2 * Copyright © 2016 Red Hat.
3 * Copyright © 2016 Bas Nieuwenhuizen
5 * based in part on anv driver which is:
6 * Copyright © 2015 Intel Corporation
8 * Permission is hereby granted, free of charge, to any person obtaining a
9 * copy of this software and associated documentation files (the "Software"),
10 * to deal in the Software without restriction, including without limitation
11 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
12 * and/or sell copies of the Software, and to permit persons to whom the
13 * Software is furnished to do so, subject to the following conditions:
15 * The above copyright notice and this permission notice (including the next
16 * paragraph) shall be included in all copies or substantial portions of the
19 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
20 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
21 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
22 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
23 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
24 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
32 #include "radv_debug.h"
33 #include "radv_private.h"
34 #include "radv_shader.h"
36 #include "util/disk_cache.h"
37 #include "util/strtod.h"
41 #include <amdgpu_drm.h>
42 #include "winsys/amdgpu/radv_amdgpu_winsys_public.h"
43 #include "ac_llvm_util.h"
44 #include "vk_format.h"
48 #include "util/build_id.h"
49 #include "util/debug.h"
50 #include "util/mesa-sha1.h"
53 radv_device_get_cache_uuid(enum radeon_family family
, void *uuid
)
56 unsigned char sha1
[20];
57 unsigned ptr_size
= sizeof(void*);
59 memset(uuid
, 0, VK_UUID_SIZE
);
60 _mesa_sha1_init(&ctx
);
62 if (!disk_cache_get_function_identifier(radv_device_get_cache_uuid
, &ctx
) ||
63 !disk_cache_get_function_identifier(LLVMInitializeAMDGPUTargetInfo
, &ctx
))
66 _mesa_sha1_update(&ctx
, &family
, sizeof(family
));
67 _mesa_sha1_update(&ctx
, &ptr_size
, sizeof(ptr_size
));
68 _mesa_sha1_final(&ctx
, sha1
);
70 memcpy(uuid
, sha1
, VK_UUID_SIZE
);
75 radv_get_driver_uuid(void *uuid
)
77 ac_compute_driver_uuid(uuid
, VK_UUID_SIZE
);
81 radv_get_device_uuid(struct radeon_info
*info
, void *uuid
)
83 ac_compute_device_uuid(info
, uuid
, VK_UUID_SIZE
);
87 radv_get_device_name(enum radeon_family family
, char *name
, size_t name_len
)
89 const char *chip_string
;
90 char llvm_string
[32] = {};
93 case CHIP_TAHITI
: chip_string
= "AMD RADV TAHITI"; break;
94 case CHIP_PITCAIRN
: chip_string
= "AMD RADV PITCAIRN"; break;
95 case CHIP_VERDE
: chip_string
= "AMD RADV CAPE VERDE"; break;
96 case CHIP_OLAND
: chip_string
= "AMD RADV OLAND"; break;
97 case CHIP_HAINAN
: chip_string
= "AMD RADV HAINAN"; break;
98 case CHIP_BONAIRE
: chip_string
= "AMD RADV BONAIRE"; break;
99 case CHIP_KAVERI
: chip_string
= "AMD RADV KAVERI"; break;
100 case CHIP_KABINI
: chip_string
= "AMD RADV KABINI"; break;
101 case CHIP_HAWAII
: chip_string
= "AMD RADV HAWAII"; break;
102 case CHIP_MULLINS
: chip_string
= "AMD RADV MULLINS"; break;
103 case CHIP_TONGA
: chip_string
= "AMD RADV TONGA"; break;
104 case CHIP_ICELAND
: chip_string
= "AMD RADV ICELAND"; break;
105 case CHIP_CARRIZO
: chip_string
= "AMD RADV CARRIZO"; break;
106 case CHIP_FIJI
: chip_string
= "AMD RADV FIJI"; break;
107 case CHIP_POLARIS10
: chip_string
= "AMD RADV POLARIS10"; break;
108 case CHIP_POLARIS11
: chip_string
= "AMD RADV POLARIS11"; break;
109 case CHIP_POLARIS12
: chip_string
= "AMD RADV POLARIS12"; break;
110 case CHIP_STONEY
: chip_string
= "AMD RADV STONEY"; break;
111 case CHIP_VEGAM
: chip_string
= "AMD RADV VEGA M"; break;
112 case CHIP_VEGA10
: chip_string
= "AMD RADV VEGA10"; break;
113 case CHIP_VEGA12
: chip_string
= "AMD RADV VEGA12"; break;
114 case CHIP_RAVEN
: chip_string
= "AMD RADV RAVEN"; break;
115 case CHIP_RAVEN2
: chip_string
= "AMD RADV RAVEN2"; break;
116 default: chip_string
= "AMD RADV unknown"; break;
119 snprintf(llvm_string
, sizeof(llvm_string
),
120 " (LLVM %i.%i.%i)", (HAVE_LLVM
>> 8) & 0xff,
121 HAVE_LLVM
& 0xff, MESA_LLVM_VERSION_PATCH
);
122 snprintf(name
, name_len
, "%s%s", chip_string
, llvm_string
);
126 radv_get_visible_vram_size(struct radv_physical_device
*device
)
128 return MIN2(device
->rad_info
.vram_size
, device
->rad_info
.vram_vis_size
);
132 radv_get_vram_size(struct radv_physical_device
*device
)
134 return device
->rad_info
.vram_size
- radv_get_visible_vram_size(device
);
138 radv_physical_device_init_mem_types(struct radv_physical_device
*device
)
140 STATIC_ASSERT(RADV_MEM_HEAP_COUNT
<= VK_MAX_MEMORY_HEAPS
);
141 uint64_t visible_vram_size
= radv_get_visible_vram_size(device
);
142 uint64_t vram_size
= radv_get_vram_size(device
);
143 int vram_index
= -1, visible_vram_index
= -1, gart_index
= -1;
144 device
->memory_properties
.memoryHeapCount
= 0;
146 vram_index
= device
->memory_properties
.memoryHeapCount
++;
147 device
->memory_properties
.memoryHeaps
[vram_index
] = (VkMemoryHeap
) {
149 .flags
= VK_MEMORY_HEAP_DEVICE_LOCAL_BIT
,
152 if (visible_vram_size
) {
153 visible_vram_index
= device
->memory_properties
.memoryHeapCount
++;
154 device
->memory_properties
.memoryHeaps
[visible_vram_index
] = (VkMemoryHeap
) {
155 .size
= visible_vram_size
,
156 .flags
= VK_MEMORY_HEAP_DEVICE_LOCAL_BIT
,
159 if (device
->rad_info
.gart_size
> 0) {
160 gart_index
= device
->memory_properties
.memoryHeapCount
++;
161 device
->memory_properties
.memoryHeaps
[gart_index
] = (VkMemoryHeap
) {
162 .size
= device
->rad_info
.gart_size
,
163 .flags
= device
->rad_info
.has_dedicated_vram
? 0 : VK_MEMORY_HEAP_DEVICE_LOCAL_BIT
,
167 STATIC_ASSERT(RADV_MEM_TYPE_COUNT
<= VK_MAX_MEMORY_TYPES
);
168 unsigned type_count
= 0;
169 if (vram_index
>= 0) {
170 device
->mem_type_indices
[type_count
] = RADV_MEM_TYPE_VRAM
;
171 device
->memory_properties
.memoryTypes
[type_count
++] = (VkMemoryType
) {
172 .propertyFlags
= VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
,
173 .heapIndex
= vram_index
,
176 if (gart_index
>= 0) {
177 device
->mem_type_indices
[type_count
] = RADV_MEM_TYPE_GTT_WRITE_COMBINE
;
178 device
->memory_properties
.memoryTypes
[type_count
++] = (VkMemoryType
) {
179 .propertyFlags
= VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT
|
180 VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
|
181 (device
->rad_info
.has_dedicated_vram
? 0 : VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
),
182 .heapIndex
= gart_index
,
185 if (visible_vram_index
>= 0) {
186 device
->mem_type_indices
[type_count
] = RADV_MEM_TYPE_VRAM_CPU_ACCESS
;
187 device
->memory_properties
.memoryTypes
[type_count
++] = (VkMemoryType
) {
188 .propertyFlags
= VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
|
189 VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT
|
190 VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
,
191 .heapIndex
= visible_vram_index
,
194 if (gart_index
>= 0) {
195 device
->mem_type_indices
[type_count
] = RADV_MEM_TYPE_GTT_CACHED
;
196 device
->memory_properties
.memoryTypes
[type_count
++] = (VkMemoryType
) {
197 .propertyFlags
= VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT
|
198 VK_MEMORY_PROPERTY_HOST_COHERENT_BIT
|
199 VK_MEMORY_PROPERTY_HOST_CACHED_BIT
|
200 (device
->rad_info
.has_dedicated_vram
? 0 : VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT
),
201 .heapIndex
= gart_index
,
204 device
->memory_properties
.memoryTypeCount
= type_count
;
208 radv_handle_env_var_force_family(struct radv_physical_device
*device
)
210 const char *family
= getenv("RADV_FORCE_FAMILY");
216 for (i
= CHIP_TAHITI
; i
< CHIP_LAST
; i
++) {
217 if (!strcmp(family
, ac_get_llvm_processor_name(i
))) {
218 /* Override family and chip_class. */
219 device
->rad_info
.family
= i
;
221 if (i
>= CHIP_VEGA10
)
222 device
->rad_info
.chip_class
= GFX9
;
223 else if (i
>= CHIP_TONGA
)
224 device
->rad_info
.chip_class
= VI
;
225 else if (i
>= CHIP_BONAIRE
)
226 device
->rad_info
.chip_class
= CIK
;
228 device
->rad_info
.chip_class
= SI
;
234 fprintf(stderr
, "radv: Unknown family: %s\n", family
);
239 radv_physical_device_init(struct radv_physical_device
*device
,
240 struct radv_instance
*instance
,
241 drmDevicePtr drm_device
)
243 const char *path
= drm_device
->nodes
[DRM_NODE_RENDER
];
245 drmVersionPtr version
;
249 fd
= open(path
, O_RDWR
| O_CLOEXEC
);
251 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
252 radv_logi("Could not open device '%s'", path
);
254 return vk_error(instance
, VK_ERROR_INCOMPATIBLE_DRIVER
);
257 version
= drmGetVersion(fd
);
261 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
262 radv_logi("Could not get the kernel driver version for device '%s'", path
);
264 return vk_errorf(instance
, VK_ERROR_INCOMPATIBLE_DRIVER
,
265 "failed to get version %s: %m", path
);
268 if (strcmp(version
->name
, "amdgpu")) {
269 drmFreeVersion(version
);
272 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
273 radv_logi("Device '%s' is not using the amdgpu kernel driver.", path
);
275 return VK_ERROR_INCOMPATIBLE_DRIVER
;
277 drmFreeVersion(version
);
279 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
280 radv_logi("Found compatible device '%s'.", path
);
282 device
->_loader_data
.loaderMagic
= ICD_LOADER_MAGIC
;
283 device
->instance
= instance
;
285 device
->ws
= radv_amdgpu_winsys_create(fd
, instance
->debug_flags
,
286 instance
->perftest_flags
);
288 result
= vk_error(instance
, VK_ERROR_INCOMPATIBLE_DRIVER
);
292 if (instance
->enabled_extensions
.KHR_display
) {
293 master_fd
= open(drm_device
->nodes
[DRM_NODE_PRIMARY
], O_RDWR
| O_CLOEXEC
);
294 if (master_fd
>= 0) {
295 uint32_t accel_working
= 0;
296 struct drm_amdgpu_info request
= {
297 .return_pointer
= (uintptr_t)&accel_working
,
298 .return_size
= sizeof(accel_working
),
299 .query
= AMDGPU_INFO_ACCEL_WORKING
302 if (drmCommandWrite(master_fd
, DRM_AMDGPU_INFO
, &request
, sizeof (struct drm_amdgpu_info
)) < 0 || !accel_working
) {
309 device
->master_fd
= master_fd
;
310 device
->local_fd
= fd
;
311 device
->ws
->query_info(device
->ws
, &device
->rad_info
);
313 radv_handle_env_var_force_family(device
);
315 radv_get_device_name(device
->rad_info
.family
, device
->name
, sizeof(device
->name
));
317 if (radv_device_get_cache_uuid(device
->rad_info
.family
, device
->cache_uuid
)) {
318 device
->ws
->destroy(device
->ws
);
319 result
= vk_errorf(instance
, VK_ERROR_INITIALIZATION_FAILED
,
320 "cannot generate UUID");
324 /* These flags affect shader compilation. */
325 uint64_t shader_env_flags
=
326 (device
->instance
->perftest_flags
& RADV_PERFTEST_SISCHED
? 0x1 : 0) |
327 (device
->instance
->debug_flags
& RADV_DEBUG_UNSAFE_MATH
? 0x2 : 0);
329 /* The gpu id is already embedded in the uuid so we just pass "radv"
330 * when creating the cache.
332 char buf
[VK_UUID_SIZE
* 2 + 1];
333 disk_cache_format_hex_id(buf
, device
->cache_uuid
, VK_UUID_SIZE
* 2);
334 device
->disk_cache
= disk_cache_create(device
->name
, buf
, shader_env_flags
);
336 if (device
->rad_info
.chip_class
< VI
||
337 device
->rad_info
.chip_class
> GFX9
)
338 fprintf(stderr
, "WARNING: radv is not a conformant vulkan implementation, testing use only.\n");
340 radv_get_driver_uuid(&device
->device_uuid
);
341 radv_get_device_uuid(&device
->rad_info
, &device
->device_uuid
);
343 if (device
->rad_info
.family
== CHIP_STONEY
||
344 device
->rad_info
.chip_class
>= GFX9
) {
345 device
->has_rbplus
= true;
346 device
->rbplus_allowed
= device
->rad_info
.family
== CHIP_STONEY
||
347 device
->rad_info
.family
== CHIP_VEGA12
||
348 device
->rad_info
.family
== CHIP_RAVEN
||
349 device
->rad_info
.family
== CHIP_RAVEN2
;
352 /* The mere presence of CLEAR_STATE in the IB causes random GPU hangs
355 device
->has_clear_state
= device
->rad_info
.chip_class
>= CIK
;
357 device
->cpdma_prefetch_writes_memory
= device
->rad_info
.chip_class
<= VI
;
359 /* Vega10/Raven need a special workaround for a hardware bug. */
360 device
->has_scissor_bug
= device
->rad_info
.family
== CHIP_VEGA10
||
361 device
->rad_info
.family
== CHIP_RAVEN
;
363 /* Out-of-order primitive rasterization. */
364 device
->has_out_of_order_rast
= device
->rad_info
.chip_class
>= VI
&&
365 device
->rad_info
.max_se
>= 2;
366 device
->out_of_order_rast_allowed
= device
->has_out_of_order_rast
&&
367 !(device
->instance
->debug_flags
& RADV_DEBUG_NO_OUT_OF_ORDER
);
369 device
->dcc_msaa_allowed
=
370 (device
->instance
->perftest_flags
& RADV_PERFTEST_DCC_MSAA
);
372 radv_physical_device_init_mem_types(device
);
373 radv_fill_device_extension_table(device
, &device
->supported_extensions
);
375 device
->bus_info
= *drm_device
->businfo
.pci
;
377 if ((device
->instance
->debug_flags
& RADV_DEBUG_INFO
))
378 ac_print_gpu_info(&device
->rad_info
);
380 /* The WSI is structured as a layer on top of the driver, so this has
381 * to be the last part of initialization (at least until we get other
384 result
= radv_init_wsi(device
);
385 if (result
!= VK_SUCCESS
) {
386 device
->ws
->destroy(device
->ws
);
387 vk_error(instance
, result
);
401 radv_physical_device_finish(struct radv_physical_device
*device
)
403 radv_finish_wsi(device
);
404 device
->ws
->destroy(device
->ws
);
405 disk_cache_destroy(device
->disk_cache
);
406 close(device
->local_fd
);
407 if (device
->master_fd
!= -1)
408 close(device
->master_fd
);
412 default_alloc_func(void *pUserData
, size_t size
, size_t align
,
413 VkSystemAllocationScope allocationScope
)
419 default_realloc_func(void *pUserData
, void *pOriginal
, size_t size
,
420 size_t align
, VkSystemAllocationScope allocationScope
)
422 return realloc(pOriginal
, size
);
426 default_free_func(void *pUserData
, void *pMemory
)
431 static const VkAllocationCallbacks default_alloc
= {
433 .pfnAllocation
= default_alloc_func
,
434 .pfnReallocation
= default_realloc_func
,
435 .pfnFree
= default_free_func
,
438 static const struct debug_control radv_debug_options
[] = {
439 {"nofastclears", RADV_DEBUG_NO_FAST_CLEARS
},
440 {"nodcc", RADV_DEBUG_NO_DCC
},
441 {"shaders", RADV_DEBUG_DUMP_SHADERS
},
442 {"nocache", RADV_DEBUG_NO_CACHE
},
443 {"shaderstats", RADV_DEBUG_DUMP_SHADER_STATS
},
444 {"nohiz", RADV_DEBUG_NO_HIZ
},
445 {"nocompute", RADV_DEBUG_NO_COMPUTE_QUEUE
},
446 {"unsafemath", RADV_DEBUG_UNSAFE_MATH
},
447 {"allbos", RADV_DEBUG_ALL_BOS
},
448 {"noibs", RADV_DEBUG_NO_IBS
},
449 {"spirv", RADV_DEBUG_DUMP_SPIRV
},
450 {"vmfaults", RADV_DEBUG_VM_FAULTS
},
451 {"zerovram", RADV_DEBUG_ZERO_VRAM
},
452 {"syncshaders", RADV_DEBUG_SYNC_SHADERS
},
453 {"nosisched", RADV_DEBUG_NO_SISCHED
},
454 {"preoptir", RADV_DEBUG_PREOPTIR
},
455 {"nodynamicbounds", RADV_DEBUG_NO_DYNAMIC_BOUNDS
},
456 {"nooutoforder", RADV_DEBUG_NO_OUT_OF_ORDER
},
457 {"info", RADV_DEBUG_INFO
},
458 {"errors", RADV_DEBUG_ERRORS
},
459 {"startup", RADV_DEBUG_STARTUP
},
460 {"checkir", RADV_DEBUG_CHECKIR
},
461 {"nothreadllvm", RADV_DEBUG_NOTHREADLLVM
},
462 {"nobinning", RADV_DEBUG_NOBINNING
},
467 radv_get_debug_option_name(int id
)
469 assert(id
< ARRAY_SIZE(radv_debug_options
) - 1);
470 return radv_debug_options
[id
].string
;
473 static const struct debug_control radv_perftest_options
[] = {
474 {"nobatchchain", RADV_PERFTEST_NO_BATCHCHAIN
},
475 {"sisched", RADV_PERFTEST_SISCHED
},
476 {"localbos", RADV_PERFTEST_LOCAL_BOS
},
477 {"dccmsaa", RADV_PERFTEST_DCC_MSAA
},
482 radv_get_perftest_option_name(int id
)
484 assert(id
< ARRAY_SIZE(radv_perftest_options
) - 1);
485 return radv_perftest_options
[id
].string
;
489 radv_handle_per_app_options(struct radv_instance
*instance
,
490 const VkApplicationInfo
*info
)
492 const char *name
= info
? info
->pApplicationName
: NULL
;
497 if (!strcmp(name
, "Talos - Linux - 32bit") ||
498 !strcmp(name
, "Talos - Linux - 64bit")) {
499 if (!(instance
->debug_flags
& RADV_DEBUG_NO_SISCHED
)) {
500 /* Force enable LLVM sisched for Talos because it looks
501 * safe and it gives few more FPS.
