ac_compute_device_uuid(info, uuid, VK_UUID_SIZE);
}
-static const char *
-get_chip_name(enum radeon_family family)
+static void
+radv_get_device_name(enum radeon_family family, char *name, size_t name_len)
{
+ const char *chip_string;
+ char llvm_string[32] = {};
+
switch (family) {
- case CHIP_TAHITI: return "AMD RADV TAHITI";
- case CHIP_PITCAIRN: return "AMD RADV PITCAIRN";
- case CHIP_VERDE: return "AMD RADV CAPE VERDE";
- case CHIP_OLAND: return "AMD RADV OLAND";
- case CHIP_HAINAN: return "AMD RADV HAINAN";
- case CHIP_BONAIRE: return "AMD RADV BONAIRE";
- case CHIP_KAVERI: return "AMD RADV KAVERI";
- case CHIP_KABINI: return "AMD RADV KABINI";
- case CHIP_HAWAII: return "AMD RADV HAWAII";
- case CHIP_MULLINS: return "AMD RADV MULLINS";
- case CHIP_TONGA: return "AMD RADV TONGA";
- case CHIP_ICELAND: return "AMD RADV ICELAND";
- case CHIP_CARRIZO: return "AMD RADV CARRIZO";
- case CHIP_FIJI: return "AMD RADV FIJI";
- case CHIP_POLARIS10: return "AMD RADV POLARIS10";
- case CHIP_POLARIS11: return "AMD RADV POLARIS11";
- case CHIP_POLARIS12: return "AMD RADV POLARIS12";
- case CHIP_STONEY: return "AMD RADV STONEY";
- case CHIP_VEGA10: return "AMD RADV VEGA";
- case CHIP_RAVEN: return "AMD RADV RAVEN";
- default: return "AMD RADV unknown";
- }
+ case CHIP_TAHITI: chip_string = "AMD RADV TAHITI"; break;
+ case CHIP_PITCAIRN: chip_string = "AMD RADV PITCAIRN"; break;
+ case CHIP_VERDE: chip_string = "AMD RADV CAPE VERDE"; break;
+ case CHIP_OLAND: chip_string = "AMD RADV OLAND"; break;
+ case CHIP_HAINAN: chip_string = "AMD RADV HAINAN"; break;
+ case CHIP_BONAIRE: chip_string = "AMD RADV BONAIRE"; break;
+ case CHIP_KAVERI: chip_string = "AMD RADV KAVERI"; break;
+ case CHIP_KABINI: chip_string = "AMD RADV KABINI"; break;
+ case CHIP_HAWAII: chip_string = "AMD RADV HAWAII"; break;
+ case CHIP_MULLINS: chip_string = "AMD RADV MULLINS"; break;
+ case CHIP_TONGA: chip_string = "AMD RADV TONGA"; break;
+ case CHIP_ICELAND: chip_string = "AMD RADV ICELAND"; break;
+ case CHIP_CARRIZO: chip_string = "AMD RADV CARRIZO"; break;
+ case CHIP_FIJI: chip_string = "AMD RADV FIJI"; break;
+ case CHIP_POLARIS10: chip_string = "AMD RADV POLARIS10"; break;
+ case CHIP_POLARIS11: chip_string = "AMD RADV POLARIS11"; break;
+ case CHIP_POLARIS12: chip_string = "AMD RADV POLARIS12"; break;
+ case CHIP_STONEY: chip_string = "AMD RADV STONEY"; break;
+ case CHIP_VEGA10: chip_string = "AMD RADV VEGA"; break;
+ case CHIP_RAVEN: chip_string = "AMD RADV RAVEN"; break;
+ default: chip_string = "AMD RADV unknown"; break;
+ }
+
+ if (HAVE_LLVM > 0) {
+ snprintf(llvm_string, sizeof(llvm_string),
+ " (LLVM %i.%i.%i)", (HAVE_LLVM >> 8) & 0xff,
+ HAVE_LLVM & 0xff, MESA_LLVM_VERSION_PATCH);
+ }
+
+ snprintf(name, name_len, "%s%s", chip_string, llvm_string);
}
static void
if (strcmp(version->name, "amdgpu")) {
drmFreeVersion(version);
close(fd);
- return vk_error(VK_ERROR_INCOMPATIBLE_DRIVER);
+ return VK_ERROR_INCOMPATIBLE_DRIVER;
}
drmFreeVersion(version);
device->local_fd = fd;
device->ws->query_info(device->ws, &device->rad_info);
- result = radv_init_wsi(device);
- if (result != VK_SUCCESS) {
