nir: get ffma support from NIR options for nir_lower_flrp
[mesa.git] / src / gallium / frontends / vallium / val_pass.c
1 /*
2 * Copyright © 2019 Red Hat.
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is 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
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
21 * IN THE SOFTWARE.
22 */
23
24 #include "val_private.h"
25
26 static void
27 val_render_pass_compile(struct val_render_pass *pass)
28 {
29 for (uint32_t i = 0; i < pass->subpass_count; i++) {
30 struct val_subpass *subpass = &pass->subpasses[i];
31
32 for (uint32_t j = 0; j < subpass->attachment_count; j++) {
33 struct val_subpass_attachment *subpass_att =
34 &subpass->attachments[j];
35 if (subpass_att->attachment == VK_ATTACHMENT_UNUSED)
36 continue;
37
38 struct val_render_pass_attachment *pass_att =
39 &pass->attachments[subpass_att->attachment];
40
41 pass_att->first_subpass_idx = UINT32_MAX;
42 }
43 }
44
45 for (uint32_t i = 0; i < pass->subpass_count; i++) {
46 struct val_subpass *subpass = &pass->subpasses[i];
47 uint32_t color_sample_count = 1, depth_sample_count = 1;
48
49 /* We don't allow depth_stencil_attachment to be non-NULL and
50 * be VK_ATTACHMENT_UNUSED. This way something can just check
51 * for NULL and be guaranteed that they have a valid
52 * attachment.
53 */
54 if (subpass->depth_stencil_attachment &&
55 subpass->depth_stencil_attachment->attachment == VK_ATTACHMENT_UNUSED)
56 subpass->depth_stencil_attachment = NULL;
57
58 if (subpass->ds_resolve_attachment &&
59 subpass->ds_resolve_attachment->attachment == VK_ATTACHMENT_UNUSED)
60 subpass->ds_resolve_attachment = NULL;
61
62 for (uint32_t j = 0; j < subpass->attachment_count; j++) {
63 struct val_subpass_attachment *subpass_att =
64 &subpass->attachments[j];
65 if (subpass_att->attachment == VK_ATTACHMENT_UNUSED)
66 continue;
67
68 struct val_render_pass_attachment *pass_att =
69 &pass->attachments[subpass_att->attachment];
70
71 if (i < pass_att->first_subpass_idx)
72 pass_att->first_subpass_idx = i;
73 pass_att->last_subpass_idx = i;
74 }
75
76 subpass->has_color_att = false;
77 for (uint32_t j = 0; j < subpass->color_count; j++) {
78 struct val_subpass_attachment *subpass_att =
79 &subpass->color_attachments[j];
80 if (subpass_att->attachment == VK_ATTACHMENT_UNUSED)
81 continue;
82
83 subpass->has_color_att = true;
84
85 struct val_render_pass_attachment *pass_att =
86 &pass->attachments[subpass_att->attachment];
87
88 color_sample_count = pass_att->samples;
89 }
90
91 if (subpass->depth_stencil_attachment) {
92 const uint32_t a =
93 subpass->depth_stencil_attachment->attachment;
94 struct val_render_pass_attachment *pass_att =
95 &pass->attachments[a];
96 depth_sample_count = pass_att->samples;
97 }
98
99 subpass->max_sample_count = MAX2(color_sample_count,
100 depth_sample_count);
101
102 /* We have to handle resolve attachments specially */
103 subpass->has_color_resolve = false;
104 if (subpass->resolve_attachments) {
105 for (uint32_t j = 0; j < subpass->color_count; j++) {
106 struct val_subpass_attachment *resolve_att =
107 &subpass->resolve_attachments[j];
108
109 if (resolve_att->attachment == VK_ATTACHMENT_UNUSED)
110 continue;
111
112 subpass->has_color_resolve = true;
113 }
114 }
115
116 for (uint32_t j = 0; j < subpass->input_count; ++j) {
117 if (subpass->input_attachments[j].attachment == VK_ATTACHMENT_UNUSED)
118 continue;
119
120 for (uint32_t k = 0; k < subpass->color_count; ++k) {
121 if (subpass->color_attachments[k].attachment == subpass->input_attachments[j].attachment) {
122 subpass->input_attachments[j].in_render_loop = true;
123 subpass->color_attachments[k].in_render_loop = true;
124 }
125 }
126
127 if (subpass->depth_stencil_attachment &&
128 subpass->depth_stencil_attachment->attachment == subpass->input_attachments[j].attachment) {
129 subpass->input_attachments[j].in_render_loop = true;
130 subpass->depth_stencil_attachment->in_render_loop = true;
131 }
132 }
133 }
134 }
135
136 static unsigned
137 val_num_subpass_attachments(const VkSubpassDescription *desc)
138 {
139 return desc->inputAttachmentCount +
140 desc->colorAttachmentCount +
141 (desc->pResolveAttachments ? desc->colorAttachmentCount : 0) +
142 (desc->pDepthStencilAttachment != NULL);
143 }
144
145 VkResult val_CreateRenderPass(
146 VkDevice _device,
147 const VkRenderPassCreateInfo* pCreateInfo,
148 const VkAllocationCallbacks* pAllocator,
149 VkRenderPass* pRenderPass)
150 {
151 VAL_FROM_HANDLE(val_device, device, _device);
152 struct val_render_pass *pass;
153 size_t size;
154 size_t attachments_offset;
155
156 assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO);
157
158 size = sizeof(*pass);
159 size += pCreateInfo->subpassCount * sizeof(pass->subpasses[0]);
160 attachments_offset = size;
161 size += pCreateInfo->attachmentCount * sizeof(pass->attachments[0]);
162
163 pass = vk_alloc2(&device->alloc, pAllocator, size, 8,
164 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
165 if (pass == NULL)
166 return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
167
168 /* Clear the subpasses along with the parent pass. This required because
169 * each array member of val_subpass must be a valid pointer if not NULL.