503 instance
->perftest_flags
|= RADV_PERFTEST_SISCHED
;
505 } else if (!strcmp(name
, "DOOM_VFR")) {
506 /* Work around a Doom VFR game bug */
507 instance
->debug_flags
|= RADV_DEBUG_NO_DYNAMIC_BOUNDS
;
511 static int radv_get_instance_extension_index(const char *name
)
513 for (unsigned i
= 0; i
< RADV_INSTANCE_EXTENSION_COUNT
; ++i
) {
514 if (strcmp(name
, radv_instance_extensions
[i
].extensionName
) == 0)
521 VkResult
radv_CreateInstance(
522 const VkInstanceCreateInfo
* pCreateInfo
,
523 const VkAllocationCallbacks
* pAllocator
,
524 VkInstance
* pInstance
)
526 struct radv_instance
*instance
;
529 assert(pCreateInfo
->sType
== VK_STRUCTURE_TYPE_INSTANCE_CREATE_INFO
);
531 uint32_t client_version
;
532 if (pCreateInfo
->pApplicationInfo
&&
533 pCreateInfo
->pApplicationInfo
->apiVersion
!= 0) {
534 client_version
= pCreateInfo
->pApplicationInfo
->apiVersion
;
536 client_version
= VK_API_VERSION_1_0
;
539 instance
= vk_zalloc2(&default_alloc
, pAllocator
, sizeof(*instance
), 8,
540 VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE
);
542 return vk_error(NULL
, VK_ERROR_OUT_OF_HOST_MEMORY
);
544 instance
->_loader_data
.loaderMagic
= ICD_LOADER_MAGIC
;
547 instance
->alloc
= *pAllocator
;
549 instance
->alloc
= default_alloc
;
551 instance
->apiVersion
= client_version
;
552 instance
->physicalDeviceCount
= -1;
554 instance
->debug_flags
= parse_debug_string(getenv("RADV_DEBUG"),
557 instance
->perftest_flags
= parse_debug_string(getenv("RADV_PERFTEST"),
558 radv_perftest_options
);
561 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
562 radv_logi("Created an instance");
564 for (uint32_t i
= 0; i
< pCreateInfo
->enabledExtensionCount
; i
++) {
565 const char *ext_name
= pCreateInfo
->ppEnabledExtensionNames
[i
];
566 int index
= radv_get_instance_extension_index(ext_name
);
568 if (index
< 0 || !radv_supported_instance_extensions
.extensions
[index
]) {
569 vk_free2(&default_alloc
, pAllocator
, instance
);
570 return vk_error(instance
, VK_ERROR_EXTENSION_NOT_PRESENT
);
573 instance
->enabled_extensions
.extensions
[index
] = true;
576 result
= vk_debug_report_instance_init(&instance
->debug_report_callbacks
);
577 if (result
!= VK_SUCCESS
) {
578 vk_free2(&default_alloc
, pAllocator
, instance
);
579 return vk_error(instance
, result
);
584 VG(VALGRIND_CREATE_MEMPOOL(instance
, 0, false));
586 radv_handle_per_app_options(instance
, pCreateInfo
->pApplicationInfo
);
588 *pInstance
= radv_instance_to_handle(instance
);
593 void radv_DestroyInstance(
594 VkInstance _instance
,
595 const VkAllocationCallbacks
* pAllocator
)
597 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
602 for (int i
= 0; i
< instance
->physicalDeviceCount
; ++i
) {
603 radv_physical_device_finish(instance
->physicalDevices
+ i
);
606 VG(VALGRIND_DESTROY_MEMPOOL(instance
));
610 vk_debug_report_instance_destroy(&instance
->debug_report_callbacks
);
612 vk_free(&instance
->alloc
, instance
);
616 radv_enumerate_devices(struct radv_instance
*instance
)
618 /* TODO: Check for more devices ? */
619 drmDevicePtr devices
[8];
620 VkResult result
= VK_ERROR_INCOMPATIBLE_DRIVER
;
623 instance
->physicalDeviceCount
= 0;
625 max_devices
= drmGetDevices2(0, devices
, ARRAY_SIZE(devices
));
627 if (instance
->debug_flags
& RADV_DEBUG_STARTUP
)
628 radv_logi("Found %d drm nodes", max_devices
);
631 return vk_error(instance
, VK_ERROR_INCOMPATIBLE_DRIVER
);
633 for (unsigned i
= 0; i
< (unsigned)max_devices
; i
++) {
634 if (devices
[i
]->available_nodes
& 1 << DRM_NODE_RENDER
&&
635 devices
[i
]->bustype
== DRM_BUS_PCI
&&
636 devices
[i
]->deviceinfo
.pci
->vendor_id
== ATI_VENDOR_ID
) {
638 result
= radv_physical_device_init(instance
->physicalDevices
+
639 instance
->physicalDeviceCount
,
642 if (result
== VK_SUCCESS
)
643 ++instance
->physicalDeviceCount
;
644 else if (result
!= VK_ERROR_INCOMPATIBLE_DRIVER
)
648 drmFreeDevices(devices
, max_devices
);
653 VkResult
radv_EnumeratePhysicalDevices(
654 VkInstance _instance
,
655 uint32_t* pPhysicalDeviceCount
,
656 VkPhysicalDevice
* pPhysicalDevices
)
658 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
661 if (instance
->physicalDeviceCount
< 0) {
662 result
= radv_enumerate_devices(instance
);
663 if (result
!= VK_SUCCESS
&&
664 result
!= VK_ERROR_INCOMPATIBLE_DRIVER
)
668 if (!pPhysicalDevices
) {
669 *pPhysicalDeviceCount
= instance
->physicalDeviceCount
;
671 *pPhysicalDeviceCount
= MIN2(*pPhysicalDeviceCount
, instance
->physicalDeviceCount
);
672 for (unsigned i
= 0; i
< *pPhysicalDeviceCount
; ++i
)
673 pPhysicalDevices
[i
] = radv_physical_device_to_handle(instance
->physicalDevices
+ i
);
676 return *pPhysicalDeviceCount
< instance
->physicalDeviceCount
? VK_INCOMPLETE
680 VkResult
radv_EnumeratePhysicalDeviceGroups(
681 VkInstance _instance
,
682 uint32_t* pPhysicalDeviceGroupCount
,
683 VkPhysicalDeviceGroupProperties
* pPhysicalDeviceGroupProperties
)
685 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
688 if (instance
->physicalDeviceCount
< 0) {
689 result
= radv_enumerate_devices(instance
);
690 if (result
!= VK_SUCCESS
&&
691 result
!= VK_ERROR_INCOMPATIBLE_DRIVER
)
695 if (!pPhysicalDeviceGroupProperties
) {
696 *pPhysicalDeviceGroupCount
= instance
->physicalDeviceCount
;
698 *pPhysicalDeviceGroupCount
= MIN2(*pPhysicalDeviceGroupCount
, instance
->physicalDeviceCount
);
699 for (unsigned i
= 0; i
< *pPhysicalDeviceGroupCount
; ++i
) {
700 pPhysicalDeviceGroupProperties
[i
].physicalDeviceCount
= 1;
701 pPhysicalDeviceGroupProperties
[i
].physicalDevices
[0] = radv_physical_device_to_handle(instance
->physicalDevices
+ i
);
702 pPhysicalDeviceGroupProperties
[i
].subsetAllocation
= false;
705 return *pPhysicalDeviceGroupCount
< instance
->physicalDeviceCount
? VK_INCOMPLETE
709 void radv_GetPhysicalDeviceFeatures(
710 VkPhysicalDevice physicalDevice
,
711 VkPhysicalDeviceFeatures
* pFeatures
)
713 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
714 memset(pFeatures
, 0, sizeof(*pFeatures
));
716 *pFeatures
= (VkPhysicalDeviceFeatures
) {
717 .robustBufferAccess
= true,
718 .fullDrawIndexUint32
= true,
719 .imageCubeArray
= true,
720 .independentBlend
= true,
721 .geometryShader
= true,
722 .tessellationShader
= true,
723 .sampleRateShading
= true,
724 .dualSrcBlend
= true,
726 .multiDrawIndirect
= true,
727 .drawIndirectFirstInstance
= true,
729 .depthBiasClamp
= true,
730 .fillModeNonSolid
= true,
735 .multiViewport
= true,
736 .samplerAnisotropy
= true,
737 .textureCompressionETC2
= pdevice
->rad_info
.chip_class
>= GFX9
||
738 pdevice
->rad_info
.family
== CHIP_STONEY
,
739 .textureCompressionASTC_LDR
= false,
740 .textureCompressionBC
= true,
741 .occlusionQueryPrecise
= true,
742 .pipelineStatisticsQuery
= true,
743 .vertexPipelineStoresAndAtomics
= true,
744 .fragmentStoresAndAtomics
= true,
745 .shaderTessellationAndGeometryPointSize
= true,
746 .shaderImageGatherExtended
= true,
747 .shaderStorageImageExtendedFormats
= true,
748 .shaderStorageImageMultisample
= pdevice
->rad_info
.chip_class
>= VI
,
749 .shaderUniformBufferArrayDynamicIndexing
= true,
750 .shaderSampledImageArrayDynamicIndexing
= true,
751 .shaderStorageBufferArrayDynamicIndexing
= true,
752 .shaderStorageImageArrayDynamicIndexing
= true,
753 .shaderStorageImageReadWithoutFormat
= true,
754 .shaderStorageImageWriteWithoutFormat
= true,
755 .shaderClipDistance
= true,
756 .shaderCullDistance
= true,
757 .shaderFloat64
= true,
759 .shaderInt16
= pdevice
->rad_info
.chip_class
>= GFX9
,
760 .sparseBinding
= true,
761 .variableMultisampleRate
= true,
762 .inheritedQueries
= true,
766 void radv_GetPhysicalDeviceFeatures2(
767 VkPhysicalDevice physicalDevice
,
768 VkPhysicalDeviceFeatures2
*pFeatures
)
770 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
771 vk_foreach_struct(ext
, pFeatures
->pNext
) {
772 switch (ext
->sType
) {
773 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTER_FEATURES
: {
774 VkPhysicalDeviceVariablePointerFeatures
*features
= (void *)ext
;
775 features
->variablePointersStorageBuffer
= true;
776 features
->variablePointers
= true;
779 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES
: {
780 VkPhysicalDeviceMultiviewFeatures
*features
= (VkPhysicalDeviceMultiviewFeatures
*)ext
;
781 features
->multiview
= true;
782 features
->multiviewGeometryShader
= true;
783 features
->multiviewTessellationShader
= true;
786 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETER_FEATURES
: {
787 VkPhysicalDeviceShaderDrawParameterFeatures
*features
=
788 (VkPhysicalDeviceShaderDrawParameterFeatures
*)ext
;
789 features
->shaderDrawParameters
= true;
792 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROTECTED_MEMORY_FEATURES
: {
793 VkPhysicalDeviceProtectedMemoryFeatures
*features
=
794 (VkPhysicalDeviceProtectedMemoryFeatures
*)ext
;
795 features
->protectedMemory
= false;
798 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES
: {
799 VkPhysicalDevice16BitStorageFeatures
*features
=
800 (VkPhysicalDevice16BitStorageFeatures
*)ext
;
801 bool enabled
= pdevice
->rad_info
.chip_class
>= VI
;
802 features
->storageBuffer16BitAccess
= enabled
;
803 features
->uniformAndStorageBuffer16BitAccess
= enabled
;
804 features
->storagePushConstant16
= enabled
;
805 features
->storageInputOutput16
= enabled
;
808 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_YCBCR_CONVERSION_FEATURES
: {
809 VkPhysicalDeviceSamplerYcbcrConversionFeatures
*features
=
810 (VkPhysicalDeviceSamplerYcbcrConversionFeatures
*)ext
;
811 features
->samplerYcbcrConversion
= false;
814 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_FEATURES_EXT
: {
815 VkPhysicalDeviceDescriptorIndexingFeaturesEXT
*features
=
816 (VkPhysicalDeviceDescriptorIndexingFeaturesEXT
*)ext
;
817 features
->shaderInputAttachmentArrayDynamicIndexing
= true;
818 features
->shaderUniformTexelBufferArrayDynamicIndexing
= true;
819 features
->shaderStorageTexelBufferArrayDynamicIndexing
= true;
820 features
->shaderUniformBufferArrayNonUniformIndexing
= false;
821 features
->shaderSampledImageArrayNonUniformIndexing
= false;
822 features
->shaderStorageBufferArrayNonUniformIndexing
= false;
823 features
->shaderStorageImageArrayNonUniformIndexing
= false;
824 features
->shaderInputAttachmentArrayNonUniformIndexing
= false;
825 features
->shaderUniformTexelBufferArrayNonUniformIndexing
= false;
826 features
->shaderStorageTexelBufferArrayNonUniformIndexing
= false;
827 features
->descriptorBindingUniformBufferUpdateAfterBind
= true;
828 features
->descriptorBindingSampledImageUpdateAfterBind
= true;
829 features
->descriptorBindingStorageImageUpdateAfterBind
= true;
830 features
->descriptorBindingStorageBufferUpdateAfterBind
= true;
831 features
->descriptorBindingUniformTexelBufferUpdateAfterBind
= true;
832 features
->descriptorBindingStorageTexelBufferUpdateAfterBind
= true;
833 features
->descriptorBindingUpdateUnusedWhilePending
= true;
834 features
->descriptorBindingPartiallyBound
= true;
835 features
->descriptorBindingVariableDescriptorCount
= true;
836 features
->runtimeDescriptorArray
= true;
839 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONDITIONAL_RENDERING_FEATURES_EXT
: {
840 VkPhysicalDeviceConditionalRenderingFeaturesEXT
*features
=
841 (VkPhysicalDeviceConditionalRenderingFeaturesEXT
*)ext
;
842 features
->conditionalRendering
= true;
843 features
->inheritedConditionalRendering
= false;
846 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_FEATURES_EXT
: {
847 VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT
*features
=
848 (VkPhysicalDeviceVertexAttributeDivisorFeaturesEXT
*)ext
;
849 features
->vertexAttributeInstanceRateDivisor
= VK_TRUE
;
850 features
->vertexAttributeInstanceRateZeroDivisor
= VK_TRUE
;
853 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT
: {
854 VkPhysicalDeviceTransformFeedbackFeaturesEXT
*features
=
855 (VkPhysicalDeviceTransformFeedbackFeaturesEXT
*)ext
;
856 features
->transformFeedback
= true;
857 features
->geometryStreams
= true;
860 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SCALAR_BLOCK_LAYOUT_FEATURES_EXT
: {
861 VkPhysicalDeviceScalarBlockLayoutFeaturesEXT
*features
=
862 (VkPhysicalDeviceScalarBlockLayoutFeaturesEXT
*)ext
;
863 features
->scalarBlockLayout
= pdevice
->rad_info
.chip_class
>= CIK
;
866 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MEMORY_PRIORITY_FEATURES_EXT
: {
867 VkPhysicalDeviceMemoryPriorityFeaturesEXT
*features
=
868 (VkPhysicalDeviceMemoryPriorityFeaturesEXT
*)ext
;
869 features
->memoryPriority
= VK_TRUE
;
872 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_BUFFER_ADDRESS_FEATURES_EXT
: {
873 VkPhysicalDeviceBufferAddressFeaturesEXT
*features
=
874 (VkPhysicalDeviceBufferAddressFeaturesEXT
*)ext
;
875 features
->bufferDeviceAddress
= true;
876 features
->bufferDeviceAddressCaptureReplay
= false;
877 features
->bufferDeviceAddressMultiDevice
= false;
884 return radv_GetPhysicalDeviceFeatures(physicalDevice
, &pFeatures
->features
);
887 void radv_GetPhysicalDeviceProperties(
888 VkPhysicalDevice physicalDevice
,
889 VkPhysicalDeviceProperties
* pProperties
)
891 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
892 VkSampleCountFlags sample_counts
= 0xf;
894 /* make sure that the entire descriptor set is addressable with a signed
895 * 32-bit int. So the sum of all limits scaled by descriptor size has to
896 * be at most 2 GiB. the combined image & samples object count as one of
897 * both. This limit is for the pipeline layout, not for the set layout, but
898 * there is no set limit, so we just set a pipeline limit. I don't think
899 * any app is going to hit this soon. */
900 size_t max_descriptor_set_size
= ((1ull << 31) - 16 * MAX_DYNAMIC_BUFFERS
) /
901 (32 /* uniform buffer, 32 due to potential space wasted on alignment */ +
902 32 /* storage buffer, 32 due to potential space wasted on alignment */ +
903 32 /* sampler, largest when combined with image */ +
904 64 /* sampled image */ +
905 64 /* storage image */);
907 VkPhysicalDeviceLimits limits
= {
908 .maxImageDimension1D
= (1 << 14),
909 .maxImageDimension2D
= (1 << 14),
910 .maxImageDimension3D
= (1 << 11),
911 .maxImageDimensionCube
= (1 << 14),
912 .maxImageArrayLayers
= (1 << 11),
913 .maxTexelBufferElements
= 128 * 1024 * 1024,
914 .maxUniformBufferRange
= UINT32_MAX
,
915 .maxStorageBufferRange
= UINT32_MAX
,
916 .maxPushConstantsSize
= MAX_PUSH_CONSTANTS_SIZE
,
917 .maxMemoryAllocationCount
= UINT32_MAX
,
918 .maxSamplerAllocationCount
= 64 * 1024,
919 .bufferImageGranularity
= 64, /* A cache line */
920 .sparseAddressSpaceSize
= 0xffffffffu
, /* buffer max size */
921 .maxBoundDescriptorSets
= MAX_SETS
,
922 .maxPerStageDescriptorSamplers
= max_descriptor_set_size
,
923 .maxPerStageDescriptorUniformBuffers
= max_descriptor_set_size
,
924 .maxPerStageDescriptorStorageBuffers
= max_descriptor_set_size
,
925 .maxPerStageDescriptorSampledImages
= max_descriptor_set_size
,
926 .maxPerStageDescriptorStorageImages
= max_descriptor_set_size
,
927 .maxPerStageDescriptorInputAttachments
= max_descriptor_set_size
,
928 .maxPerStageResources
= max_descriptor_set_size
,
929 .maxDescriptorSetSamplers
= max_descriptor_set_size
,
930 .maxDescriptorSetUniformBuffers
= max_descriptor_set_size
,
931 .maxDescriptorSetUniformBuffersDynamic
= MAX_DYNAMIC_UNIFORM_BUFFERS
,
932 .maxDescriptorSetStorageBuffers
= max_descriptor_set_size
,
933 .maxDescriptorSetStorageBuffersDynamic
= MAX_DYNAMIC_STORAGE_BUFFERS
,
934 .maxDescriptorSetSampledImages
= max_descriptor_set_size
,
935 .maxDescriptorSetStorageImages
= max_descriptor_set_size
,
936 .maxDescriptorSetInputAttachments
= max_descriptor_set_size
,
937 .maxVertexInputAttributes
= 32,
938 .maxVertexInputBindings
= 32,
939 .maxVertexInputAttributeOffset
= 2047,
940 .maxVertexInputBindingStride
= 2048,
941 .maxVertexOutputComponents
= 128,
942 .maxTessellationGenerationLevel
= 64,
943 .maxTessellationPatchSize
= 32,
944 .maxTessellationControlPerVertexInputComponents
= 128,
945 .maxTessellationControlPerVertexOutputComponents
= 128,
946 .maxTessellationControlPerPatchOutputComponents
= 120,
947 .maxTessellationControlTotalOutputComponents
= 4096,
948 .maxTessellationEvaluationInputComponents
= 128,
949 .maxTessellationEvaluationOutputComponents
= 128,
950 .maxGeometryShaderInvocations
= 127,
951 .maxGeometryInputComponents
= 64,
952 .maxGeometryOutputComponents
= 128,
953 .maxGeometryOutputVertices
= 256,
954 .maxGeometryTotalOutputComponents
= 1024,
955 .maxFragmentInputComponents
= 128,
956 .maxFragmentOutputAttachments
= 8,
957 .maxFragmentDualSrcAttachments
= 1,
958 .maxFragmentCombinedOutputResources
= 8,
959 .maxComputeSharedMemorySize
= 32768,
960 .maxComputeWorkGroupCount
= { 65535, 65535, 65535 },
961 .maxComputeWorkGroupInvocations
= 2048,
962 .maxComputeWorkGroupSize
= {
967 .subPixelPrecisionBits
= 8,
968 .subTexelPrecisionBits
= 8,
969 .mipmapPrecisionBits
= 8,
970 .maxDrawIndexedIndexValue
= UINT32_MAX
,
971 .maxDrawIndirectCount
= UINT32_MAX
,
972 .maxSamplerLodBias
= 16,
973 .maxSamplerAnisotropy
= 16,
974 .maxViewports
= MAX_VIEWPORTS
,
975 .maxViewportDimensions
= { (1 << 14), (1 << 14) },
976 .viewportBoundsRange
= { INT16_MIN
, INT16_MAX
},
977 .viewportSubPixelBits
= 8,
978 .minMemoryMapAlignment
= 4096, /* A page */
979 .minTexelBufferOffsetAlignment
= 1,
980 .minUniformBufferOffsetAlignment
= 4,
981 .minStorageBufferOffsetAlignment
= 4,
982 .minTexelOffset
= -32,
983 .maxTexelOffset
= 31,
984 .minTexelGatherOffset
= -32,
985 .maxTexelGatherOffset
= 31,
986 .minInterpolationOffset
= -2,
987 .maxInterpolationOffset
= 2,
988 .subPixelInterpolationOffsetBits
= 8,
989 .maxFramebufferWidth
= (1 << 14),
990 .maxFramebufferHeight
= (1 << 14),
991 .maxFramebufferLayers
= (1 << 10),
992 .framebufferColorSampleCounts
= sample_counts
,
993 .framebufferDepthSampleCounts
= sample_counts
,
994 .framebufferStencilSampleCounts
= sample_counts
,
995 .framebufferNoAttachmentsSampleCounts
= sample_counts
,
996 .maxColorAttachments
= MAX_RTS
,
997 .sampledImageColorSampleCounts
= sample_counts
,
998 .sampledImageIntegerSampleCounts
= VK_SAMPLE_COUNT_1_BIT
,
999 .sampledImageDepthSampleCounts
= sample_counts
,
1000 .sampledImageStencilSampleCounts
= sample_counts
,
1001 .storageImageSampleCounts
= pdevice
->rad_info
.chip_class
>= VI
? sample_counts
: VK_SAMPLE_COUNT_1_BIT
,
1002 .maxSampleMaskWords
= 1,
1003 .timestampComputeAndGraphics
= true,
1004 .timestampPeriod
= 1000000.0 / pdevice
->rad_info
.clock_crystal_freq
,
1005 .maxClipDistances
= 8,
1006 .maxCullDistances
= 8,
1007 .maxCombinedClipAndCullDistances
= 8,
1008 .discreteQueuePriorities
= 2,
1009 .pointSizeRange
= { 0.125, 255.875 },
1010 .lineWidthRange
= { 0.0, 7.9921875 },
1011 .pointSizeGranularity
= (1.0 / 8.0),
1012 .lineWidthGranularity
= (1.0 / 128.0),
1013 .strictLines
= false, /* FINISHME */
1014 .standardSampleLocations
= true,
1015 .optimalBufferCopyOffsetAlignment
= 128,
1016 .optimalBufferCopyRowPitchAlignment
= 128,
1017 .nonCoherentAtomSize
= 64,
1020 *pProperties
= (VkPhysicalDeviceProperties
) {
1021 .apiVersion
= radv_physical_device_api_version(pdevice
),
1022 .driverVersion
= vk_get_driver_version(),
1023 .vendorID
= ATI_VENDOR_ID
,
1024 .deviceID
= pdevice
->rad_info
.pci_id
,
1025 .deviceType
= pdevice
->rad_info
.has_dedicated_vram
? VK_PHYSICAL_DEVICE_TYPE_DISCRETE_GPU
: VK_PHYSICAL_DEVICE_TYPE_INTEGRATED_GPU
,
1027 .sparseProperties
= {0},
1030 strcpy(pProperties
->deviceName
, pdevice
->name
);
1031 memcpy(pProperties
->pipelineCacheUUID
, pdevice
->cache_uuid
, VK_UUID_SIZE
);
1034 void radv_GetPhysicalDeviceProperties2(
1035 VkPhysicalDevice physicalDevice
,
1036 VkPhysicalDeviceProperties2
*pProperties
)
1038 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
1039 radv_GetPhysicalDeviceProperties(physicalDevice
, &pProperties
->properties
);
1041 vk_foreach_struct(ext
, pProperties
->pNext
) {
1042 switch (ext
->sType
) {
1043 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR
: {
1044 VkPhysicalDevicePushDescriptorPropertiesKHR
*properties
=
1045 (VkPhysicalDevicePushDescriptorPropertiesKHR
*) ext
;
1046 properties
->maxPushDescriptors
= MAX_PUSH_DESCRIPTORS
;
1049 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ID_PROPERTIES
: {
1050 VkPhysicalDeviceIDProperties
*properties
= (VkPhysicalDeviceIDProperties
*)ext
;
1051 memcpy(properties
->driverUUID
, pdevice
->driver_uuid
, VK_UUID_SIZE
);
1052 memcpy(properties
->deviceUUID
, pdevice
->device_uuid
, VK_UUID_SIZE
);
1053 properties
->deviceLUIDValid
= false;
1056 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_PROPERTIES
: {
1057 VkPhysicalDeviceMultiviewProperties
*properties
= (VkPhysicalDeviceMultiviewProperties
*)ext
;
1058 properties
->maxMultiviewViewCount
= MAX_VIEWS
;
1059 properties
->maxMultiviewInstanceIndex
= INT_MAX
;
1062 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_POINT_CLIPPING_PROPERTIES
: {
1063 VkPhysicalDevicePointClippingProperties
*properties
=
1064 (VkPhysicalDevicePointClippingProperties
*)ext
;
1065 properties
->pointClippingBehavior
= VK_POINT_CLIPPING_BEHAVIOR_ALL_CLIP_PLANES
;
1068 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DISCARD_RECTANGLE_PROPERTIES_EXT
: {
1069 VkPhysicalDeviceDiscardRectanglePropertiesEXT
*properties
=
1070 (VkPhysicalDeviceDiscardRectanglePropertiesEXT
*)ext
;
1071 properties
->maxDiscardRectangles
= MAX_DISCARD_RECTANGLES
;
1074 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_EXTERNAL_MEMORY_HOST_PROPERTIES_EXT
: {
1075 VkPhysicalDeviceExternalMemoryHostPropertiesEXT
*properties
=
1076 (VkPhysicalDeviceExternalMemoryHostPropertiesEXT
*) ext
;
1077 properties
->minImportedHostPointerAlignment
= 4096;
1080 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SUBGROUP_PROPERTIES
: {
1081 VkPhysicalDeviceSubgroupProperties
*properties
=
1082 (VkPhysicalDeviceSubgroupProperties
*)ext
;
1083 properties
->subgroupSize
= 64;
1084 properties
->supportedStages
= VK_SHADER_STAGE_ALL
;
1085 properties
->supportedOperations
=
1086 VK_SUBGROUP_FEATURE_BASIC_BIT
|
1087 VK_SUBGROUP_FEATURE_BALLOT_BIT
|
1088 VK_SUBGROUP_FEATURE_QUAD_BIT
|
1089 VK_SUBGROUP_FEATURE_VOTE_BIT
;
1090 if (pdevice
->rad_info
.chip_class
>= VI
) {
1091 properties
->supportedOperations
|=
1092 VK_SUBGROUP_FEATURE_ARITHMETIC_BIT
|
1093 VK_SUBGROUP_FEATURE_SHUFFLE_BIT
|
1094 VK_SUBGROUP_FEATURE_SHUFFLE_RELATIVE_BIT
;
1096 properties
->quadOperationsInAllStages
= true;
1099 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_3_PROPERTIES
: {
1100 VkPhysicalDeviceMaintenance3Properties
*properties
=
1101 (VkPhysicalDeviceMaintenance3Properties
*)ext
;
1102 /* Make sure everything is addressable by a signed 32-bit int, and
1103 * our largest descriptors are 96 bytes. */
1104 properties
->maxPerSetDescriptors
= (1ull << 31) / 96;
1105 /* Our buffer size fields allow only this much */
1106 properties
->maxMemoryAllocationSize
= 0xFFFFFFFFull
;
1109 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_FILTER_MINMAX_PROPERTIES_EXT
: {
1110 VkPhysicalDeviceSamplerFilterMinmaxPropertiesEXT
*properties
=
1111 (VkPhysicalDeviceSamplerFilterMinmaxPropertiesEXT
*)ext
;
1112 /* GFX6-8 only support single channel min/max filter. */
1113 properties
->filterMinmaxImageComponentMapping
= pdevice
->rad_info
.chip_class
>= GFX9
;
1114 properties
->filterMinmaxSingleComponentFormats
= true;
1117 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_CORE_PROPERTIES_AMD
