* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
- * IN THE SOFTWARE.
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
*/
#include "tu_private.h"
-#include "util/debug.h"
-#include "util/disk_cache.h"
-#include "util/strtod.h"
-#include "vk_format.h"
-#include "vk_util.h"
+
#include <fcntl.h>
+#include <libsync.h>
#include <stdbool.h>
#include <string.h>
+#include <sys/mman.h>
#include <sys/sysinfo.h>
#include <unistd.h>
#include <xf86drm.h>
+#include "compiler/glsl_types.h"
+#include "util/debug.h"
+#include "util/disk_cache.h"
+#include "vk_format.h"
+#include "vk_util.h"
+
+#include "drm-uapi/msm_drm.h"
+
static int
tu_device_get_cache_uuid(uint16_t family, void *uuid)
{
return -1;
memcpy(uuid, &mesa_timestamp, 4);
- memcpy((char *)uuid + 4, &f, 2);
- snprintf((char *)uuid + 6, VK_UUID_SIZE - 10, "tu");
+ memcpy((char *) uuid + 4, &f, 2);
+ snprintf((char *) uuid + 6, VK_UUID_SIZE - 10, "tu");
return 0;
}
tu_get_driver_uuid(void *uuid)
{
memset(uuid, 0, VK_UUID_SIZE);
+ snprintf(uuid, VK_UUID_SIZE, "freedreno");
}
static void
tu_get_device_uuid(void *uuid)
{
- stub();
+ memset(uuid, 0, VK_UUID_SIZE);
+}
+
+static VkResult
+tu_bo_init(struct tu_device *dev,
+ struct tu_bo *bo,
+ uint32_t gem_handle,
+ uint64_t size)
+{
+ uint64_t iova = tu_gem_info_iova(dev, gem_handle);
+ if (!iova)
+ return VK_ERROR_OUT_OF_DEVICE_MEMORY;
+
+ *bo = (struct tu_bo) {
+ .gem_handle = gem_handle,
+ .size = size,
+ .iova = iova,
+ };
+
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_bo_init_new(struct tu_device *dev, struct tu_bo *bo, uint64_t size)
+{
+ /* TODO: Choose better flags. As of 2018-11-12, freedreno/drm/msm_bo.c
+ * always sets `flags = MSM_BO_WC`, and we copy that behavior here.
+ */
+ uint32_t gem_handle = tu_gem_new(dev, size, MSM_BO_WC);
+ if (!gem_handle)
+ return vk_error(dev->instance, VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ VkResult result = tu_bo_init(dev, bo, gem_handle, size);
+ if (result != VK_SUCCESS) {
+ tu_gem_close(dev, gem_handle);
+ return vk_error(dev->instance, result);
+ }
+
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_bo_init_dmabuf(struct tu_device *dev,
+ struct tu_bo *bo,
+ uint64_t size,
+ int fd)
+{
+ uint32_t gem_handle = tu_gem_import_dmabuf(dev, fd, size);
+ if (!gem_handle)
+ return vk_error(dev->instance, VK_ERROR_INVALID_EXTERNAL_HANDLE);
+
+ VkResult result = tu_bo_init(dev, bo, gem_handle, size);
+ if (result != VK_SUCCESS) {
+ tu_gem_close(dev, gem_handle);
+ return vk_error(dev->instance, result);
+ }
+
+ return VK_SUCCESS;
+}
+
+int
+tu_bo_export_dmabuf(struct tu_device *dev, struct tu_bo *bo)
+{
+ return tu_gem_export_dmabuf(dev, bo->gem_handle);
+}
+
+VkResult
+tu_bo_map(struct tu_device *dev, struct tu_bo *bo)
+{
+ if (bo->map)
+ return VK_SUCCESS;
+
+ uint64_t offset = tu_gem_info_offset(dev, bo->gem_handle);
+ if (!offset)
+ return vk_error(dev->instance, VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ /* TODO: Should we use the wrapper os_mmap() like Freedreno does? */
+ void *map = mmap(0, bo->size, PROT_READ | PROT_WRITE, MAP_SHARED,
+ dev->physical_device->local_fd, offset);
+ if (map == MAP_FAILED)
+ return vk_error(dev->instance, VK_ERROR_MEMORY_MAP_FAILED);
+
+ bo->map = map;
+ return VK_SUCCESS;
+}
+
+void
+tu_bo_finish(struct tu_device *dev, struct tu_bo *bo)
+{
+ assert(bo->gem_handle);
+
+ if (bo->map)
+ munmap(bo->map, bo->size);
+
+ tu_gem_close(dev, bo->gem_handle);
}
static VkResult
drmDevicePtr drm_device)
{
const char *path = drm_device->nodes[DRM_NODE_RENDER];
- VkResult result;
+ VkResult result = VK_SUCCESS;
drmVersionPtr version;
int fd;
int master_fd = -1;
- struct fd_pipe *tmp_pipe = NULL;
- uint64_t val;
fd = open(path, O_RDWR | O_CLOEXEC);
if (fd < 0) {
- if (instance->debug_flags & TU_DEBUG_STARTUP)
- tu_logi("Could not open device '%s'", path);
-
- return vk_error(instance, VK_ERROR_INCOMPATIBLE_DRIVER);
+ return vk_errorf(instance, VK_ERROR_INCOMPATIBLE_DRIVER,
+ "failed to open device %s", path);
}
+ /* Version 1.3 added MSM_INFO_IOVA. */
+ const int min_version_major = 1;
+ const int min_version_minor = 3;
+
version = drmGetVersion(fd);
if (!version) {
close(fd);
-
- if (instance->debug_flags & TU_DEBUG_STARTUP)
- tu_logi("Could not get the kernel driver version for device '%s'",
- path);
-
- return vk_errorf(instance,
- VK_ERROR_INCOMPATIBLE_DRIVER,
- "failed to get version %s: %m",
+ return vk_errorf(instance, VK_ERROR_INCOMPATIBLE_DRIVER,
+ "failed to query kernel driver version for device %s",
path);
}
if (strcmp(version->name, "msm")) {
drmFreeVersion(version);
- if (master_fd != -1)
- close(master_fd);
close(fd);
+ return vk_errorf(instance, VK_ERROR_INCOMPATIBLE_DRIVER,
+ "device %s does not use the msm kernel driver", path);
+ }
- if (instance->debug_flags & TU_DEBUG_STARTUP)
- tu_logi("Device '%s' is not using the msm kernel driver.", path);
-
- return VK_ERROR_INCOMPATIBLE_DRIVER;
+ if (version->version_major != min_version_major ||
+ version->version_minor < min_version_minor) {
+ result = vk_errorf(instance, VK_ERROR_INCOMPATIBLE_DRIVER,
+ "kernel driver for device %s has version %d.%d, "
+ "but Vulkan requires version >= %d.%d",
+ path, version->version_major, version->version_minor,
+ min_version_major, min_version_minor);
+ drmFreeVersion(version);
+ close(fd);
+ return result;
}
+
drmFreeVersion(version);
if (instance->debug_flags & TU_DEBUG_STARTUP)
strncpy(device->path, path, ARRAY_SIZE(device->path));
if (instance->enabled_extensions.KHR_display) {
- master_fd = open(drm_device->nodes[DRM_NODE_PRIMARY], O_RDWR | O_CLOEXEC);
+ master_fd =
+ open(drm_device->nodes[DRM_NODE_PRIMARY], O_RDWR | O_CLOEXEC);