- device->ws->destroy(device->ws);
- goto fail;
- }
- device->name = get_chip_name(device->rad_info.family);
+ radv_get_device_name(device->rad_info.family, device->name, sizeof(device->name));
if (radv_device_get_cache_uuid(device->rad_info.family, device->cache_uuid)) {
- radv_finish_wsi(device);
device->ws->destroy(device->ws);
result = vk_errorf(VK_ERROR_INITIALIZATION_FAILED,
"cannot generate UUID");
*/
device->has_clear_state = device->rad_info.chip_class >= CIK;
+ device->cpdma_prefetch_writes_memory = device->rad_info.chip_class <= VI;
+
+ /* Vega10/Raven need a special workaround for a hardware bug. */
+ device->has_scissor_bug = device->rad_info.family == CHIP_VEGA10 ||
+ device->rad_info.family == CHIP_RAVEN;
+
radv_physical_device_init_mem_types(device);
+
+ result = radv_init_wsi(device);
+ if (result != VK_SUCCESS) {
+ device->ws->destroy(device->ws);
+ goto fail;
+ }
+
return VK_SUCCESS;
fail:
static const struct debug_control radv_perftest_options[] = {
{"nobatchchain", RADV_PERFTEST_NO_BATCHCHAIN},
{"sisched", RADV_PERFTEST_SISCHED},
+ {"localbos", RADV_PERFTEST_LOCAL_BOS},
+ {"binning", RADV_PERFTEST_BINNING},
{NULL, 0}
};
properties->pointClippingBehavior = VK_POINT_CLIPPING_BEHAVIOR_ALL_CLIP_PLANES_KHR;
break;
}
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DISCARD_RECTANGLE_PROPERTIES_EXT: {
+ VkPhysicalDeviceDiscardRectanglePropertiesEXT *properties =
+ (VkPhysicalDeviceDiscardRectanglePropertiesEXT*)ext;
+ properties->maxDiscardRectangles = MAX_DISCARD_RECTANGLES;
+ break;
+ }
default:
break;
}
return RADEON_CTX_PRIORITY_MEDIUM;
switch(pObj->globalPriority) {
- case VK_QUEUE_GLOBAL_PRIORITY_REALTIME:
+ case VK_QUEUE_GLOBAL_PRIORITY_REALTIME_EXT:
return RADEON_CTX_PRIORITY_REALTIME;
- case VK_QUEUE_GLOBAL_PRIORITY_HIGH:
+ case VK_QUEUE_GLOBAL_PRIORITY_HIGH_EXT:
return RADEON_CTX_PRIORITY_HIGH;
- case VK_QUEUE_GLOBAL_PRIORITY_MEDIUM:
+ case VK_QUEUE_GLOBAL_PRIORITY_MEDIUM_EXT:
return RADEON_CTX_PRIORITY_MEDIUM;
- case VK_QUEUE_GLOBAL_PRIORITY_LOW:
+ case VK_QUEUE_GLOBAL_PRIORITY_LOW_EXT:
return RADEON_CTX_PRIORITY_LOW;
default:
unreachable("Illegal global priority value");
static int
radv_queue_init(struct radv_device *device, struct radv_queue *queue,
- int queue_family_index, int idx,
+ uint32_t queue_family_index, int idx,
const VkDeviceQueueGlobalPriorityCreateInfoEXT *global_priority)
{
queue->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
}
}
+ device->pbb_allowed = device->physical_device->rad_info.chip_class >= GFX9 &&
+ (device->instance->perftest_flags & RADV_PERFTEST_BINNING);
+
+ /* Disabled and not implemented for now. */
+ device->dfsm_allowed = device->pbb_allowed && false;
+
+
#if HAVE_LLVM < 0x0400
device->llvm_supports_spill = false;
#else
device->scratch_waves = MAX2(32 * physical_device->rad_info.num_good_compute_units,
max_threads_per_block / 64);
+ device->dispatch_initiator = S_00B800_COMPUTE_SHADER_EN(1) |
+ S_00B800_FORCE_START_AT_000(1);
+
+ if (device->physical_device->rad_info.chip_class >= CIK) {
+ /* If the KMD allows it (there is a KMD hw register for it),
+ * allow launching waves out-of-order.
