Home | History | Annotate | Download | only in i965
      1 /*
      2  Copyright (C) Intel Corp.  2006.  All Rights Reserved.
      3  Intel funded Tungsten Graphics (http://www.tungstengraphics.com) to
      4  develop this 3D driver.
      5 
      6  Permission is hereby granted, free of charge, to any person obtaining
      7  a copy of this software and associated documentation files (the
      8  "Software"), to deal in the Software without restriction, including
      9  without limitation the rights to use, copy, modify, merge, publish,
     10  distribute, sublicense, and/or sell copies of the Software, and to
     11  permit persons to whom the Software is furnished to do so, subject to
     12  the following conditions:
     13 
     14  The above copyright notice and this permission notice (including the
     15  next paragraph) shall be included in all copies or substantial
     16  portions of the Software.
     17 
     18  THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
     19  EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
     20  MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
     21  IN NO EVENT SHALL THE COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS BE
     22  LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION
     23  OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION
     24  WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
     25 
     26  **********************************************************************/
     27  /*
     28   * Authors:
     29   *   Keith Whitwell <keith (at) tungstengraphics.com>
     30   */
     31 
     32 #include "brw_context.h"
     33 #include "brw_state.h"
     34 #include "brw_defines.h"
     35 
     36 static void
     37 brw_upload_clip_unit(struct brw_context *brw)
     38 {
     39    struct intel_context *intel = &brw->intel;
     40    struct gl_context *ctx = &intel->ctx;
     41    struct brw_clip_unit_state *clip;
     42 
     43    clip = brw_state_batch(brw, AUB_TRACE_CLIP_STATE,
     44 			  sizeof(*clip), 32, &brw->clip.state_offset);
     45    memset(clip, 0, sizeof(*clip));
     46 
     47    /* BRW_NEW_PROGRAM_CACHE | CACHE_NEW_CLIP_PROG */
     48    clip->thread0.grf_reg_count = (ALIGN(brw->clip.prog_data->total_grf, 16) /
     49 				 16 - 1);
     50    clip->thread0.kernel_start_pointer =
     51       brw_program_reloc(brw,
     52 			brw->clip.state_offset +
     53 			offsetof(struct brw_clip_unit_state, thread0),
     54 			brw->clip.prog_offset +
     55 			(clip->thread0.grf_reg_count << 1)) >> 6;
     56 
     57    clip->thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754;
     58    clip->thread1.single_program_flow = 1;
     59 
     60    clip->thread3.urb_entry_read_length = brw->clip.prog_data->urb_read_length;
     61    clip->thread3.const_urb_entry_read_length =
     62       brw->clip.prog_data->curb_read_length;
     63 
     64    /* BRW_NEW_CURBE_OFFSETS */
     65    clip->thread3.const_urb_entry_read_offset = brw->curbe.clip_start * 2;
     66    clip->thread3.dispatch_grf_start_reg = 1;
     67    clip->thread3.urb_entry_read_offset = 0;
     68 
     69    /* BRW_NEW_URB_FENCE */
     70    clip->thread4.nr_urb_entries = brw->urb.nr_clip_entries;
     71    clip->thread4.urb_entry_allocation_size = brw->urb.vsize - 1;
     72    /* If we have enough clip URB entries to run two threads, do so.
     73     */
     74    if (brw->urb.nr_clip_entries >= 10) {
     75       /* Half of the URB entries go to each thread, and it has to be an
     76        * even number.
     77        */
     78       assert(brw->urb.nr_clip_entries % 2 == 0);
     79 
     80       /* Although up to 16 concurrent Clip threads are allowed on Ironlake,
     81        * only 2 threads can output VUEs at a time.
     82        */
     83       if (intel->gen == 5)
     84          clip->thread4.max_threads = 16 - 1;
     85       else
     86          clip->thread4.max_threads = 2 - 1;
     87    } else {
     88       assert(brw->urb.nr_clip_entries >= 5);
     89       clip->thread4.max_threads = 1 - 1;
     90    }
     91 
     92    if (unlikely(INTEL_DEBUG & DEBUG_STATS))
     93       clip->thread4.stats_enable = 1;
     94 
     95    clip->clip5.userclip_enable_flags = 0x7f;
     96    clip->clip5.userclip_must_clip = 1;
     97    clip->clip5.guard_band_enable = 0;
     98    /* _NEW_TRANSFORM */
     99    if (!ctx->Transform.DepthClamp)
    100       clip->clip5.viewport_z_clip_enable = 1;
    101    clip->clip5.viewport_xy_clip_enable = 1;
    102    clip->clip5.vertex_position_space = BRW_CLIP_NDCSPACE;
    103    clip->clip5.api_mode = BRW_CLIP_API_OGL;
    104    clip->clip5.clip_mode = brw->clip.prog_data->clip_mode;
    105 
    106    if (intel->is_g4x)
    107       clip->clip5.negative_w_clip_test = 1;
    108 
    109    clip->clip6.clipper_viewport_state_ptr = 0;
    110    clip->viewport_xmin = -1;
    111    clip->viewport_xmax = 1;
    112    clip->viewport_ymin = -1;
    113    clip->viewport_ymax = 1;
    114 
    115    brw->state.dirty.cache |= CACHE_NEW_CLIP_UNIT;
    116 }
    117 
    118 const struct brw_tracked_state brw_clip_unit = {
    119    .dirty = {
    120       .mesa  = _NEW_TRANSFORM,
    121       .brw   = (BRW_NEW_BATCH |
    122 		BRW_NEW_PROGRAM_CACHE |
    123 		BRW_NEW_CURBE_OFFSETS |
    124 		BRW_NEW_URB_FENCE),
    125       .cache = CACHE_NEW_CLIP_PROG
    126    },
    127    .emit = brw_upload_clip_unit,
    128 };
    129