This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH, middle-end]: Fix mode-switching MODE_EXIT check with __builtin_apply/__builtin_return

> This sequence breaks assumption in mode-switching.c, that final
> function return register load operates in MODE_EXIT mode and triggere
> following code:
> 		    for (j = n_entities - 1; j >= 0; j--)
> 		      {
> 			int e = entity_map[j];
> 			int mode = MODE_NEEDED (e, return_copy);
> 			if (mode != num_modes[e] && mode != MODE_EXIT (e))
> 			  break;
> 		      }
> As discussed above, modes of loads, generated from __builtin_apply
> have no connection to function return mode. mode-switching.c does
> detect __builtin_apply situation and raises maybe_builtin_apply flag,
> but doesn't use it to short-circuit wrong check. In proposed patch, we
> detect this situation and raise force_late_switch in the same way, as
> SH4 does for its "late" fpscr emission.

If I understand correctly, we need to insert the vzeroupper because the 
function returns double in SSE registers but we generate an OImode load 
instead of a DFmode load because of the __builtin_return.  So we're in the 
forced_late_switch case but we fail to recognize the tweaked return value load 
since the number of registers doesn't match.

If so, I'd rather add another special case, like for the SH4, instead of a 
generic bypass for maybe_builtin_apply, something along the lines of:

	/* For the x86 with AVX, we might be using a larger load for a value
	   returned in SSE registers and we want to put the final mode switch
	   after this return value copy.  */
	if (copy_start == ret_start
	    && nregs == hard_regno_nregs[ret_start][GET_MODE (ret_reg)]
	    && copy_num >= nregs
	    && OBJECT_P (SET_SRC (return_copy_pat)))
	  forced_late_switch = 1;

> 2012-11-04  Vladimir Yakovlev  <>
> 	    Uros Bizjak  <>
>         * mode-switching.c (create_pre_exit): Added code for
> maybe_builtin_apply case.

Present tense in ChangeLog.

Eric Botcazou

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]