: {
1118 VkPhysicalDeviceShaderCorePropertiesAMD
*properties
=
1119 (VkPhysicalDeviceShaderCorePropertiesAMD
*)ext
;
1121 /* Shader engines. */
1122 properties
->shaderEngineCount
=
1123 pdevice
->rad_info
.max_se
;
1124 properties
->shaderArraysPerEngineCount
=
1125 pdevice
->rad_info
.max_sh_per_se
;
1126 properties
->computeUnitsPerShaderArray
=
1127 pdevice
->rad_info
.num_good_cu_per_sh
;
1128 properties
->simdPerComputeUnit
= 4;
1129 properties
->wavefrontsPerSimd
=
1130 pdevice
->rad_info
.family
== CHIP_TONGA
||
1131 pdevice
->rad_info
.family
== CHIP_ICELAND
||
1132 pdevice
->rad_info
.family
== CHIP_POLARIS10
||
1133 pdevice
->rad_info
.family
== CHIP_POLARIS11
||
1134 pdevice
->rad_info
.family
== CHIP_POLARIS12
||
1135 pdevice
->rad_info
.family
== CHIP_VEGAM
? 8 : 10;
1136 properties
->wavefrontSize
= 64;
1139 properties
->sgprsPerSimd
=
1140 ac_get_num_physical_sgprs(pdevice
->rad_info
.chip_class
);
1141 properties
->minSgprAllocation
=
1142 pdevice
->rad_info
.chip_class
>= VI
? 16 : 8;
1143 properties
->maxSgprAllocation
=
1144 pdevice
->rad_info
.family
== CHIP_TONGA
||
1145 pdevice
->rad_info
.family
== CHIP_ICELAND
? 96 : 104;
1146 properties
->sgprAllocationGranularity
=
1147 pdevice
->rad_info
.chip_class
>= VI
? 16 : 8;
1150 properties
->vgprsPerSimd
= RADV_NUM_PHYSICAL_VGPRS
;
1151 properties
->minVgprAllocation
= 4;
1152 properties
->maxVgprAllocation
= 256;
1153 properties
->vgprAllocationGranularity
= 4;
1156 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VERTEX_ATTRIBUTE_DIVISOR_PROPERTIES_EXT
: {
1157 VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT
*properties
=
1158 (VkPhysicalDeviceVertexAttributeDivisorPropertiesEXT
*)ext
;
1159 properties
->maxVertexAttribDivisor
= UINT32_MAX
;
1162 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_PROPERTIES_EXT
: {
1163 VkPhysicalDeviceDescriptorIndexingPropertiesEXT
*properties
=
1164 (VkPhysicalDeviceDescriptorIndexingPropertiesEXT
*)ext
;
1165 properties
->maxUpdateAfterBindDescriptorsInAllPools
= UINT32_MAX
/ 64;
1166 properties
->shaderUniformBufferArrayNonUniformIndexingNative
= false;
1167 properties
->shaderSampledImageArrayNonUniformIndexingNative
= false;
1168 properties
->shaderStorageBufferArrayNonUniformIndexingNative
= false;
1169 properties
->shaderStorageImageArrayNonUniformIndexingNative
= false;
1170 properties
->shaderInputAttachmentArrayNonUniformIndexingNative
= false;
1171 properties
->robustBufferAccessUpdateAfterBind
= false;
1172 properties
->quadDivergentImplicitLod
= false;
1174 size_t max_descriptor_set_size
= ((1ull << 31) - 16 * MAX_DYNAMIC_BUFFERS
) /
1175 (32 /* uniform buffer, 32 due to potential space wasted on alignment */ +
1176 32 /* storage buffer, 32 due to potential space wasted on alignment */ +
1177 32 /* sampler, largest when combined with image */ +
1178 64 /* sampled image */ +
1179 64 /* storage image */);
1180 properties
->maxPerStageDescriptorUpdateAfterBindSamplers
= max_descriptor_set_size
;
1181 properties
->maxPerStageDescriptorUpdateAfterBindUniformBuffers
= max_descriptor_set_size
;
1182 properties
->maxPerStageDescriptorUpdateAfterBindStorageBuffers
= max_descriptor_set_size
;
1183 properties
->maxPerStageDescriptorUpdateAfterBindSampledImages
= max_descriptor_set_size
;
1184 properties
->maxPerStageDescriptorUpdateAfterBindStorageImages
= max_descriptor_set_size
;
1185 properties
->maxPerStageDescriptorUpdateAfterBindInputAttachments
= max_descriptor_set_size
;
1186 properties
->maxPerStageUpdateAfterBindResources
= max_descriptor_set_size
;
1187 properties
->maxDescriptorSetUpdateAfterBindSamplers
= max_descriptor_set_size
;
1188 properties
->maxDescriptorSetUpdateAfterBindUniformBuffers
= max_descriptor_set_size
;
1189 properties
->maxDescriptorSetUpdateAfterBindUniformBuffersDynamic
= MAX_DYNAMIC_UNIFORM_BUFFERS
;
1190 properties
->maxDescriptorSetUpdateAfterBindStorageBuffers
= max_descriptor_set_size
;
1191 properties
->maxDescriptorSetUpdateAfterBindStorageBuffersDynamic
= MAX_DYNAMIC_STORAGE_BUFFERS
;
1192 properties
->maxDescriptorSetUpdateAfterBindSampledImages
= max_descriptor_set_size
;
1193 properties
->maxDescriptorSetUpdateAfterBindStorageImages
= max_descriptor_set_size
;
1194 properties
->maxDescriptorSetUpdateAfterBindInputAttachments
= max_descriptor_set_size
;
1197 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROTECTED_MEMORY_PROPERTIES
: {
1198 VkPhysicalDeviceProtectedMemoryProperties
*properties
=
1199 (VkPhysicalDeviceProtectedMemoryProperties
*)ext
;
1200 properties
->protectedNoFault
= false;
1203 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONSERVATIVE_RASTERIZATION_PROPERTIES_EXT
: {
1204 VkPhysicalDeviceConservativeRasterizationPropertiesEXT
*properties
=
1205 (VkPhysicalDeviceConservativeRasterizationPropertiesEXT
*)ext
;
1206 properties
->primitiveOverestimationSize
= 0;
1207 properties
->maxExtraPrimitiveOverestimationSize
= 0;
1208 properties
->extraPrimitiveOverestimationSizeGranularity
= 0;
1209 properties
->primitiveUnderestimation
= VK_FALSE
;
1210 properties
->conservativePointAndLineRasterization
= VK_FALSE
;
1211 properties
->degenerateTrianglesRasterized
= VK_FALSE
;
1212 properties
->degenerateLinesRasterized
= VK_FALSE
;
1213 properties
->fullyCoveredFragmentShaderInputVariable
= VK_FALSE
;
1214 properties
->conservativeRasterizationPostDepthCoverage
= VK_FALSE
;
1217 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PCI_BUS_INFO_PROPERTIES_EXT
: {
1218 VkPhysicalDevicePCIBusInfoPropertiesEXT
*properties
=
1219 (VkPhysicalDevicePCIBusInfoPropertiesEXT
*)ext
;
1220 properties
->pciDomain
= pdevice
->bus_info
.domain
;
1221 properties
->pciBus
= pdevice
->bus_info
.bus
;
1222 properties
->pciDevice
= pdevice
->bus_info
.dev
;
1223 properties
->pciFunction
= pdevice
->bus_info
.func
;
1226 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DRIVER_PROPERTIES_KHR
: {
1227 VkPhysicalDeviceDriverPropertiesKHR
*driver_props
=
1228 (VkPhysicalDeviceDriverPropertiesKHR
*) ext
;
1230 driver_props
->driverID
= VK_DRIVER_ID_MESA_RADV_KHR
;
1231 memset(driver_props
->driverName
, 0, VK_MAX_DRIVER_NAME_SIZE_KHR
);
1232 strcpy(driver_props
->driverName
, "radv");
1234 memset(driver_props
->driverInfo
, 0, VK_MAX_DRIVER_INFO_SIZE_KHR
);
1235 snprintf(driver_props
->driverInfo
, VK_MAX_DRIVER_INFO_SIZE_KHR
,
1236 "Mesa " PACKAGE_VERSION MESA_GIT_SHA1
1238 (HAVE_LLVM
>> 8) & 0xff, HAVE_LLVM
& 0xff,
1239 MESA_LLVM_VERSION_PATCH
);
1241 driver_props
->conformanceVersion
= (VkConformanceVersionKHR
) {
1249 case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT
: {
1250 VkPhysicalDeviceTransformFeedbackPropertiesEXT
*properties
=
1251 (VkPhysicalDeviceTransformFeedbackPropertiesEXT
*)ext
;
1252 properties
->maxTransformFeedbackStreams
= MAX_SO_STREAMS
;
1253 properties
->maxTransformFeedbackBuffers
= MAX_SO_BUFFERS
;
1254 properties
->maxTransformFeedbackBufferSize
= UINT32_MAX
;
1255 properties
->maxTransformFeedbackStreamDataSize
= 512;
1256 properties
->maxTransformFeedbackBufferDataSize
= UINT32_MAX
;
1257 properties
->maxTransformFeedbackBufferDataStride
= 512;
1258 properties
->transformFeedbackQueries
= true;
1259 properties
->transformFeedbackStreamsLinesTriangles
= false;
1260 properties
->transformFeedbackRasterizationStreamSelect
= false;
1261 properties
->transformFeedbackDraw
= true;
1270 static void radv_get_physical_device_queue_family_properties(
1271 struct radv_physical_device
* pdevice
,
1273 VkQueueFamilyProperties
** pQueueFamilyProperties
)
1275 int num_queue_families
= 1;
1277 if (pdevice
->rad_info
.num_compute_rings
> 0 &&
1278 !(pdevice
->instance
->debug_flags
& RADV_DEBUG_NO_COMPUTE_QUEUE
))
1279 num_queue_families
++;
1281 if (pQueueFamilyProperties
== NULL
) {
1282 *pCount
= num_queue_families
;
1291 *pQueueFamilyProperties
[idx
] = (VkQueueFamilyProperties
) {
1292 .queueFlags
= VK_QUEUE_GRAPHICS_BIT
|
1293 VK_QUEUE_COMPUTE_BIT
|
1294 VK_QUEUE_TRANSFER_BIT
|
1295 VK_QUEUE_SPARSE_BINDING_BIT
,
1297 .timestampValidBits
= 64,
1298 .minImageTransferGranularity
= (VkExtent3D
) { 1, 1, 1 },
1303 if (pdevice
->rad_info
.num_compute_rings
> 0 &&
1304 !(pdevice
->instance
->debug_flags
& RADV_DEBUG_NO_COMPUTE_QUEUE
)) {
1305 if (*pCount
> idx
) {
1306 *pQueueFamilyProperties
[idx
] = (VkQueueFamilyProperties
) {
1307 .queueFlags
= VK_QUEUE_COMPUTE_BIT
|
1308 VK_QUEUE_TRANSFER_BIT
|
1309 VK_QUEUE_SPARSE_BINDING_BIT
,
1310 .queueCount
= pdevice
->rad_info
.num_compute_rings
,
1311 .timestampValidBits
= 64,
1312 .minImageTransferGranularity
= (VkExtent3D
) { 1, 1, 1 },
1320 void radv_GetPhysicalDeviceQueueFamilyProperties(
1321 VkPhysicalDevice physicalDevice
,
1323 VkQueueFamilyProperties
* pQueueFamilyProperties
)
1325 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
1326 if (!pQueueFamilyProperties
) {
1327 radv_get_physical_device_queue_family_properties(pdevice
, pCount
, NULL
);
1330 VkQueueFamilyProperties
*properties
[] = {
1331 pQueueFamilyProperties
+ 0,
1332 pQueueFamilyProperties
+ 1,
1333 pQueueFamilyProperties
+ 2,
1335 radv_get_physical_device_queue_family_properties(pdevice
, pCount
, properties
);
1336 assert(*pCount
<= 3);
1339 void radv_GetPhysicalDeviceQueueFamilyProperties2(
1340 VkPhysicalDevice physicalDevice
,
1342 VkQueueFamilyProperties2
*pQueueFamilyProperties
)
1344 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
1345 if (!pQueueFamilyProperties
) {
1346 radv_get_physical_device_queue_family_properties(pdevice
, pCount
, NULL
);
1349 VkQueueFamilyProperties
*properties
[] = {
1350 &pQueueFamilyProperties
[0].queueFamilyProperties
,
1351 &pQueueFamilyProperties
[1].queueFamilyProperties
,
1352 &pQueueFamilyProperties
[2].queueFamilyProperties
,
1354 radv_get_physical_device_queue_family_properties(pdevice
, pCount
, properties
);
1355 assert(*pCount
<= 3);
1358 void radv_GetPhysicalDeviceMemoryProperties(
1359 VkPhysicalDevice physicalDevice
,
1360 VkPhysicalDeviceMemoryProperties
*pMemoryProperties
)
1362 RADV_FROM_HANDLE(radv_physical_device
, physical_device
, physicalDevice
);
1364 *pMemoryProperties
= physical_device
->memory_properties
;
1368 radv_get_memory_budget_properties(VkPhysicalDevice physicalDevice
,
1369 VkPhysicalDeviceMemoryBudgetPropertiesEXT
*memoryBudget
)
1371 RADV_FROM_HANDLE(radv_physical_device
, device
, physicalDevice
);
1372 VkPhysicalDeviceMemoryProperties
*memory_properties
= &device
->memory_properties
;
1373 uint64_t visible_vram_size
= radv_get_visible_vram_size(device
);
1374 uint64_t vram_size
= radv_get_vram_size(device
);
1375 uint64_t gtt_size
= device
->rad_info
.gart_size
;
1376 uint64_t heap_budget
, heap_usage
;
1378 /* For all memory heaps, the computation of budget is as follow:
1379 * heap_budget = heap_size - global_heap_usage + app_heap_usage
1381 * The Vulkan spec 1.1.97 says that the budget should include any
1382 * currently allocated device memory.
1384 * Note that the application heap usages are not really accurate (eg.
1385 * in presence of shared buffers).
1388 heap_usage
= device
->ws
->query_value(device
->ws
,
1389 RADEON_ALLOCATED_VRAM
);
1391 heap_budget
= vram_size
-
1392 device
->ws
->query_value(device
->ws
, RADEON_VRAM_USAGE
) +
1395 memoryBudget
->heapBudget
[RADV_MEM_HEAP_VRAM
] = heap_budget
;
1396 memoryBudget
->heapUsage
[RADV_MEM_HEAP_VRAM
] = heap_usage
;
1399 if (visible_vram_size
) {
1400 heap_usage
= device
->ws
->query_value(device
->ws
,
1401 RADEON_ALLOCATED_VRAM_VIS
);
1403 heap_budget
= visible_vram_size
-
1404 device
->ws
->query_value(device
->ws
, RADEON_VRAM_VIS_USAGE
) +
1407 memoryBudget
->heapBudget
[RADV_MEM_HEAP_VRAM_CPU_ACCESS
] = heap_budget
;
1408 memoryBudget
->heapUsage
[RADV_MEM_HEAP_VRAM_CPU_ACCESS
] = heap_usage
;
1412 heap_usage
= device
->ws
->query_value(device
->ws
,
1413 RADEON_ALLOCATED_GTT
);
1415 heap_budget
= gtt_size
-
1416 device
->ws
->query_value(device
->ws
, RADEON_GTT_USAGE
) +
1419 memoryBudget
->heapBudget
[RADV_MEM_HEAP_GTT
] = heap_budget
;
1420 memoryBudget
->heapUsage
[RADV_MEM_HEAP_GTT
] = heap_usage
;
1423 /* The heapBudget and heapUsage values must be zero for array elements
1424 * greater than or equal to
1425 * VkPhysicalDeviceMemoryProperties::memoryHeapCount.
1427 for (uint32_t i
= memory_properties
->memoryHeapCount
; i
< VK_MAX_MEMORY_HEAPS
; i
++) {
1428 memoryBudget
->heapBudget
[i
] = 0;
1429 memoryBudget
->heapUsage
[i
] = 0;
1433 void radv_GetPhysicalDeviceMemoryProperties2(
1434 VkPhysicalDevice physicalDevice
,
1435 VkPhysicalDeviceMemoryProperties2
*pMemoryProperties
)
1437 radv_GetPhysicalDeviceMemoryProperties(physicalDevice
,
1438 &pMemoryProperties
->memoryProperties
);
1440 VkPhysicalDeviceMemoryBudgetPropertiesEXT
*memory_budget
=
1441 vk_find_struct(pMemoryProperties
->pNext
,
1442 PHYSICAL_DEVICE_MEMORY_BUDGET_PROPERTIES_EXT
);
1444 radv_get_memory_budget_properties(physicalDevice
, memory_budget
);
1447 VkResult
radv_GetMemoryHostPointerPropertiesEXT(
1449 VkExternalMemoryHandleTypeFlagBits handleType
,
1450 const void *pHostPointer
,
1451 VkMemoryHostPointerPropertiesEXT
*pMemoryHostPointerProperties
)
1453 RADV_FROM_HANDLE(radv_device
, device
, _device
);
1457 case VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT
: {
1458 const struct radv_physical_device
*physical_device
= device
->physical_device
;
1459 uint32_t memoryTypeBits
= 0;
1460 for (int i
= 0; i
< physical_device
->memory_properties
.memoryTypeCount
; i
++) {
1461 if (physical_device
->mem_type_indices
[i
] == RADV_MEM_TYPE_GTT_CACHED
) {
1462 memoryTypeBits
= (1 << i
);
1466 pMemoryHostPointerProperties
->memoryTypeBits
= memoryTypeBits
;
1470 return VK_ERROR_INVALID_EXTERNAL_HANDLE
;
1474 static enum radeon_ctx_priority
1475 radv_get_queue_global_priority(const VkDeviceQueueGlobalPriorityCreateInfoEXT
*pObj
)
1477 /* Default to MEDIUM when a specific global priority isn't requested */
1479 return RADEON_CTX_PRIORITY_MEDIUM
;
1481 switch(pObj
->globalPriority
) {
1482 case VK_QUEUE_GLOBAL_PRIORITY_REALTIME_EXT
:
1483 return RADEON_CTX_PRIORITY_REALTIME
;
1484 case VK_QUEUE_GLOBAL_PRIORITY_HIGH_EXT
:
1485 return RADEON_CTX_PRIORITY_HIGH
;
1486 case VK_QUEUE_GLOBAL_PRIORITY_MEDIUM_EXT
:
1487 return RADEON_CTX_PRIORITY_MEDIUM
;
1488 case VK_QUEUE_GLOBAL_PRIORITY_LOW_EXT
:
1489 return RADEON_CTX_PRIORITY_LOW
;
1491 unreachable("Illegal global priority value");
1492 return RADEON_CTX_PRIORITY_INVALID
;
1497 radv_queue_init(struct radv_device
*device
, struct radv_queue
*queue
,
1498 uint32_t queue_family_index
, int idx
,
1499 VkDeviceQueueCreateFlags flags
,
1500 const VkDeviceQueueGlobalPriorityCreateInfoEXT
*global_priority
)
1502 queue
->_loader_data
.loaderMagic
= ICD_LOADER_MAGIC
;
1503 queue
->device
= device
;
1504 queue
->queue_family_index
= queue_family_index
;
1505 queue
->queue_idx
= idx
;
1506 queue
->priority
= radv_get_queue_global_priority(global_priority
);
1507 queue
->flags
= flags
;
1509 queue
->hw_ctx
= device
->ws
->ctx_create(device
->ws
, queue
->priority
);
1511 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
1517 radv_queue_finish(struct radv_queue
*queue
)
1520 queue
->device
->ws
->ctx_destroy(queue
->hw_ctx
);
1522 if (queue
->initial_full_flush_preamble_cs
)
1523 queue
->device
->ws
->cs_destroy(queue
->initial_full_flush_preamble_cs
);
1524 if (queue
->initial_preamble_cs
)
1525 queue
->device
->ws
->cs_destroy(queue
->initial_preamble_cs
);
1526 if (queue
->continue_preamble_cs
)
1527 queue
->device
->ws
->cs_destroy(queue
->continue_preamble_cs
);
1528 if (queue
->descriptor_bo
)
1529 queue
->device
->ws
->buffer_destroy(queue
->descriptor_bo
);
1530 if (queue
->scratch_bo
)
1531 queue
->device
->ws
->buffer_destroy(queue
->scratch_bo
);
1532 if (queue
->esgs_ring_bo
)
1533 queue
->device
->ws
->buffer_destroy(queue
->esgs_ring_bo
);
1534 if (queue
->gsvs_ring_bo
)
1535 queue
->device
->ws
->buffer_destroy(queue
->gsvs_ring_bo
);
1536 if (queue
->tess_rings_bo
)
1537 queue
->device
->ws
->buffer_destroy(queue
->tess_rings_bo
);
1538 if (queue
->compute_scratch_bo
)
1539 queue
->device
->ws
->buffer_destroy(queue
->compute_scratch_bo
);
1543 radv_bo_list_init(struct radv_bo_list
*bo_list
)
1545 pthread_mutex_init(&bo_list
->mutex
, NULL
);
1546 bo_list
->list
.count
= bo_list
->capacity
= 0;
1547 bo_list
->list
.bos
= NULL
;
1551 radv_bo_list_finish(struct radv_bo_list
*bo_list
)
1553 free(bo_list
->list
.bos
);
1554 pthread_mutex_destroy(&bo_list
->mutex
);
1557 static VkResult
radv_bo_list_add(struct radv_device
*device
,
1558 struct radeon_winsys_bo
*bo
)
1560 struct radv_bo_list
*bo_list
= &device
->bo_list
;
1565 if (unlikely(!device
->use_global_bo_list
))
1568 pthread_mutex_lock(&bo_list
->mutex
);
1569 if (bo_list
->list
.count
== bo_list
->capacity
) {
1570 unsigned capacity
= MAX2(4, bo_list
->capacity
* 2);
1571 void *data
= realloc(bo_list
->list
.bos
, capacity
* sizeof(struct radeon_winsys_bo
*));
1574 pthread_mutex_unlock(&bo_list
->mutex
);
1575 return VK_ERROR_OUT_OF_HOST_MEMORY
;
1578 bo_list
->list
.bos
= (struct radeon_winsys_bo
**)data
;
1579 bo_list
->capacity
= capacity
;
1582 bo_list
->list
.bos
[bo_list
->list
.count
++] = bo
;
1583 pthread_mutex_unlock(&bo_list
->mutex
);
1587 static void radv_bo_list_remove(struct radv_device
*device
,
1588 struct radeon_winsys_bo
*bo
)
1590 struct radv_bo_list
*bo_list
= &device
->bo_list
;
1595 if (unlikely(!device
->use_global_bo_list
))
1598 pthread_mutex_lock(&bo_list
->mutex
);
1599 for(unsigned i
= 0; i
< bo_list
->list
.count
; ++i
) {
1600 if (bo_list
->list
.bos
[i
] == bo
) {
1601 bo_list
->list
.bos
[i
] = bo_list
->list
.bos
[bo_list
->list
.count
- 1];
1602 --bo_list
->list
.count
;
1606 pthread_mutex_unlock(&bo_list
->mutex
);
1610 radv_device_init_gs_info(struct radv_device
*device
)
1612 device
->gs_table_depth
= ac_get_gs_table_depth(device
->physical_device
->rad_info
.chip_class
,
1613 device
->physical_device
->rad_info
.family
);
1616 static int radv_get_device_extension_index(const char *name
)
1618 for (unsigned i
= 0; i
< RADV_DEVICE_EXTENSION_COUNT
; ++i
) {
1619 if (strcmp(name
, radv_device_extensions
[i
].extensionName
) == 0)
1626 radv_get_int_debug_option(const char *name
, int default_value
)
1633 result
= default_value
;
1637 result
= strtol(str
, &endptr
, 0);
1638 if (str
== endptr
) {
1639 /* No digits founs. */
1640 result
= default_value
;
1647 VkResult
radv_CreateDevice(
1648 VkPhysicalDevice physicalDevice
,
1649 const VkDeviceCreateInfo
* pCreateInfo
,
1650 const VkAllocationCallbacks
* pAllocator
,
1653 RADV_FROM_HANDLE(radv_physical_device
, physical_device
, physicalDevice
);
1655 struct radv_device
*device
;
1657 bool keep_shader_info
= false;
1659 /* Check enabled features */
1660 if (pCreateInfo
->pEnabledFeatures
) {
1661 VkPhysicalDeviceFeatures supported_features
;
1662 radv_GetPhysicalDeviceFeatures(physicalDevice
, &supported_features
);
1663 VkBool32
*supported_feature
= (VkBool32
*)&supported_features
;
1664 VkBool32
*enabled_feature
= (VkBool32
*)pCreateInfo
->pEnabledFeatures
;
1665 unsigned num_features
= sizeof(VkPhysicalDeviceFeatures
) / sizeof(VkBool32
);
1666 for (uint32_t i
= 0; i
< num_features
; i
++) {
1667 if (enabled_feature
[i
] && !supported_feature
[i
])
1668 return vk_error(physical_device
->instance
, VK_ERROR_FEATURE_NOT_PRESENT
);
1672 device
= vk_zalloc2(&physical_device
->instance
->alloc
, pAllocator
,
1674 VK_SYSTEM_ALLOCATION_SCOPE_DEVICE
);
1676 return vk_error(physical_device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
1678 device
->_loader_data
.loaderMagic
= ICD_LOADER_MAGIC
;
1679 device
->instance
= physical_device
->instance
;
1680 device
->physical_device
= physical_device
;
1682 device
->ws
= physical_device
->ws
;
1684 device
->alloc
= *pAllocator
;
1686 device
->alloc
= physical_device
->instance
->alloc
;
1688 for (uint32_t i
= 0; i
< pCreateInfo
->enabledExtensionCount
; i
++) {
1689 const char *ext_name
= pCreateInfo
->ppEnabledExtensionNames
[i
];
1690 int index
= radv_get_device_extension_index(ext_name
);
1691 if (index
< 0 || !physical_device
->supported_extensions
.extensions
[index
]) {
1692 vk_free(&device
->alloc
, device
);
1693 return vk_error(physical_device
->instance
, VK_ERROR_EXTENSION_NOT_PRESENT
);
1696 device
->enabled_extensions
.extensions
[index
] = true;
1699 keep_shader_info
= device
->enabled_extensions
.AMD_shader_info
;
1701 /* With update after bind we can't attach bo's to the command buffer
1702 * from the descriptor set anymore, so we have to use a global BO list.
1704 device
->use_global_bo_list
=
1705 device
->enabled_extensions
.EXT_descriptor_indexing
||
1706 device
->enabled_extensions
.EXT_buffer_device_address
;
1708 mtx_init(&device
->shader_slab_mutex
, mtx_plain
);
1709 list_inithead(&device
->shader_slabs
);
1711 radv_bo_list_init(&device
->bo_list
);
1713 for (unsigned i
= 0; i
< pCreateInfo
->queueCreateInfoCount
; i
++) {
1714 const VkDeviceQueueCreateInfo
*queue_create
= &pCreateInfo
->pQueueCreateInfos
[i
];
1715 uint32_t qfi
= queue_create
->queueFamilyIndex
;
1716 const VkDeviceQueueGlobalPriorityCreateInfoEXT
*global_priority
=
1717 vk_find_struct_const(queue_create
->pNext
, DEVICE_QUEUE_GLOBAL_PRIORITY_CREATE_INFO_EXT
);
1719 assert(!global_priority
|| device
->physical_device
->rad_info
.has_ctx_priority
);
1721 device
->queues
[qfi
] = vk_alloc(&device
->alloc
,
1722 queue_create
->queueCount
* sizeof(struct radv_queue
), 8, VK_SYSTEM_ALLOCATION_SCOPE_DEVICE
);
1723 if (!device
->queues
[qfi
]) {
1724 result
= VK_ERROR_OUT_OF_HOST_MEMORY
;
1728 memset(device
->queues
[qfi
], 0, queue_create
->queueCount
* sizeof(struct radv_queue
));
1730 device
->queue_count
[qfi
] = queue_create
->queueCount
;
1732 for (unsigned q
= 0; q
< queue_create
->queueCount
; q
++) {
1733 result
= radv_queue_init(device
, &device
->queues
[qfi
][q
],
1734 qfi
, q
, queue_create
->flags
,
1736 if (result
!= VK_SUCCESS
)
1741 device
->pbb_allowed
= device
->physical_device
->rad_info
.chip_class
>= GFX9
&&
1742 !(device
->instance
->debug_flags
& RADV_DEBUG_NOBINNING
);
1744 /* Disabled and not implemented for now. */
1745 device
->dfsm_allowed
= device
->pbb_allowed
&&
1746 (device
->physical_device
->rad_info
.family
== CHIP_RAVEN
||
1747 device
->physical_device
->rad_info
.family
== CHIP_RAVEN2
);
1750 device
->always_use_syncobj
= device
->physical_device
->rad_info
.has_syncobj_wait_for_submit
;
1753 /* The maximum number of scratch waves. Scratch space isn't divided
1754 * evenly between CUs. The number is only a function of the number of CUs.
1755 * We can decrease the constant to decrease the scratch buffer size.
1757 * sctx->scratch_waves must be >= the maximum possible size of
1758 * 1 threadgroup, so that the hw doesn't hang from being unable
1761 * The recommended value is 4 per CU at most. Higher numbers don't
1762 * bring much benefit, but they still occupy chip resources (think
1763 * async compute). I've seen ~2% performance difference between 4 and 32.
1765 uint32_t max_threads_per_block
= 2048;
1766 device
->scratch_waves
= MAX2(32 * physical_device
->rad_info
.num_good_compute_units
,
1767 max_threads_per_block
/ 64);
1769 device
->dispatch_initiator
= S_00B800_COMPUTE_SHADER_EN(1);
1771 if (device
->physical_device
->rad_info
.chip_class
>= CIK
) {
1772 /* If the KMD allows it (there is a KMD hw register for it),
1773 * allow launching waves out-of-order.
