The el_match function on lines 2393-2615 of backend/el.c doesn't elaborate cases for simd. I'm not going to make a pull request because I don't have time to work through the lengthy process of isolating a particular bug, so, here they are for anyone that needs to tweak their own dmd:

	case TYfloat4:
	case TYdouble2:
	case TYschar16:
	case TYuchar16:
	case TYshort8:
	case TYushort8:
	case TYlong4:
	case TYulong4:
	case TYllong2:
	case TYullong2:
		if(n1->EV.Vcent.msw != n2->EV.Vcent.msw || n1->EV.Vcent.lsw != n2->EV.Vcent.lsw)
			goto nomatch;
			break;

"Etienne"  wrote in message news:m9lvn5$28cr$1@digitalmars.com... 

> The el_match function on lines 2393-2615 of backend/el.c doesn't elaborate cases for simd. I'm not going to make a pull request because I don't have time to work through the lengthy process of isolating a particular bug, so, here they are for anyone that needs to tweak their own dmd:
> 
> case TYfloat4:
> case TYdouble2:
> case TYschar16:
> case TYuchar16:
> case TYshort8:
> case TYushort8:
> case TYlong4:
> case TYulong4:
> case TYllong2:
> case TYullong2:
> if(n1->EV.Vcent.msw != n2->EV.Vcent.msw || n1->EV.Vcent.lsw != n2->EV.Vcent.lsw)
> goto nomatch;
> break;

Is this in bugzilla somewhere?