1 /*
2 * Copyright © 2022 Imagination Technologies Ltd.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a copy
5 * of this software and associated documentation files (the "Software"), to deal
6 * in the Software without restriction, including without limitation the rights
7 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
8 * copies of the Software, and to permit persons to whom the Software is
9 * furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
18 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
20 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
21 * SOFTWARE.
22 */
23
24 #include <stdbool.h>
25 #include <stdint.h>
26 #include <stddef.h>
27 #include <vulkan/vulkan.h>
28
29 #include "pvr_job_common.h"
30 #include "pvr_job_context.h"
31 #include "pvr_job_transfer.h"
32 #include "pvr_private.h"
33 #include "pvr_winsys.h"
34 #include "util/list.h"
35 #include "util/macros.h"
36 #include "vk_sync.h"
37
38 /* FIXME: Implement gpu based transfer support. */
pvr_transfer_job_submit(struct pvr_device * device,struct pvr_transfer_ctx * ctx,struct pvr_sub_cmd_transfer * sub_cmd,struct vk_sync ** waits,uint32_t wait_count,uint32_t * stage_flags,struct vk_sync * signal_sync)39 VkResult pvr_transfer_job_submit(struct pvr_device *device,
40 struct pvr_transfer_ctx *ctx,
41 struct pvr_sub_cmd_transfer *sub_cmd,
42 struct vk_sync **waits,
43 uint32_t wait_count,
44 uint32_t *stage_flags,
45 struct vk_sync *signal_sync)
46 {
47 /* Wait for transfer semaphores here before doing any transfers. */
48 for (uint32_t i = 0U; i < wait_count; i++) {
49 if (stage_flags[i] & PVR_PIPELINE_STAGE_TRANSFER_BIT) {
50 VkResult result = vk_sync_wait(&device->vk,
51 waits[i],
52 0U,
53 VK_SYNC_WAIT_COMPLETE,
54 UINT64_MAX);
55 if (result != VK_SUCCESS)
56 return result;
57
58 stage_flags[i] &= ~PVR_PIPELINE_STAGE_TRANSFER_BIT;
59 }
60 }
61
62 list_for_each_entry_safe (struct pvr_transfer_cmd,
63 transfer_cmd,
64 &sub_cmd->transfer_cmds,
65 link) {
66 bool src_mapped = false;
67 bool dst_mapped = false;
68 void *src_addr;
69 void *dst_addr;
70 void *ret_ptr;
71
72 /* Map if bo is not mapped. */
73 if (!transfer_cmd->src->vma->bo->map) {
74 src_mapped = true;
75 ret_ptr = device->ws->ops->buffer_map(transfer_cmd->src->vma->bo);
76 if (!ret_ptr)
77 return vk_error(device, VK_ERROR_MEMORY_MAP_FAILED);
78 }
79
80 if (!transfer_cmd->dst->vma->bo->map) {
81 dst_mapped = true;
82 ret_ptr = device->ws->ops->buffer_map(transfer_cmd->dst->vma->bo);
83 if (!ret_ptr)
84 return vk_error(device, VK_ERROR_MEMORY_MAP_FAILED);
85 }
86
87 src_addr =
88 transfer_cmd->src->vma->bo->map + transfer_cmd->src->vma->bo_offset;
89 dst_addr =
90 transfer_cmd->dst->vma->bo->map + transfer_cmd->dst->vma->bo_offset;
91
92 for (uint32_t i = 0; i < transfer_cmd->region_count; i++) {
93 VkBufferCopy2 *region = &transfer_cmd->regions[i];
94
95 memcpy(dst_addr + region->dstOffset,
96 src_addr + region->srcOffset,
97 region->size);
98 }
99
100 if (src_mapped)
101 device->ws->ops->buffer_unmap(transfer_cmd->src->vma->bo);
102
103 if (dst_mapped)
104 device->ws->ops->buffer_unmap(transfer_cmd->dst->vma->bo);
105 }
106
107 /* Given we are doing CPU based copy, completion fence should always be
108 * signaled. This should be fixed when GPU based copy is implemented.
109 */
110 return vk_sync_signal(&device->vk, signal_sync, 0);
111 }
112