1775 device
->dispatch_initiator
|= S_00B800_ORDER_MODE(1);
1778 radv_device_init_gs_info(device
);
1780 device
->tess_offchip_block_dw_size
=
1781 device
->physical_device
->rad_info
.family
== CHIP_HAWAII
? 4096 : 8192;
1782 device
->has_distributed_tess
=
1783 device
->physical_device
->rad_info
.chip_class
>= VI
&&
1784 device
->physical_device
->rad_info
.max_se
>= 2;
1786 if (getenv("RADV_TRACE_FILE")) {
1787 const char *filename
= getenv("RADV_TRACE_FILE");
1789 keep_shader_info
= true;
1791 if (!radv_init_trace(device
))
1794 fprintf(stderr
, "*****************************************************************************\n");
1795 fprintf(stderr
, "* WARNING: RADV_TRACE_FILE is costly and should only be used for debugging! *\n");
1796 fprintf(stderr
, "*****************************************************************************\n");
1798 fprintf(stderr
, "Trace file will be dumped to %s\n", filename
);
1799 radv_dump_enabled_options(device
, stderr
);
1802 device
->keep_shader_info
= keep_shader_info
;
1804 result
= radv_device_init_meta(device
);
1805 if (result
!= VK_SUCCESS
)
1808 radv_device_init_msaa(device
);
1810 for (int family
= 0; family
< RADV_MAX_QUEUE_FAMILIES
; ++family
) {
1811 device
->empty_cs
[family
] = device
->ws
->cs_create(device
->ws
, family
);
1813 case RADV_QUEUE_GENERAL
:
1814 radeon_emit(device
->empty_cs
[family
], PKT3(PKT3_CONTEXT_CONTROL
, 1, 0));
1815 radeon_emit(device
->empty_cs
[family
], CONTEXT_CONTROL_LOAD_ENABLE(1));
1816 radeon_emit(device
->empty_cs
[family
], CONTEXT_CONTROL_SHADOW_ENABLE(1));
1818 case RADV_QUEUE_COMPUTE
:
1819 radeon_emit(device
->empty_cs
[family
], PKT3(PKT3_NOP
, 0, 0));
1820 radeon_emit(device
->empty_cs
[family
], 0);
1823 device
->ws
->cs_finalize(device
->empty_cs
[family
]);
1826 if (device
->physical_device
->rad_info
.chip_class
>= CIK
)
1827 cik_create_gfx_config(device
);
1829 VkPipelineCacheCreateInfo ci
;
1830 ci
.sType
= VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO
;
1833 ci
.pInitialData
= NULL
;
1834 ci
.initialDataSize
= 0;
1836 result
= radv_CreatePipelineCache(radv_device_to_handle(device
),
1838 if (result
!= VK_SUCCESS
)
1841 device
->mem_cache
= radv_pipeline_cache_from_handle(pc
);
1843 device
->force_aniso
=
1844 MIN2(16, radv_get_int_debug_option("RADV_TEX_ANISO", -1));
1845 if (device
->force_aniso
>= 0) {
1846 fprintf(stderr
, "radv: Forcing anisotropy filter to %ix\n",
1847 1 << util_logbase2(device
->force_aniso
));
1850 *pDevice
= radv_device_to_handle(device
);
1854 radv_device_finish_meta(device
);
1856 radv_bo_list_finish(&device
->bo_list
);
1858 if (device
->trace_bo
)
1859 device
->ws
->buffer_destroy(device
->trace_bo
);
1861 if (device
->gfx_init
)
1862 device
->ws
->buffer_destroy(device
->gfx_init
);
1864 for (unsigned i
= 0; i
< RADV_MAX_QUEUE_FAMILIES
; i
++) {
1865 for (unsigned q
= 0; q
< device
->queue_count
[i
]; q
++)
1866 radv_queue_finish(&device
->queues
[i
][q
]);
1867 if (device
->queue_count
[i
])
1868 vk_free(&device
->alloc
, device
->queues
[i
]);
1871 vk_free(&device
->alloc
, device
);
1875 void radv_DestroyDevice(
1877 const VkAllocationCallbacks
* pAllocator
)
1879 RADV_FROM_HANDLE(radv_device
, device
, _device
);
1884 if (device
->trace_bo
)
1885 device
->ws
->buffer_destroy(device
->trace_bo
);
1887 if (device
->gfx_init
)
1888 device
->ws
->buffer_destroy(device
->gfx_init
);
1890 for (unsigned i
= 0; i
< RADV_MAX_QUEUE_FAMILIES
; i
++) {
1891 for (unsigned q
= 0; q
< device
->queue_count
[i
]; q
++)
1892 radv_queue_finish(&device
->queues
[i
][q
]);
1893 if (device
->queue_count
[i
])
1894 vk_free(&device
->alloc
, device
->queues
[i
]);
1895 if (device
->empty_cs
[i
])
1896 device
->ws
->cs_destroy(device
->empty_cs
[i
]);
1898 radv_device_finish_meta(device
);
1900 VkPipelineCache pc
= radv_pipeline_cache_to_handle(device
->mem_cache
);
1901 radv_DestroyPipelineCache(radv_device_to_handle(device
), pc
, NULL
);
1903 radv_destroy_shader_slabs(device
);
1905 radv_bo_list_finish(&device
->bo_list
);
1906 vk_free(&device
->alloc
, device
);
1909 VkResult
radv_EnumerateInstanceLayerProperties(
1910 uint32_t* pPropertyCount
,
1911 VkLayerProperties
* pProperties
)
1913 if (pProperties
== NULL
) {
1914 *pPropertyCount
= 0;
1918 /* None supported at this time */
1919 return vk_error(NULL
, VK_ERROR_LAYER_NOT_PRESENT
);
1922 VkResult
radv_EnumerateDeviceLayerProperties(
1923 VkPhysicalDevice physicalDevice
,
1924 uint32_t* pPropertyCount
,
1925 VkLayerProperties
* pProperties
)
1927 if (pProperties
== NULL
) {
1928 *pPropertyCount
= 0;
1932 /* None supported at this time */
1933 return vk_error(NULL
, VK_ERROR_LAYER_NOT_PRESENT
);
1936 void radv_GetDeviceQueue2(
1938 const VkDeviceQueueInfo2
* pQueueInfo
,
1941 RADV_FROM_HANDLE(radv_device
, device
, _device
);
1942 struct radv_queue
*queue
;
1944 queue
= &device
->queues
[pQueueInfo
->queueFamilyIndex
][pQueueInfo
->queueIndex
];
1945 if (pQueueInfo
->flags
!= queue
->flags
) {
1946 /* From the Vulkan 1.1.70 spec:
1948 * "The queue returned by vkGetDeviceQueue2 must have the same
1949 * flags value from this structure as that used at device
1950 * creation time in a VkDeviceQueueCreateInfo instance. If no
1951 * matching flags were specified at device creation time then
1952 * pQueue will return VK_NULL_HANDLE."
1954 *pQueue
= VK_NULL_HANDLE
;
1958 *pQueue
= radv_queue_to_handle(queue
);
1961 void radv_GetDeviceQueue(
1963 uint32_t queueFamilyIndex
,
1964 uint32_t queueIndex
,
1967 const VkDeviceQueueInfo2 info
= (VkDeviceQueueInfo2
) {
1968 .sType
= VK_STRUCTURE_TYPE_DEVICE_QUEUE_INFO_2
,
1969 .queueFamilyIndex
= queueFamilyIndex
,
1970 .queueIndex
= queueIndex
1973 radv_GetDeviceQueue2(_device
, &info
, pQueue
);
1977 fill_geom_tess_rings(struct radv_queue
*queue
,
1979 bool add_sample_positions
,
1980 uint32_t esgs_ring_size
,
1981 struct radeon_winsys_bo
*esgs_ring_bo
,
1982 uint32_t gsvs_ring_size
,
1983 struct radeon_winsys_bo
*gsvs_ring_bo
,
1984 uint32_t tess_factor_ring_size
,
1985 uint32_t tess_offchip_ring_offset
,
1986 uint32_t tess_offchip_ring_size
,
1987 struct radeon_winsys_bo
*tess_rings_bo
)
1989 uint32_t *desc
= &map
[4];
1992 uint64_t esgs_va
= radv_buffer_get_va(esgs_ring_bo
);
1994 /* stride 0, num records - size, add tid, swizzle, elsize4,
1997 desc
[1] = S_008F04_BASE_ADDRESS_HI(esgs_va
>> 32) |
1998 S_008F04_STRIDE(0) |
1999 S_008F04_SWIZZLE_ENABLE(true);
2000 desc
[2] = esgs_ring_size
;
2001 desc
[3] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2002 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2003 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2004 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2005 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2006 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2007 S_008F0C_ELEMENT_SIZE(1) |
2008 S_008F0C_INDEX_STRIDE(3) |
2009 S_008F0C_ADD_TID_ENABLE(true);
2011 /* GS entry for ES->GS ring */
2012 /* stride 0, num records - size, elsize0,
2015 desc
[5] = S_008F04_BASE_ADDRESS_HI(esgs_va
>> 32)|
2016 S_008F04_STRIDE(0) |
2017 S_008F04_SWIZZLE_ENABLE(false);
2018 desc
[6] = esgs_ring_size
;
2019 desc
[7] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2020 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2021 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2022 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2023 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2024 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2025 S_008F0C_ELEMENT_SIZE(0) |
2026 S_008F0C_INDEX_STRIDE(0) |
2027 S_008F0C_ADD_TID_ENABLE(false);
2033 uint64_t gsvs_va
= radv_buffer_get_va(gsvs_ring_bo
);
2035 /* VS entry for GS->VS ring */
2036 /* stride 0, num records - size, elsize0,
2039 desc
[1] = S_008F04_BASE_ADDRESS_HI(gsvs_va
>> 32)|
2040 S_008F04_STRIDE(0) |
2041 S_008F04_SWIZZLE_ENABLE(false);
2042 desc
[2] = gsvs_ring_size
;
2043 desc
[3] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2044 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2045 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2046 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2047 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2048 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2049 S_008F0C_ELEMENT_SIZE(0) |
2050 S_008F0C_INDEX_STRIDE(0) |
2051 S_008F0C_ADD_TID_ENABLE(false);
2053 /* stride gsvs_itemsize, num records 64
2054 elsize 4, index stride 16 */
2055 /* shader will patch stride and desc[2] */
2057 desc
[5] = S_008F04_BASE_ADDRESS_HI(gsvs_va
>> 32)|
2058 S_008F04_STRIDE(0) |
2059 S_008F04_SWIZZLE_ENABLE(true);
2061 desc
[7] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2062 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2063 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2064 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2065 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2066 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2067 S_008F0C_ELEMENT_SIZE(1) |
2068 S_008F0C_INDEX_STRIDE(1) |
2069 S_008F0C_ADD_TID_ENABLE(true);
2074 if (tess_rings_bo
) {
2075 uint64_t tess_va
= radv_buffer_get_va(tess_rings_bo
);
2076 uint64_t tess_offchip_va
= tess_va
+ tess_offchip_ring_offset
;
2079 desc
[1] = S_008F04_BASE_ADDRESS_HI(tess_va
>> 32) |
2080 S_008F04_STRIDE(0) |
2081 S_008F04_SWIZZLE_ENABLE(false);
2082 desc
[2] = tess_factor_ring_size
;
2083 desc
[3] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2084 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2085 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2086 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2087 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2088 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2089 S_008F0C_ELEMENT_SIZE(0) |
2090 S_008F0C_INDEX_STRIDE(0) |
2091 S_008F0C_ADD_TID_ENABLE(false);
2093 desc
[4] = tess_offchip_va
;
2094 desc
[5] = S_008F04_BASE_ADDRESS_HI(tess_offchip_va
>> 32) |
2095 S_008F04_STRIDE(0) |
2096 S_008F04_SWIZZLE_ENABLE(false);
2097 desc
[6] = tess_offchip_ring_size
;
2098 desc
[7] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X
) |
2099 S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y
) |
2100 S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z
) |
2101 S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W
) |
2102 S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT
) |
2103 S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32
) |
2104 S_008F0C_ELEMENT_SIZE(0) |
2105 S_008F0C_INDEX_STRIDE(0) |
2106 S_008F0C_ADD_TID_ENABLE(false);
2111 if (add_sample_positions
) {
2112 /* add sample positions after all rings */
2113 memcpy(desc
, queue
->device
->sample_locations_1x
, 8);
2115 memcpy(desc
, queue
->device
->sample_locations_2x
, 16);
2117 memcpy(desc
, queue
->device
->sample_locations_4x
, 32);
2119 memcpy(desc
, queue
->device
->sample_locations_8x
, 64);
2124 radv_get_hs_offchip_param(struct radv_device
*device
, uint32_t *max_offchip_buffers_p
)
2126 bool double_offchip_buffers
= device
->physical_device
->rad_info
.chip_class
>= CIK
&&
2127 device
->physical_device
->rad_info
.family
!= CHIP_CARRIZO
&&
2128 device
->physical_device
->rad_info
.family
!= CHIP_STONEY
;
2129 unsigned max_offchip_buffers_per_se
= double_offchip_buffers
? 128 : 64;
2130 unsigned max_offchip_buffers
;
2131 unsigned offchip_granularity
;
2132 unsigned hs_offchip_param
;
2136 * This must be one less than the maximum number due to a hw limitation.
2137 * Various hardware bugs in SI, CIK, and GFX9 need this.
2140 * Vega10 should limit max_offchip_buffers to 508 (4 * 127).
2141 * Gfx7 should limit max_offchip_buffers to 508
2142 * Gfx6 should limit max_offchip_buffers to 126 (2 * 63)
2144 * Follow AMDVLK here.
2146 if (device
->physical_device
->rad_info
.family
== CHIP_VEGA10
||
2147 device
->physical_device
->rad_info
.chip_class
== CIK
||
2148 device
->physical_device
->rad_info
.chip_class
== SI
)
2149 --max_offchip_buffers_per_se
;
2151 max_offchip_buffers
= max_offchip_buffers_per_se
*
2152 device
->physical_device
->rad_info
.max_se
;
2154 /* Hawaii has a bug with offchip buffers > 256 that can be worked
2155 * around by setting 4K granularity.
2157 if (device
->tess_offchip_block_dw_size
== 4096) {
2158 assert(device
->physical_device
->rad_info
.family
== CHIP_HAWAII
);
2159 offchip_granularity
= V_03093C_X_4K_DWORDS
;
2161 assert(device
->tess_offchip_block_dw_size
== 8192);
2162 offchip_granularity
= V_03093C_X_8K_DWORDS
;
2165 switch (device
->physical_device
->rad_info
.chip_class
) {
2167 max_offchip_buffers
= MIN2(max_offchip_buffers
, 126);
2173 max_offchip_buffers
= MIN2(max_offchip_buffers
, 508);
2177 *max_offchip_buffers_p
= max_offchip_buffers
;
2178 if (device
->physical_device
->rad_info
.chip_class
>= CIK
) {
2179 if (device
->physical_device
->rad_info
.chip_class
>= VI
)
2180 --max_offchip_buffers
;
2182 S_03093C_OFFCHIP_BUFFERING(max_offchip_buffers
) |
2183 S_03093C_OFFCHIP_GRANULARITY(offchip_granularity
);
2186 S_0089B0_OFFCHIP_BUFFERING(max_offchip_buffers
);
2188 return hs_offchip_param
;
2192 radv_emit_gs_ring_sizes(struct radv_queue
*queue
, struct radeon_cmdbuf
*cs
,
2193 struct radeon_winsys_bo
*esgs_ring_bo
,
2194 uint32_t esgs_ring_size
,
2195 struct radeon_winsys_bo
*gsvs_ring_bo
,
2196 uint32_t gsvs_ring_size
)
2198 if (!esgs_ring_bo
&& !gsvs_ring_bo
)
2202 radv_cs_add_buffer(queue
->device
->ws
, cs
, esgs_ring_bo
);
2205 radv_cs_add_buffer(queue
->device
->ws
, cs
, gsvs_ring_bo
);
2207 if (queue
->device
->physical_device
->rad_info
.chip_class
>= CIK
) {
2208 radeon_set_uconfig_reg_seq(cs
, R_030900_VGT_ESGS_RING_SIZE
, 2);
2209 radeon_emit(cs
, esgs_ring_size
>> 8);
2210 radeon_emit(cs
, gsvs_ring_size
>> 8);
2212 radeon_set_config_reg_seq(cs
, R_0088C8_VGT_ESGS_RING_SIZE
, 2);
2213 radeon_emit(cs
, esgs_ring_size
>> 8);
2214 radeon_emit(cs
, gsvs_ring_size
>> 8);
2219 radv_emit_tess_factor_ring(struct radv_queue
*queue
, struct radeon_cmdbuf
*cs
,
2220 unsigned hs_offchip_param
, unsigned tf_ring_size
,
2221 struct radeon_winsys_bo
*tess_rings_bo
)
2228 tf_va
= radv_buffer_get_va(tess_rings_bo
);
2230 radv_cs_add_buffer(queue
->device
->ws
, cs
, tess_rings_bo
);
2232 if (queue
->device
->physical_device
->rad_info
.chip_class
>= CIK
) {
2233 radeon_set_uconfig_reg(cs
, R_030938_VGT_TF_RING_SIZE
,
2234 S_030938_SIZE(tf_ring_size
/ 4));
2235 radeon_set_uconfig_reg(cs
, R_030940_VGT_TF_MEMORY_BASE
,
2237 if (queue
->device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
2238 radeon_set_uconfig_reg(cs
, R_030944_VGT_TF_MEMORY_BASE_HI
,
2239 S_030944_BASE_HI(tf_va
>> 40));
2241 radeon_set_uconfig_reg(cs
, R_03093C_VGT_HS_OFFCHIP_PARAM
,
2244 radeon_set_config_reg(cs
, R_008988_VGT_TF_RING_SIZE
,
2245 S_008988_SIZE(tf_ring_size
/ 4));
2246 radeon_set_config_reg(cs
, R_0089B8_VGT_TF_MEMORY_BASE
,
2248 radeon_set_config_reg(cs
, R_0089B0_VGT_HS_OFFCHIP_PARAM
,
2254 radv_emit_compute_scratch(struct radv_queue
*queue
, struct radeon_cmdbuf
*cs
,
2255 struct radeon_winsys_bo
*compute_scratch_bo
)
2257 uint64_t scratch_va
;
2259 if (!compute_scratch_bo
)
2262 scratch_va
= radv_buffer_get_va(compute_scratch_bo
);
2264 radv_cs_add_buffer(queue
->device
->ws
, cs
, compute_scratch_bo
);
2266 radeon_set_sh_reg_seq(cs
, R_00B900_COMPUTE_USER_DATA_0
, 2);
2267 radeon_emit(cs
, scratch_va
);
2268 radeon_emit(cs
, S_008F04_BASE_ADDRESS_HI(scratch_va
>> 32) |
2269 S_008F04_SWIZZLE_ENABLE(1));
2273 radv_emit_global_shader_pointers(struct radv_queue
*queue
,
2274 struct radeon_cmdbuf
*cs
,
2275 struct radeon_winsys_bo
*descriptor_bo
)
2282 va
= radv_buffer_get_va(descriptor_bo
);
2284 radv_cs_add_buffer(queue
->device
->ws
, cs
, descriptor_bo
);
2286 if (queue
->device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
2287 uint32_t regs
[] = {R_00B030_SPI_SHADER_USER_DATA_PS_0
,
2288 R_00B130_SPI_SHADER_USER_DATA_VS_0
,
2289 R_00B208_SPI_SHADER_USER_DATA_ADDR_LO_GS
,
2290 R_00B408_SPI_SHADER_USER_DATA_ADDR_LO_HS
};
2292 for (int i
= 0; i
< ARRAY_SIZE(regs
); ++i
) {
2293 radv_emit_shader_pointer(queue
->device
, cs
, regs
[i
],
2297 uint32_t regs
[] = {R_00B030_SPI_SHADER_USER_DATA_PS_0
,
2298 R_00B130_SPI_SHADER_USER_DATA_VS_0
,
2299 R_00B230_SPI_SHADER_USER_DATA_GS_0
,
2300 R_00B330_SPI_SHADER_USER_DATA_ES_0
,
2301 R_00B430_SPI_SHADER_USER_DATA_HS_0
,
2302 R_00B530_SPI_SHADER_USER_DATA_LS_0
};
2304 for (int i
= 0; i
< ARRAY_SIZE(regs
); ++i
) {
2305 radv_emit_shader_pointer(queue
->device
, cs
, regs
[i
],
2312 radv_init_graphics_state(struct radeon_cmdbuf
*cs
, struct radv_queue
*queue
)
2314 struct radv_device
*device
= queue
->device
;
2316 if (device
->gfx_init
) {
2317 uint64_t va
= radv_buffer_get_va(device
->gfx_init
);
2319 radeon_emit(cs
, PKT3(PKT3_INDIRECT_BUFFER_CIK
, 2, 0));
2320 radeon_emit(cs
, va
);
2321 radeon_emit(cs
, va
>> 32);
2322 radeon_emit(cs
, device
->gfx_init_size_dw
& 0xffff);
2324 radv_cs_add_buffer(device
->ws
, cs
, device
->gfx_init
);
2326 struct radv_physical_device
*physical_device
= device
->physical_device
;
2327 si_emit_graphics(physical_device
, cs
);
2332 radv_init_compute_state(struct radeon_cmdbuf
*cs
, struct radv_queue
*queue
)
2334 struct radv_physical_device
*physical_device
= queue
->device
->physical_device
;
2335 si_emit_compute(physical_device
, cs
);
2339 radv_get_preamble_cs(struct radv_queue
*queue
,
2340 uint32_t scratch_size
,
2341 uint32_t compute_scratch_size
,
2342 uint32_t esgs_ring_size
,
2343 uint32_t gsvs_ring_size
,
2344 bool needs_tess_rings
,
2345 bool needs_sample_positions
,
2346 struct radeon_cmdbuf
**initial_full_flush_preamble_cs
,
2347 struct radeon_cmdbuf
**initial_preamble_cs
,
2348 struct radeon_cmdbuf
**continue_preamble_cs
)
2350 struct radeon_winsys_bo
*scratch_bo
= NULL
;
2351 struct radeon_winsys_bo
*descriptor_bo
= NULL
;
2352 struct radeon_winsys_bo
*compute_scratch_bo
= NULL
;
2353 struct radeon_winsys_bo
*esgs_ring_bo
= NULL
;
2354 struct radeon_winsys_bo
*gsvs_ring_bo
= NULL
;
2355 struct radeon_winsys_bo
*tess_rings_bo
= NULL
;
2356 struct radeon_cmdbuf
*dest_cs
[3] = {0};
2357 bool add_tess_rings
= false, add_sample_positions
= false;
2358 unsigned tess_factor_ring_size
= 0, tess_offchip_ring_size
= 0;
2359 unsigned max_offchip_buffers
;
2360 unsigned hs_offchip_param
= 0;
2361 unsigned tess_offchip_ring_offset
;
2362 uint32_t ring_bo_flags
= RADEON_FLAG_NO_CPU_ACCESS
| RADEON_FLAG_NO_INTERPROCESS_SHARING
;
2363 if (!queue
->has_tess_rings
) {
2364 if (needs_tess_rings
)
2365 add_tess_rings
= true;
2367 if (!queue
->has_sample_positions
) {
2368 if (needs_sample_positions
)
2369 add_sample_positions
= true;
2371 tess_factor_ring_size
= 32768 * queue
->device
->physical_device
->rad_info
.max_se
;
2372 hs_offchip_param
= radv_get_hs_offchip_param(queue
->device
,
2373 &max_offchip_buffers
);
2374 tess_offchip_ring_offset
= align(tess_factor_ring_size
, 64 * 1024);
2375 tess_offchip_ring_size
= max_offchip_buffers
*
2376 queue
->device
->tess_offchip_block_dw_size
* 4;
2378 if (scratch_size
<= queue