if (master_fd >= 0) {
/* TODO: free master_fd is accel is not working? */
}
device->master_fd = master_fd;
device->local_fd = fd;
- device->drm_device = fd_device_new_dup(fd);
- if (!device->drm_device) {
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "could not create the libdrm device");
- goto fail;
- }
-
- tmp_pipe = fd_pipe_new(device->drm_device, FD_PIPE_3D);
- if (!tmp_pipe) {
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "could not open the 3D pipe");
+ if (tu_drm_get_gpu_id(device, &device->gpu_id)) {
+ if (instance->debug_flags & TU_DEBUG_STARTUP)
+ tu_logi("Could not query the GPU ID");
+ result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED,
+ "could not get GPU ID");
goto fail;
}
- if (fd_pipe_get_param(tmp_pipe, FD_GPU_ID, &val)) {
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "could not get GPU ID");
+ if (tu_drm_get_gmem_size(device, &device->gmem_size)) {
+ if (instance->debug_flags & TU_DEBUG_STARTUP)
+ tu_logi("Could not query the GMEM size");
+ result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED,
+ "could not get GMEM size");
goto fail;
}
- device->gpu_id = val;
- if (fd_pipe_get_param(tmp_pipe, FD_GMEM_SIZE, &val)) {
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "could not get GMEM size");
+ if (tu_drm_get_gmem_base(device, &device->gmem_base)) {
+ if (instance->debug_flags & TU_DEBUG_STARTUP)
+ tu_logi("Could not query the GMEM size");
+ result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED,
+ "could not get GMEM size");
goto fail;
}
- device->gmem_size = val;
-
- fd_pipe_del(tmp_pipe);
- tmp_pipe = NULL;
memset(device->name, 0, sizeof(device->name));
sprintf(device->name, "FD%d", device->gpu_id);
- switch(device->gpu_id) {
- case 530:
+ switch (device->gpu_id) {
+ case 618:
+ device->magic.RB_UNKNOWN_8E04_blit = 0x00100000;
+ device->ccu_offset_gmem = 0x7c000; /* 0x7e000 in some cases? */
+ device->ccu_offset_bypass = 0x10000;
+ device->magic.PC_UNKNOWN_9805 = 0x0;
+ device->magic.SP_UNKNOWN_A0F8 = 0x0;
+ break;
+ case 630:
+ case 640:
+ device->magic.RB_UNKNOWN_8E04_blit = 0x01000000;
+ device->ccu_offset_gmem = 0xf8000;
+ device->ccu_offset_bypass = 0x20000;
+ device->magic.PC_UNKNOWN_9805 = 0x1;
+ device->magic.SP_UNKNOWN_A0F8 = 0x1;
break;
default:
- if (instance->debug_flags & TU_DEBUG_STARTUP)
- tu_logi("Device '%s' is not supported.", device->name);
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "unsupported device");
+ result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED,
+ "device %s is unsupported", device->name);
goto fail;
}
if (tu_device_get_cache_uuid(device->gpu_id, device->cache_uuid)) {
- result = vk_errorf(
- instance, VK_ERROR_INITIALIZATION_FAILED, "cannot generate UUID");
+ result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED,
+ "cannot generate UUID");
goto fail;
}
disk_cache_format_hex_id(buf, device->cache_uuid, VK_UUID_SIZE * 2);
device->disk_cache = disk_cache_create(device->name, buf, 0);
- fprintf(stderr,
- "WARNING: tu is not a conformant vulkan implementation, "
- "testing use only.\n");
+ fprintf(stderr, "WARNING: tu is not a conformant vulkan implementation, "
+ "testing use only.\n");
tu_get_driver_uuid(&device->device_uuid);
tu_get_device_uuid(&device->device_uuid);
goto fail;
}
+ result = tu_wsi_init(device);
+ if (result != VK_SUCCESS) {
+ vk_error(instance, result);
+ goto fail;
+ }
+
return VK_SUCCESS;
fail:
- if (tmp_pipe)
- fd_pipe_del(tmp_pipe);
- if (device->drm_device)
- fd_device_del(device->drm_device);
close(fd);
if (master_fd != -1)
close(master_fd);
static void
tu_physical_device_finish(struct tu_physical_device *device)
{
+ tu_wsi_finish(device);
+
disk_cache_destroy(device->disk_cache);
close(device->local_fd);
if (device->master_fd != -1)
close(device->master_fd);
}
-static void *
+static VKAPI_ATTR void *
default_alloc_func(void *pUserData,
size_t size,
size_t align,
return malloc(size);
}
-static void *
+static VKAPI_ATTR void *
default_realloc_func(void *pUserData,
void *pOriginal,
size_t size,
return realloc(pOriginal, size);
}
-static void
+static VKAPI_ATTR void
default_free_func(void *pUserData, void *pMemory)
{
free(pMemory);
.pfnFree = default_free_func,
};
-static const struct debug_control tu_debug_options[] = { { "startup",
- TU_DEBUG_STARTUP },
- { NULL, 0 } };
+static const struct debug_control tu_debug_options[] = {
+ { "startup", TU_DEBUG_STARTUP },
+ { "nir", TU_DEBUG_NIR },
+ { "ir3", TU_DEBUG_IR3 },
+ { "nobin", TU_DEBUG_NOBIN },
+ { "sysmem", TU_DEBUG_SYSMEM },
+ { "forcebin", TU_DEBUG_FORCEBIN },
+ { NULL, 0 }
+};
const char *
tu_get_debug_option_name(int id)
tu_EnumerateInstanceVersion(&client_version);
}
- instance = vk_zalloc2(&default_alloc,
- pAllocator,
- sizeof(*instance),
- 8,
+ instance = vk_zalloc2(&default_alloc, pAllocator, sizeof(*instance), 8,
VK_SYSTEM_ALLOCATION_SCOPE_INSTANCE);
if (!instance)
return vk_error(NULL, VK_ERROR_OUT_OF_HOST_MEMORY);
instance->physical_device_count = -1;
instance->debug_flags =
- parse_debug_string(getenv("TU_DEBUG"), tu_debug_options);
+ parse_debug_string(getenv("TU_DEBUG"), tu_debug_options);
if (instance->debug_flags & TU_DEBUG_STARTUP)
tu_logi("Created an instance");
return vk_error(instance, result);
}
- _mesa_locale_init();
+ glsl_type_singleton_init_or_ref();
VG(VALGRIND_CREATE_MEMPOOL(instance, 0, false));
VG(VALGRIND_DESTROY_MEMPOOL(instance));
- _mesa_locale_fini();
+ glsl_type_singleton_decref();
vk_debug_report_instance_destroy(&instance->debug_report_callbacks);
if (max_devices < 1)
return vk_error(instance, VK_ERROR_INCOMPATIBLE_DRIVER);
- for (unsigned i = 0; i < (unsigned)max_devices; i++) {