+ */
+ device->dispatch_initiator |= S_00B800_ORDER_MODE(1);
+ }
+
radv_device_init_gs_info(device);
device->tess_offchip_block_dw_size =
size,
4096,
RADEON_DOMAIN_VRAM,
- RADEON_FLAG_CPU_ACCESS|RADEON_FLAG_NO_INTERPROCESS_SHARING);
+ RADEON_FLAG_CPU_ACCESS |
+ RADEON_FLAG_NO_INTERPROCESS_SHARING |
+ RADEON_FLAG_READ_ONLY);
if (!descriptor_bo)
goto fail;
} else
static VkResult radv_alloc_sem_counts(struct radv_winsys_sem_counts *counts,
int num_sems,
const VkSemaphore *sems,
+ VkFence _fence,
bool reset_temp)
{
int syncobj_idx = 0, sem_idx = 0;
- if (num_sems == 0)
+ if (num_sems == 0 && _fence == VK_NULL_HANDLE)
return VK_SUCCESS;
+
for (uint32_t i = 0; i < num_sems; i++) {
RADV_FROM_HANDLE(radv_semaphore, sem, sems[i]);
counts->sem_count++;
}
+ if (_fence != VK_NULL_HANDLE) {
+ RADV_FROM_HANDLE(radv_fence, fence, _fence);
+ if (fence->temp_syncobj || fence->syncobj)
+ counts->syncobj_count++;
+ }
+
if (counts->syncobj_count) {
counts->syncobj = (uint32_t *)malloc(sizeof(uint32_t) * counts->syncobj_count);
if (!counts->syncobj)
}
}
+ if (_fence != VK_NULL_HANDLE) {
+ RADV_FROM_HANDLE(radv_fence, fence, _fence);
+ if (fence->temp_syncobj)
+ counts->syncobj[syncobj_idx++] = fence->temp_syncobj;
+ else if (fence->syncobj)
+ counts->syncobj[syncobj_idx++] = fence->syncobj;
+ }
+
return VK_SUCCESS;
}
int num_wait_sems,
const VkSemaphore *wait_sems,
int num_signal_sems,
- const VkSemaphore *signal_sems)
+ const VkSemaphore *signal_sems,
+ VkFence fence)
{
VkResult ret;
memset(sem_info, 0, sizeof(*sem_info));
- ret = radv_alloc_sem_counts(&sem_info->wait, num_wait_sems, wait_sems, true);
+ ret = radv_alloc_sem_counts(&sem_info->wait, num_wait_sems, wait_sems, VK_NULL_HANDLE, true);
if (ret)
return ret;
- ret = radv_alloc_sem_counts(&sem_info->signal, num_signal_sems, signal_sems, false);
+ ret = radv_alloc_sem_counts(&sem_info->signal, num_signal_sems, signal_sems, fence, false);
if (ret)
radv_free_sem_info(sem_info);
pSubmits[i].waitSemaphoreCount,
pSubmits[i].pWaitSemaphores,
pSubmits[i].signalSemaphoreCount,
- pSubmits[i].pSignalSemaphores);
+ pSubmits[i].pSignalSemaphores,
+ _fence);
if (result != VK_SUCCESS)
return result;
cs_array[j] = cmd_buffer->cs;
if ((cmd_buffer->usage_flags & VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT))
can_patch = false;
+
+ cmd_buffer->status = RADV_CMD_BUFFER_STATUS_PENDING;
}
for (uint32_t j = 0; j < pSubmits[i].commandBufferCount; j += advance) {
if (fence) {
if (!fence_emitted) {
- struct radv_winsys_sem_info sem_info = {0};
+ struct radv_winsys_sem_info sem_info;
+
+ result = radv_alloc_sem_info(&sem_info, 0, NULL, 0, NULL,
+ _fence);
+ if (result != VK_SUCCESS)
+ return result;
+
ret = queue->device->ws->cs_submit(ctx, queue->queue_idx,
&queue->device->empty_cs[queue->queue_family_index],
1, NULL, NULL, &sem_info,
false, base_fence);
+ radv_free_sem_info(&sem_info);
}
fence->submitted = true;
}
pFD);
}
-VkResult radv_alloc_memory(VkDevice _device,
- const VkMemoryAllocateInfo* pAllocateInfo,
- const VkAllocationCallbacks* pAllocator,
- enum radv_mem_flags_bits mem_flags,
- VkDeviceMemory* pMem)
+static VkResult radv_alloc_memory(struct radv_device *device,
+ const VkMemoryAllocateInfo* pAllocateInfo,
+ const VkAllocationCallbacks* pAllocator,
+ VkDeviceMemory* pMem)
{
- RADV_FROM_HANDLE(radv_device, device, _device);
struct radv_device_memory *mem;
VkResult result;
enum radeon_bo_domain domain;
vk_find_struct_const(pAllocateInfo->pNext, IMPORT_MEMORY_FD_INFO_KHR);
const VkMemoryDedicatedAllocateInfoKHR *dedicate_info =
vk_find_struct_const(pAllocateInfo->pNext, MEMORY_DEDICATED_ALLOCATE_INFO_KHR);
+ const VkExportMemoryAllocateInfoKHR *export_info =
