[Mesa-dev] [RFC PATCH 09/10] i965/clip: Use clip distances for all user clipping

[Chris Forbes]

This is what I started with, but brw_ENDIF destroys the flags.

On Tue, Aug 6, 2013 at 9:40 AM, Paul Berry <stereotype441 at gmail.com> wrote:
> On 3 August 2013 19:59, Chris Forbes <chrisf at ijw.co.nz> wrote:
>>
>> Signed-off-by: Chris Forbes <chrisf at ijw.co.nz>
>> ---
>>  src/mesa/drivers/dri/i965/brw_clip_tri.c | 7 +++++--
>>  1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/src/mesa/drivers/dri/i965/brw_clip_tri.c
>> b/src/mesa/drivers/dri/i965/brw_clip_tri.c
>> index bd3ffb3..4d2eb0a 100644
>> --- a/src/mesa/drivers/dri/i965/brw_clip_tri.c
>> +++ b/src/mesa/drivers/dri/i965/brw_clip_tri.c
>> @@ -241,16 +241,19 @@ load_clip_distance(struct brw_clip_compile *c,
>> struct brw_indirect vtx,
>>     brw_AND(p, vec1(brw_null_reg()), c->reg.vertex_src_mask,
>> brw_imm_ud(1));
>>     brw_IF(p, BRW_EXECUTE_1);
>>     {
>> -      brw_MOV(p, dst, deref_4f(vtx, clip_offset));
>> +      struct brw_indirect temp_ptr = brw_indirect(7, 0);
>> +      brw_ADD(p, get_addr_reg(temp_ptr), get_addr_reg(vtx),
>> c->reg.clipdistance_offset);
>> +      brw_MOV(p, vec1(dst), deref_1f(temp_ptr, 0));
>>     }
>>     brw_ELSE(p);
>>     {
>>        brw_MOV(p, dst, deref_4f(vtx, hpos_offset));
>> +      brw_DP4(p, dst, dst, c->reg.plane_equation);
>>     }
>>     brw_ENDIF(p);
>>
>>     brw_set_conditionalmod(p, cond);
>> -   brw_DP4(p, dst, dst, c->reg.plane_equation);
>> +   brw_CMP(p, brw_null_reg(), cond, vec1(dst), brw_imm_f(0.0f));
>>  }
>
>
> I think you can use one less instruction if you do this instead:
>
>    brw_IF(p, BRW_EXECUTE_1);
>    {
>
>       struct brw_indirect temp_ptr = brw_indirect(7, 0);
>       brw_ADD(p, get_addr_reg(temp_ptr), get_addr_reg(vtx),
> c->reg.clipdistance_offset);
>       brw_set_conditionalmod(p, cond);
>
>       brw_MOV(p, vec1(dst), deref_1f(temp_ptr, 0));
>    }
>    brw_ELSE(p);
>    {
>       brw_MOV(p, dst, deref_4f(vtx, hpos_offset));
>       brw_set_conditionalmod(p, cond);
>
>       brw_DP4(p, dst, dst, c->reg.plane_equation);
>    }
>    brw_ENDIF(p);
>
> But I haven't tried it so it's possible that it doesn't work for some
> non-obvious reason :)
>
> In any case, this patch is:
>
> Reviewed-by: Paul Berry <stereotype441 at gmail.com>