+ for (unsigned i = 0; i < (unsigned) max_devices; i++) {
if (devices[i]->available_nodes & 1 << DRM_NODE_RENDER &&
devices[i]->bustype == DRM_BUS_PLATFORM) {
- result = tu_physical_device_init(instance->physical_devices +
- instance->physical_device_count,
- instance,
- devices[i]);
+ result = tu_physical_device_init(
+ instance->physical_devices + instance->physical_device_count,
+ instance, devices[i]);
if (result == VK_SUCCESS)
++instance->physical_device_count;
else if (result != VK_ERROR_INCOMPATIBLE_DRIVER)
}
for (uint32_t i = 0; i < instance->physical_device_count; ++i) {
- vk_outarray_append(&out, p) {
+ vk_outarray_append(&out, p)
+ {
*p = tu_physical_device_to_handle(instance->physical_devices + i);
}
-
}
return vk_outarray_status(&out);
VkPhysicalDeviceGroupProperties *pPhysicalDeviceGroupProperties)
{
TU_FROM_HANDLE(tu_instance, instance, _instance);
- VK_OUTARRAY_MAKE(out, pPhysicalDeviceGroupProperties, pPhysicalDeviceGroupCount);
+ VK_OUTARRAY_MAKE(out, pPhysicalDeviceGroupProperties,
+ pPhysicalDeviceGroupCount);
VkResult result;
if (instance->physical_device_count < 0) {
}
for (uint32_t i = 0; i < instance->physical_device_count; ++i) {
- vk_outarray_append(&out, p) {
+ vk_outarray_append(&out, p)
+ {
p->physicalDeviceCount = 1;
p->physicalDevices[0] =
- tu_physical_device_to_handle(instance->physical_devices + i);
+ tu_physical_device_to_handle(instance->physical_devices + i);
p->subsetAllocation = false;
}
}
{
memset(pFeatures, 0, sizeof(*pFeatures));
- *pFeatures = (VkPhysicalDeviceFeatures){
+ *pFeatures = (VkPhysicalDeviceFeatures) {
.robustBufferAccess = false,
- .fullDrawIndexUint32 = false,
- .imageCubeArray = false,
- .independentBlend = false,
- .geometryShader = false,
+ .fullDrawIndexUint32 = true,
+ .imageCubeArray = true,
+ .independentBlend = true,
+ .geometryShader = true,
.tessellationShader = false,
- .sampleRateShading = false,
- .dualSrcBlend = false,
- .logicOp = false,
+ .sampleRateShading = true,
+ .dualSrcBlend = true,
+ .logicOp = true,
.multiDrawIndirect = false,
.drawIndirectFirstInstance = false,
- .depthClamp = false,
+ .depthClamp = true,
.depthBiasClamp = false,
.fillModeNonSolid = false,
.depthBounds = false,
.largePoints = false,
.alphaToOne = false,
.multiViewport = false,
- .samplerAnisotropy = false,
- .textureCompressionETC2 = false,
- .textureCompressionASTC_LDR = false,
- .textureCompressionBC = false,
- .occlusionQueryPrecise = false,
+ .samplerAnisotropy = true,
+ .textureCompressionETC2 = true,
+ .textureCompressionASTC_LDR = true,
+ .textureCompressionBC = true,
+ .occlusionQueryPrecise = true,
.pipelineStatisticsQuery = false,
.vertexPipelineStoresAndAtomics = false,
.fragmentStoresAndAtomics = false,
void
tu_GetPhysicalDeviceFeatures2(VkPhysicalDevice physicalDevice,
- VkPhysicalDeviceFeatures2KHR *pFeatures)
+ VkPhysicalDeviceFeatures2 *pFeatures)
{
vk_foreach_struct(ext, pFeatures->pNext)
{
switch (ext->sType) {
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTER_FEATURES_KHR: {
- VkPhysicalDeviceVariablePointerFeaturesKHR *features = (void *)ext;
- features->variablePointersStorageBuffer = false;
- features->variablePointers = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES_KHR: {
- VkPhysicalDeviceMultiviewFeaturesKHR *features =
- (VkPhysicalDeviceMultiviewFeaturesKHR *)ext;
- features->multiview = false;
- features->multiviewGeometryShader = false;
- features->multiviewTessellationShader = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETER_FEATURES: {
- VkPhysicalDeviceShaderDrawParameterFeatures *features =
- (VkPhysicalDeviceShaderDrawParameterFeatures *)ext;
- features->shaderDrawParameters = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROTECTED_MEMORY_FEATURES: {
- VkPhysicalDeviceProtectedMemoryFeatures *features =
- (VkPhysicalDeviceProtectedMemoryFeatures *)ext;
- features->protectedMemory = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES: {
- VkPhysicalDevice16BitStorageFeatures *features =
- (VkPhysicalDevice16BitStorageFeatures *)ext;
- features->storageBuffer16BitAccess = false;
- features->uniformAndStorageBuffer16BitAccess = false;
- features->storagePushConstant16 = false;
- features->storageInputOutput16 = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_YCBCR_CONVERSION_FEATURES: {
- VkPhysicalDeviceSamplerYcbcrConversionFeatures *features =
- (VkPhysicalDeviceSamplerYcbcrConversionFeatures *)ext;
- features->samplerYcbcrConversion = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_FEATURES_EXT: {
- VkPhysicalDeviceDescriptorIndexingFeaturesEXT *features =
- (VkPhysicalDeviceDescriptorIndexingFeaturesEXT *)ext;
- features->shaderInputAttachmentArrayDynamicIndexing = false;
- features->shaderUniformTexelBufferArrayDynamicIndexing = false;
- features->shaderStorageTexelBufferArrayDynamicIndexing = false;
- features->shaderUniformBufferArrayNonUniformIndexing = false;
- features->shaderSampledImageArrayNonUniformIndexing = false;
- features->shaderStorageBufferArrayNonUniformIndexing = false;
- features->shaderStorageImageArrayNonUniformIndexing = false;
- features->shaderInputAttachmentArrayNonUniformIndexing = false;
- features->shaderUniformTexelBufferArrayNonUniformIndexing = false;
- features->shaderStorageTexelBufferArrayNonUniformIndexing = false;
- features->descriptorBindingUniformBufferUpdateAfterBind = false;
- features->descriptorBindingSampledImageUpdateAfterBind = false;
- features->descriptorBindingStorageImageUpdateAfterBind = false;
- features->descriptorBindingStorageBufferUpdateAfterBind = false;
- features->descriptorBindingUniformTexelBufferUpdateAfterBind = false;