->scratch_size
&&
2379 compute_scratch_size
<= queue
->compute_scratch_size
&&
2380 esgs_ring_size
<= queue
->esgs_ring_size
&&
2381 gsvs_ring_size
<= queue
->gsvs_ring_size
&&
2382 !add_tess_rings
&& !add_sample_positions
&&
2383 queue
->initial_preamble_cs
) {
2384 *initial_full_flush_preamble_cs
= queue
->initial_full_flush_preamble_cs
;
2385 *initial_preamble_cs
= queue
->initial_preamble_cs
;
2386 *continue_preamble_cs
= queue
->continue_preamble_cs
;
2387 if (!scratch_size
&& !compute_scratch_size
&& !esgs_ring_size
&& !gsvs_ring_size
)
2388 *continue_preamble_cs
= NULL
;
2392 if (scratch_size
> queue
->scratch_size
) {
2393 scratch_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2398 RADV_BO_PRIORITY_SCRATCH
);
2402 scratch_bo
= queue
->scratch_bo
;
2404 if (compute_scratch_size
> queue
->compute_scratch_size
) {
2405 compute_scratch_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2406 compute_scratch_size
,
2410 RADV_BO_PRIORITY_SCRATCH
);
2411 if (!compute_scratch_bo
)
2415 compute_scratch_bo
= queue
->compute_scratch_bo
;
2417 if (esgs_ring_size
> queue
->esgs_ring_size
) {
2418 esgs_ring_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2423 RADV_BO_PRIORITY_SCRATCH
);
2427 esgs_ring_bo
= queue
->esgs_ring_bo
;
2428 esgs_ring_size
= queue
->esgs_ring_size
;
2431 if (gsvs_ring_size
> queue
->gsvs_ring_size
) {
2432 gsvs_ring_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2437 RADV_BO_PRIORITY_SCRATCH
);
2441 gsvs_ring_bo
= queue
->gsvs_ring_bo
;
2442 gsvs_ring_size
= queue
->gsvs_ring_size
;
2445 if (add_tess_rings
) {
2446 tess_rings_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2447 tess_offchip_ring_offset
+ tess_offchip_ring_size
,
2451 RADV_BO_PRIORITY_SCRATCH
);
2455 tess_rings_bo
= queue
->tess_rings_bo
;
2458 if (scratch_bo
!= queue
->scratch_bo
||
2459 esgs_ring_bo
!= queue
->esgs_ring_bo
||
2460 gsvs_ring_bo
!= queue
->gsvs_ring_bo
||
2461 tess_rings_bo
!= queue
->tess_rings_bo
||
2462 add_sample_positions
) {
2464 if (gsvs_ring_bo
|| esgs_ring_bo
||
2465 tess_rings_bo
|| add_sample_positions
) {
2466 size
= 112; /* 2 dword + 2 padding + 4 dword * 6 */
2467 if (add_sample_positions
)
2468 size
+= 128; /* 64+32+16+8 = 120 bytes */
2470 else if (scratch_bo
)
2471 size
= 8; /* 2 dword */
2473 descriptor_bo
= queue
->device
->ws
->buffer_create(queue
->device
->ws
,
2477 RADEON_FLAG_CPU_ACCESS
|
2478 RADEON_FLAG_NO_INTERPROCESS_SHARING
|
2479 RADEON_FLAG_READ_ONLY
,
2480 RADV_BO_PRIORITY_DESCRIPTOR
);
2484 descriptor_bo
= queue
->descriptor_bo
;
2486 if (descriptor_bo
!= queue
->descriptor_bo
) {
2487 uint32_t *map
= (uint32_t*)queue
->device
->ws
->buffer_map(descriptor_bo
);
2490 uint64_t scratch_va
= radv_buffer_get_va(scratch_bo
);
2491 uint32_t rsrc1
= S_008F04_BASE_ADDRESS_HI(scratch_va
>> 32) |
2492 S_008F04_SWIZZLE_ENABLE(1);
2493 map
[0] = scratch_va
;
2497 if (esgs_ring_bo
|| gsvs_ring_bo
|| tess_rings_bo
|| add_sample_positions
)
2498 fill_geom_tess_rings(queue
, map
, add_sample_positions
,
2499 esgs_ring_size
, esgs_ring_bo
,
2500 gsvs_ring_size
, gsvs_ring_bo
,
2501 tess_factor_ring_size
,
2502 tess_offchip_ring_offset
,
2503 tess_offchip_ring_size
,
2506 queue
->device
->ws
->buffer_unmap(descriptor_bo
);
2509 for(int i
= 0; i
< 3; ++i
) {
2510 struct radeon_cmdbuf
*cs
= NULL
;
2511 cs
= queue
->device
->ws
->cs_create(queue
->device
->ws
,
2512 queue
->queue_family_index
? RING_COMPUTE
: RING_GFX
);
2519 radv_cs_add_buffer(queue
->device
->ws
, cs
, scratch_bo
);
2521 /* Emit initial configuration. */
2522 switch (queue
->queue_family_index
) {
2523 case RADV_QUEUE_GENERAL
:
2524 radv_init_graphics_state(cs
, queue
);
2526 case RADV_QUEUE_COMPUTE
:
2527 radv_init_compute_state(cs
, queue
);
2529 case RADV_QUEUE_TRANSFER
:
2533 if (esgs_ring_bo
|| gsvs_ring_bo
|| tess_rings_bo
) {
2534 radeon_emit(cs
, PKT3(PKT3_EVENT_WRITE
, 0, 0));
2535 radeon_emit(cs
, EVENT_TYPE(V_028A90_VS_PARTIAL_FLUSH
) | EVENT_INDEX(4));
2536 radeon_emit(cs
, PKT3(PKT3_EVENT_WRITE
, 0, 0));
2537 radeon_emit(cs
, EVENT_TYPE(V_028A90_VGT_FLUSH
) | EVENT_INDEX(0));
2540 radv_emit_gs_ring_sizes(queue
, cs
, esgs_ring_bo
, esgs_ring_size
,
2541 gsvs_ring_bo
, gsvs_ring_size
);
2542 radv_emit_tess_factor_ring(queue
, cs
, hs_offchip_param
,
2543 tess_factor_ring_size
, tess_rings_bo
);
2544 radv_emit_global_shader_pointers(queue
, cs
, descriptor_bo
);
2545 radv_emit_compute_scratch(queue
, cs
, compute_scratch_bo
);
2548 si_cs_emit_cache_flush(cs
,
2549 queue
->device
->physical_device
->rad_info
.chip_class
,
2551 queue
->queue_family_index
== RING_COMPUTE
&&
2552 queue
->device
->physical_device
->rad_info
.chip_class
>= CIK
,
2553 (queue
->queue_family_index
== RADV_QUEUE_COMPUTE
? RADV_CMD_FLAG_CS_PARTIAL_FLUSH
: (RADV_CMD_FLAG_CS_PARTIAL_FLUSH
| RADV_CMD_FLAG_PS_PARTIAL_FLUSH
)) |
2554 RADV_CMD_FLAG_INV_ICACHE
|
2555 RADV_CMD_FLAG_INV_SMEM_L1
|
2556 RADV_CMD_FLAG_INV_VMEM_L1
|
2557 RADV_CMD_FLAG_INV_GLOBAL_L2
|
2558 RADV_CMD_FLAG_START_PIPELINE_STATS
, 0);
2559 } else if (i
== 1) {
2560 si_cs_emit_cache_flush(cs
,
2561 queue
->device
->physical_device
->rad_info
.chip_class
,
2563 queue
->queue_family_index
== RING_COMPUTE
&&
2564 queue
->device
->physical_device
->rad_info
.chip_class
>= CIK
,
2565 RADV_CMD_FLAG_INV_ICACHE
|
2566 RADV_CMD_FLAG_INV_SMEM_L1
|
2567 RADV_CMD_FLAG_INV_VMEM_L1
|
2568 RADV_CMD_FLAG_INV_GLOBAL_L2
|
2569 RADV_CMD_FLAG_START_PIPELINE_STATS
, 0);
2572 if (!queue
->device
->ws
->cs_finalize(cs
))
2576 if (queue
->initial_full_flush_preamble_cs
)
2577 queue
->device
->ws
->cs_destroy(queue
->initial_full_flush_preamble_cs
);
2579 if (queue
->initial_preamble_cs
)
2580 queue
->device
->ws
->cs_destroy(queue
->initial_preamble_cs
);
2582 if (queue
->continue_preamble_cs
)
2583 queue
->device
->ws
->cs_destroy(queue
->continue_preamble_cs
);
2585 queue
->initial_full_flush_preamble_cs
= dest_cs
[0];
2586 queue
->initial_preamble_cs
= dest_cs
[1];
2587 queue
->continue_preamble_cs
= dest_cs
[2];
2589 if (scratch_bo
!= queue
->scratch_bo
) {
2590 if (queue
->scratch_bo
)
2591 queue
->device
->ws
->buffer_destroy(queue
->scratch_bo
);
2592 queue
->scratch_bo
= scratch_bo
;
2593 queue
->scratch_size
= scratch_size
;
2596 if (compute_scratch_bo
!= queue
->compute_scratch_bo
) {
2597 if (queue
->compute_scratch_bo
)
2598 queue
->device
->ws
->buffer_destroy(queue
->compute_scratch_bo
);
2599 queue
->compute_scratch_bo
= compute_scratch_bo
;
2600 queue
->compute_scratch_size
= compute_scratch_size
;
2603 if (esgs_ring_bo
!= queue
->esgs_ring_bo
) {
2604 if (queue
->esgs_ring_bo
)
2605 queue
->device
->ws
->buffer_destroy(queue
->esgs_ring_bo
);
2606 queue
->esgs_ring_bo
= esgs_ring_bo
;
2607 queue
->esgs_ring_size
= esgs_ring_size
;
2610 if (gsvs_ring_bo
!= queue
->gsvs_ring_bo
) {
2611 if (queue
->gsvs_ring_bo
)
2612 queue
->device
->ws
->buffer_destroy(queue
->gsvs_ring_bo
);
2613 queue
->gsvs_ring_bo
= gsvs_ring_bo
;
2614 queue
->gsvs_ring_size
= gsvs_ring_size
;
2617 if (tess_rings_bo
!= queue
->tess_rings_bo
) {
2618 queue
->tess_rings_bo
= tess_rings_bo
;
2619 queue
->has_tess_rings
= true;
2622 if (descriptor_bo
!= queue
->descriptor_bo
) {
2623 if (queue
->descriptor_bo
)
2624 queue
->device
->ws
->buffer_destroy(queue
->descriptor_bo
);
2626 queue
->descriptor_bo
= descriptor_bo
;
2629 if (add_sample_positions
)
2630 queue
->has_sample_positions
= true;
2632 *initial_full_flush_preamble_cs
= queue
->initial_full_flush_preamble_cs
;
2633 *initial_preamble_cs
= queue
->initial_preamble_cs
;
2634 *continue_preamble_cs
= queue
->continue_preamble_cs
;
2635 if (!scratch_size
&& !compute_scratch_size
&& !esgs_ring_size
&& !gsvs_ring_size
)
2636 *continue_preamble_cs
= NULL
;
2639 for (int i
= 0; i
< ARRAY_SIZE(dest_cs
); ++i
)
2641 queue
->device
->ws
->cs_destroy(dest_cs
[i
]);
2642 if (descriptor_bo
&& descriptor_bo
!= queue
->descriptor_bo
)
2643 queue
->device
->ws
->buffer_destroy(descriptor_bo
);
2644 if (scratch_bo
&& scratch_bo
!= queue
->scratch_bo
)
2645 queue
->device
->ws
->buffer_destroy(scratch_bo
);
2646 if (compute_scratch_bo
&& compute_scratch_bo
!= queue
->compute_scratch_bo
)
2647 queue
->device
->ws
->buffer_destroy(compute_scratch_bo
);
2648 if (esgs_ring_bo
&& esgs_ring_bo
!= queue
->esgs_ring_bo
)
2649 queue
->device
->ws
->buffer_destroy(esgs_ring_bo
);
2650 if (gsvs_ring_bo
&& gsvs_ring_bo
!= queue
->gsvs_ring_bo
)
2651 queue
->device
->ws
->buffer_destroy(gsvs_ring_bo
);
2652 if (tess_rings_bo
&& tess_rings_bo
!= queue
->tess_rings_bo
)
2653 queue
->device
->ws
->buffer_destroy(tess_rings_bo
);
2654 return vk_error(queue
->device
->instance
, VK_ERROR_OUT_OF_DEVICE_MEMORY
);
2657 static VkResult
radv_alloc_sem_counts(struct radv_instance
*instance
,
2658 struct radv_winsys_sem_counts
*counts
,
2660 const VkSemaphore
*sems
,
2664 int syncobj_idx
= 0, sem_idx
= 0;
2666 if (num_sems
== 0 && _fence
== VK_NULL_HANDLE
)
2669 for (uint32_t i
= 0; i
< num_sems
; i
++) {
2670 RADV_FROM_HANDLE(radv_semaphore
, sem
, sems
[i
]);
2672 if (sem
->temp_syncobj
|| sem
->syncobj
)
2673 counts
->syncobj_count
++;
2675 counts
->sem_count
++;
2678 if (_fence
!= VK_NULL_HANDLE
) {
2679 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
2680 if (fence
->temp_syncobj
|| fence
->syncobj
)
2681 counts
->syncobj_count
++;
2684 if (counts
->syncobj_count
) {
2685 counts
->syncobj
= (uint32_t *)malloc(sizeof(uint32_t) * counts
->syncobj_count
);
2686 if (!counts
->syncobj
)
2687 return vk_error(instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
2690 if (counts
->sem_count
) {
2691 counts
->sem
= (struct radeon_winsys_sem
**)malloc(sizeof(struct radeon_winsys_sem
*) * counts
->sem_count
);
2693 free(counts
->syncobj
);
2694 return vk_error(instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
2698 for (uint32_t i
= 0; i
< num_sems
; i
++) {
2699 RADV_FROM_HANDLE(radv_semaphore
, sem
, sems
[i
]);
2701 if (sem
->temp_syncobj
) {
2702 counts
->syncobj
[syncobj_idx
++] = sem
->temp_syncobj
;
2704 else if (sem
->syncobj
)
2705 counts
->syncobj
[syncobj_idx
++] = sem
->syncobj
;
2708 counts
->sem
[sem_idx
++] = sem
->sem
;
2712 if (_fence
!= VK_NULL_HANDLE
) {
2713 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
2714 if (fence
->temp_syncobj
)
2715 counts
->syncobj
[syncobj_idx
++] = fence
->temp_syncobj
;
2716 else if (fence
->syncobj
)
2717 counts
->syncobj
[syncobj_idx
++] = fence
->syncobj
;
2724 radv_free_sem_info(struct radv_winsys_sem_info
*sem_info
)
2726 free(sem_info
->wait
.syncobj
);
2727 free(sem_info
->wait
.sem
);
2728 free(sem_info
->signal
.syncobj
);
2729 free(sem_info
->signal
.sem
);
2733 static void radv_free_temp_syncobjs(struct radv_device
*device
,
2735 const VkSemaphore
*sems
)
2737 for (uint32_t i
= 0; i
< num_sems
; i
++) {
2738 RADV_FROM_HANDLE(radv_semaphore
, sem
, sems
[i
]);
2740 if (sem
->temp_syncobj
) {
2741 device
->ws
->destroy_syncobj(device
->ws
, sem
->temp_syncobj
);
2742 sem
->temp_syncobj
= 0;
2748 radv_alloc_sem_info(struct radv_instance
*instance
,
2749 struct radv_winsys_sem_info
*sem_info
,
2751 const VkSemaphore
*wait_sems
,
2752 int num_signal_sems
,
2753 const VkSemaphore
*signal_sems
,
2757 memset(sem_info
, 0, sizeof(*sem_info
));
2759 ret
= radv_alloc_sem_counts(instance
, &sem_info
->wait
, num_wait_sems
, wait_sems
, VK_NULL_HANDLE
, true);
2762 ret
= radv_alloc_sem_counts(instance
, &sem_info
->signal
, num_signal_sems
, signal_sems
, fence
, false);
2764 radv_free_sem_info(sem_info
);
2766 /* caller can override these */
2767 sem_info
->cs_emit_wait
= true;
2768 sem_info
->cs_emit_signal
= true;
2772 /* Signals fence as soon as all the work currently put on queue is done. */
2773 static VkResult
radv_signal_fence(struct radv_queue
*queue
,
2774 struct radv_fence
*fence
)
2778 struct radv_winsys_sem_info sem_info
;
2780 result
= radv_alloc_sem_info(queue
->device
->instance
, &sem_info
, 0, NULL
, 0, NULL
,
2781 radv_fence_to_handle(fence
));
2782 if (result
!= VK_SUCCESS
)
2785 ret
= queue
->device
->ws
->cs_submit(queue
->hw_ctx
, queue
->queue_idx
,
2786 &queue
->device
->empty_cs
[queue
->queue_family_index
],
2787 1, NULL
, NULL
, &sem_info
, NULL
,
2788 false, fence
->fence
);
2789 radv_free_sem_info(&sem_info
);
2792 return vk_error(queue
->device
->instance
, VK_ERROR_DEVICE_LOST
);
2797 VkResult
radv_QueueSubmit(
2799 uint32_t submitCount
,
2800 const VkSubmitInfo
* pSubmits
,
2803 RADV_FROM_HANDLE(radv_queue
, queue
, _queue
);
2804 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
2805 struct radeon_winsys_fence
*base_fence
= fence
? fence
->fence
: NULL
;
2806 struct radeon_winsys_ctx
*ctx
= queue
->hw_ctx
;
2808 uint32_t max_cs_submission
= queue
->device
->trace_bo
? 1 : UINT32_MAX
;
2809 uint32_t scratch_size
= 0;
2810 uint32_t compute_scratch_size
= 0;
2811 uint32_t esgs_ring_size
= 0, gsvs_ring_size
= 0;
2812 struct radeon_cmdbuf
*initial_preamble_cs
= NULL
, *initial_flush_preamble_cs
= NULL
, *continue_preamble_cs
= NULL
;
2814 bool fence_emitted
= false;
2815 bool tess_rings_needed
= false;
2816 bool sample_positions_needed
= false;
2818 /* Do this first so failing to allocate scratch buffers can't result in
2819 * partially executed submissions. */
2820 for (uint32_t i
= 0; i
< submitCount
; i
++) {
2821 for (uint32_t j
= 0; j
< pSubmits
[i
].commandBufferCount
; j
++) {
2822 RADV_FROM_HANDLE(radv_cmd_buffer
, cmd_buffer
,
2823 pSubmits
[i
].pCommandBuffers
[j
]);
2825 scratch_size
= MAX2(scratch_size
, cmd_buffer
->scratch_size_needed
);
2826 compute_scratch_size
= MAX2(compute_scratch_size
,
2827 cmd_buffer
->compute_scratch_size_needed
);
2828 esgs_ring_size
= MAX2(esgs_ring_size
, cmd_buffer
->esgs_ring_size_needed
);
2829 gsvs_ring_size
= MAX2(gsvs_ring_size
, cmd_buffer
->gsvs_ring_size_needed
);
2830 tess_rings_needed
|= cmd_buffer
->tess_rings_needed
;
2831 sample_positions_needed
|= cmd_buffer
->sample_positions_needed
;
2835 result
= radv_get_preamble_cs(queue
, scratch_size
, compute_scratch_size
,
2836 esgs_ring_size
, gsvs_ring_size
, tess_rings_needed
,
2837 sample_positions_needed
, &initial_flush_preamble_cs
,
2838 &initial_preamble_cs
, &continue_preamble_cs
);
2839 if (result
!= VK_SUCCESS
)
2842 for (uint32_t i
= 0; i
< submitCount
; i
++) {
2843 struct radeon_cmdbuf
**cs_array
;
2844 bool do_flush
= !i
|| pSubmits
[i
].pWaitDstStageMask
;
2845 bool can_patch
= true;
2847 struct radv_winsys_sem_info sem_info
;
2849 result
= radv_alloc_sem_info(queue
->device
->instance
,
2851 pSubmits
[i
].waitSemaphoreCount
,
2852 pSubmits
[i
].pWaitSemaphores
,
2853 pSubmits
[i
].signalSemaphoreCount
,
2854 pSubmits
[i
].pSignalSemaphores
,
2856 if (result
!= VK_SUCCESS
)
2859 if (!pSubmits
[i
].commandBufferCount
) {
2860 if (pSubmits
[i
].waitSemaphoreCount
|| pSubmits
[i
].signalSemaphoreCount
) {
2861 ret
= queue
->device
->ws
->cs_submit(ctx
, queue
->queue_idx
,
2862 &queue
->device
->empty_cs
[queue
->queue_family_index
],
2867 radv_loge("failed to submit CS %d\n", i
);
2870 fence_emitted
= true;
2872 radv_free_sem_info(&sem_info
);
2876 cs_array
= malloc(sizeof(struct radeon_cmdbuf
*) *
2877 (pSubmits
[i
].commandBufferCount
));
2879 for (uint32_t j
= 0; j
< pSubmits
[i
].commandBufferCount
; j
++) {
2880 RADV_FROM_HANDLE(radv_cmd_buffer
, cmd_buffer
,
2881 pSubmits
[i
].pCommandBuffers
[j
]);
2882 assert(cmd_buffer
->level
== VK_COMMAND_BUFFER_LEVEL_PRIMARY
);
2884 cs_array
[j
] = cmd_buffer
->cs
;
2885 if ((cmd_buffer
->usage_flags
& VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT
))
2888 cmd_buffer
->status
= RADV_CMD_BUFFER_STATUS_PENDING
;
2891 for (uint32_t j
= 0; j
< pSubmits
[i
].commandBufferCount
; j
+= advance
) {
2892 struct radeon_cmdbuf
*initial_preamble
= (do_flush
&& !j
) ? initial_flush_preamble_cs
: initial_preamble_cs
;
2893 const struct radv_winsys_bo_list
*bo_list
= NULL
;
2895 advance
= MIN2(max_cs_submission
,
2896 pSubmits
[i
].commandBufferCount
- j
);
2898 if (queue
->device
->trace_bo
)
2899 *queue
->device
->trace_id_ptr
= 0;
2901 sem_info
.cs_emit_wait
= j
== 0;
2902 sem_info
.cs_emit_signal
= j
+ advance
== pSubmits
[i
].commandBufferCount
;
2904 if (unlikely(queue
->device
->use_global_bo_list
)) {
2905 pthread_mutex_lock(&queue
->device
->bo_list
.mutex
);
2906 bo_list
= &queue
->device
->bo_list
.list
;
2909 ret
= queue
->device
->ws
->cs_submit(ctx
, queue
->queue_idx
, cs_array
+ j
,
2910 advance
, initial_preamble
, continue_preamble_cs
,
2912 can_patch
, base_fence
);
2914 if (unlikely(queue
->device
->use_global_bo_list
))
2915 pthread_mutex_unlock(&queue
->device
->bo_list
.mutex
);
2918 radv_loge("failed to submit CS %d\n", i
);
2921 fence_emitted
= true;
2922 if (queue
->device
->trace_bo
) {
2923 radv_check_gpu_hangs(queue
, cs_array
[j
]);
2927 radv_free_temp_syncobjs(queue
->device
,
2928 pSubmits
[i
].waitSemaphoreCount
,
2929 pSubmits
[i
].pWaitSemaphores
);
2930 radv_free_sem_info(&sem_info
);
2935 if (!fence_emitted
) {
2936 result
= radv_signal_fence(queue
, fence
);
2937 if (result
!= VK_SUCCESS
)
2940 fence
->submitted
= true;
2946 VkResult
radv_QueueWaitIdle(
2949 RADV_FROM_HANDLE(radv_queue
, queue
, _queue
);
2951 queue
->device
->ws
->ctx_wait_idle(queue
->hw_ctx
,
2952 radv_queue_family_to_ring(queue
->queue_family_index
),
2957 VkResult
radv_DeviceWaitIdle(
2960 RADV_FROM_HANDLE(radv_device
, device
, _device
);
2962 for (unsigned i
= 0; i
< RADV_MAX_QUEUE_FAMILIES
; i
++) {
2963 for (unsigned q
= 0; q
< device
->queue_count
[i
]; q
++) {
2964 radv_QueueWaitIdle(radv_queue_to_handle(&device
->queues
[i
][q
]));
2970 VkResult
radv_EnumerateInstanceExtensionProperties(
2971 const char* pLayerName
,
2972 uint32_t* pPropertyCount
,
2973 VkExtensionProperties
* pProperties
)
2975 VK_OUTARRAY_MAKE(out
, pProperties
, pPropertyCount
);
2977 for (int i
= 0; i
< RADV_INSTANCE_EXTENSION_COUNT
; i
++) {
2978 if (radv_supported_instance_extensions
.extensions
[i
]) {
2979 vk_outarray_append(&out
, prop
) {
2980 *prop
= radv_instance_extensions
[i
];
2985 return vk_outarray_status(&out
);
2988 VkResult
radv_EnumerateDeviceExtensionProperties(
2989 VkPhysicalDevice physicalDevice
,
2990 const char* pLayerName
,
2991 uint32_t* pPropertyCount
,
2992 VkExtensionProperties
* pProperties
)
2994 RADV_FROM_HANDLE(radv_physical_device
, device
, physicalDevice
);
2995 VK_OUTARRAY_MAKE(out
, pProperties
, pPropertyCount
);
2997 for (int i
= 0; i
< RADV_DEVICE_EXTENSION_COUNT
; i
++) {
2998 if (device
->supported_extensions
.extensions
[i
]) {
2999 vk_outarray_append(&out
, prop
) {
3000 *prop
= radv_device_extensions
[i
];
3005 return vk_outarray_status(&out
);
3008 PFN_vkVoidFunction
radv_GetInstanceProcAddr(
3009 VkInstance _instance
,
3012 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
3014 return radv_lookup_entrypoint_checked(pName
,
3015 instance
? instance
->apiVersion
: 0,
3016 instance
? &instance
->enabled_extensions
: NULL
,
3020 /* The loader wants us to expose a second GetInstanceProcAddr function
3021 * to work around certain LD_PRELOAD issues seen in apps.
