[Beignet] [PATCH] Fix a HSW regression.

[Zhigang Gong]

LGTM, just pushed, Thanks.

On Mon, Oct 13, 2014 at 02:48:06PM +0800, Yang Rong wrote:
> HSW's JMPI instruction has 32bits JIP, can't merge JIP and UIP as other jump instruction.
> 
> Signed-off-by: Yang Rong <rong.r.yang at intel.com>
> ---
>  backend/src/backend/gen75_encoder.cpp | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/backend/src/backend/gen75_encoder.cpp b/backend/src/backend/gen75_encoder.cpp
> index 10a9b5c..6a357c2 100644
> --- a/backend/src/backend/gen75_encoder.cpp
> +++ b/backend/src/backend/gen75_encoder.cpp
> @@ -270,11 +270,13 @@ namespace gbe
>        }
>      }
>  
> -    if (insn.header.opcode == GEN_OPCODE_JMPI) {
> +    if (insn.header.opcode != GEN_OPCODE_JMPI)
> +      this->setSrc1(&insn, GenRegister::immd((jip & 0xffff) | (uip<<16)));
> +    else if (insn.header.opcode == GEN_OPCODE_JMPI) {
>        //jumpDistance'unit is Qword, and the HSW's JMPI offset of jmpi is in byte, so multi 8
>        jip = (jip - 2) * 8;
> +      this->setSrc1(&insn, GenRegister::immd(jip));
>      }
> -    this->setSrc1(&insn, GenRegister::immd((jip & 0xffff) | uip<<16));
>      return;
>    }
>  } /* End of the name space. */
> -- 
> 1.8.3.2
> 
> _______________________________________________
> Beignet mailing list
> Beignet at lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/beignet