[Mesa-dev] [PATCH v3 06/19] i965/gen6_depth_state.c: Remove (gen < 6) code paths

[Jordan Justen]

On Fri, Aug 1, 2014 at 9:28 AM, Pohjolainen, Topi
<topi.pohjolainen at intel.com> wrote:
> On Fri, Aug 01, 2014 at 09:18:00AM -0700, Jordan Justen wrote:
>> On Fri, Aug 1, 2014 at 2:28 AM, Pohjolainen, Topi
>> <topi.pohjolainen at intel.com> wrote:
>> > On Fri, Aug 01, 2014 at 12:53:36AM -0700, Jordan Justen wrote:
>> >> Since this code was branched from brw_misc_state.c, it had support for
>> >> gen < 6. We can now remove this.
>> >>
>> >> Signed-off-by: Jordan Justen <jordan.l.justen at intel.com>
>> >> ---
>> >>  src/mesa/drivers/dri/i965/gen6_depth_state.c | 40 +++++++++-------------------
>> >>  1 file changed, 13 insertions(+), 27 deletions(-)
>> >>
>> >> diff --git a/src/mesa/drivers/dri/i965/gen6_depth_state.c b/src/mesa/drivers/dri/i965/gen6_depth_state.c
>> >> index d37aae8..2cb60b3 100644
>> >> --- a/src/mesa/drivers/dri/i965/gen6_depth_state.c
>> >> +++ b/src/mesa/drivers/dri/i965/gen6_depth_state.c
>> >> @@ -63,16 +63,8 @@ gen6_emit_depth_stencil_hiz(struct brw_context *brw,
>> >>        intel_emit_depth_stall_flushes(brw);
>> >>     }
>> >>
>> >> -   unsigned int len;
>> >> -   if (brw->gen >= 6)
>> >> -      len = 7;
>> >> -   else if (brw->is_g4x || brw->gen == 5)
>> >> -      len = 6;
>> >> -   else
>> >> -      len = 5;
>> >> -
>> >> -   BEGIN_BATCH(len);
>> >> -   OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (len - 2));
>> >> +   BEGIN_BATCH(7);
>> >> +   OUT_BATCH(_3DSTATE_DEPTH_BUFFER << 16 | (7 - 2));
>> >>     OUT_BATCH((depth_mt ? depth_mt->pitch - 1 : 0) |
>> >>               (depthbuffer_format << 18) |
>> >>               ((enable_hiz_ss ? 1 : 0) << 21) | /* separate stencil enable */
>> >> @@ -94,13 +86,9 @@ gen6_emit_depth_stencil_hiz(struct brw_context *brw,
>> >>               ((height + tile_y - 1) << 19));
>> >>     OUT_BATCH(0);
>> >>
>> >> -   if (brw->is_g4x || brw->gen >= 5)
>> >> -      OUT_BATCH(tile_x | (tile_y << 16));
>> >> -   else
>> >> -      assert(tile_x == 0 && tile_y == 0);
>> >> +   OUT_BATCH(tile_x | (tile_y << 16));
>> >>
>> >> -   if (brw->gen >= 6)
>> >> -      OUT_BATCH(0);
>> >> +   OUT_BATCH(0);
>> >>
>> >>     ADVANCE_BATCH();
>> >>
>> >> @@ -162,15 +150,13 @@ gen6_emit_depth_stencil_hiz(struct brw_context *brw,
>> >>      *     3DSTATE_CLEAR_PARAMS packet must follow the DEPTH_BUFFER_STATE packet
>> >>      *     when HiZ is enabled and the DEPTH_BUFFER_STATE changes.
>> >>      */
>> >> -   if (brw->gen >= 6 || hiz) {
>> >> -      if (brw->gen == 6)
>> >> -      intel_emit_post_sync_nonzero_flush(brw);
>> >> -
>> >> -      BEGIN_BATCH(2);
>> >> -      OUT_BATCH(_3DSTATE_CLEAR_PARAMS << 16 |
>> >> -             GEN5_DEPTH_CLEAR_VALID |
>> >> -             (2 - 2));
>> >> -      OUT_BATCH(depth_mt ? depth_mt->depth_clear_value : 0);
>> >> -      ADVANCE_BATCH();
>> >> -   }
>> >> +   if (brw->gen == 6)
>> >
>> > This path is not going to be used with anything else than gen6, right? With
>> > this removed:
>> >
>> > Reviewed-by: Topi Pohjolainen <topi.pohjolainen at intel.com>
>> >
>> > I guess we need an equivalent patch for the gen < 6 path (removing gen6
>> > logic).
>>
>> You mentioned potentially combining some gen6-gen8 paths, so I figured
>> I'd leave the (gen == 6) checks. But, I notice in your new series, you
>> kept gen6 separate. So, should we make these paths gen6 only?
>
> I think we probably should do that, there is quite a bit delta to gen7 and
> merging these two needs more thought. It would be clearer in the mean time.

Ok, I'll make patches 02 & 06 assume gen == 6 rather than gen > 6. If
we merge the gen6+ paths we can re-add the gen == 6 checks as needed.

-Jordan