src/gallium/drivers/panfrost/pan_attributes.c

   1 /*
   2  * Copyright (C) 2018-2019 Alyssa Rosenzweig
   3  * Copyright (C) 2019 Collabora, Ltd.
   4  *
   5  * Permission is hereby granted, free of charge, to any person obtaining a
   6  * copy of this software and associated documentation files (the "Software"),
   7  * to deal in the Software without restriction, including without limitation
   8  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   9  * and/or sell copies of the Software, and to permit persons to whom the
  10  * Software is furnished to do so, subject to the following conditions:
  11  *
  12  * The above copyright notice and this permission notice (including the next
  13  * paragraph) shall be included in all copies or substantial portions of the
  14  * Software.
  15  *
  16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  17  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  18  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  19  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  20  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
  21  * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  22  * SOFTWARE.
  23  *
  24  */
  25
  26 #include "pan_bo.h"
  27 #include "pan_context.h"
  28
  29 void
  30 panfrost_emit_vertex_data(struct panfrost_batch *batch)
  31 {
  32         struct panfrost_context *ctx = batch->ctx;
  33         struct panfrost_vertex_state *so = ctx->vertex;
  34
  35         /* Staged mali_attr, and index into them. i =/= k, depending on the
  36          * vertex buffer mask and instancing. Twice as much room is allocated,
  37          * for a worst case of NPOT_DIVIDEs which take up extra slot */
  38         union mali_attr attrs[PIPE_MAX_ATTRIBS * 2];
  39         unsigned k = 0;
  40
  41         unsigned vertex_count = ctx->vertex_count;
  42         unsigned instanced_count = ctx->instance_count;
  43
  44         for (unsigned i = 0; i < so->num_elements; ++i) {
  45                 /* We map a mali_attr to be 1:1 with the mali_attr_meta, which
  46                  * means duplicating some vertex buffers (who cares? aside from
  47                  * maybe some caching implications but I somehow doubt that
  48                  * matters) */
  49
  50                 struct pipe_vertex_element *elem = &so->pipe[i];
  51                 unsigned vbi = elem->vertex_buffer_index;
  52
  53                 /* The exception to 1:1 mapping is that we can have multiple
  54                  * entries (NPOT divisors), so we fixup anyways */
  55
  56                 so->hw[i].index = k;
  57
  58                 if (!(ctx->vb_mask & (1 << vbi))) continue;
  59
  60                 struct pipe_vertex_buffer *buf = &ctx->vertex_buffers[vbi];
  61                 struct panfrost_resource *rsrc = (struct panfrost_resource *) (buf->buffer.resource);
  62
  63                 if (!rsrc) continue;
  64
  65                 /* Align to 64 bytes by masking off the lower bits. This
  66                  * will be adjusted back when we fixup the src_offset in
  67                  * mali_attr_meta */
  68
  69                 mali_ptr raw_addr = rsrc->bo->gpu + buf->buffer_offset;
  70                 mali_ptr addr = raw_addr & ~63;
  71                 unsigned chopped_addr = raw_addr - addr;
  72
  73                 /* Add a dependency of the batch on the vertex buffer */
  74                 panfrost_batch_add_bo(batch, rsrc->bo,
  75                                       PAN_BO_ACCESS_SHARED |
  76                                       PAN_BO_ACCESS_READ |
  77                                       PAN_BO_ACCESS_VERTEX_TILER);
  78
  79                 /* Set common fields */
  80                 attrs[k].elements = addr;
  81                 attrs[k].stride = buf->stride;
  82
  83                 /* Since we advanced the base pointer, we shrink the buffer
  84                  * size */
  85                 attrs[k].size = rsrc->base.width0 - buf->buffer_offset;
  86
  87                 /* We need to add the extra size we masked off (for
  88                  * correctness) so the data doesn't get clamped away */
  89                 attrs[k].size += chopped_addr;
  90
  91                 /* For non-instancing make sure we initialize */
  92                 attrs[k].shift = attrs[k].extra_flags = 0;
  93
  94                 /* Instancing uses a dramatically different code path than
  95                  * linear, so dispatch for the actual emission now that the
  96                  * common code is finished */
  97
  98                 unsigned divisor = elem->instance_divisor;
  99
 100                 if (divisor && instanced_count == 1) {
 101                         /* Silly corner case where there's a divisor(=1) but
 102                          * there's no legitimate instancing. So we want *every*
 103                          * attribute to be the same. So set stride to zero so
 104                          * we don't go anywhere. */
 105
 106                         attrs[k].size = attrs[k].stride + chopped_addr;
 107                         attrs[k].stride = 0;
 108                         attrs[k++].elements |= MALI_ATTR_LINEAR;
 109                 } else if (instanced_count <= 1) {
 110                         /* Normal, non-instanced attributes */
 111                         attrs[k++].elements |= MALI_ATTR_LINEAR;
 112                 } else {
 113                         unsigned instance_shift = batch->ctx->payloads[PIPE_SHADER_FRAGMENT].instance_shift;
 114                         unsigned instance_odd = batch->ctx->payloads[PIPE_SHADER_FRAGMENT].instance_odd;
 115
 116                         k += panfrost_vertex_instanced(batch->ctx->padded_count,
 117                                         instance_shift, instance_odd, divisor, &attrs[k]);
 118                 }
 119         }
 120
 121         /* Upload whatever we emitted and go */
 122
 123         ctx->payloads[PIPE_SHADER_VERTEX].postfix.attributes =
 124                 panfrost_upload_transient(batch, attrs, k * sizeof(union mali_attr));
 125 }
 126
 127