+ vk_find_struct_const(pAllocateInfo->pNext, EXPORT_MEMORY_ALLOCATE_INFO_KHR);
+
+ const struct wsi_memory_allocate_info *wsi_info =
+ vk_find_struct_const(pAllocateInfo->pNext, WSI_MEMORY_ALLOCATE_INFO_MESA);
mem = vk_alloc2(&device->alloc, pAllocator, sizeof(*mem), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (mem == NULL)
return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ if (wsi_info && wsi_info->implicit_sync)
+ flags |= RADEON_FLAG_IMPLICIT_SYNC;
+
if (dedicate_info) {
mem->image = radv_image_from_handle(dedicate_info->image);
mem->buffer = radv_buffer_from_handle(dedicate_info->buffer);
if (import_info) {
assert(import_info->handleType ==
- VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT_KHR);
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT_KHR ||
+ import_info->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT);
mem->bo = device->ws->buffer_from_fd(device->ws, import_info->fd,
NULL, NULL);
if (!mem->bo) {
if (mem_type_index == RADV_MEM_TYPE_GTT_WRITE_COMBINE)
flags |= RADEON_FLAG_GTT_WC;
- if (mem_flags & RADV_MEM_IMPLICIT_SYNC)
- flags |= RADEON_FLAG_IMPLICIT_SYNC;
-
- if (!dedicate_info && !import_info)
+ if (!dedicate_info && !import_info && (!export_info || !export_info->handleTypes))
flags |= RADEON_FLAG_NO_INTERPROCESS_SHARING;
mem->bo = device->ws->buffer_create(device->ws, alloc_size, device->physical_device->rad_info.max_alignment,
const VkAllocationCallbacks* pAllocator,
VkDeviceMemory* pMem)
{
- return radv_alloc_memory(_device, pAllocateInfo, pAllocator, 0, pMem);
+ RADV_FROM_HANDLE(radv_device, device, _device);
+ return radv_alloc_memory(device, pAllocateInfo, pAllocator, pMem);
}
void radv_FreeMemory(
pBindInfo[i].waitSemaphoreCount,
pBindInfo[i].pWaitSemaphores,
pBindInfo[i].signalSemaphoreCount,
- pBindInfo[i].pSignalSemaphores);
+ pBindInfo[i].pSignalSemaphores,
+ _fence);
if (result != VK_SUCCESS)
return result;
VkFence* pFence)
{
RADV_FROM_HANDLE(radv_device, device, _device);
+ const VkExportFenceCreateInfoKHR *export =
+ vk_find_struct_const(pCreateInfo->pNext, EXPORT_FENCE_CREATE_INFO_KHR);
+ VkExternalFenceHandleTypeFlagsKHR handleTypes =
+ export ? export->handleTypes : 0;
+
struct radv_fence *fence = vk_alloc2(&device->alloc, pAllocator,
sizeof(*fence), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
fence->submitted = false;
fence->signalled = !!(pCreateInfo->flags & VK_FENCE_CREATE_SIGNALED_BIT);
- fence->fence = device->ws->create_fence();
- if (!fence->fence) {
- vk_free2(&device->alloc, pAllocator, fence);
- return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ fence->temp_syncobj = 0;
+ if (handleTypes) {
+ int ret = device->ws->create_syncobj(device->ws, &fence->syncobj);
+ if (ret) {
+ vk_free2(&device->alloc, pAllocator, fence);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+ if (pCreateInfo->flags & VK_FENCE_CREATE_SIGNALED_BIT) {
+ device->ws->signal_syncobj(device->ws, fence->syncobj);
+ }
+ fence->fence = NULL;
+ } else {
+ fence->fence = device->ws->create_fence();
+ if (!fence->fence) {
+ vk_free2(&device->alloc, pAllocator, fence);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
+ }
+ fence->syncobj = 0;
}
*pFence = radv_fence_to_handle(fence);
if (!fence)
return;
- device->ws->destroy_fence(fence->fence);
+
+ if (fence->temp_syncobj)
+ device->ws->destroy_syncobj(device->ws, fence->temp_syncobj);
+ if (fence->syncobj)
+ device->ws->destroy_syncobj(device->ws, fence->syncobj);
+ if (fence->fence)
+ device->ws->destroy_fence(fence->fence);
vk_free2(&device->alloc, pAllocator, fence);
}
RADV_FROM_HANDLE(radv_fence, fence, pFences[i]);
bool expired = false;
+ if (fence->temp_syncobj) {
