This sounds a bit too trivial. OTOH, I had problem with people not
putting one space after the "//" as comment.
This happens almost always with China trained engrs.
Please add that space. It helps to keep formats consistent and a bit
more easy reading
Sun
On Tue, Aug 31, 2010 at 10:55 PM, <s...@open64.net> wrote:
> Author: zhuqing
> Date: 2010-09-01 01:55:28 -0400 (Wed, 01 Sep 2010)
> New Revision: 3332
>
> Modified:
> trunk/osprey/wgen/wgen_expr.cxx
> Log:
> add intrinsic for _mm_extract_pi16
>
> Modified: trunk/osprey/wgen/wgen_expr.cxx
> ===================================================================
> --- trunk/osprey/wgen/wgen_expr.cxx 2010-09-01 05:42:02 UTC (rev 3331)
> +++ trunk/osprey/wgen/wgen_expr.cxx 2010-09-01 05:55:28 UTC (rev 3332)
> @@ -3967,7 +3967,29 @@
> case GSBI_IX86_BUILTIN_VEC_EXT_V4SI:
> *iopc = INTRN_VEC_EXT_V2SI;
> break;
> - case GSBI_IX86_BUILTIN_VEC_EXT_V2DI:
> + case GSBI_IX86_BUILTIN_VEC_EXT_V4HI:
> + //add intrinsic for _mm_extract_pi16
> + if (WN_operator(arg1) != OPR_INTCONST)
> + Fail_FmtAssertion ("selector must be an integer constant in the range
> 0..3");
> + switch (WN_const_val(arg1)){
> + case 0:
> + *iopc = INTRN_PEXTRW0;
> + break;
> + case 1:
> + *iopc = INTRN_PEXTRW1;
> + break;
> + case 2:
> + *iopc = INTRN_PEXTRW2;
> + break;
> + case 3:
> + *iopc = INTRN_PEXTRW3;
> + break;
> + default:
> + Fail_FmtAssertion ("selector must be an integer constant in the range
> 0..3" );
> + }
> + break;
> +
> + case GSBI_IX86_BUILTIN_VEC_EXT_V2DI:
> *iopc = INTRN_VEC_EXT_V2SI;
> break;
> case GSBI_IX86_BUILTIN_VEC_EXT_V2DF:
> @@ -9294,7 +9316,18 @@
> for (i = 0, list = gs_tree_operand (exp, 1);
> list;
> i++, list = gs_tree_chain (list)) {
> - arg_wn = WGEN_Expand_Expr (gs_tree_value (list));
> + //if intrinsic equals _mm_extract_pi16,
> + //the second param is fixed to const 0,1,2,3, so
> + //there is no need to generate the second arg node.
> + if (i ==1 && (iopc == INTRN_PEXTRW0 ||
> + iopc == INTRN_PEXTRW1 ||
> + iopc == INTRN_PEXTRW2 ||
> + iopc == INTRN_PEXTRW3)) {
> + num_args = 1;
> + break;
> + }
> + arg_wn = WGEN_Expand_Expr (gs_tree_value (list));
> +
> #ifdef KEY // bug 11286
> if (i == 1 && TARGET_64BIT &&
> (gs_decl_function_code(func) == GSBI_BUILT_IN_POWI ||
>
>
> ------------------------------------------------------------------------------
> This SF.net Dev2Dev email is sponsored by:
>
> Show off your parallel programming skills.
> Enter the Intel(R) Threading Challenge 2010.
> http://p.sf.net/sfu/intel-thread-sfd
> _______________________________________________
> Open64-devel mailing list
> Open64-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/open64-devel
>
------------------------------------------------------------------------------
This SF.net Dev2Dev email is sponsored by:
Show off your parallel programming skills.
Enter the Intel(R) Threading Challenge 2010.
http://p.sf.net/sfu/intel-thread-sfd
_______________________________________________
Open64-devel mailing list
Open64-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/open64-devel