- features->descriptorBindingStorageTexelBufferUpdateAfterBind = false;
- features->descriptorBindingUpdateUnusedWhilePending = false;
- features->descriptorBindingPartiallyBound = false;
- features->descriptorBindingVariableDescriptorCount = false;
- features->runtimeDescriptorArray = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONDITIONAL_RENDERING_FEATURES_EXT: {
- VkPhysicalDeviceConditionalRenderingFeaturesEXT *features =
- (VkPhysicalDeviceConditionalRenderingFeaturesEXT *)ext;
- features->conditionalRendering = false;
- features->inheritedConditionalRendering = false;
- break;
- }
- default:
- break;
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTERS_FEATURES: {
+ VkPhysicalDeviceVariablePointersFeatures *features = (void *) ext;
+ features->variablePointersStorageBuffer = false;
+ features->variablePointers = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_FEATURES: {
+ VkPhysicalDeviceMultiviewFeatures *features =
+ (VkPhysicalDeviceMultiviewFeatures *) ext;
+ features->multiview = false;
+ features->multiviewGeometryShader = false;
+ features->multiviewTessellationShader = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETERS_FEATURES: {
+ VkPhysicalDeviceShaderDrawParametersFeatures *features =
+ (VkPhysicalDeviceShaderDrawParametersFeatures *) ext;
+ features->shaderDrawParameters = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PROTECTED_MEMORY_FEATURES: {
+ VkPhysicalDeviceProtectedMemoryFeatures *features =
+ (VkPhysicalDeviceProtectedMemoryFeatures *) ext;
+ features->protectedMemory = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES: {
+ VkPhysicalDevice16BitStorageFeatures *features =
+ (VkPhysicalDevice16BitStorageFeatures *) ext;
+ features->storageBuffer16BitAccess = false;
+ features->uniformAndStorageBuffer16BitAccess = false;
+ features->storagePushConstant16 = false;
+ features->storageInputOutput16 = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_YCBCR_CONVERSION_FEATURES: {
+ VkPhysicalDeviceSamplerYcbcrConversionFeatures *features =
+ (VkPhysicalDeviceSamplerYcbcrConversionFeatures *) ext;
+ features->samplerYcbcrConversion = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_DESCRIPTOR_INDEXING_FEATURES_EXT: {
+ VkPhysicalDeviceDescriptorIndexingFeaturesEXT *features =
+ (VkPhysicalDeviceDescriptorIndexingFeaturesEXT *) ext;
+ features->shaderInputAttachmentArrayDynamicIndexing = false;
+ features->shaderUniformTexelBufferArrayDynamicIndexing = false;
+ features->shaderStorageTexelBufferArrayDynamicIndexing = false;
+ features->shaderUniformBufferArrayNonUniformIndexing = false;
+ features->shaderSampledImageArrayNonUniformIndexing = false;
+ features->shaderStorageBufferArrayNonUniformIndexing = false;
+ features->shaderStorageImageArrayNonUniformIndexing = false;
+ features->shaderInputAttachmentArrayNonUniformIndexing = false;
+ features->shaderUniformTexelBufferArrayNonUniformIndexing = false;
+ features->shaderStorageTexelBufferArrayNonUniformIndexing = false;
+ features->descriptorBindingUniformBufferUpdateAfterBind = false;
+ features->descriptorBindingSampledImageUpdateAfterBind = false;
+ features->descriptorBindingStorageImageUpdateAfterBind = false;
+ features->descriptorBindingStorageBufferUpdateAfterBind = false;
+ features->descriptorBindingUniformTexelBufferUpdateAfterBind = false;
+ features->descriptorBindingStorageTexelBufferUpdateAfterBind = false;
+ features->descriptorBindingUpdateUnusedWhilePending = false;
+ features->descriptorBindingPartiallyBound = false;
+ features->descriptorBindingVariableDescriptorCount = false;
+ features->runtimeDescriptorArray = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_CONDITIONAL_RENDERING_FEATURES_EXT: {
+ VkPhysicalDeviceConditionalRenderingFeaturesEXT *features =
+ (VkPhysicalDeviceConditionalRenderingFeaturesEXT *) ext;
+ features->conditionalRendering = false;
+ features->inheritedConditionalRendering = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_FEATURES_EXT: {
+ VkPhysicalDeviceTransformFeedbackFeaturesEXT *features =
+ (VkPhysicalDeviceTransformFeedbackFeaturesEXT *) ext;
+ features->transformFeedback = true;
+ features->geometryStreams = false;
+ break;
+ }
+ default:
+ break;
}
}
return tu_GetPhysicalDeviceFeatures(physicalDevice, &pFeatures->features);
VkPhysicalDeviceProperties *pProperties)
{
TU_FROM_HANDLE(tu_physical_device, pdevice, physicalDevice);
- VkSampleCountFlags sample_counts = 0xf;
-
- /* make sure that the entire descriptor set is addressable with a signed
- * 32-bit int. So the sum of all limits scaled by descriptor size has to
- * be at most 2 GiB. the combined image & samples object count as one of
- * both. This limit is for the pipeline layout, not for the set layout, but
- * there is no set limit, so we just set a pipeline limit. I don't think
- * any app is going to hit this soon. */
- size_t max_descriptor_set_size =
- ((1ull << 31) - 16 * MAX_DYNAMIC_BUFFERS) /
- (32 /* uniform buffer, 32 due to potential space wasted on alignment */ +
- 32 /* storage buffer, 32 due to potential space wasted on alignment */ +
- 32 /* sampler, largest when combined with image */ +
- 64 /* sampled image */ + 64 /* storage image */);
+ VkSampleCountFlags sample_counts =
+ VK_SAMPLE_COUNT_1_BIT | VK_SAMPLE_COUNT_2_BIT | VK_SAMPLE_COUNT_4_BIT;
+
+ /* I have no idea what the maximum size is, but the hardware supports very
+ * large numbers of descriptors (at least 2^16). This limit is based on
+ * CP_LOAD_STATE6, which has a 28-bit field for the DWORD offset, so that
+ * we don't have to think about what to do if that overflows, but really
+ * nothing is likely to get close to this.
