src/gallium/drivers/panfrost/pan_allocate.c

   1 /*
   2  * © Copyright 2018 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 <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29 #include <assert.h>
  30 #include <panfrost-misc.h>
  31 #include <panfrost-job.h>
  32 #include "pan_context.h"
  33
  34 /* TODO: What does this actually have to be? */
  35 #define ALIGNMENT 128
  36
  37 /* Allocate a new transient slab */
  38
  39 static struct panfrost_bo *
  40 panfrost_create_slab(struct panfrost_screen *screen, unsigned *index)
  41 {
  42         /* Allocate a new slab on the screen */
  43
  44         struct panfrost_bo **new =
  45                 util_dynarray_grow(&screen->transient_bo,
  46                                 struct panfrost_bo *, 1);
  47
  48         struct panfrost_bo *alloc = panfrost_drm_create_bo(screen, TRANSIENT_SLAB_SIZE, 0);
  49
  50         *new = alloc;
  51
  52         /* Return the BO as well as the index we just added */
  53
  54         *index = util_dynarray_num_elements(&screen->transient_bo, void *) - 1;
  55         return alloc;
  56 }
  57
  58 /* Transient command stream pooling: command stream uploads try to simply copy
  59  * into whereever we left off. If there isn't space, we allocate a new entry
  60  * into the pool and copy there */
  61
  62 struct panfrost_transfer
  63 panfrost_allocate_transient(struct panfrost_context *ctx, size_t sz)
  64 {
  65         struct panfrost_screen *screen = pan_screen(ctx->base.screen);
  66         struct panfrost_batch *batch = panfrost_get_batch_for_fbo(ctx);
  67
  68         /* Pad the size */
  69         sz = ALIGN_POT(sz, ALIGNMENT);
  70
  71         /* Find or create a suitable BO */
  72         struct panfrost_bo *bo = NULL;
  73
  74         unsigned offset = 0;
  75         bool update_offset = false;
  76
  77         pthread_mutex_lock(&screen->transient_lock);
  78         bool has_current = batch->transient_indices.size;
  79         bool fits_in_current = (batch->transient_offset + sz) < TRANSIENT_SLAB_SIZE;
  80
  81         if (likely(has_current && fits_in_current)) {
  82                 /* We can reuse the topmost BO, so get it */
  83                 unsigned idx = util_dynarray_top(&batch->transient_indices, unsigned);
  84                 bo = pan_bo_for_index(screen, idx);
  85
  86                 /* Use the specified offset */
  87                 offset = batch->transient_offset;
  88                 update_offset = true;
  89         } else if (sz < TRANSIENT_SLAB_SIZE) {
  90                 /* We can't reuse the topmost BO, but we can get a new one.
  91                  * First, look for a free slot */
  92
  93                 unsigned count = util_dynarray_num_elements(&screen->transient_bo, void *);
  94                 unsigned index = 0;
  95
  96                 unsigned free = __bitset_ffs(
  97                                 screen->free_transient,
  98                                 count / BITSET_WORDBITS);
  99
 100                 if (likely(free)) {
 101                         /* Use this one */
 102                         index = free - 1;
 103
 104                         /* It's ours, so no longer free */
 105                         BITSET_CLEAR(screen->free_transient, index);
 106
 107                         /* Grab the BO */
 108                         bo = pan_bo_for_index(screen, index);
 109                 } else {
 110                         /* Otherwise, create a new BO */
 111                         bo = panfrost_create_slab(screen, &index);
 112                 }
 113
 114                 panfrost_batch_add_bo(batch, bo);
 115
 116                 /* Remember we created this */
 117                 util_dynarray_append(&batch->transient_indices, unsigned, index);
 118
 119                 update_offset = true;
 120         } else {
 121                 /* Create a new BO and reference it */
 122                 bo = panfrost_drm_create_bo(screen, ALIGN_POT(sz, 4096), 0);
 123                 panfrost_batch_add_bo(batch, bo);
 124
 125                 /* Creating a BO adds a reference, and then the job adds a
 126                  * second one. So we need to pop back one reference */
 127                 panfrost_bo_unreference(&screen->base, bo);
 128         }
 129
 130         struct panfrost_transfer ret = {
 131                 .cpu = bo->cpu + offset,
 132                 .gpu = bo->gpu + offset,
 133         };
 134
 135         if (update_offset)
 136                 batch->transient_offset = offset + sz;
 137         pthread_mutex_unlock(&screen->transient_lock);
 138
 139         return ret;
 140
 141 }
 142
 143 mali_ptr
 144 panfrost_upload_transient(struct panfrost_context *ctx, const void *data, size_t sz)
 145 {
 146         struct panfrost_transfer transfer = panfrost_allocate_transient(ctx, sz);
 147         memcpy(transfer.cpu, data, sz);
 148         return transfer.gpu;
 149 }