+ if (!device->ws->wait_syncobj(device->ws, fence->temp_syncobj, timeout))
+ return VK_TIMEOUT;
+ continue;
+ }
+
+ if (fence->syncobj) {
+ if (!device->ws->wait_syncobj(device->ws, fence->syncobj, timeout))
+ return VK_TIMEOUT;
+ continue;
+ }
+
if (fence->signalled)
continue;
return VK_SUCCESS;
}
-VkResult radv_ResetFences(VkDevice device,
+VkResult radv_ResetFences(VkDevice _device,
uint32_t fenceCount,
const VkFence *pFences)
{
+ RADV_FROM_HANDLE(radv_device, device, _device);
+
for (unsigned i = 0; i < fenceCount; ++i) {
RADV_FROM_HANDLE(radv_fence, fence, pFences[i]);
fence->submitted = fence->signalled = false;
+
+ /* Per spec, we first restore the permanent payload, and then reset, so
+ * having a temp syncobj should not skip resetting the permanent syncobj. */
+ if (fence->temp_syncobj) {
+ device->ws->destroy_syncobj(device->ws, fence->temp_syncobj);
+ fence->temp_syncobj = 0;
+ }
+
+ if (fence->syncobj) {
+ device->ws->reset_syncobj(device->ws, fence->syncobj);
+ }
}
return VK_SUCCESS;
RADV_FROM_HANDLE(radv_device, device, _device);
RADV_FROM_HANDLE(radv_fence, fence, _fence);
+ if (fence->temp_syncobj) {
+ bool success = device->ws->wait_syncobj(device->ws, fence->temp_syncobj, 0);
+ return success ? VK_SUCCESS : VK_NOT_READY;
+ }
+
+ if (fence->syncobj) {
+ bool success = device->ws->wait_syncobj(device->ws, fence->syncobj, 0);
+ return success ? VK_SUCCESS : VK_NOT_READY;
+ }
+
if (fence->signalled)
return VK_SUCCESS;
if (!fence->submitted)
return VK_NOT_READY;
-
if (!device->ws->fence_wait(device->ws, fence->fence, false, 0))
return VK_NOT_READY;
/* create a syncobject if we are going to export this semaphore */
if (handleTypes) {
assert (device->physical_device->rad_info.has_syncobj);
- assert (handleTypes == VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR);
int ret = device->ws->create_syncobj(device->ws, &sem->syncobj);
if (ret) {
vk_free2(&device->alloc, pAllocator, sem);
return image->surface.u.legacy.tiling_index[level];
}
-static uint32_t radv_surface_layer_count(struct radv_image_view *iview)
+static uint32_t radv_surface_max_layer_count(struct radv_image_view *iview)
{
- return iview->type == VK_IMAGE_VIEW_TYPE_3D ? iview->extent.depth : iview->layer_count;
+ return iview->type == VK_IMAGE_VIEW_TYPE_3D ? iview->extent.depth : (iview->base_layer + iview->layer_count);
}
static void
cb->cb_color_cmask_slice = iview->image->cmask.slice_tile_max;
cb->cb_color_attrib |= S_028C74_TILE_MODE_INDEX(tile_mode_index);
- cb->micro_tile_mode = iview->image->surface.micro_tile_mode;
if (iview->image->fmask.size) {
if (device->physical_device->rad_info.chip_class >= CIK)
cb->cb_dcc_base = va >> 8;
cb->cb_dcc_base |= iview->image->surface.tile_swizzle;
- uint32_t max_slice = radv_surface_layer_count(iview);
+ uint32_t max_slice = radv_surface_max_layer_count(iview) - 1;
cb->cb_color_view = S_028C6C_SLICE_START(iview->base_layer) |
- S_028C6C_SLICE_MAX(iview->base_layer + max_slice - 1);
+ S_028C6C_SLICE_MAX(max_slice);
if (iview->image->info.samples > 1) {
unsigned log_samples = util_logbase2(iview->image->info.samples);
cb->cb_color_info |= S_028C70_DCC_ENABLE(1);
if (device->physical_device->rad_info.chip_class >= VI) {
- unsigned max_uncompressed_block_size = 2;
+ unsigned max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_256B;
+ unsigned min_compressed_block_size = V_028C78_MIN_BLOCK_SIZE_32B;
+ unsigned independent_64b_blocks = 0;
+ unsigned max_compressed_block_size;
+
+ /* amdvlk: [min-compressed-block-size] should be set to 32 for dGPU and
+ 64 for APU because all of our APUs to date use DIMMs which have
+ a request granularity size of 64B while all other chips have a
+ 32B request size */