+ */
+ const size_t max_descriptor_set_size = (1 << 28) / A6XX_TEX_CONST_DWORDS;
VkPhysicalDeviceLimits limits = {
.maxImageDimension1D = (1 << 14),
.maxImageDimensionCube = (1 << 14),
.maxImageArrayLayers = (1 << 11),
.maxTexelBufferElements = 128 * 1024 * 1024,
- .maxUniformBufferRange = UINT32_MAX,
- .maxStorageBufferRange = UINT32_MAX,
+ .maxUniformBufferRange = MAX_UNIFORM_BUFFER_RANGE,
+ .maxStorageBufferRange = MAX_STORAGE_BUFFER_RANGE,
.maxPushConstantsSize = MAX_PUSH_CONSTANTS_SIZE,
.maxMemoryAllocationCount = UINT32_MAX,
.maxSamplerAllocationCount = 64 * 1024,
.maxPerStageDescriptorStorageBuffers = max_descriptor_set_size,
.maxPerStageDescriptorSampledImages = max_descriptor_set_size,
.maxPerStageDescriptorStorageImages = max_descriptor_set_size,
- .maxPerStageDescriptorInputAttachments = max_descriptor_set_size,
+ .maxPerStageDescriptorInputAttachments = MAX_RTS,
.maxPerStageResources = max_descriptor_set_size,
.maxDescriptorSetSamplers = max_descriptor_set_size,
.maxDescriptorSetUniformBuffers = max_descriptor_set_size,
.maxDescriptorSetStorageBuffersDynamic = MAX_DYNAMIC_STORAGE_BUFFERS,
.maxDescriptorSetSampledImages = max_descriptor_set_size,
.maxDescriptorSetStorageImages = max_descriptor_set_size,
- .maxDescriptorSetInputAttachments = max_descriptor_set_size,
+ .maxDescriptorSetInputAttachments = MAX_RTS,
.maxVertexInputAttributes = 32,
.maxVertexInputBindings = 32,
- .maxVertexInputAttributeOffset = 2047,
+ .maxVertexInputAttributeOffset = 4095,
.maxVertexInputBindingStride = 2048,
.maxVertexOutputComponents = 128,
.maxTessellationGenerationLevel = 64,
.maxTessellationControlTotalOutputComponents = 4096,
.maxTessellationEvaluationInputComponents = 128,
.maxTessellationEvaluationOutputComponents = 128,
- .maxGeometryShaderInvocations = 127,
+ .maxGeometryShaderInvocations = 32,
.maxGeometryInputComponents = 64,
.maxGeometryOutputComponents = 128,
.maxGeometryOutputVertices = 256,
.maxGeometryTotalOutputComponents = 1024,
- .maxFragmentInputComponents = 128,
+ .maxFragmentInputComponents = 124,
.maxFragmentOutputAttachments = 8,
.maxFragmentDualSrcAttachments = 1,
.maxFragmentCombinedOutputResources = 8,
.maxComputeWorkGroupCount = { 65535, 65535, 65535 },
.maxComputeWorkGroupInvocations = 2048,
.maxComputeWorkGroupSize = { 2048, 2048, 2048 },
- .subPixelPrecisionBits = 4 /* FIXME */,
+ .subPixelPrecisionBits = 8,
.subTexelPrecisionBits = 4 /* FIXME */,
.mipmapPrecisionBits = 4 /* FIXME */,
.maxDrawIndexedIndexValue = UINT32_MAX,
.viewportBoundsRange = { INT16_MIN, INT16_MAX },
.viewportSubPixelBits = 8,
.minMemoryMapAlignment = 4096, /* A page */
- .minTexelBufferOffsetAlignment = 1,
- .minUniformBufferOffsetAlignment = 4,
- .minStorageBufferOffsetAlignment = 4,
+ .minTexelBufferOffsetAlignment = 64,
+ .minUniformBufferOffsetAlignment = 64,
+ .minStorageBufferOffsetAlignment = 64,
.minTexelOffset = -32,
.maxTexelOffset = 31,
.minTexelGatherOffset = -32,
.storageImageSampleCounts = VK_SAMPLE_COUNT_1_BIT,
.maxSampleMaskWords = 1,
.timestampComputeAndGraphics = true,
- .timestampPeriod = 1,
+ .timestampPeriod = 1000000000.0 / 19200000.0, /* CP_ALWAYS_ON_COUNTER is fixed 19.2MHz */
.maxClipDistances = 8,
.maxCullDistances = 8,
.maxCombinedClipAndCullDistances = 8,
.nonCoherentAtomSize = 64,
};
- *pProperties = (VkPhysicalDeviceProperties){
+ *pProperties = (VkPhysicalDeviceProperties) {
.apiVersion = tu_physical_device_api_version(pdevice),
.driverVersion = vk_get_driver_version(),
.vendorID = 0, /* TODO */
void
tu_GetPhysicalDeviceProperties2(VkPhysicalDevice physicalDevice,
- VkPhysicalDeviceProperties2KHR *pProperties)
+ VkPhysicalDeviceProperties2 *pProperties)
{
TU_FROM_HANDLE(tu_physical_device, pdevice, physicalDevice);
tu_GetPhysicalDeviceProperties(physicalDevice, &pProperties->properties);
vk_foreach_struct(ext, pProperties->pNext)
{
switch (ext->sType) {
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR: {
- VkPhysicalDevicePushDescriptorPropertiesKHR *properties =
- (VkPhysicalDevicePushDescriptorPropertiesKHR *)ext;
- properties->maxPushDescriptors = MAX_PUSH_DESCRIPTORS;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ID_PROPERTIES_KHR: {
- VkPhysicalDeviceIDPropertiesKHR *properties =
- (VkPhysicalDeviceIDPropertiesKHR *)ext;
- memcpy(properties->driverUUID, pdevice->driver_uuid, VK_UUID_SIZE);
- memcpy(properties->deviceUUID, pdevice->device_uuid, VK_UUID_SIZE);
- properties->deviceLUIDValid = false;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_PROPERTIES_KHR: {
- VkPhysicalDeviceMultiviewPropertiesKHR *properties =
- (VkPhysicalDeviceMultiviewPropertiesKHR *)ext;
- properties->maxMultiviewViewCount = MAX_VIEWS;
- properties->maxMultiviewInstanceIndex = INT_MAX;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_POINT_CLIPPING_PROPERTIES_KHR: {
- VkPhysicalDevicePointClippingPropertiesKHR *properties =
- (VkPhysicalDevicePointClippingPropertiesKHR *)ext;
- properties->pointClippingBehavior =
- VK_POINT_CLIPPING_BEHAVIOR_ALL_CLIP_PLANES_KHR;
- break;
- }
- case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_3_PROPERTIES: {
- VkPhysicalDeviceMaintenance3Properties *properties =
- (VkPhysicalDeviceMaintenance3Properties *)ext;
- /* Make sure everything is addressable by a signed 32-bit int, and