3024 VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL
vk_icdGetInstanceProcAddr(
3025 VkInstance instance
,
3029 VKAPI_ATTR PFN_vkVoidFunction VKAPI_CALL
vk_icdGetInstanceProcAddr(
3030 VkInstance instance
,
3033 return radv_GetInstanceProcAddr(instance
, pName
);
3036 PFN_vkVoidFunction
radv_GetDeviceProcAddr(
3040 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3042 return radv_lookup_entrypoint_checked(pName
,
3043 device
->instance
->apiVersion
,
3044 &device
->instance
->enabled_extensions
,
3045 &device
->enabled_extensions
);
3048 bool radv_get_memory_fd(struct radv_device
*device
,
3049 struct radv_device_memory
*memory
,
3052 struct radeon_bo_metadata metadata
;
3054 if (memory
->image
) {
3055 radv_init_metadata(device
, memory
->image
, &metadata
);
3056 device
->ws
->buffer_set_metadata(memory
->bo
, &metadata
);
3059 return device
->ws
->buffer_get_fd(device
->ws
, memory
->bo
,
3063 static VkResult
radv_alloc_memory(struct radv_device
*device
,
3064 const VkMemoryAllocateInfo
* pAllocateInfo
,
3065 const VkAllocationCallbacks
* pAllocator
,
3066 VkDeviceMemory
* pMem
)
3068 struct radv_device_memory
*mem
;
3070 enum radeon_bo_domain domain
;
3072 enum radv_mem_type mem_type_index
= device
->physical_device
->mem_type_indices
[pAllocateInfo
->memoryTypeIndex
];
3074 assert(pAllocateInfo
->sType
== VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO
);
3076 if (pAllocateInfo
->allocationSize
== 0) {
3077 /* Apparently, this is allowed */
3078 *pMem
= VK_NULL_HANDLE
;
3082 const VkImportMemoryFdInfoKHR
*import_info
=
3083 vk_find_struct_const(pAllocateInfo
->pNext
, IMPORT_MEMORY_FD_INFO_KHR
);
3084 const VkMemoryDedicatedAllocateInfo
*dedicate_info
=
3085 vk_find_struct_const(pAllocateInfo
->pNext
, MEMORY_DEDICATED_ALLOCATE_INFO
);
3086 const VkExportMemoryAllocateInfo
*export_info
=
3087 vk_find_struct_const(pAllocateInfo
->pNext
, EXPORT_MEMORY_ALLOCATE_INFO
);
3088 const VkImportMemoryHostPointerInfoEXT
*host_ptr_info
=
3089 vk_find_struct_const(pAllocateInfo
->pNext
, IMPORT_MEMORY_HOST_POINTER_INFO_EXT
);
3091 const struct wsi_memory_allocate_info
*wsi_info
=
3092 vk_find_struct_const(pAllocateInfo
->pNext
, WSI_MEMORY_ALLOCATE_INFO_MESA
);
3094 mem
= vk_alloc2(&device
->alloc
, pAllocator
, sizeof(*mem
), 8,
3095 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
3097 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3099 if (wsi_info
&& wsi_info
->implicit_sync
)
3100 flags
|= RADEON_FLAG_IMPLICIT_SYNC
;
3102 if (dedicate_info
) {
3103 mem
->image
= radv_image_from_handle(dedicate_info
->image
);
3104 mem
->buffer
= radv_buffer_from_handle(dedicate_info
->buffer
);
3110 float priority_float
= 0.5;
3111 const struct VkMemoryPriorityAllocateInfoEXT
*priority_ext
=
3112 vk_find_struct_const(pAllocateInfo
->pNext
,
3113 MEMORY_PRIORITY_ALLOCATE_INFO_EXT
);
3115 priority_float
= priority_ext
->priority
;
3117 unsigned priority
= MIN2(RADV_BO_PRIORITY_APPLICATION_MAX
- 1,
3118 (int)(priority_float
* RADV_BO_PRIORITY_APPLICATION_MAX
));
3120 mem
->user_ptr
= NULL
;
3123 assert(import_info
->handleType
==
3124 VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT
||
3125 import_info
->handleType
==
3126 VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT
);
3127 mem
->bo
= device
->ws
->buffer_from_fd(device
->ws
, import_info
->fd
,
3128 priority
, NULL
, NULL
);
3130 result
= VK_ERROR_INVALID_EXTERNAL_HANDLE
;
3133 close(import_info
->fd
);
3135 } else if (host_ptr_info
) {
3136 assert(host_ptr_info
->handleType
== VK_EXTERNAL_MEMORY_HANDLE_TYPE_HOST_ALLOCATION_BIT_EXT
);
3137 assert(mem_type_index
== RADV_MEM_TYPE_GTT_CACHED
);
3138 mem
->bo
= device
->ws
->buffer_from_ptr(device
->ws
, host_ptr_info
->pHostPointer
,
3139 pAllocateInfo
->allocationSize
,
3142 result
= VK_ERROR_INVALID_EXTERNAL_HANDLE
;
3145 mem
->user_ptr
= host_ptr_info
->pHostPointer
;
3148 uint64_t alloc_size
= align_u64(pAllocateInfo
->allocationSize
, 4096);
3149 if (mem_type_index
== RADV_MEM_TYPE_GTT_WRITE_COMBINE
||
3150 mem_type_index
== RADV_MEM_TYPE_GTT_CACHED
)
3151 domain
= RADEON_DOMAIN_GTT
;
3153 domain
= RADEON_DOMAIN_VRAM
;
3155 if (mem_type_index
== RADV_MEM_TYPE_VRAM
)
3156 flags
|= RADEON_FLAG_NO_CPU_ACCESS
;
3158 flags
|= RADEON_FLAG_CPU_ACCESS
;
3160 if (mem_type_index
== RADV_MEM_TYPE_GTT_WRITE_COMBINE
)
3161 flags
|= RADEON_FLAG_GTT_WC
;
3163 if (!dedicate_info
&& !import_info
&& (!export_info
|| !export_info
->handleTypes
))
3164 flags
|= RADEON_FLAG_NO_INTERPROCESS_SHARING
;
3166 mem
->bo
= device
->ws
->buffer_create(device
->ws
, alloc_size
, device
->physical_device
->rad_info
.max_alignment
,
3167 domain
, flags
, priority
);
3170 result
= VK_ERROR_OUT_OF_DEVICE_MEMORY
;
3173 mem
->type_index
= mem_type_index
;
3176 result
= radv_bo_list_add(device
, mem
->bo
);
3177 if (result
!= VK_SUCCESS
)
3180 *pMem
= radv_device_memory_to_handle(mem
);
3185 device
->ws
->buffer_destroy(mem
->bo
);
3187 vk_free2(&device
->alloc
, pAllocator
, mem
);
3192 VkResult
radv_AllocateMemory(
3194 const VkMemoryAllocateInfo
* pAllocateInfo
,
3195 const VkAllocationCallbacks
* pAllocator
,
3196 VkDeviceMemory
* pMem
)
3198 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3199 return radv_alloc_memory(device
, pAllocateInfo
, pAllocator
, pMem
);
3202 void radv_FreeMemory(
3204 VkDeviceMemory _mem
,
3205 const VkAllocationCallbacks
* pAllocator
)
3207 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3208 RADV_FROM_HANDLE(radv_device_memory
, mem
, _mem
);
3213 radv_bo_list_remove(device
, mem
->bo
);
3214 device
->ws
->buffer_destroy(mem
->bo
);
3217 vk_free2(&device
->alloc
, pAllocator
, mem
);
3220 VkResult
radv_MapMemory(
3222 VkDeviceMemory _memory
,
3223 VkDeviceSize offset
,
3225 VkMemoryMapFlags flags
,
3228 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3229 RADV_FROM_HANDLE(radv_device_memory
, mem
, _memory
);
3237 *ppData
= mem
->user_ptr
;
3239 *ppData
= device
->ws
->buffer_map(mem
->bo
);
3246 return vk_error(device
->instance
, VK_ERROR_MEMORY_MAP_FAILED
);
3249 void radv_UnmapMemory(
3251 VkDeviceMemory _memory
)
3253 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3254 RADV_FROM_HANDLE(radv_device_memory
, mem
, _memory
);
3259 if (mem
->user_ptr
== NULL
)
3260 device
->ws
->buffer_unmap(mem
->bo
);
3263 VkResult
radv_FlushMappedMemoryRanges(
3265 uint32_t memoryRangeCount
,
3266 const VkMappedMemoryRange
* pMemoryRanges
)
3271 VkResult
radv_InvalidateMappedMemoryRanges(
3273 uint32_t memoryRangeCount
,
3274 const VkMappedMemoryRange
* pMemoryRanges
)
3279 void radv_GetBufferMemoryRequirements(
3282 VkMemoryRequirements
* pMemoryRequirements
)
3284 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3285 RADV_FROM_HANDLE(radv_buffer
, buffer
, _buffer
);
3287 pMemoryRequirements
->memoryTypeBits
= (1u << device
->physical_device
->memory_properties
.memoryTypeCount
) - 1;
3289 if (buffer
->flags
& VK_BUFFER_CREATE_SPARSE_BINDING_BIT
)
3290 pMemoryRequirements
->alignment
= 4096;
3292 pMemoryRequirements
->alignment
= 16;
3294 pMemoryRequirements
->size
= align64(buffer
->size
, pMemoryRequirements
->alignment
);
3297 void radv_GetBufferMemoryRequirements2(
3299 const VkBufferMemoryRequirementsInfo2
*pInfo
,
3300 VkMemoryRequirements2
*pMemoryRequirements
)
3302 radv_GetBufferMemoryRequirements(device
, pInfo
->buffer
,
3303 &pMemoryRequirements
->memoryRequirements
);
3304 RADV_FROM_HANDLE(radv_buffer
, buffer
, pInfo
->buffer
);
3305 vk_foreach_struct(ext
, pMemoryRequirements
->pNext
) {
3306 switch (ext
->sType
) {
3307 case VK_STRUCTURE_TYPE_MEMORY_DEDICATED_REQUIREMENTS
: {
3308 VkMemoryDedicatedRequirements
*req
=
3309 (VkMemoryDedicatedRequirements
*) ext
;
3310 req
->requiresDedicatedAllocation
= buffer
->shareable
;
3311 req
->prefersDedicatedAllocation
= req
->requiresDedicatedAllocation
;
3320 void radv_GetImageMemoryRequirements(
3323 VkMemoryRequirements
* pMemoryRequirements
)
3325 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3326 RADV_FROM_HANDLE(radv_image
, image
, _image
);
3328 pMemoryRequirements
->memoryTypeBits
= (1u << device
->physical_device
->memory_properties
.memoryTypeCount
) - 1;
3330 pMemoryRequirements
->size
= image
->size
;
3331 pMemoryRequirements
->alignment
= image
->alignment
;
3334 void radv_GetImageMemoryRequirements2(
3336 const VkImageMemoryRequirementsInfo2
*pInfo
,
3337 VkMemoryRequirements2
*pMemoryRequirements
)
3339 radv_GetImageMemoryRequirements(device
, pInfo
->image
,
3340 &pMemoryRequirements
->memoryRequirements
);
3342 RADV_FROM_HANDLE(radv_image
, image
, pInfo
->image
);
3344 vk_foreach_struct(ext
, pMemoryRequirements
->pNext
) {
3345 switch (ext
->sType
) {
3346 case VK_STRUCTURE_TYPE_MEMORY_DEDICATED_REQUIREMENTS
: {
3347 VkMemoryDedicatedRequirements
*req
=
3348 (VkMemoryDedicatedRequirements
*) ext
;
3349 req
->requiresDedicatedAllocation
= image
->shareable
;
3350 req
->prefersDedicatedAllocation
= req
->requiresDedicatedAllocation
;
3359 void radv_GetImageSparseMemoryRequirements(
3362 uint32_t* pSparseMemoryRequirementCount
,
3363 VkSparseImageMemoryRequirements
* pSparseMemoryRequirements
)
3368 void radv_GetImageSparseMemoryRequirements2(
3370 const VkImageSparseMemoryRequirementsInfo2
*pInfo
,
3371 uint32_t* pSparseMemoryRequirementCount
,
3372 VkSparseImageMemoryRequirements2
*pSparseMemoryRequirements
)
3377 void radv_GetDeviceMemoryCommitment(
3379 VkDeviceMemory memory
,
3380 VkDeviceSize
* pCommittedMemoryInBytes
)
3382 *pCommittedMemoryInBytes
= 0;
3385 VkResult
radv_BindBufferMemory2(VkDevice device
,
3386 uint32_t bindInfoCount
,
3387 const VkBindBufferMemoryInfo
*pBindInfos
)
3389 for (uint32_t i
= 0; i
< bindInfoCount
; ++i
) {
3390 RADV_FROM_HANDLE(radv_device_memory
, mem
, pBindInfos
[i
].memory
);
3391 RADV_FROM_HANDLE(radv_buffer
, buffer
, pBindInfos
[i
].buffer
);
3394 buffer
->bo
= mem
->bo
;
3395 buffer
->offset
= pBindInfos
[i
].memoryOffset
;
3403 VkResult
radv_BindBufferMemory(
3406 VkDeviceMemory memory
,
3407 VkDeviceSize memoryOffset
)
3409 const VkBindBufferMemoryInfo info
= {
3410 .sType
= VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO
,
3413 .memoryOffset
= memoryOffset
3416 return radv_BindBufferMemory2(device
, 1, &info
);
3419 VkResult
radv_BindImageMemory2(VkDevice device
,
3420 uint32_t bindInfoCount
,
3421 const VkBindImageMemoryInfo
*pBindInfos
)
3423 for (uint32_t i
= 0; i
< bindInfoCount
; ++i
) {
3424 RADV_FROM_HANDLE(radv_device_memory
, mem
, pBindInfos
[i
].memory
);
3425 RADV_FROM_HANDLE(radv_image
, image
, pBindInfos
[i
].image
);
3428 image
->bo
= mem
->bo
;
3429 image
->offset
= pBindInfos
[i
].memoryOffset
;
3439 VkResult
radv_BindImageMemory(
3442 VkDeviceMemory memory
,
3443 VkDeviceSize memoryOffset
)
3445 const VkBindImageMemoryInfo info
= {
3446 .sType
= VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO
,
3449 .memoryOffset
= memoryOffset
3452 return radv_BindImageMemory2(device
, 1, &info
);
3457 radv_sparse_buffer_bind_memory(struct radv_device
*device
,
3458 const VkSparseBufferMemoryBindInfo
*bind
)
3460 RADV_FROM_HANDLE(radv_buffer
, buffer
, bind
->buffer
);
3462 for (uint32_t i
= 0; i
< bind
->bindCount
; ++i
) {
3463 struct radv_device_memory
*mem
= NULL
;
3465 if (bind
->pBinds
[i
].memory
!= VK_NULL_HANDLE
)
3466 mem
= radv_device_memory_from_handle(bind
->pBinds
[i
].memory
);
3468 device
->ws
->buffer_virtual_bind(buffer
->bo
,
3469 bind
->pBinds
[i
].resourceOffset
,
3470 bind
->pBinds
[i
].size
,
3471 mem
? mem
->bo
: NULL
,
3472 bind
->pBinds
[i
].memoryOffset
);
3477 radv_sparse_image_opaque_bind_memory(struct radv_device
*device
,
3478 const VkSparseImageOpaqueMemoryBindInfo
*bind
)
3480 RADV_FROM_HANDLE(radv_image
, image
, bind
->image
);
3482 for (uint32_t i
= 0; i
< bind
->bindCount
; ++i
) {
3483 struct radv_device_memory
*mem
= NULL
;
3485 if (bind
->pBinds
[i
].memory
!= VK_NULL_HANDLE
)
3486 mem
= radv_device_memory_from_handle(bind
->pBinds
[i
].memory
);
3488 device
->ws
->buffer_virtual_bind(image
->bo
,
3489 bind
->pBinds
[i
].resourceOffset
,
3490 bind
->pBinds
[i
].size
,
3491 mem
? mem
->bo
: NULL
,
3492 bind
->pBinds
[i
].memoryOffset
);
3496 VkResult
radv_QueueBindSparse(
3498 uint32_t bindInfoCount
,
3499 const VkBindSparseInfo
* pBindInfo
,
3502 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
3503 RADV_FROM_HANDLE(radv_queue
, queue
, _queue
);
3504 struct radeon_winsys_fence
*base_fence
= fence
? fence
->fence
: NULL
;
3505 bool fence_emitted
= false;
3509 for (uint32_t i
= 0; i
< bindInfoCount
; ++i
) {
3510 struct radv_winsys_sem_info sem_info
;
3511 for (uint32_t j
= 0; j
< pBindInfo
[i
].bufferBindCount
; ++j
) {
3512 radv_sparse_buffer_bind_memory(queue
->device
,
3513 pBindInfo
[i
].pBufferBinds
+ j
);
3516 for (uint32_t j
= 0; j
< pBindInfo
[i
].imageOpaqueBindCount
; ++j
) {
3517 radv_sparse_image_opaque_bind_memory(queue
->device
,
3518 pBindInfo
[i
].pImageOpaqueBinds
+ j
);
3522 result
= radv_alloc_sem_info(queue
->device
->instance
,
3524 pBindInfo
[i
].waitSemaphoreCount
,
3525 pBindInfo
[i
].pWaitSemaphores
,
3526 pBindInfo
[i
].signalSemaphoreCount
,
3527 pBindInfo
[i
].pSignalSemaphores
,
3529 if (result
!= VK_SUCCESS
)
3532 if (pBindInfo
[i
].waitSemaphoreCount
|| pBindInfo
[i
].signalSemaphoreCount
) {
3533 ret
= queue
->device
->ws
->cs_submit(queue
->hw_ctx
, queue
->queue_idx
,
3534 &queue
->device
->empty_cs
[queue
->queue_family_index
],
3539 radv_loge("failed to submit CS %d\n", i
);
3543 fence_emitted
= true;
3545 fence
->submitted
= true;
3548 radv_free_sem_info(&sem_info
);
3553 if (!fence_emitted
) {
3554 result
= radv_signal_fence(queue
, fence
);
3555 if (result
!= VK_SUCCESS
)
3558 fence
->submitted
= true;
3564 VkResult
radv_CreateFence(
3566 const VkFenceCreateInfo
* pCreateInfo
,
3567 const VkAllocationCallbacks
* pAllocator
,
3570 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3571 const VkExportFenceCreateInfo
*export
=
3572 vk_find_struct_const(pCreateInfo
->pNext
, EXPORT_FENCE_CREATE_INFO
);
3573 VkExternalFenceHandleTypeFlags handleTypes
=
3574 export
? export
->handleTypes
: 0;
3576 struct radv_fence
*fence
= vk_alloc2(&device
->alloc
, pAllocator
,
3578 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
3581 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3583 fence
->fence_wsi
= NULL
;
3584 fence
->submitted
= false;
3585 fence
->signalled
= !!(pCreateInfo
->flags
& VK_FENCE_CREATE_SIGNALED_BIT
);
3586 fence
->temp_syncobj
= 0;
3587 if (device
->always_use_syncobj
|| handleTypes
) {
3588 int ret
= device
->ws
->create_syncobj(device
->ws
, &fence
->syncobj
);
3590 vk_free2(&device
->alloc
, pAllocator
, fence
);
3591 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3593 if (pCreateInfo
->flags
& VK_FENCE_CREATE_SIGNALED_BIT
) {
3594 device
->ws
->signal_syncobj(device
->ws
, fence
->syncobj
);
3596 fence
->fence
= NULL
;
3598 fence
->fence
= device
->ws
->create_fence();
3599 if (!fence
->fence
) {
3600 vk_free2(&device
->alloc
, pAllocator
, fence
);
3601 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3606 *pFence
= radv_fence_to_handle(fence
);
3611 void radv_DestroyFence(
3614 const VkAllocationCallbacks
* pAllocator
)
3616 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3617 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
3622 if (fence
->temp_syncobj
)
3623 device
->ws
->destroy_syncobj(device
->ws
, fence
->temp_syncobj
);
3625 device
->ws
->destroy_syncobj(device
->ws
, fence
->syncobj
);
3627 device
->ws
->destroy_fence(fence
->fence
);
3628 if (fence
->fence_wsi
)
3629 fence
->fence_wsi
->destroy(fence
->fence_wsi
);
3630 vk_free2(&device
->alloc
, pAllocator
, fence
);
3634 static uint64_t radv_get_current_time()
3637 clock_gettime(CLOCK_MONOTONIC
, &tv
);
3638 return tv
.tv_nsec
+ tv
.tv_sec
*1000000000ull;
3641 static uint64_t radv_get_absolute_timeout(uint64_t timeout
)
3643 uint64_t current_time
= radv_get_current_time();
3645 timeout
= MIN2(UINT64_MAX
- current_time
, timeout
);
3647 return current_time
+ timeout
;
3651 static bool radv_all_fences_plain_and_submitted(uint32_t fenceCount
, const VkFence
*pFences
)
3653 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3654 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3655 if (fence
->fence
== NULL
|| fence
->syncobj
||
3656 fence
->temp_syncobj
||
3657 (!fence
->signalled
&& !fence
->submitted
))
3663 static bool radv_all_fences_syncobj(uint32_t fenceCount
, const VkFence
*pFences
)
3665 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3666 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3667 if (fence
->syncobj
== 0 && fence
->temp_syncobj
== 0)
3673 VkResult
radv_WaitForFences(
3675 uint32_t fenceCount
,
3676 const VkFence
* pFences
,
3680 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3681 timeout
= radv_get_absolute_timeout(timeout
);
3683 if (device
->always_use_syncobj
&&
3684 radv_all_fences_syncobj(fenceCount
, pFences
))
3686 uint32_t *handles
= malloc(sizeof(uint32_t) * fenceCount
);
3688 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3690 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3691 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3692 handles
[i
] = fence
->temp_syncobj
? fence
->temp_syncobj
: fence
->syncobj
;
3695 bool success
= device
->ws
->wait_syncobj(device
->ws
, handles
, fenceCount
, waitAll
, timeout
);
3698 return success
? VK_SUCCESS
: VK_TIMEOUT
;
3701 if (!waitAll
&& fenceCount
> 1) {
3702 /* Not doing this by default for waitAll, due to needing to allocate twice. */
3703 if (device
->physical_device
->rad_info
.drm_minor
>= 10 && radv_all_fences_plain_and_submitted(fenceCount
, pFences
)) {
3704 uint32_t wait_count
= 0;
3705 struct radeon_winsys_fence
**fences
= malloc(sizeof(struct radeon_winsys_fence
*) * fenceCount
);
3707 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3709 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3710 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3712 if (fence
->signalled
) {
3717 fences
[wait_count
++] = fence
->fence
;
3720 bool success
= device
->ws
->fences_wait(device
->ws
, fences
, wait_count
,
3721 waitAll
, timeout
- radv_get_current_time());
3724 return success
? VK_SUCCESS
: VK_TIMEOUT
;
3727 while(radv_get_current_time() <= timeout
) {
3728 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3729 if (radv_GetFenceStatus(_device
, pFences
[i
]) == VK_SUCCESS
)
3736 for (uint32_t i
= 0; i
< fenceCount
; ++i
) {
3737 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3738 bool expired
= false;
3740 if (fence
->temp_syncobj
) {
3741 if (!device
->ws
->wait_syncobj(device
->ws
, &fence
->temp_syncobj
, 1, true, timeout
))
3746 if (fence
->syncobj
) {
3747 if (!device
->ws
->wait_syncobj(device
->ws
, &fence
->syncobj
, 1, true, timeout
))
3752 if (fence
->signalled
)
3756 if (!fence
->submitted
) {
3757 while(radv_get_current_time() <= timeout
&&
3761 if (!fence
->submitted
)
3764 /* Recheck as it may have been set by
3765 * submitting operations. */
3767 if (fence
->signalled
)
3771 expired
= device
->ws
->fence_wait(device
->ws
,
3778 if (fence
->fence_wsi
) {
3779 VkResult result
= fence
->fence_wsi
->wait(fence
->fence_wsi
, timeout
);
3780 if (result
!= VK_SUCCESS
)
3784 fence
->signalled
= true;
3790 VkResult
radv_ResetFences(VkDevice _device
,
3791 uint32_t fenceCount
,
3792 const VkFence
*pFences
)
3794 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3796 for (unsigned i
= 0; i
< fenceCount
; ++i
) {
3797 RADV_FROM_HANDLE(radv_fence
, fence
, pFences
[i
]);
3798 fence
->submitted
= fence
->signalled
= false;
3800 /* Per spec, we first restore the permanent payload, and then reset, so
3801 * having a temp syncobj should not skip resetting the permanent syncobj. */
3802 if (fence
->temp_syncobj
) {
3803 device
->ws
->destroy_syncobj(device
->ws
, fence
->temp_syncobj
);
3804 fence
->temp_syncobj
= 0;
3807 if (fence
->syncobj
) {
3808 device
->ws
->reset_syncobj(device
->ws
, fence
->syncobj
);
3815 VkResult
radv_GetFenceStatus(VkDevice _device
, VkFence _fence
)
3817 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3818 RADV_FROM_HANDLE(radv_fence
, fence
, _fence
);
3820 if (fence
->temp_syncobj
) {
3821 bool success
= device
->ws
->wait_syncobj(device
->ws
, &fence
->temp_syncobj
, 1, true, 0);
3822 return success
? VK_SUCCESS
: VK_NOT_READY
;
3825 if (fence
->syncobj
) {
3826 bool success
= device
->ws
->wait_syncobj(device
->ws
, &fence
->syncobj
, 1, true, 0);
3827 return success
? VK_SUCCESS
: VK_NOT_READY
;
3830 if (fence
->signalled
)
3832 if (!fence
->submitted
)
3833 return VK_NOT_READY
;
3835 if (!device
->ws
->fence_wait(device
->ws
, fence
->fence
, false, 0))
3836 return VK_NOT_READY
;
3838 if (fence
->fence_wsi
) {
3839 VkResult result
= fence
->fence_wsi
->wait(fence
->fence_wsi
, 0);
3841 if (result
!= VK_SUCCESS
) {
3842 if (result
== VK_TIMEOUT
)
3843 return VK_NOT_READY
;
3851 // Queue semaphore functions
3853 VkResult
radv_CreateSemaphore(
3855 const VkSemaphoreCreateInfo
* pCreateInfo
,
3856 const VkAllocationCallbacks
* pAllocator
,
3857 VkSemaphore
* pSemaphore
)
3859 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3860 const VkExportSemaphoreCreateInfo
*export
=
3861 vk_find_struct_const(pCreateInfo
->pNext
, EXPORT_SEMAPHORE_CREATE_INFO
);
3862 VkExternalSemaphoreHandleTypeFlags handleTypes
=
3863 export
? export
->handleTypes
: 0;
3865 struct radv_semaphore
*sem
= vk_alloc2(&device
->alloc
, pAllocator
,
3867 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
3869 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3871 sem
->temp_syncobj
= 0;
3872 /* create a syncobject if we are going to export this semaphore */
3873 if (device
->always_use_syncobj
|| handleTypes
) {
3874 assert (device
->physical_device
->rad_info
.has_syncobj
);
3875 int ret
= device
->ws
->create_syncobj(device
->ws
, &sem
->syncobj
);
3877 vk_free2(&device
->alloc
, pAllocator
, sem
);
3878 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3882 sem
->sem
= device
->ws
->create_sem(device
->ws
);
3884 vk_free2(&device
->alloc
, pAllocator
, sem
);
3885 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3890 *pSemaphore
= radv_semaphore_to_handle(sem
);
3894 void radv_DestroySemaphore(
3896 VkSemaphore _semaphore
,
3897 const VkAllocationCallbacks
* pAllocator
)
3899 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3900 RADV_FROM_HANDLE(radv_semaphore
, sem
, _semaphore
);
3905 device
->ws
->destroy_syncobj(device
->ws
, sem
->syncobj
);
3907 device
->ws
->destroy_sem(sem
->sem
);
3908 vk_free2(&device
->alloc
, pAllocator
, sem
);
3911 VkResult
radv_CreateEvent(
3913 const VkEventCreateInfo
* pCreateInfo
,
3914 const VkAllocationCallbacks
* pAllocator
,
3917 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3918 struct radv_event
*event
= vk_alloc2(&device
->alloc
, pAllocator
,
3920 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
3923 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
3925 event
->bo
= device
->ws
->buffer_create(device
->ws
, 8, 8,
3927 RADEON_FLAG_VA_UNCACHED
| RADEON_FLAG_CPU_ACCESS
| RADEON_FLAG_NO_INTERPROCESS_SHARING
,
3928 RADV_BO_PRIORITY_FENCE
);
3930 vk_free2(&device
->alloc
, pAllocator
, event
);
3931 return vk_error(device
->instance
, VK_ERROR_OUT_OF_DEVICE_MEMORY
);
3934 event
->map
= (uint64_t*)device
->ws
->buffer_map(event
->bo
);
3936 *pEvent
= radv_event_to_handle(event
);
3941 void radv_DestroyEvent(
3944 const VkAllocationCallbacks
* pAllocator
)
3946 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3947 RADV_FROM_HANDLE(radv_event
, event
, _event
);
3951 device
->ws
->buffer_destroy(event
->bo
);
3952 vk_free2(&device
->alloc
, pAllocator
, event
);
3955 VkResult
radv_GetEventStatus(
3959 RADV_FROM_HANDLE(radv_event
, event
, _event
);
3961 if (*event
->map
== 1)
3962 return VK_EVENT_SET
;
3963 return VK_EVENT_RESET
;
3966 VkResult
radv_SetEvent(
3970 RADV_FROM_HANDLE(radv_event
, event
, _event
);
3976 VkResult
radv_ResetEvent(
3980 RADV_FROM_HANDLE(radv_event
, event
, _event
);
3986 VkResult
radv_CreateBuffer(
3988 const VkBufferCreateInfo
* pCreateInfo
,
3989 const VkAllocationCallbacks
* pAllocator
,
3992 RADV_FROM_HANDLE(radv_device
, device
, _device
);
3993 struct radv_buffer
*buffer
;
3995 assert(pCreateInfo
->sType
== VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO
);
3997 buffer
= vk_alloc2(&device
->alloc
, pAllocator
, sizeof(*buffer
), 8,
3998 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
4000 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
4002 buffer
->size
= pCreateInfo
->size
;
4003 buffer
->usage
= pCreateInfo
->usage
;
4006 buffer
->flags
= pCreateInfo
->flags
;
4008 buffer
->shareable
= vk_find_struct_const(pCreateInfo
->pNext
,
4009 EXTERNAL_MEMORY_BUFFER_CREATE_INFO
) != NULL
;
4011 if (pCreateInfo
->flags
& VK_BUFFER_CREATE_SPARSE_BINDING_BIT
) {
4012 buffer
->bo
= device
->ws
->buffer_create(device
->ws
,
4013 align64(buffer
->size
, 4096),
4014 4096, 0, RADEON_FLAG_VIRTUAL
,
4015 RADV_BO_PRIORITY_VIRTUAL
);
4017 vk_free2(&device
->alloc
, pAllocator
, buffer
);
4018 return vk_error(device
->instance
, VK_ERROR_OUT_OF_DEVICE_MEMORY
);
4022 *pBuffer
= radv_buffer_to_handle(buffer
);
4027 void radv_DestroyBuffer(
4030 const VkAllocationCallbacks
* pAllocator
)
4032 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4033 RADV_FROM_HANDLE(radv_buffer
, buffer
, _buffer
);
4038 if (buffer
->flags
& VK_BUFFER_CREATE_SPARSE_BINDING_BIT
)
4039 device
->ws
->buffer_destroy(buffer
->bo
);
4041 vk_free2(&device
->alloc
, pAllocator
, buffer
);
4044 VkDeviceAddress
radv_GetBufferDeviceAddressEXT(
4046 const VkBufferDeviceAddressInfoEXT
* pInfo
)
4048 RADV_FROM_HANDLE(radv_buffer
, buffer
, pInfo
->buffer
);
4049 return radv_buffer_get_va(buffer
->bo
) + buffer
->offset
;
4053 static inline unsigned
4054 si_tile_mode_index(const struct radv_image
*image
, unsigned level
, bool stencil
)
4057 return image
->surface
.u
.legacy
.stencil_tiling_index
[level
];
4059 return image
->surface
.u
.legacy
.tiling_index
[level
];
4062 static uint32_t radv_surface_max_layer_count(struct radv_image_view
*iview
)
4064 return iview
->type
== VK_IMAGE_VIEW_TYPE_3D
? iview
->extent
.depth
: (iview
->base_layer
+ iview
->layer_count
);
4068 radv_init_dcc_control_reg(struct radv_device
*device
,
4069 struct radv_image_view
*iview
)
4071 unsigned max_uncompressed_block_size
= V_028C78_MAX_BLOCK_SIZE_256B
;
4072 unsigned min_compressed_block_size
= V_028C78_MIN_BLOCK_SIZE_32B
;
4073 unsigned max_compressed_block_size
;
4074 unsigned independent_64b_blocks
;
4076 if (!radv_image_has_dcc(iview
->image
))
4079 if (iview
->image
->info
.samples
> 1) {
4080 if (iview
->image
->surface
.bpe
== 1)
4081 max_uncompressed_block_size
= V_028C78_MAX_BLOCK_SIZE_64B
;
4082 else if (iview
->image
->surface
.bpe
== 2)
4083 max_uncompressed_block_size
= V_028C78_MAX_BLOCK_SIZE_128B
;
4086 if (!device
->physical_device
->rad_info
.has_dedicated_vram
) {
4087 /* amdvlk: [min-compressed-block-size] should be set to 32 for
4088 * dGPU and 64 for APU because all of our APUs to date use
4089 * DIMMs which have a request granularity size of 64B while all
4090 * other chips have a 32B request size.
