src/gallium/drivers/freedreno/a4xx/fd4_query.c

   1 /* -*- mode: C; c-file-style: "k&r"; tab-width 4; indent-tabs-mode: t; -*- */
   2
   3 /*
   4  * Copyright (C) 2014 Rob Clark <robclark@freedesktop.org>
   5  *
   6  * Permission is hereby granted, free of charge, to any person obtaining a
   7  * copy of this software and associated documentation files (the "Software"),
   8  * to deal in the Software without restriction, including without limitation
   9  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
  10  * and/or sell copies of the Software, and to permit persons to whom the
  11  * Software is furnished to do so, subject to the following conditions:
  12  *
  13  * The above copyright notice and this permission notice (including the next
  14  * paragraph) shall be included in all copies or substantial portions of the
  15  * Software.
  16  *
  17  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  18  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  19  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  20  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  21  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  22  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  23  * SOFTWARE.
  24  *
  25  * Authors:
  26  *    Rob Clark <robclark@freedesktop.org>
  27  */
  28
  29 #include "freedreno_query_hw.h"
  30 #include "freedreno_context.h"
  31 #include "freedreno_util.h"
  32
  33 #include "fd4_query.h"
  34 #include "fd4_draw.h"
  35 #include "fd4_format.h"
  36
  37
  38 struct fd_rb_samp_ctrs {
  39         uint64_t ctr[16];
  40 };
  41
  42 /*
  43  * Occlusion Query:
  44  *
  45  * OCCLUSION_COUNTER and OCCLUSION_PREDICATE differ only in how they
  46  * interpret results
  47  */
  48
  49 static struct fd_hw_sample *
  50 occlusion_get_sample(struct fd_context *ctx, struct fd_ringbuffer *ring)
  51 {
  52         struct fd_hw_sample *samp =
  53                         fd_hw_sample_init(ctx, sizeof(struct fd_rb_samp_ctrs));
  54
  55         /* low bits of sample addr should be zero (since they are control
  56          * flags in RB_SAMPLE_COUNT_CONTROL):
  57          */
  58         debug_assert((samp->offset & 0x3) == 0);
  59
  60         /* Set RB_SAMPLE_COUNT_ADDR to samp->offset plus value of
  61          * HW_QUERY_BASE_REG register:
  62          */
  63         OUT_PKT3(ring, CP_SET_CONSTANT, 3);
  64         OUT_RING(ring, CP_REG(REG_A4XX_RB_SAMPLE_COUNT_CONTROL) | 0x80000000);
  65         OUT_RING(ring, HW_QUERY_BASE_REG);
  66         OUT_RING(ring, A4XX_RB_SAMPLE_COUNT_CONTROL_COPY |
  67                         samp->offset);
  68
  69         OUT_PKT3(ring, CP_DRAW_INDX_OFFSET, 3);
  70         OUT_RING(ring, DRAW4(DI_PT_POINTLIST_PSIZE, DI_SRC_SEL_AUTO_INDEX,
  71                                                 INDEX4_SIZE_32_BIT, USE_VISIBILITY));
  72         OUT_RING(ring, 1);             /* NumInstances */
  73         OUT_RING(ring, 0);             /* NumIndices */
  74
  75         fd_event_write(ctx, ring, ZPASS_DONE);
  76
  77         return samp;
  78 }
  79
  80 static uint64_t
  81 count_samples(const struct fd_rb_samp_ctrs *start,
  82                 const struct fd_rb_samp_ctrs *end)
  83 {
  84         return end->ctr[0] - start->ctr[0];
  85 }
  86
  87 static void
  88 occlusion_counter_accumulate_result(struct fd_context *ctx,
  89                 const void *start, const void *end,
  90                 union pipe_query_result *result)
  91 {
  92         uint64_t n = count_samples(start, end);
  93         result->u64 += n;
  94 }
  95
  96 static void
  97 occlusion_predicate_accumulate_result(struct fd_context *ctx,
  98                 const void *start, const void *end,
  99                 union pipe_query_result *result)
 100 {
 101         uint64_t n = count_samples(start, end);
 102         result->b |= (n > 0);
 103 }
 104
 105 static const struct fd_hw_sample_provider occlusion_counter = {
 106                 .query_type = PIPE_QUERY_OCCLUSION_COUNTER,
 107                 .active = FD_STAGE_DRAW,
 108                 .get_sample = occlusion_get_sample,
 109                 .accumulate_result = occlusion_counter_accumulate_result,
 110 };
 111
 112 static const struct fd_hw_sample_provider occlusion_predicate = {
 113                 .query_type = PIPE_QUERY_OCCLUSION_PREDICATE,
 114                 .active = FD_STAGE_DRAW,
 115                 .get_sample = occlusion_get_sample,
 116                 .accumulate_result = occlusion_predicate_accumulate_result,
 117 };
 118
 119 void fd4_query_context_init(struct pipe_context *pctx)
 120 {
 121         fd_hw_query_register_provider(pctx, &occlusion_counter);
 122         fd_hw_query_register_provider(pctx, &occlusion_predicate);
 123 }