- * our largest descriptors are 96 bytes. */
- properties->maxPerSetDescriptors = (1ull << 31) / 96;
- /* Our buffer size fields allow only this much */
- properties->maxMemoryAllocationSize = 0xFFFFFFFFull;
- break;
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_PUSH_DESCRIPTOR_PROPERTIES_KHR: {
+ VkPhysicalDevicePushDescriptorPropertiesKHR *properties =
+ (VkPhysicalDevicePushDescriptorPropertiesKHR *) ext;
+ properties->maxPushDescriptors = MAX_PUSH_DESCRIPTORS;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ID_PROPERTIES: {
+ VkPhysicalDeviceIDProperties *properties =
+ (VkPhysicalDeviceIDProperties *) ext;
+ memcpy(properties->driverUUID, pdevice->driver_uuid, VK_UUID_SIZE);
+ memcpy(properties->deviceUUID, pdevice->device_uuid, VK_UUID_SIZE);
+ properties->deviceLUIDValid = false;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MULTIVIEW_PROPERTIES: {
+ VkPhysicalDeviceMultiviewProperties *properties =
+ (VkPhysicalDeviceMultiviewProperties *) ext;
+ properties->maxMultiviewViewCount = MAX_VIEWS;
+ properties->maxMultiviewInstanceIndex = INT_MAX;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_POINT_CLIPPING_PROPERTIES: {
+ VkPhysicalDevicePointClippingProperties *properties =
+ (VkPhysicalDevicePointClippingProperties *) ext;
+ properties->pointClippingBehavior =
+ VK_POINT_CLIPPING_BEHAVIOR_ALL_CLIP_PLANES;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MAINTENANCE_3_PROPERTIES: {
+ VkPhysicalDeviceMaintenance3Properties *properties =
+ (VkPhysicalDeviceMaintenance3Properties *) ext;
+ /* Make sure everything is addressable by a signed 32-bit int, and
+ * our largest descriptors are 96 bytes. */
+ properties->maxPerSetDescriptors = (1ull << 31) / 96;
+ /* Our buffer size fields allow only this much */
+ properties->maxMemoryAllocationSize = 0xFFFFFFFFull;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_TRANSFORM_FEEDBACK_PROPERTIES_EXT: {
+ VkPhysicalDeviceTransformFeedbackPropertiesEXT *properties =
+ (VkPhysicalDeviceTransformFeedbackPropertiesEXT *)ext;
+
+ properties->maxTransformFeedbackStreams = IR3_MAX_SO_STREAMS;
+ properties->maxTransformFeedbackBuffers = IR3_MAX_SO_BUFFERS;
+ properties->maxTransformFeedbackBufferSize = UINT32_MAX;
+ properties->maxTransformFeedbackStreamDataSize = 512;
+ properties->maxTransformFeedbackBufferDataSize = 512;
+ properties->maxTransformFeedbackBufferDataStride = 512;
+ properties->transformFeedbackQueries = true;
+ properties->transformFeedbackStreamsLinesTriangles = false;
+ properties->transformFeedbackRasterizationStreamSelect = false;
+ properties->transformFeedbackDraw = true;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLE_LOCATIONS_PROPERTIES_EXT: {
+ VkPhysicalDeviceSampleLocationsPropertiesEXT *properties =
+ (VkPhysicalDeviceSampleLocationsPropertiesEXT *)ext;
+ properties->sampleLocationSampleCounts = 0;
+ if (pdevice->supported_extensions.EXT_sample_locations) {
+ properties->sampleLocationSampleCounts =
+ VK_SAMPLE_COUNT_1_BIT | VK_SAMPLE_COUNT_2_BIT | VK_SAMPLE_COUNT_4_BIT;
}
- default:
- break;
+ properties->maxSampleLocationGridSize = (VkExtent2D) { 1 , 1 };
+ properties->sampleLocationCoordinateRange[0] = 0.0f;
+ properties->sampleLocationCoordinateRange[1] = 0.9375f;
+ properties->sampleLocationSubPixelBits = 4;
+ properties->variableSampleLocations = true;
+ break;
+ }
+ case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SAMPLER_FILTER_MINMAX_PROPERTIES: {
+ VkPhysicalDeviceSamplerFilterMinmaxProperties *properties =
+ (VkPhysicalDeviceSamplerFilterMinmaxProperties *)ext;
+ properties->filterMinmaxImageComponentMapping = true;
+ properties->filterMinmaxSingleComponentFormats = true;
+ break;
+ }
+
+ default:
+ break;
}
}
}
-static const VkQueueFamilyProperties
-tu_queue_family_properties = {
- .queueFlags = VK_QUEUE_GRAPHICS_BIT |
- VK_QUEUE_COMPUTE_BIT |
- VK_QUEUE_TRANSFER_BIT,
+static const VkQueueFamilyProperties tu_queue_family_properties = {
+ .queueFlags =
+ VK_QUEUE_GRAPHICS_BIT | VK_QUEUE_COMPUTE_BIT | VK_QUEUE_TRANSFER_BIT,
.queueCount = 1,
- .timestampValidBits = 64,
- .minImageTransferGranularity = (VkExtent3D) { 1, 1, 1 },
+ .timestampValidBits = 48,
+ .minImageTransferGranularity = { 1, 1, 1 },
};
void
{
VK_OUTARRAY_MAKE(out, pQueueFamilyProperties, pQueueFamilyPropertyCount);
- vk_outarray_append(&out, p) {
- *p = tu_queue_family_properties;
- }
+ vk_outarray_append(&out, p) { *p = tu_queue_family_properties; }
}
void
tu_GetPhysicalDeviceQueueFamilyProperties2(
VkPhysicalDevice physicalDevice,
uint32_t *pQueueFamilyPropertyCount,
- VkQueueFamilyProperties2KHR *pQueueFamilyProperties)
+ VkQueueFamilyProperties2 *pQueueFamilyProperties)
{
VK_OUTARRAY_MAKE(out, pQueueFamilyProperties, pQueueFamilyPropertyCount);
- vk_outarray_append(&out, p) {
+ vk_outarray_append(&out, p)
+ {
p->queueFamilyProperties = tu_queue_family_properties;
}
}
struct sysinfo info;
sysinfo(&info);
- uint64_t total_ram = (uint64_t)info.totalram * (uint64_t)info.mem_unit;
+ uint64_t total_ram = (uint64_t) info.totalram * (uint64_t) info.mem_unit;
/* We don't want to burn too much ram with the GPU. If the user has 4GiB
* or less, we use at most half. If they have more than 4GiB, we use 3/4.
