch to fix an ICE involving sign-extract of mmx expression

2005-09-22 Thread Andrew Pinski
On Sep 22, 2005, at 4:21 PM, Fariborz Jahanian wrote: Following patch avoids this problem. If this is OK, I will submit a patch when fsf mainline is unfrozen. The FSF mainline is not frozen, only the 4.0 release branch. Thanks, Andrew Pinski

[RFC] patch to fix an ICE involving sign-extract of mmx expression

2005-09-22 Thread Fariborz Jahanian
In a given test case with 128 bit mmx intrinsics, routine make_compound_operation (in combine.c) attempts to do a sign-extract of the middle 64bit of the 128 bit (TImode) register. Pattern we have is: (lshiftrt:TI (ashift:TI (subreg:TI (reg/v:V2DI 75 [ vu16YPrediction3 ]) 0) (const_int 32