4092 min_compressed_block_size
= V_028C78_MIN_BLOCK_SIZE_64B
;
4095 if (iview
->image
->usage
& (VK_IMAGE_USAGE_SAMPLED_BIT
|
4096 VK_IMAGE_USAGE_TRANSFER_SRC_BIT
|
4097 VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT
)) {
4098 /* If this DCC image is potentially going to be used in texture
4099 * fetches, we need some special settings.
4101 independent_64b_blocks
= 1;
4102 max_compressed_block_size
= V_028C78_MAX_BLOCK_SIZE_64B
;
4104 /* MAX_UNCOMPRESSED_BLOCK_SIZE must be >=
4105 * MAX_COMPRESSED_BLOCK_SIZE. Set MAX_COMPRESSED_BLOCK_SIZE as
4106 * big as possible for better compression state.
4108 independent_64b_blocks
= 0;
4109 max_compressed_block_size
= max_uncompressed_block_size
;
4112 return S_028C78_MAX_UNCOMPRESSED_BLOCK_SIZE(max_uncompressed_block_size
) |
4113 S_028C78_MAX_COMPRESSED_BLOCK_SIZE(max_compressed_block_size
) |
4114 S_028C78_MIN_COMPRESSED_BLOCK_SIZE(min_compressed_block_size
) |
4115 S_028C78_INDEPENDENT_64B_BLOCKS(independent_64b_blocks
);
4119 radv_initialise_color_surface(struct radv_device
*device
,
4120 struct radv_color_buffer_info
*cb
,
4121 struct radv_image_view
*iview
)
4123 const struct vk_format_description
*desc
;
4124 unsigned ntype
, format
, swap
, endian
;
4125 unsigned blend_clamp
= 0, blend_bypass
= 0;
4127 const struct radeon_surf
*surf
= &iview
->image
->surface
;
4129 desc
= vk_format_description(iview
->vk_format
);
4131 memset(cb
, 0, sizeof(*cb
));
4133 /* Intensity is implemented as Red, so treat it that way. */
4134 cb
->cb_color_attrib
= S_028C74_FORCE_DST_ALPHA_1(desc
->swizzle
[3] == VK_SWIZZLE_1
);
4136 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
;
4138 cb
->cb_color_base
= va
>> 8;
4140 if (device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
4141 struct gfx9_surf_meta_flags meta
;
4142 if (iview
->image
->dcc_offset
)
4143 meta
= iview
->image
->surface
.u
.gfx9
.dcc
;
4145 meta
= iview
->image
->surface
.u
.gfx9
.cmask
;
4147 cb
->cb_color_attrib
|= S_028C74_COLOR_SW_MODE(iview
->image
->surface
.u
.gfx9
.surf
.swizzle_mode
) |
4148 S_028C74_FMASK_SW_MODE(iview
->image
->surface
.u
.gfx9
.fmask
.swizzle_mode
) |
4149 S_028C74_RB_ALIGNED(meta
.rb_aligned
) |
4150 S_028C74_PIPE_ALIGNED(meta
.pipe_aligned
);
4152 cb
->cb_color_base
+= iview
->image
->surface
.u
.gfx9
.surf_offset
>> 8;
4153 cb
->cb_color_base
|= iview
->image
->surface
.tile_swizzle
;
4155 const struct legacy_surf_level
*level_info
= &surf
->u
.legacy
.level
[iview
->base_mip
];
4156 unsigned pitch_tile_max
, slice_tile_max
, tile_mode_index
;
4158 cb
->cb_color_base
+= level_info
->offset
>> 8;
4159 if (level_info
->mode
== RADEON_SURF_MODE_2D
)
4160 cb
->cb_color_base
|= iview
->image
->surface
.tile_swizzle
;
4162 pitch_tile_max
= level_info
->nblk_x
/ 8 - 1;
4163 slice_tile_max
= (level_info
->nblk_x
* level_info
->nblk_y
) / 64 - 1;
4164 tile_mode_index
= si_tile_mode_index(iview
->image
, iview
->base_mip
, false);
4166 cb
->cb_color_pitch
= S_028C64_TILE_MAX(pitch_tile_max
);
4167 cb
->cb_color_slice
= S_028C68_TILE_MAX(slice_tile_max
);
4168 cb
->cb_color_cmask_slice
= iview
->image
->cmask
.slice_tile_max
;
4170 cb
->cb_color_attrib
|= S_028C74_TILE_MODE_INDEX(tile_mode_index
);
4172 if (radv_image_has_fmask(iview
->image
)) {
4173 if (device
->physical_device
->rad_info
.chip_class
>= CIK
)
4174 cb
->cb_color_pitch
|= S_028C64_FMASK_TILE_MAX(iview
->image
->fmask
.pitch_in_pixels
/ 8 - 1);
4175 cb
->cb_color_attrib
|= S_028C74_FMASK_TILE_MODE_INDEX(iview
->image
->fmask
.tile_mode_index
);
4176 cb
->cb_color_fmask_slice
= S_028C88_TILE_MAX(iview
->image
->fmask
.slice_tile_max
);
4178 /* This must be set for fast clear to work without FMASK. */
4179 if (device
->physical_device
->rad_info
.chip_class
>= CIK
)
4180 cb
->cb_color_pitch
|= S_028C64_FMASK_TILE_MAX(pitch_tile_max
);
4181 cb
->cb_color_attrib
|= S_028C74_FMASK_TILE_MODE_INDEX(tile_mode_index
);
4182 cb
->cb_color_fmask_slice
= S_028C88_TILE_MAX(slice_tile_max
);
4186 /* CMASK variables */
4187 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
;
4188 va
+= iview
->image
->cmask
.offset
;
4189 cb
->cb_color_cmask
= va
>> 8;
4191 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
;
4192 va
+= iview
->image
->dcc_offset
;
4193 cb
->cb_dcc_base
= va
>> 8;
4194 cb
->cb_dcc_base
|= iview
->image
->surface
.tile_swizzle
;
4196 uint32_t max_slice
= radv_surface_max_layer_count(iview
) - 1;
4197 cb
->cb_color_view
= S_028C6C_SLICE_START(iview
->base_layer
) |
4198 S_028C6C_SLICE_MAX(max_slice
);
4200 if (iview
->image
->info
.samples
> 1) {
4201 unsigned log_samples
= util_logbase2(iview
->image
->info
.samples
);
4203 cb
->cb_color_attrib
|= S_028C74_NUM_SAMPLES(log_samples
) |
4204 S_028C74_NUM_FRAGMENTS(log_samples
);
4207 if (radv_image_has_fmask(iview
->image
)) {
4208 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
+ iview
->image
->fmask
.offset
;
4209 cb
->cb_color_fmask
= va
>> 8;
4210 cb
->cb_color_fmask
|= iview
->image
->fmask
.tile_swizzle
;
4212 cb
->cb_color_fmask
= cb
->cb_color_base
;
4215 ntype
= radv_translate_color_numformat(iview
->vk_format
,
4217 vk_format_get_first_non_void_channel(iview
->vk_format
));
4218 format
= radv_translate_colorformat(iview
->vk_format
);
4219 if (format
== V_028C70_COLOR_INVALID
|| ntype
== ~0u)
4220 radv_finishme("Illegal color\n");
4221 swap
= radv_translate_colorswap(iview
->vk_format
, FALSE
);
4222 endian
= radv_colorformat_endian_swap(format
);
4224 /* blend clamp should be set for all NORM/SRGB types */
4225 if (ntype
== V_028C70_NUMBER_UNORM
||
4226 ntype
== V_028C70_NUMBER_SNORM
||
4227 ntype
== V_028C70_NUMBER_SRGB
)
4230 /* set blend bypass according to docs if SINT/UINT or
4231 8/24 COLOR variants */
4232 if (ntype
== V_028C70_NUMBER_UINT
|| ntype
== V_028C70_NUMBER_SINT
||
4233 format
== V_028C70_COLOR_8_24
|| format
== V_028C70_COLOR_24_8
||
4234 format
== V_028C70_COLOR_X24_8_32_FLOAT
) {
4239 if ((ntype
== V_028C70_NUMBER_UINT
|| ntype
== V_028C70_NUMBER_SINT
) &&
4240 (format
== V_028C70_COLOR_8
||
4241 format
== V_028C70_COLOR_8_8
||
4242 format
== V_028C70_COLOR_8_8_8_8
))
4243 ->color_is_int8
= true;
4245 cb
->cb_color_info
= S_028C70_FORMAT(format
) |
4246 S_028C70_COMP_SWAP(swap
) |
4247 S_028C70_BLEND_CLAMP(blend_clamp
) |
4248 S_028C70_BLEND_BYPASS(blend_bypass
) |
4249 S_028C70_SIMPLE_FLOAT(1) |
4250 S_028C70_ROUND_MODE(ntype
!= V_028C70_NUMBER_UNORM
&&
4251 ntype
!= V_028C70_NUMBER_SNORM
&&
4252 ntype
!= V_028C70_NUMBER_SRGB
&&
4253 format
!= V_028C70_COLOR_8_24
&&
4254 format
!= V_028C70_COLOR_24_8
) |
4255 S_028C70_NUMBER_TYPE(ntype
) |
4256 S_028C70_ENDIAN(endian
);
4257 if (radv_image_has_fmask(iview
->image
)) {
4258 cb
->cb_color_info
|= S_028C70_COMPRESSION(1);
4259 if (device
->physical_device
->rad_info
.chip_class
== SI
) {
4260 unsigned fmask_bankh
= util_logbase2(iview
->image
->fmask
.bank_height
);
4261 cb
->cb_color_attrib
|= S_028C74_FMASK_BANK_HEIGHT(fmask_bankh
);
4265 if (radv_image_has_cmask(iview
->image
) &&
4266 !(device
->instance
->debug_flags
& RADV_DEBUG_NO_FAST_CLEARS
))
4267 cb
->cb_color_info
|= S_028C70_FAST_CLEAR(1);
4269 if (radv_dcc_enabled(iview
->image
, iview
->base_mip
))
4270 cb
->cb_color_info
|= S_028C70_DCC_ENABLE(1);
4272 cb
->cb_dcc_control
= radv_init_dcc_control_reg(device
, iview
);
4274 /* This must be set for fast clear to work without FMASK. */
4275 if (!radv_image_has_fmask(iview
->image
) &&
4276 device
->physical_device
->rad_info
.chip_class
== SI
) {
4277 unsigned bankh
= util_logbase2(iview
->image
->surface
.u
.legacy
.bankh
);
4278 cb
->cb_color_attrib
|= S_028C74_FMASK_BANK_HEIGHT(bankh
);
4281 if (device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
4282 unsigned mip0_depth
= iview
->image
->type
== VK_IMAGE_TYPE_3D
?
4283 (iview
->extent
.depth
- 1) : (iview
->image
->info
.array_size
- 1);
4285 cb
->cb_color_view
|= S_028C6C_MIP_LEVEL(iview
->base_mip
);
4286 cb
->cb_color_attrib
|= S_028C74_MIP0_DEPTH(mip0_depth
) |
4287 S_028C74_RESOURCE_TYPE(iview
->image
->surface
.u
.gfx9
.resource_type
);
4288 cb
->cb_color_attrib2
= S_028C68_MIP0_WIDTH(iview
->extent
.width
- 1) |
4289 S_028C68_MIP0_HEIGHT(iview
->extent
.height
- 1) |
4290 S_028C68_MAX_MIP(iview
->image
->info
.levels
- 1);
4295 radv_calc_decompress_on_z_planes(struct radv_device
*device
,
4296 struct radv_image_view
*iview
)
4298 unsigned max_zplanes
= 0;
4300 assert(radv_image_is_tc_compat_htile(iview
->image
));
4302 if (device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
4303 /* Default value for 32-bit depth surfaces. */
4306 if (iview
->vk_format
== VK_FORMAT_D16_UNORM
&&
4307 iview
->image
->info
.samples
> 1)
4310 max_zplanes
= max_zplanes
+ 1;
4312 if (iview
->vk_format
== VK_FORMAT_D16_UNORM
) {
4313 /* Do not enable Z plane compression for 16-bit depth
4314 * surfaces because isn't supported on GFX8. Only
4315 * 32-bit depth surfaces are supported by the hardware.
4316 * This allows to maintain shader compatibility and to
4317 * reduce the number of depth decompressions.
4321 if (iview
->image
->info
.samples
<= 1)
4323 else if (iview
->image
->info
.samples
<= 4)
4334 radv_initialise_ds_surface(struct radv_device
*device
,
4335 struct radv_ds_buffer_info
*ds
,
4336 struct radv_image_view
*iview
)
4338 unsigned level
= iview
->base_mip
;
4339 unsigned format
, stencil_format
;
4340 uint64_t va
, s_offs
, z_offs
;
4341 bool stencil_only
= false;
4342 memset(ds
, 0, sizeof(*ds
));
4343 switch (iview
->image
->vk_format
) {
4344 case VK_FORMAT_D24_UNORM_S8_UINT
:
4345 case VK_FORMAT_X8_D24_UNORM_PACK32
:
4346 ds
->pa_su_poly_offset_db_fmt_cntl
= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(-24);
4347 ds
->offset_scale
= 2.0f
;
4349 case VK_FORMAT_D16_UNORM
:
4350 case VK_FORMAT_D16_UNORM_S8_UINT
:
4351 ds
->pa_su_poly_offset_db_fmt_cntl
= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(-16);
4352 ds
->offset_scale
= 4.0f
;
4354 case VK_FORMAT_D32_SFLOAT
:
4355 case VK_FORMAT_D32_SFLOAT_S8_UINT
:
4356 ds
->pa_su_poly_offset_db_fmt_cntl
= S_028B78_POLY_OFFSET_NEG_NUM_DB_BITS(-23) |
4357 S_028B78_POLY_OFFSET_DB_IS_FLOAT_FMT(1);
4358 ds
->offset_scale
= 1.0f
;
4360 case VK_FORMAT_S8_UINT
:
4361 stencil_only
= true;
4367 format
= radv_translate_dbformat(iview
->image
->vk_format
);
4368 stencil_format
= iview
->image
->surface
.has_stencil
?
4369 V_028044_STENCIL_8
: V_028044_STENCIL_INVALID
;
4371 uint32_t max_slice
= radv_surface_max_layer_count(iview
) - 1;
4372 ds
->db_depth_view
= S_028008_SLICE_START(iview
->base_layer
) |
4373 S_028008_SLICE_MAX(max_slice
);
4375 ds
->db_htile_data_base
= 0;
4376 ds
->db_htile_surface
= 0;
4378 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
;
4379 s_offs
= z_offs
= va
;
4381 if (device
->physical_device
->rad_info
.chip_class
>= GFX9
) {
4382 assert(iview
->image
->surface
.u
.gfx9
.surf_offset
== 0);
4383 s_offs
+= iview
->image
->surface
.u
.gfx9
.stencil_offset
;
4385 ds
->db_z_info
= S_028038_FORMAT(format
) |
4386 S_028038_NUM_SAMPLES(util_logbase2(iview
->image
->info
.samples
)) |
4387 S_028038_SW_MODE(iview
->image
->surface
.u
.gfx9
.surf
.swizzle_mode
) |
4388 S_028038_MAXMIP(iview
->image
->info
.levels
- 1) |
4389 S_028038_ZRANGE_PRECISION(1);
4390 ds
->db_stencil_info
= S_02803C_FORMAT(stencil_format
) |
4391 S_02803C_SW_MODE(iview
->image
->surface
.u
.gfx9
.stencil
.swizzle_mode
);
4393 ds
->db_z_info2
= S_028068_EPITCH(iview
->image
->surface
.u
.gfx9
.surf
.epitch
);
4394 ds
->db_stencil_info2
= S_02806C_EPITCH(iview
->image
->surface
.u
.gfx9
.stencil
.epitch
);
4395 ds
->db_depth_view
|= S_028008_MIPID(level
);
4397 ds
->db_depth_size
= S_02801C_X_MAX(iview
->image
->info
.width
- 1) |
4398 S_02801C_Y_MAX(iview
->image
->info
.height
- 1);
4400 if (radv_htile_enabled(iview
->image
, level
)) {
4401 ds
->db_z_info
|= S_028038_TILE_SURFACE_ENABLE(1);
4403 if (radv_image_is_tc_compat_htile(iview
->image
)) {
4404 unsigned max_zplanes
=
4405 radv_calc_decompress_on_z_planes(device
, iview
);
4407 ds
->db_z_info
|= S_028038_DECOMPRESS_ON_N_ZPLANES(max_zplanes
) |
4408 S_028038_ITERATE_FLUSH(1);
4409 ds
->db_stencil_info
|= S_02803C_ITERATE_FLUSH(1);
4412 if (!iview
->image
->surface
.has_stencil
)
4413 /* Use all of the htile_buffer for depth if there's no stencil. */
4414 ds
->db_stencil_info
|= S_02803C_TILE_STENCIL_DISABLE(1);
4415 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
+
4416 iview
->image
->htile_offset
;
4417 ds
->db_htile_data_base
= va
>> 8;
4418 ds
->db_htile_surface
= S_028ABC_FULL_CACHE(1) |
4419 S_028ABC_PIPE_ALIGNED(iview
->image
->surface
.u
.gfx9
.htile
.pipe_aligned
) |
4420 S_028ABC_RB_ALIGNED(iview
->image
->surface
.u
.gfx9
.htile
.rb_aligned
);
4423 const struct legacy_surf_level
*level_info
= &iview
->image
->surface
.u
.legacy
.level
[level
];
4426 level_info
= &iview
->image
->surface
.u
.legacy
.stencil_level
[level
];
4428 z_offs
+= iview
->image
->surface
.u
.legacy
.level
[level
].offset
;
4429 s_offs
+= iview
->image
->surface
.u
.legacy
.stencil_level
[level
].offset
;
4431 ds
->db_depth_info
= S_02803C_ADDR5_SWIZZLE_MASK(!radv_image_is_tc_compat_htile(iview
->image
));
4432 ds
->db_z_info
= S_028040_FORMAT(format
) | S_028040_ZRANGE_PRECISION(1);
4433 ds
->db_stencil_info
= S_028044_FORMAT(stencil_format
);
4435 if (iview
->image
->info
.samples
> 1)
4436 ds
->db_z_info
|= S_028040_NUM_SAMPLES(util_logbase2(iview
->image
->info
.samples
));
4438 if (device
->physical_device
->rad_info
.chip_class
>= CIK
) {
4439 struct radeon_info
*info
= &device
->physical_device
->rad_info
;
4440 unsigned tiling_index
= iview
->image
->surface
.u
.legacy
.tiling_index
[level
];
4441 unsigned stencil_index
= iview
->image
->surface
.u
.legacy
.stencil_tiling_index
[level
];
4442 unsigned macro_index
= iview
->image
->surface
.u
.legacy
.macro_tile_index
;
4443 unsigned tile_mode
= info
->si_tile_mode_array
[tiling_index
];
4444 unsigned stencil_tile_mode
= info
->si_tile_mode_array
[stencil_index
];
4445 unsigned macro_mode
= info
->cik_macrotile_mode_array
[macro_index
];
4448 tile_mode
= stencil_tile_mode
;
4450 ds
->db_depth_info
|=
4451 S_02803C_ARRAY_MODE(G_009910_ARRAY_MODE(tile_mode
)) |
4452 S_02803C_PIPE_CONFIG(G_009910_PIPE_CONFIG(tile_mode
)) |
4453 S_02803C_BANK_WIDTH(G_009990_BANK_WIDTH(macro_mode
)) |
4454 S_02803C_BANK_HEIGHT(G_009990_BANK_HEIGHT(macro_mode
)) |
4455 S_02803C_MACRO_TILE_ASPECT(G_009990_MACRO_TILE_ASPECT(macro_mode
)) |
4456 S_02803C_NUM_BANKS(G_009990_NUM_BANKS(macro_mode
));
4457 ds
->db_z_info
|= S_028040_TILE_SPLIT(G_009910_TILE_SPLIT(tile_mode
));
4458 ds
->db_stencil_info
|= S_028044_TILE_SPLIT(G_009910_TILE_SPLIT(stencil_tile_mode
));
4460 unsigned tile_mode_index
= si_tile_mode_index(iview
->image
, level
, false);
4461 ds
->db_z_info
|= S_028040_TILE_MODE_INDEX(tile_mode_index
);
4462 tile_mode_index
= si_tile_mode_index(iview
->image
, level
, true);
4463 ds
->db_stencil_info
|= S_028044_TILE_MODE_INDEX(tile_mode_index
);
4465 ds
->db_z_info
|= S_028040_TILE_MODE_INDEX(tile_mode_index
);
4468 ds
->db_depth_size
= S_028058_PITCH_TILE_MAX((level_info
->nblk_x
/ 8) - 1) |
4469 S_028058_HEIGHT_TILE_MAX((level_info
->nblk_y
/ 8) - 1);
4470 ds
->db_depth_slice
= S_02805C_SLICE_TILE_MAX((level_info
->nblk_x
* level_info
->nblk_y
) / 64 - 1);
4472 if (radv_htile_enabled(iview
->image
, level
)) {
4473 ds
->db_z_info
|= S_028040_TILE_SURFACE_ENABLE(1);
4475 if (!iview
->image
->surface
.has_stencil
&&
4476 !radv_image_is_tc_compat_htile(iview
->image
))
4477 /* Use all of the htile_buffer for depth if there's no stencil. */
4478 ds
->db_stencil_info
|= S_028044_TILE_STENCIL_DISABLE(1);
4480 va
= radv_buffer_get_va(iview
->bo
) + iview
->image
->offset
+
4481 iview
->image
->htile_offset
;
4482 ds
->db_htile_data_base
= va
>> 8;
4483 ds
->db_htile_surface
= S_028ABC_FULL_CACHE(1);
4485 if (radv_image_is_tc_compat_htile(iview
->image
)) {
4486 unsigned max_zplanes
=
4487 radv_calc_decompress_on_z_planes(device
, iview
);
4489 ds
->db_htile_surface
|= S_028ABC_TC_COMPATIBLE(1);
4490 ds
->db_z_info
|= S_028040_DECOMPRESS_ON_N_ZPLANES(max_zplanes
);
4495 ds
->db_z_read_base
= ds
->db_z_write_base
= z_offs
>> 8;
4496 ds
->db_stencil_read_base
= ds
->db_stencil_write_base
= s_offs
>> 8;
4499 VkResult
radv_CreateFramebuffer(
4501 const VkFramebufferCreateInfo
* pCreateInfo
,
4502 const VkAllocationCallbacks
* pAllocator
,
4503 VkFramebuffer
* pFramebuffer
)
4505 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4506 struct radv_framebuffer
*framebuffer
;
4508 assert(pCreateInfo
->sType
== VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO
);
4510 size_t size
= sizeof(*framebuffer
) +
4511 sizeof(struct radv_attachment_info
) * pCreateInfo
->attachmentCount
;
4512 framebuffer
= vk_alloc2(&device
->alloc
, pAllocator
, size
, 8,
4513 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
4514 if (framebuffer
== NULL
)
4515 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
4517 framebuffer
->attachment_count
= pCreateInfo
->attachmentCount
;
4518 framebuffer
->width
= pCreateInfo
->width
;
4519 framebuffer
->height
= pCreateInfo
->height
;
4520 framebuffer
->layers
= pCreateInfo
->layers
;
4521 for (uint32_t i
= 0; i
< pCreateInfo
->attachmentCount
; i
++) {
4522 VkImageView _iview
= pCreateInfo
->pAttachments
[i
];
4523 struct radv_image_view
*iview
= radv_image_view_from_handle(_iview
);
4524 framebuffer
->attachments
[i
].attachment
= iview
;
4525 if (iview
->aspect_mask
& VK_IMAGE_ASPECT_COLOR_BIT
) {
4526 radv_initialise_color_surface(device
, &framebuffer
->attachments
[i
].cb
, iview
);
4527 } else if (iview
->aspect_mask
& (VK_IMAGE_ASPECT_DEPTH_BIT
| VK_IMAGE_ASPECT_STENCIL_BIT
)) {
4528 radv_initialise_ds_surface(device
, &framebuffer
->attachments
[i
].ds
, iview
);
4530 framebuffer
->width
= MIN2(framebuffer
->width
, iview
->extent
.width
);
4531 framebuffer
->height
= MIN2(framebuffer
->height
, iview
->extent
.height
);
4532 framebuffer
->layers
= MIN2(framebuffer
->layers
, radv_surface_max_layer_count(iview
));
4535 *pFramebuffer
= radv_framebuffer_to_handle(framebuffer
);
4539 void radv_DestroyFramebuffer(
4542 const VkAllocationCallbacks
* pAllocator
)
4544 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4545 RADV_FROM_HANDLE(radv_framebuffer
, fb
, _fb
);
4549 vk_free2(&device
->alloc
, pAllocator
, fb
);
4552 static unsigned radv_tex_wrap(VkSamplerAddressMode address_mode
)
4554 switch (address_mode
) {
4555 case VK_SAMPLER_ADDRESS_MODE_REPEAT
:
4556 return V_008F30_SQ_TEX_WRAP
;
4557 case VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT
:
4558 return V_008F30_SQ_TEX_MIRROR
;
4559 case VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE
:
4560 return V_008F30_SQ_TEX_CLAMP_LAST_TEXEL
;
4561 case VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER
:
4562 return V_008F30_SQ_TEX_CLAMP_BORDER
;
4563 case VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE
:
4564 return V_008F30_SQ_TEX_MIRROR_ONCE_LAST_TEXEL
;
4566 unreachable("illegal tex wrap mode");
4572 radv_tex_compare(VkCompareOp op
)
4575 case VK_COMPARE_OP_NEVER
:
4576 return V_008F30_SQ_TEX_DEPTH_COMPARE_NEVER
;
4577 case VK_COMPARE_OP_LESS
:
4578 return V_008F30_SQ_TEX_DEPTH_COMPARE_LESS
;
4579 case VK_COMPARE_OP_EQUAL
:
4580 return V_008F30_SQ_TEX_DEPTH_COMPARE_EQUAL
;
4581 case VK_COMPARE_OP_LESS_OR_EQUAL
:
4582 return V_008F30_SQ_TEX_DEPTH_COMPARE_LESSEQUAL
;
4583 case VK_COMPARE_OP_GREATER
:
4584 return V_008F30_SQ_TEX_DEPTH_COMPARE_GREATER
;
4585 case VK_COMPARE_OP_NOT_EQUAL
:
4586 return V_008F30_SQ_TEX_DEPTH_COMPARE_NOTEQUAL
;
4587 case VK_COMPARE_OP_GREATER_OR_EQUAL
:
4588 return V_008F30_SQ_TEX_DEPTH_COMPARE_GREATEREQUAL
;
4589 case VK_COMPARE_OP_ALWAYS
:
4590 return V_008F30_SQ_TEX_DEPTH_COMPARE_ALWAYS
;
4592 unreachable("illegal compare mode");
4598 radv_tex_filter(VkFilter filter
, unsigned max_ansio
)
4601 case VK_FILTER_NEAREST
:
4602 return (max_ansio
> 1 ? V_008F38_SQ_TEX_XY_FILTER_ANISO_POINT
:
4603 V_008F38_SQ_TEX_XY_FILTER_POINT
);
4604 case VK_FILTER_LINEAR
:
4605 return (max_ansio
> 1 ? V_008F38_SQ_TEX_XY_FILTER_ANISO_BILINEAR
:
4606 V_008F38_SQ_TEX_XY_FILTER_BILINEAR
);
4607 case VK_FILTER_CUBIC_IMG
:
4609 fprintf(stderr
, "illegal texture filter");
4615 radv_tex_mipfilter(VkSamplerMipmapMode mode
)
4618 case VK_SAMPLER_MIPMAP_MODE_NEAREST
:
4619 return V_008F38_SQ_TEX_Z_FILTER_POINT
;
4620 case VK_SAMPLER_MIPMAP_MODE_LINEAR
:
4621 return V_008F38_SQ_TEX_Z_FILTER_LINEAR
;