pMemoryProperties->memoryHeaps[0].flags = VK_MEMORY_HEAP_DEVICE_LOCAL_BIT;
pMemoryProperties->memoryTypeCount = 1;
- pMemoryProperties->memoryTypes[0].propertyFlags = VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
- VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
- VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
+ pMemoryProperties->memoryTypes[0].propertyFlags =
+ VK_MEMORY_PROPERTY_DEVICE_LOCAL_BIT |
+ VK_MEMORY_PROPERTY_HOST_VISIBLE_BIT |
+ VK_MEMORY_PROPERTY_HOST_COHERENT_BIT;
pMemoryProperties->memoryTypes[0].heapIndex = 0;
}
void
tu_GetPhysicalDeviceMemoryProperties2(
VkPhysicalDevice physicalDevice,
- VkPhysicalDeviceMemoryProperties2KHR *pMemoryProperties)
+ VkPhysicalDeviceMemoryProperties2 *pMemoryProperties)
{
return tu_GetPhysicalDeviceMemoryProperties(
- physicalDevice, &pMemoryProperties->memoryProperties);
+ physicalDevice, &pMemoryProperties->memoryProperties);
}
-static int
+static VkResult
tu_queue_init(struct tu_device *device,
struct tu_queue *queue,
uint32_t queue_family_index,
queue->queue_idx = idx;
queue->flags = flags;
+ int ret = tu_drm_submitqueue_new(device, 0, &queue->msm_queue_id);
+ if (ret)
+ return VK_ERROR_INITIALIZATION_FAILED;
+
+ tu_fence_init(&queue->submit_fence, false);
+
return VK_SUCCESS;
}
static void
tu_queue_finish(struct tu_queue *queue)
{
+ tu_fence_finish(&queue->submit_fence);
+ tu_drm_submitqueue_close(queue->device, queue->msm_queue_id);
}
static int
return -1;
}
+struct PACKED bcolor_entry {
+ uint32_t fp32[4];
+ uint16_t ui16[4];
+ int16_t si16[4];
+ uint16_t fp16[4];
+ uint16_t rgb565;
+ uint16_t rgb5a1;
+ uint16_t rgba4;
+ uint8_t __pad0[2];
+ uint8_t ui8[4];
+ int8_t si8[4];
+ uint32_t rgb10a2;
+ uint32_t z24; /* also s8? */
+ uint16_t srgb[4]; /* appears to duplicate fp16[], but clamped, used for srgb */
+ uint8_t __pad1[56];
+} border_color[] = {
+ [VK_BORDER_COLOR_FLOAT_TRANSPARENT_BLACK] = {},
+ [VK_BORDER_COLOR_INT_TRANSPARENT_BLACK] = {},
+ [VK_BORDER_COLOR_FLOAT_OPAQUE_BLACK] = {
+ .fp32[3] = 0x3f800000,
+ .ui16[3] = 0xffff,
+ .si16[3] = 0x7fff,
+ .fp16[3] = 0x3c00,
+ .rgb5a1 = 0x8000,
+ .rgba4 = 0xf000,
+ .ui8[3] = 0xff,
+ .si8[3] = 0x7f,
+ .rgb10a2 = 0xc0000000,
+ .srgb[3] = 0x3c00,
+ },
+ [VK_BORDER_COLOR_INT_OPAQUE_BLACK] = {
+ .fp32[3] = 1,
+ .fp16[3] = 1,
+ },
+ [VK_BORDER_COLOR_FLOAT_OPAQUE_WHITE] = {
+ .fp32[0 ... 3] = 0x3f800000,
+ .ui16[0 ... 3] = 0xffff,
+ .si16[0 ... 3] = 0x7fff,
+ .fp16[0 ... 3] = 0x3c00,
+ .rgb565 = 0xffff,
+ .rgb5a1 = 0xffff,
+ .rgba4 = 0xffff,
+ .ui8[0 ... 3] = 0xff,
+ .si8[0 ... 3] = 0x7f,
+ .rgb10a2 = 0xffffffff,
+ .z24 = 0xffffff,
+ .srgb[0 ... 3] = 0x3c00,
+ },
+ [VK_BORDER_COLOR_INT_OPAQUE_WHITE] = {
+ .fp32[0 ... 3] = 1,
+ .fp16[0 ... 3] = 1,
+ },
+};
+
+
VkResult
tu_CreateDevice(VkPhysicalDevice physicalDevice,
const VkDeviceCreateInfo *pCreateInfo,
if (pCreateInfo->pEnabledFeatures) {
VkPhysicalDeviceFeatures supported_features;
tu_GetPhysicalDeviceFeatures(physicalDevice, &supported_features);
- VkBool32 *supported_feature = (VkBool32 *)&supported_features;
- VkBool32 *enabled_feature = (VkBool32 *)pCreateInfo->pEnabledFeatures;
+ VkBool32 *supported_feature = (VkBool32 *) &supported_features;
+ VkBool32 *enabled_feature = (VkBool32 *) pCreateInfo->pEnabledFeatures;
unsigned num_features =
- sizeof(VkPhysicalDeviceFeatures) / sizeof(VkBool32);
+ sizeof(VkPhysicalDeviceFeatures) / sizeof(VkBool32);
for (uint32_t i = 0; i < num_features; i++) {
if (enabled_feature[i] && !supported_feature[i])
return vk_error(physical_device->instance,
}
}
- device = vk_zalloc2(&physical_device->instance->alloc,
- pAllocator,
- sizeof(*device),
- 8,
- VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
+ device = vk_zalloc2(&physical_device->instance->alloc, pAllocator,
+ sizeof(*device), 8, VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
if (!device)
return vk_error(physical_device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
for (unsigned i = 0; i < pCreateInfo->queueCreateInfoCount; i++) {
const VkDeviceQueueCreateInfo *queue_create =