+ if (!device->physical_device->rad_info.has_dedicated_vram)
+ min_compressed_block_size = V_028C78_MIN_BLOCK_SIZE_64B;
+
if (iview->image->info.samples > 1) {
if (iview->image->surface.bpe == 1)
- max_uncompressed_block_size = 0;
+ max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_64B;
else if (iview->image->surface.bpe == 2)
- max_uncompressed_block_size = 1;
+ max_uncompressed_block_size = V_028C78_MAX_BLOCK_SIZE_128B;
}
+ if (iview->image->usage & (VK_IMAGE_USAGE_SAMPLED_BIT | VK_IMAGE_USAGE_TRANSFER_SRC_BIT |
+ VK_IMAGE_USAGE_INPUT_ATTACHMENT_BIT)) {
+ independent_64b_blocks = 1;
+ max_compressed_block_size = V_028C78_MAX_BLOCK_SIZE_64B;
+ } else
+ max_compressed_block_size = max_uncompressed_block_size;
+
cb->cb_dcc_control = S_028C78_MAX_UNCOMPRESSED_BLOCK_SIZE(max_uncompressed_block_size) |
- S_028C78_INDEPENDENT_64B_BLOCKS(1);
+ S_028C78_MAX_COMPRESSED_BLOCK_SIZE(max_compressed_block_size) |
+ S_028C78_MIN_COMPRESSED_BLOCK_SIZE(min_compressed_block_size) |
+ S_028C78_INDEPENDENT_64B_BLOCKS(independent_64b_blocks);
}
/* This must be set for fast clear to work without FMASK. */
cb->cb_color_attrib2 = S_028C68_MIP0_WIDTH(iview->extent.width - 1) |
S_028C68_MIP0_HEIGHT(iview->extent.height - 1) |
S_028C68_MAX_MIP(iview->image->info.levels - 1);
-
- cb->gfx9_epitch = S_0287A0_EPITCH(iview->image->surface.u.gfx9.surf.epitch);
-
}
}
stencil_format = iview->image->surface.has_stencil ?
V_028044_STENCIL_8 : V_028044_STENCIL_INVALID;
- uint32_t max_slice = radv_surface_layer_count(iview);
+ uint32_t max_slice = radv_surface_max_layer_count(iview) - 1;
ds->db_depth_view = S_028008_SLICE_START(iview->base_layer) |
- S_028008_SLICE_MAX(iview->base_layer + max_slice - 1);
+ S_028008_SLICE_MAX(max_slice);
ds->db_htile_data_base = 0;
ds->db_htile_surface = 0;
}
framebuffer->width = MIN2(framebuffer->width, iview->extent.width);
framebuffer->height = MIN2(framebuffer->height, iview->extent.height);
- framebuffer->layers = MIN2(framebuffer->layers, radv_surface_layer_count(iview));
+ framebuffer->layers = MIN2(framebuffer->layers, radv_surface_max_layer_count(iview));
}
*pFramebuffer = radv_framebuffer_to_handle(framebuffer);
S_008F38_XY_MIN_FILTER(radv_tex_filter(pCreateInfo->minFilter, max_aniso)) |
S_008F38_MIP_FILTER(radv_tex_mipfilter(pCreateInfo->mipmapMode)) |
S_008F38_MIP_POINT_PRECLAMP(0) |
- S_008F38_DISABLE_LSB_CEIL(1) |
+ S_008F38_DISABLE_LSB_CEIL(device->physical_device->rad_info.chip_class <= VI) |
S_008F38_FILTER_PREC_FIX(1) |
S_008F38_ANISO_OVERRIDE(is_vi));
sampler->state[3] = (S_008F3C_BORDER_COLOR_PTR(0) |
assert(pGetFdInfo->sType == VK_STRUCTURE_TYPE_MEMORY_GET_FD_INFO_KHR);
- /* We support only one handle type. */
+ /* At the moment, we support only the below handle types. */
assert(pGetFdInfo->handleType ==
- VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT_KHR);
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT_KHR ||
+ pGetFdInfo->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT);
bool ret = radv_get_memory_fd(device, memory, pFD);
if (ret == false)
int fd,
VkMemoryFdPropertiesKHR *pMemoryFdProperties)
{
- /* The valid usage section for this function says:
- *
- * "handleType must not be one of the handle types defined as opaque."
- *
- * Since we only handle opaque handles for now, there are no FD properties.
- */
- return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+ switch (handleType) {
+ case VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT:
+ pMemoryFdProperties->memoryTypeBits = (1 << RADV_MEM_TYPE_COUNT) - 1;
+ return VK_SUCCESS;
+
+ default:
+ /* The valid usage section for this function says:
+ *
+ * "handleType must not be one of the handle types defined as
+ * opaque."