4623 return V_008F38_SQ_TEX_Z_FILTER_NONE
;
4628 radv_tex_bordercolor(VkBorderColor bcolor
)
4631 case VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK
:
4632 case VK_BORDER_COLOR_INT_TRANSPARENT_BLACK
:
4633 return V_008F3C_SQ_TEX_BORDER_COLOR_TRANS_BLACK
;
4634 case VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK
:
4635 case VK_BORDER_COLOR_INT_OPAQUE_BLACK
:
4636 return V_008F3C_SQ_TEX_BORDER_COLOR_OPAQUE_BLACK
;
4637 case VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE
:
4638 case VK_BORDER_COLOR_INT_OPAQUE_WHITE
:
4639 return V_008F3C_SQ_TEX_BORDER_COLOR_OPAQUE_WHITE
;
4647 radv_tex_aniso_filter(unsigned filter
)
4661 radv_tex_filter_mode(VkSamplerReductionModeEXT mode
)
4664 case VK_SAMPLER_REDUCTION_MODE_WEIGHTED_AVERAGE_EXT
:
4665 return V_008F30_SQ_IMG_FILTER_MODE_BLEND
;
4666 case VK_SAMPLER_REDUCTION_MODE_MIN_EXT
:
4667 return V_008F30_SQ_IMG_FILTER_MODE_MIN
;
4668 case VK_SAMPLER_REDUCTION_MODE_MAX_EXT
:
4669 return V_008F30_SQ_IMG_FILTER_MODE_MAX
;
4677 radv_get_max_anisotropy(struct radv_device
*device
,
4678 const VkSamplerCreateInfo
*pCreateInfo
)
4680 if (device
->force_aniso
>= 0)
4681 return device
->force_aniso
;
4683 if (pCreateInfo
->anisotropyEnable
&&
4684 pCreateInfo
->maxAnisotropy
> 1.0f
)
4685 return (uint32_t)pCreateInfo
->maxAnisotropy
;
4691 radv_init_sampler(struct radv_device
*device
,
4692 struct radv_sampler
*sampler
,
4693 const VkSamplerCreateInfo
*pCreateInfo
)
4695 uint32_t max_aniso
= radv_get_max_anisotropy(device
, pCreateInfo
);
4696 uint32_t max_aniso_ratio
= radv_tex_aniso_filter(max_aniso
);
4697 bool is_vi
= (device
->physical_device
->rad_info
.chip_class
>= VI
);
4698 unsigned filter_mode
= V_008F30_SQ_IMG_FILTER_MODE_BLEND
;
4700 const struct VkSamplerReductionModeCreateInfoEXT
*sampler_reduction
=
4701 vk_find_struct_const(pCreateInfo
->pNext
,
4702 SAMPLER_REDUCTION_MODE_CREATE_INFO_EXT
);
4703 if (sampler_reduction
)
4704 filter_mode
= radv_tex_filter_mode(sampler_reduction
->reductionMode
);
4706 sampler
->state
[0] = (S_008F30_CLAMP_X(radv_tex_wrap(pCreateInfo
->addressModeU
)) |
4707 S_008F30_CLAMP_Y(radv_tex_wrap(pCreateInfo
->addressModeV
)) |
4708 S_008F30_CLAMP_Z(radv_tex_wrap(pCreateInfo
->addressModeW
)) |
4709 S_008F30_MAX_ANISO_RATIO(max_aniso_ratio
) |
4710 S_008F30_DEPTH_COMPARE_FUNC(radv_tex_compare(pCreateInfo
->compareOp
)) |
4711 S_008F30_FORCE_UNNORMALIZED(pCreateInfo
->unnormalizedCoordinates
? 1 : 0) |
4712 S_008F30_ANISO_THRESHOLD(max_aniso_ratio
>> 1) |
4713 S_008F30_ANISO_BIAS(max_aniso_ratio
) |
4714 S_008F30_DISABLE_CUBE_WRAP(0) |
4715 S_008F30_COMPAT_MODE(is_vi
) |
4716 S_008F30_FILTER_MODE(filter_mode
));
4717 sampler
->state
[1] = (S_008F34_MIN_LOD(S_FIXED(CLAMP(pCreateInfo
->minLod
, 0, 15), 8)) |
4718 S_008F34_MAX_LOD(S_FIXED(CLAMP(pCreateInfo
->maxLod
, 0, 15), 8)) |
4719 S_008F34_PERF_MIP(max_aniso_ratio
? max_aniso_ratio
+ 6 : 0));
4720 sampler
->state
[2] = (S_008F38_LOD_BIAS(S_FIXED(CLAMP(pCreateInfo
->mipLodBias
, -16, 16), 8)) |
4721 S_008F38_XY_MAG_FILTER(radv_tex_filter(pCreateInfo
->magFilter
, max_aniso
)) |
4722 S_008F38_XY_MIN_FILTER(radv_tex_filter(pCreateInfo
->minFilter
, max_aniso
)) |
4723 S_008F38_MIP_FILTER(radv_tex_mipfilter(pCreateInfo
->mipmapMode
)) |
4724 S_008F38_MIP_POINT_PRECLAMP(0) |
4725 S_008F38_DISABLE_LSB_CEIL(device
->physical_device
->rad_info
.chip_class
<= VI
) |
4726 S_008F38_FILTER_PREC_FIX(1) |
4727 S_008F38_ANISO_OVERRIDE(is_vi
));
4728 sampler
->state
[3] = (S_008F3C_BORDER_COLOR_PTR(0) |
4729 S_008F3C_BORDER_COLOR_TYPE(radv_tex_bordercolor(pCreateInfo
->borderColor
)));
4732 VkResult
radv_CreateSampler(
4734 const VkSamplerCreateInfo
* pCreateInfo
,
4735 const VkAllocationCallbacks
* pAllocator
,
4736 VkSampler
* pSampler
)
4738 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4739 struct radv_sampler
*sampler
;
4741 assert(pCreateInfo
->sType
== VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO
);
4743 sampler
= vk_alloc2(&device
->alloc
, pAllocator
, sizeof(*sampler
), 8,
4744 VK_SYSTEM_ALLOCATION_SCOPE_OBJECT
);
4746 return vk_error(device
->instance
, VK_ERROR_OUT_OF_HOST_MEMORY
);
4748 radv_init_sampler(device
, sampler
, pCreateInfo
);
4749 *pSampler
= radv_sampler_to_handle(sampler
);
4754 void radv_DestroySampler(
4757 const VkAllocationCallbacks
* pAllocator
)
4759 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4760 RADV_FROM_HANDLE(radv_sampler
, sampler
, _sampler
);
4764 vk_free2(&device
->alloc
, pAllocator
, sampler
);
4767 /* vk_icd.h does not declare this function, so we declare it here to
4768 * suppress Wmissing-prototypes.
4770 PUBLIC VKAPI_ATTR VkResult VKAPI_CALL
4771 vk_icdNegotiateLoaderICDInterfaceVersion(uint32_t *pSupportedVersion
);
4773 PUBLIC VKAPI_ATTR VkResult VKAPI_CALL
4774 vk_icdNegotiateLoaderICDInterfaceVersion(uint32_t *pSupportedVersion
)
4776 /* For the full details on loader interface versioning, see
4777 * <https://github.com/KhronosGroup/Vulkan-LoaderAndValidationLayers/blob/master/loader/LoaderAndLayerInterface.md>.
4778 * What follows is a condensed summary, to help you navigate the large and
4779 * confusing official doc.
4781 * - Loader interface v0 is incompatible with later versions. We don't
4784 * - In loader interface v1:
4785 * - The first ICD entrypoint called by the loader is
4786 * vk_icdGetInstanceProcAddr(). The ICD must statically expose this
4788 * - The ICD must statically expose no other Vulkan symbol unless it is
4789 * linked with -Bsymbolic.
4790 * - Each dispatchable Vulkan handle created by the ICD must be
4791 * a pointer to a struct whose first member is VK_LOADER_DATA. The
4792 * ICD must initialize VK_LOADER_DATA.loadMagic to ICD_LOADER_MAGIC.
4793 * - The loader implements vkCreate{PLATFORM}SurfaceKHR() and
4794 * vkDestroySurfaceKHR(). The ICD must be capable of working with
4795 * such loader-managed surfaces.
4797 * - Loader interface v2 differs from v1 in:
4798 * - The first ICD entrypoint called by the loader is
4799 * vk_icdNegotiateLoaderICDInterfaceVersion(). The ICD must
4800 * statically expose this entrypoint.
4802 * - Loader interface v3 differs from v2 in:
4803 * - The ICD must implement vkCreate{PLATFORM}SurfaceKHR(),
4804 * vkDestroySurfaceKHR(), and other API which uses VKSurfaceKHR,
4805 * because the loader no longer does so.
4807 *pSupportedVersion
= MIN2(*pSupportedVersion
, 3u);
4811 VkResult
radv_GetMemoryFdKHR(VkDevice _device
,
4812 const VkMemoryGetFdInfoKHR
*pGetFdInfo
,
4815 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4816 RADV_FROM_HANDLE(radv_device_memory
, memory
, pGetFdInfo
->memory
);
4818 assert(pGetFdInfo
->sType
== VK_STRUCTURE_TYPE_MEMORY_GET_FD_INFO_KHR
);
4820 /* At the moment, we support only the below handle types. */
4821 assert(pGetFdInfo
->handleType
==
4822 VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT
||
4823 pGetFdInfo
->handleType
==
4824 VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT
);
4826 bool ret
= radv_get_memory_fd(device
, memory
, pFD
);
4828 return vk_error(device
->instance
, VK_ERROR_OUT_OF_DEVICE_MEMORY
);
4832 VkResult
radv_GetMemoryFdPropertiesKHR(VkDevice _device
,
4833 VkExternalMemoryHandleTypeFlagBits handleType
,
4835 VkMemoryFdPropertiesKHR
*pMemoryFdProperties
)
4837 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4839 switch (handleType
) {
4840 case VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT
:
4841 pMemoryFdProperties
->memoryTypeBits
= (1 << RADV_MEM_TYPE_COUNT
) - 1;
4845 /* The valid usage section for this function says:
4847 * "handleType must not be one of the handle types defined as
4850 * So opaque handle types fall into the default "unsupported" case.
4852 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
4856 static VkResult
radv_import_opaque_fd(struct radv_device
*device
,
4860 uint32_t syncobj_handle
= 0;
4861 int ret
= device
->ws
->import_syncobj(device
->ws
, fd
, &syncobj_handle
);
4863 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
4866 device
->ws
->destroy_syncobj(device
->ws
, *syncobj
);
4868 *syncobj
= syncobj_handle
;
4874 static VkResult
radv_import_sync_fd(struct radv_device
*device
,
4878 /* If we create a syncobj we do it locally so that if we have an error, we don't
4879 * leave a syncobj in an undetermined state in the fence. */
4880 uint32_t syncobj_handle
= *syncobj
;
4881 if (!syncobj_handle
) {
4882 int ret
= device
->ws
->create_syncobj(device
->ws
, &syncobj_handle
);
4884 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
4889 device
->ws
->signal_syncobj(device
->ws
, syncobj_handle
);
4891 int ret
= device
->ws
->import_syncobj_from_sync_file(device
->ws
, syncobj_handle
, fd
);
4893 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
4896 *syncobj
= syncobj_handle
;
4903 VkResult
radv_ImportSemaphoreFdKHR(VkDevice _device
,
4904 const VkImportSemaphoreFdInfoKHR
*pImportSemaphoreFdInfo
)
4906 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4907 RADV_FROM_HANDLE(radv_semaphore
, sem
, pImportSemaphoreFdInfo
->semaphore
);
4908 uint32_t *syncobj_dst
= NULL
;
4910 if (pImportSemaphoreFdInfo
->flags
& VK_SEMAPHORE_IMPORT_TEMPORARY_BIT
) {
4911 syncobj_dst
= &sem
->temp_syncobj
;
4913 syncobj_dst
= &sem
->syncobj
;
4916 switch(pImportSemaphoreFdInfo
->handleType
) {
4917 case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
:
4918 return radv_import_opaque_fd(device
, pImportSemaphoreFdInfo
->fd
, syncobj_dst
);
4919 case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT
:
4920 return radv_import_sync_fd(device
, pImportSemaphoreFdInfo
->fd
, syncobj_dst
);
4922 unreachable("Unhandled semaphore handle type");
4926 VkResult
radv_GetSemaphoreFdKHR(VkDevice _device
,
4927 const VkSemaphoreGetFdInfoKHR
*pGetFdInfo
,
4930 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4931 RADV_FROM_HANDLE(radv_semaphore
, sem
, pGetFdInfo
->semaphore
);
4933 uint32_t syncobj_handle
;
4935 if (sem
->temp_syncobj
)
4936 syncobj_handle
= sem
->temp_syncobj
;
4938 syncobj_handle
= sem
->syncobj
;
4940 switch(pGetFdInfo
->handleType
) {
4941 case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
:
4942 ret
= device
->ws
->export_syncobj(device
->ws
, syncobj_handle
, pFd
);
4944 case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT
:
4945 ret
= device
->ws
->export_syncobj_to_sync_file(device
->ws
, syncobj_handle
, pFd
);
4947 if (sem
->temp_syncobj
) {
4948 close (sem
->temp_syncobj
);
4949 sem
->temp_syncobj
= 0;
4951 device
->ws
->reset_syncobj(device
->ws
, syncobj_handle
);
4956 unreachable("Unhandled semaphore handle type");
4960 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
4964 void radv_GetPhysicalDeviceExternalSemaphoreProperties(
4965 VkPhysicalDevice physicalDevice
,
4966 const VkPhysicalDeviceExternalSemaphoreInfo
*pExternalSemaphoreInfo
,
4967 VkExternalSemaphoreProperties
*pExternalSemaphoreProperties
)
4969 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
4971 /* Require has_syncobj_wait_for_submit for the syncobj signal ioctl introduced at virtually the same time */
4972 if (pdevice
->rad_info
.has_syncobj_wait_for_submit
&&
4973 (pExternalSemaphoreInfo
->handleType
== VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
||
4974 pExternalSemaphoreInfo
->handleType
== VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT
)) {
4975 pExternalSemaphoreProperties
->exportFromImportedHandleTypes
= VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
| VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT
;
4976 pExternalSemaphoreProperties
->compatibleHandleTypes
= VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
| VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT
;
4977 pExternalSemaphoreProperties
->externalSemaphoreFeatures
= VK_EXTERNAL_SEMAPHORE_FEATURE_EXPORTABLE_BIT
|
4978 VK_EXTERNAL_SEMAPHORE_FEATURE_IMPORTABLE_BIT
;
4979 } else if (pExternalSemaphoreInfo
->handleType
== VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
) {
4980 pExternalSemaphoreProperties
->exportFromImportedHandleTypes
= VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
;
4981 pExternalSemaphoreProperties
->compatibleHandleTypes
= VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT
;
4982 pExternalSemaphoreProperties
->externalSemaphoreFeatures
= VK_EXTERNAL_SEMAPHORE_FEATURE_EXPORTABLE_BIT
|
4983 VK_EXTERNAL_SEMAPHORE_FEATURE_IMPORTABLE_BIT
;
4985 pExternalSemaphoreProperties
->exportFromImportedHandleTypes
= 0;
4986 pExternalSemaphoreProperties
->compatibleHandleTypes
= 0;
4987 pExternalSemaphoreProperties
->externalSemaphoreFeatures
= 0;
4991 VkResult
radv_ImportFenceFdKHR(VkDevice _device
,
4992 const VkImportFenceFdInfoKHR
*pImportFenceFdInfo
)
4994 RADV_FROM_HANDLE(radv_device
, device
, _device
);
4995 RADV_FROM_HANDLE(radv_fence
, fence
, pImportFenceFdInfo
->fence
);
4996 uint32_t *syncobj_dst
= NULL
;
4999 if (pImportFenceFdInfo
->flags
& VK_FENCE_IMPORT_TEMPORARY_BIT
) {
5000 syncobj_dst
= &fence
->temp_syncobj
;
5002 syncobj_dst
= &fence
->syncobj
;
5005 switch(pImportFenceFdInfo
->handleType
) {
5006 case VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT
:
5007 return radv_import_opaque_fd(device
, pImportFenceFdInfo
->fd
, syncobj_dst
);
5008 case VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT
:
5009 return radv_import_sync_fd(device
, pImportFenceFdInfo
->fd
, syncobj_dst
);
5011 unreachable("Unhandled fence handle type");
5015 VkResult
radv_GetFenceFdKHR(VkDevice _device
,
5016 const VkFenceGetFdInfoKHR
*pGetFdInfo
,
5019 RADV_FROM_HANDLE(radv_device
, device
, _device
);
5020 RADV_FROM_HANDLE(radv_fence
, fence
, pGetFdInfo
->fence
);
5022 uint32_t syncobj_handle
;
5024 if (fence
->temp_syncobj
)
5025 syncobj_handle
= fence
->temp_syncobj
;
5027 syncobj_handle
= fence
->syncobj
;
5029 switch(pGetFdInfo
->handleType
) {
5030 case VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT
:
5031 ret
= device
->ws
->export_syncobj(device
->ws
, syncobj_handle
, pFd
);
5033 case VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT
:
5034 ret
= device
->ws
->export_syncobj_to_sync_file(device
->ws
, syncobj_handle
, pFd
);
5036 if (fence
->temp_syncobj
) {
5037 close (fence
->temp_syncobj
);
5038 fence
->temp_syncobj
= 0;
5040 device
->ws
->reset_syncobj(device
->ws
, syncobj_handle
);
5045 unreachable("Unhandled fence handle type");
5049 return vk_error(device
->instance
, VK_ERROR_INVALID_EXTERNAL_HANDLE
);
5053 void radv_GetPhysicalDeviceExternalFenceProperties(
5054 VkPhysicalDevice physicalDevice
,
5055 const VkPhysicalDeviceExternalFenceInfo
*pExternalFenceInfo
,
5056 VkExternalFenceProperties
*pExternalFenceProperties
)
5058 RADV_FROM_HANDLE(radv_physical_device
, pdevice
, physicalDevice
);
5060 if (pdevice
->rad_info
.has_syncobj_wait_for_submit
&&
5061 (pExternalFenceInfo
->handleType
== VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT
||
5062 pExternalFenceInfo
->handleType
== VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT
)) {
5063 pExternalFenceProperties
->exportFromImportedHandleTypes
= VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT
| VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT
;
5064 pExternalFenceProperties
->compatibleHandleTypes
= VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT
| VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT
;
5065 pExternalFenceProperties
->externalFenceFeatures
= VK_EXTERNAL_FENCE_FEATURE_EXPORTABLE_BIT
|
5066 VK_EXTERNAL_SEMAPHORE_FEATURE_IMPORTABLE_BIT
;
5068 pExternalFenceProperties
->exportFromImportedHandleTypes
= 0;
5069 pExternalFenceProperties
->compatibleHandleTypes
= 0;
5070 pExternalFenceProperties
->externalFenceFeatures
= 0;
5075 radv_CreateDebugReportCallbackEXT(VkInstance _instance
,
5076 const VkDebugReportCallbackCreateInfoEXT
* pCreateInfo
,
5077 const VkAllocationCallbacks
* pAllocator
,
5078 VkDebugReportCallbackEXT
* pCallback
)
5080 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
5081 return vk_create_debug_report_callback(&instance
->debug_report_callbacks
,
5082 pCreateInfo
, pAllocator
, &instance
->alloc
,
5087 radv_DestroyDebugReportCallbackEXT(VkInstance _instance
,
5088 VkDebugReportCallbackEXT _callback
,
5089 const VkAllocationCallbacks
* pAllocator
)
5091 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
5092 vk_destroy_debug_report_callback(&instance
->debug_report_callbacks
,
5093 _callback
, pAllocator
, &instance
->alloc
);
5097 radv_DebugReportMessageEXT(VkInstance _instance
,
5098 VkDebugReportFlagsEXT flags
,
5099 VkDebugReportObjectTypeEXT objectType
,
5102 int32_t messageCode
,
5103 const char* pLayerPrefix
,
5104 const char* pMessage
)
5106 RADV_FROM_HANDLE(radv_instance
, instance
, _instance
);
5107 vk_debug_report(&instance
->debug_report_callbacks
, flags
, objectType
,
5108 object
, location
, messageCode
, pLayerPrefix
, pMessage
);
5112 radv_GetDeviceGroupPeerMemoryFeatures(
5115 uint32_t localDeviceIndex
,
5116 uint32_t remoteDeviceIndex
,
5117 VkPeerMemoryFeatureFlags
* pPeerMemoryFeatures
)
5119 assert(localDeviceIndex
== remoteDeviceIndex
);
5121 *pPeerMemoryFeatures
= VK_PEER_MEMORY_FEATURE_COPY_SRC_BIT
|
5122 VK_PEER_MEMORY_FEATURE_COPY_DST_BIT
|
5123 VK_PEER_MEMORY_FEATURE_GENERIC_SRC_BIT
|
5124 VK_PEER_MEMORY_FEATURE_GENERIC_DST_BIT
;
5127 static const VkTimeDomainEXT radv_time_domains
[] = {
5128 VK_TIME_DOMAIN_DEVICE_EXT
,
5129 VK_TIME_DOMAIN_CLOCK_MONOTONIC_EXT
,
5130 VK_TIME_DOMAIN_CLOCK_MONOTONIC_RAW_EXT
,
5133 VkResult
radv_GetPhysicalDeviceCalibrateableTimeDomainsEXT(
5134 VkPhysicalDevice physicalDevice
,
5135 uint32_t *pTimeDomainCount
,
5136 VkTimeDomainEXT
*pTimeDomains
)
5139 VK_OUTARRAY_MAKE(out
, pTimeDomains
, pTimeDomainCount
);
5141 for (d
= 0; d
< ARRAY_SIZE(radv_time_domains
); d
++) {
5142 vk_outarray_append(&out
, i
) {
5143 *i
= radv_time_domains
[d
];
5147 return vk_outarray_status(&out
);
5151 radv_clock_gettime(clockid_t clock_id
)
5153 struct timespec current
;
5156 ret
= clock_gettime(clock_id
, ¤t
);
5157 if (ret
< 0 && clock_id
== CLOCK_MONOTONIC_RAW
)
5158 ret
= clock_gettime(CLOCK_MONOTONIC
, ¤t
);
5162 return (uint64_t) current
.tv_sec
* 1000000000ULL + current
.tv_nsec
;
5165 VkResult
radv_GetCalibratedTimestampsEXT(
5167 uint32_t timestampCount
,
5168 const VkCalibratedTimestampInfoEXT
*pTimestampInfos
,
5169 uint64_t *pTimestamps
,
5170 uint64_t *pMaxDeviation
)
5172 RADV_FROM_HANDLE(radv_device
, device
, _device
);
5173 uint32_t clock_crystal_freq
= device
->physical_device
->rad_info
.clock_crystal_freq
;
5175 uint64_t begin
, end
;
5176 uint64_t max_clock_period
= 0;
5178 begin
= radv_clock_gettime(CLOCK_MONOTONIC_RAW
);
5180 for (d
= 0; d
< timestampCount
; d
++) {
5181 switch (pTimestampInfos
[d
].timeDomain
) {
5182 case VK_TIME_DOMAIN_DEVICE_EXT
:
5183 pTimestamps
[d
] = device
->ws
->query_value(device
->ws
,
5185 uint64_t device_period
= DIV_ROUND_UP(1000000, clock_crystal_freq
);
5186 max_clock_period
= MAX2(max_clock_period
, device_period
);
5188 case VK_TIME_DOMAIN_CLOCK_MONOTONIC_EXT
:
5189 pTimestamps
[d
] = radv_clock_gettime(CLOCK_MONOTONIC
);
5190 max_clock_period
= MAX2(max_clock_period
, 1);
5193 case VK_TIME_DOMAIN_CLOCK_MONOTONIC_RAW_EXT
:
5194 pTimestamps
[d
] = begin
;
5202 end
= radv_clock_gettime(CLOCK_MONOTONIC_RAW
);
5205 * The maximum deviation is the sum of the interval over which we
5206 * perform the sampling and the maximum period of any sampled
5207 * clock. That's because the maximum skew between any two sampled
5208 * clock edges is when the sampled clock with the largest period is
5209 * sampled at the end of that period but right at the beginning of the
5210 * sampling interval and some other clock is sampled right at the
5211 * begining of its sampling period and right at the end of the
5212 * sampling interval. Let's assume the GPU has the longest clock
5213 * period and that the application is sampling GPU and monotonic:
5216 * w x y z 0 1 2 3 4 5 6 7 8 9 a b c d e f
5217 * Raw -_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-
5221 * GPU -----_____-----_____-----_____-----_____
5224 * x y z 0 1 2 3 4 5 6 7 8 9 a b c
5225 * Monotonic -_-_-_-_-_-_-_-_-_-_-_-_-_-_-_-
5227 * Interval <----------------->
5228 * Deviation <-------------------------->
5232 * m = read(monotonic) 2
5235 * We round the sample interval up by one tick to cover sampling error
5236 * in the interval clock
5239 uint64_t sample_interval
= end
- begin
+ 1;
5241 *pMaxDeviation
= sample_interval
+ max_clock_period
;