170 */
171 memset(pass, 0, size);
172
173 vk_object_base_init(&device->vk, &pass->base,
174 VK_OBJECT_TYPE_RENDER_PASS);
175 pass->attachment_count = pCreateInfo->attachmentCount;
176 pass->subpass_count = pCreateInfo->subpassCount;
177 pass->attachments = (void *) pass + attachments_offset;
178
179 for (uint32_t i = 0; i < pCreateInfo->attachmentCount; i++) {
180 struct val_render_pass_attachment *att = &pass->attachments[i];
181
182 att->format = pCreateInfo->pAttachments[i].format;
183 att->samples = pCreateInfo->pAttachments[i].samples;
184 att->load_op = pCreateInfo->pAttachments[i].loadOp;
185 att->stencil_load_op = pCreateInfo->pAttachments[i].stencilLoadOp;
186 att->final_layout = pCreateInfo->pAttachments[i].finalLayout;
187 att->first_subpass_idx = UINT32_MAX;
188 }
189
190 uint32_t subpass_attachment_count = 0;
191 for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
192 subpass_attachment_count += val_num_subpass_attachments(&pCreateInfo->pSubpasses[i]);
193 }
194
195 if (subpass_attachment_count) {
196 pass->subpass_attachments =
197 vk_alloc2(&device->alloc, pAllocator,
198 subpass_attachment_count * sizeof(struct val_subpass_attachment), 8,
199 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
200 if (pass->subpass_attachments == NULL) {
201 vk_free2(&device->alloc, pAllocator, pass);
202 return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
203 }
204 } else
205 pass->subpass_attachments = NULL;
206
207 struct val_subpass_attachment *p = pass->subpass_attachments;
208 for (uint32_t i = 0; i < pCreateInfo->subpassCount; i++) {
209 const VkSubpassDescription *desc = &pCreateInfo->pSubpasses[i];
210 struct val_subpass *subpass = &pass->subpasses[i];
211
212 subpass->input_count = desc->inputAttachmentCount;
213 subpass->color_count = desc->colorAttachmentCount;
214 subpass->attachment_count = val_num_subpass_attachments(desc);
215 subpass->attachments = p;
216
217 if (desc->inputAttachmentCount > 0) {
218 subpass->input_attachments = p;
219 p += desc->inputAttachmentCount;
220
221 for (uint32_t j = 0; j < desc->inputAttachmentCount; j++) {
222 subpass->input_attachments[j] = (struct val_subpass_attachment) {
223 .attachment = desc->pInputAttachments[j].attachment,
224 .layout = desc->pInputAttachments[j].layout,
225 };
226 }
227 }
228
229 if (desc->colorAttachmentCount > 0) {
230 subpass->color_attachments = p;
231 p += desc->colorAttachmentCount;
232
233 for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
234 subpass->color_attachments[j] = (struct val_subpass_attachment) {
235 .attachment = desc->pColorAttachments[j].attachment,
236 .layout = desc->pColorAttachments[j].layout,
237 };
238 }
239 }
240
241 if (desc->pResolveAttachments) {
242 subpass->resolve_attachments = p;
243 p += desc->colorAttachmentCount;
244
245 for (uint32_t j = 0; j < desc->colorAttachmentCount; j++) {
246 subpass->resolve_attachments[j] = (struct val_subpass_attachment) {
247 .attachment = desc->pResolveAttachments[j].attachment,
248 .layout = desc->pResolveAttachments[j].layout,
249 };
250 }
251 }
252
253 if (desc->pDepthStencilAttachment) {
254 subpass->depth_stencil_attachment = p++;
255
256 *subpass->depth_stencil_attachment = (struct val_subpass_attachment) {
257 .attachment = desc->pDepthStencilAttachment->attachment,
258 .layout = desc->pDepthStencilAttachment->layout,
259 };
260 }
261 }
262
263 val_render_pass_compile(pass);
264 *pRenderPass = val_render_pass_to_handle(pass);
265
266 return VK_SUCCESS;
267 }
268
269 void val_DestroyRenderPass(
270 VkDevice _device,
271 VkRenderPass _pass,
272 const VkAllocationCallbacks* pAllocator)
273 {
274 VAL_FROM_HANDLE(val_device, device, _device);
275 VAL_FROM_HANDLE(val_render_pass, pass, _pass);
276
277 if (!_pass)
278 return;
279 vk_object_base_finish(&pass->base);
280 vk_free2(&device->alloc, pAllocator, pass->subpass_attachments);
281 vk_free2(&device->alloc, pAllocator, pass);
282 }
283
284 void val_GetRenderAreaGranularity(
285 VkDevice device,
286 VkRenderPass renderPass,
287 VkExtent2D* pGranularity)
288 {
289 *pGranularity = (VkExtent2D) { 1, 1 };
290 }