+ *
+ * So opaque handle types fall into the default "unsupported" case.
+ */
+ return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+ }
+}
+
+static VkResult radv_import_opaque_fd(struct radv_device *device,
+ int fd,
+ uint32_t *syncobj)
+{
+ uint32_t syncobj_handle = 0;
+ int ret = device->ws->import_syncobj(device->ws, fd, &syncobj_handle);
+ if (ret != 0)
+ return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+
+ if (*syncobj)
+ device->ws->destroy_syncobj(device->ws, *syncobj);
+
+ *syncobj = syncobj_handle;
+ close(fd);
+
+ return VK_SUCCESS;
+}
+
+static VkResult radv_import_sync_fd(struct radv_device *device,
+ int fd,
+ uint32_t *syncobj)
+{
+ /* If we create a syncobj we do it locally so that if we have an error, we don't
+ * leave a syncobj in an undetermined state in the fence. */
+ uint32_t syncobj_handle = *syncobj;
+ if (!syncobj_handle) {
+ int ret = device->ws->create_syncobj(device->ws, &syncobj_handle);
+ if (ret) {
+ return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+ }
+ }
+
+ if (fd == -1) {
+ device->ws->signal_syncobj(device->ws, syncobj_handle);
+ } else {
+ int ret = device->ws->import_syncobj_from_sync_file(device->ws, syncobj_handle, fd);
+ if (ret != 0)
+ return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+ }
+
+ *syncobj = syncobj_handle;
+ if (fd != -1)
+ close(fd);
+
+ return VK_SUCCESS;
}
VkResult radv_ImportSemaphoreFdKHR(VkDevice _device,
{
RADV_FROM_HANDLE(radv_device, device, _device);
RADV_FROM_HANDLE(radv_semaphore, sem, pImportSemaphoreFdInfo->semaphore);
- uint32_t syncobj_handle = 0;
uint32_t *syncobj_dst = NULL;
- assert(pImportSemaphoreFdInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR);
-
- int ret = device->ws->import_syncobj(device->ws, pImportSemaphoreFdInfo->fd, &syncobj_handle);
- if (ret != 0)
- return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
if (pImportSemaphoreFdInfo->flags & VK_SEMAPHORE_IMPORT_TEMPORARY_BIT_KHR) {
syncobj_dst = &sem->temp_syncobj;
syncobj_dst = &sem->syncobj;
}
- if (*syncobj_dst)
- device->ws->destroy_syncobj(device->ws, *syncobj_dst);
-
- *syncobj_dst = syncobj_handle;
- close(pImportSemaphoreFdInfo->fd);
- return VK_SUCCESS;
+ switch(pImportSemaphoreFdInfo->handleType) {
+ case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR:
+ return radv_import_opaque_fd(device, pImportSemaphoreFdInfo->fd, syncobj_dst);
+ case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT_KHR:
+ return radv_import_sync_fd(device, pImportSemaphoreFdInfo->fd, syncobj_dst);
+ default:
+ unreachable("Unhandled semaphore handle type");
+ }
}
VkResult radv_GetSemaphoreFdKHR(VkDevice _device,
int ret;
uint32_t syncobj_handle;
- assert(pGetFdInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR);
if (sem->temp_syncobj)
syncobj_handle = sem->temp_syncobj;
else
syncobj_handle = sem->syncobj;
- ret = device->ws->export_syncobj(device->ws, syncobj_handle, pFd);
+
+ switch(pGetFdInfo->handleType) {
+ case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR:
+ ret = device->ws->export_syncobj(device->ws, syncobj_handle, pFd);
+ break;
+ case VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT_KHR:
+ ret = device->ws->export_syncobj_to_sync_file(device->ws, syncobj_handle, pFd);
+ break;
+ default:
+ unreachable("Unhandled semaphore handle type");
+ }
+
if (ret)
return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
return VK_SUCCESS;
const VkPhysicalDeviceExternalSemaphoreInfoKHR* pExternalSemaphoreInfo,
VkExternalSemaphorePropertiesKHR* pExternalSemaphoreProperties)
{
- if (pExternalSemaphoreInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR) {
+ RADV_FROM_HANDLE(radv_physical_device, pdevice, physicalDevice);
+
+ /* Require has_syncobj_wait_for_submit for the syncobj signal ioctl introduced at virtually the same time */
+ if (pdevice->rad_info.has_syncobj_wait_for_submit &&
+ (pExternalSemaphoreInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR ||
+ pExternalSemaphoreInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT_KHR)) {
+ pExternalSemaphoreProperties->exportFromImportedHandleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR | VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT_KHR;
+ pExternalSemaphoreProperties->compatibleHandleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR | VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_SYNC_FD_BIT_KHR;
+ pExternalSemaphoreProperties->externalSemaphoreFeatures = VK_EXTERNAL_SEMAPHORE_FEATURE_EXPORTABLE_BIT_KHR |
+ VK_EXTERNAL_SEMAPHORE_FEATURE_IMPORTABLE_BIT_KHR;
+ } else if (pExternalSemaphoreInfo->handleType == VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR) {
pExternalSemaphoreProperties->exportFromImportedHandleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR;
pExternalSemaphoreProperties->compatibleHandleTypes = VK_EXTERNAL_SEMAPHORE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR;
pExternalSemaphoreProperties->externalSemaphoreFeatures = VK_EXTERNAL_SEMAPHORE_FEATURE_EXPORTABLE_BIT_KHR |
pExternalSemaphoreProperties->externalSemaphoreFeatures = 0;
}
}
+
+VkResult radv_ImportFenceFdKHR(VkDevice _device,
+ const VkImportFenceFdInfoKHR *pImportFenceFdInfo)
+{
+ RADV_FROM_HANDLE(radv_device, device, _device);
+ RADV_FROM_HANDLE(radv_fence, fence, pImportFenceFdInfo->fence);
+ uint32_t *syncobj_dst = NULL;
+
+
+ if (pImportFenceFdInfo->flags & VK_FENCE_IMPORT_TEMPORARY_BIT_KHR) {
+ syncobj_dst = &fence->temp_syncobj;
+ } else {
+ syncobj_dst = &fence->syncobj;
+ }
+
+ switch(pImportFenceFdInfo->handleType) {
+ case VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR:
+ return radv_import_opaque_fd(device, pImportFenceFdInfo->fd, syncobj_dst);
+ case VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT_KHR:
+ return radv_import_sync_fd(device, pImportFenceFdInfo->fd, syncobj_dst);
+ default:
+ unreachable("Unhandled fence handle type");
+ }
+}
+
+VkResult radv_GetFenceFdKHR(VkDevice _device,
+ const VkFenceGetFdInfoKHR *pGetFdInfo,
+ int *pFd)
+{
+ RADV_FROM_HANDLE(radv_device, device, _device);
+ RADV_FROM_HANDLE(radv_fence, fence, pGetFdInfo->fence);
+ int ret;
+ uint32_t syncobj_handle;
+
+ if (fence->temp_syncobj)
+ syncobj_handle = fence->temp_syncobj;
+ else
+ syncobj_handle = fence->syncobj;
+
+ switch(pGetFdInfo->handleType) {
+ case VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR:
+ ret = device->ws->export_syncobj(device->ws, syncobj_handle, pFd);
+ break;
+ case VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT_KHR:
+ ret = device->ws->export_syncobj_to_sync_file(device->ws, syncobj_handle, pFd);
+ break;
+ default:
+ unreachable("Unhandled fence handle type");
+ }
+
+ if (ret)
+ return vk_error(VK_ERROR_INVALID_EXTERNAL_HANDLE_KHR);
+ return VK_SUCCESS;
+}
+
+void radv_GetPhysicalDeviceExternalFencePropertiesKHR(
+ VkPhysicalDevice physicalDevice,
+ const VkPhysicalDeviceExternalFenceInfoKHR* pExternalFenceInfo,
+ VkExternalFencePropertiesKHR* pExternalFenceProperties)
+{
+ RADV_FROM_HANDLE(radv_physical_device, pdevice, physicalDevice);
+
+ if (pdevice->rad_info.has_syncobj_wait_for_submit &&
+ (pExternalFenceInfo->handleType == VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR ||
+ pExternalFenceInfo->handleType == VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT_KHR)) {
+ pExternalFenceProperties->exportFromImportedHandleTypes = VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR | VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT_KHR;
+ pExternalFenceProperties->compatibleHandleTypes = VK_EXTERNAL_FENCE_HANDLE_TYPE_OPAQUE_FD_BIT_KHR | VK_EXTERNAL_FENCE_HANDLE_TYPE_SYNC_FD_BIT_KHR;
+ pExternalFenceProperties->externalFenceFeatures = VK_EXTERNAL_FENCE_FEATURE_EXPORTABLE_BIT_KHR |
+ VK_EXTERNAL_SEMAPHORE_FEATURE_IMPORTABLE_BIT_KHR;
+ } else {
+ pExternalFenceProperties->exportFromImportedHandleTypes = 0;
+ pExternalFenceProperties->compatibleHandleTypes = 0;
+ pExternalFenceProperties->externalFenceFeatures = 0;
+ }
+}
projects / mesa.git / blobdiff