- &pCreateInfo->pQueueCreateInfos[i];
+ &pCreateInfo->pQueueCreateInfos[i];
uint32_t qfi = queue_create->queueFamilyIndex;
- device->queues[qfi] =
- vk_alloc(&device->alloc,
- queue_create->queueCount * sizeof(struct tu_queue),
- 8,
- VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
+ device->queues[qfi] = vk_alloc(
+ &device->alloc, queue_create->queueCount * sizeof(struct tu_queue),
+ 8, VK_SYSTEM_ALLOCATION_SCOPE_DEVICE);
if (!device->queues[qfi]) {
result = VK_ERROR_OUT_OF_HOST_MEMORY;
- goto fail;
+ goto fail_queues;
}
- memset(device->queues[qfi],
- 0,
+ memset(device->queues[qfi], 0,
queue_create->queueCount * sizeof(struct tu_queue));
device->queue_count[qfi] = queue_create->queueCount;
for (unsigned q = 0; q < queue_create->queueCount; q++) {
- result = tu_queue_init(
- device, &device->queues[qfi][q], qfi, q, queue_create->flags);
+ result = tu_queue_init(device, &device->queues[qfi][q], qfi, q,
+ queue_create->flags);
if (result != VK_SUCCESS)
- goto fail;
+ goto fail_queues;
}
}
+ device->compiler = ir3_compiler_create(NULL, physical_device->gpu_id);
+ if (!device->compiler)
+ goto fail_queues;
+
+#define VSC_DATA_SIZE(pitch) ((pitch) * 32 + 0x100) /* extra size to store VSC_SIZE */
+#define VSC_DATA2_SIZE(pitch) ((pitch) * 32)
+
+ device->vsc_data_pitch = 0x440 * 4;
+ device->vsc_data2_pitch = 0x1040 * 4;
+
+ result = tu_bo_init_new(device, &device->vsc_data, VSC_DATA_SIZE(device->vsc_data_pitch));
+ if (result != VK_SUCCESS)
+ goto fail_vsc_data;
+
+ result = tu_bo_init_new(device, &device->vsc_data2, VSC_DATA2_SIZE(device->vsc_data2_pitch));
+ if (result != VK_SUCCESS)
+ goto fail_vsc_data2;
+
+ STATIC_ASSERT(sizeof(struct bcolor_entry) == 128);
+ result = tu_bo_init_new(device, &device->border_color, sizeof(border_color));
+ if (result != VK_SUCCESS)
+ goto fail_border_color;
+
+ result = tu_bo_map(device, &device->border_color);
+ if (result != VK_SUCCESS)
+ goto fail_border_color_map;
+
+ memcpy(device->border_color.map, border_color, sizeof(border_color));
+
VkPipelineCacheCreateInfo ci;
ci.sType = VK_STRUCTURE_TYPE_PIPELINE_CACHE_CREATE_INFO;
ci.pNext = NULL;
ci.initialDataSize = 0;
VkPipelineCache pc;
result =
- tu_CreatePipelineCache(tu_device_to_handle(device), &ci, NULL, &pc);
+ tu_CreatePipelineCache(tu_device_to_handle(device), &ci, NULL, &pc);
if (result != VK_SUCCESS)
- goto fail;
+ goto fail_pipeline_cache;
device->mem_cache = tu_pipeline_cache_from_handle(pc);
*pDevice = tu_device_to_handle(device);
return VK_SUCCESS;
-fail:
+fail_pipeline_cache:
+fail_border_color_map:
+ tu_bo_finish(device, &device->border_color);
+
+fail_border_color:
+ tu_bo_finish(device, &device->vsc_data2);
+
+fail_vsc_data2:
+ tu_bo_finish(device, &device->vsc_data);
+
+fail_vsc_data:
+ ralloc_free(device->compiler);
+
+fail_queues:
for (unsigned i = 0; i < TU_MAX_QUEUE_FAMILIES; i++) {
for (unsigned q = 0; q < device->queue_count[i]; q++)
tu_queue_finish(&device->queues[i][q]);
if (!device)
return;
+ tu_bo_finish(device, &device->vsc_data);
+ tu_bo_finish(device, &device->vsc_data2);
+
for (unsigned i = 0; i < TU_MAX_QUEUE_FAMILIES; i++) {
for (unsigned q = 0; q < device->queue_count[i]; q++)
tu_queue_finish(&device->queues[i][q]);
vk_free(&device->alloc, device->queues[i]);
}
+ /* the compiler does not use pAllocator */
+ ralloc_free(device->compiler);
+
VkPipelineCache pc = tu_pipeline_cache_to_handle(device->mem_cache);
tu_DestroyPipelineCache(tu_device_to_handle(device), pc, NULL);
struct tu_queue *queue;
queue =
- &device->queues[pQueueInfo->queueFamilyIndex][pQueueInfo->queueIndex];
+ &device->queues[pQueueInfo->queueFamilyIndex][pQueueInfo->queueIndex];
if (pQueueInfo->flags != queue->flags) {
/* From the Vulkan 1.1.70 spec:
*
VkQueue *pQueue)
{
const VkDeviceQueueInfo2 info =
- (VkDeviceQueueInfo2){.sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_INFO_2,
- .queueFamilyIndex = queueFamilyIndex,
- .queueIndex = queueIndex };
+ (VkDeviceQueueInfo2) { .sType = VK_STRUCTURE_TYPE_DEVICE_QUEUE_INFO_2,
+ .queueFamilyIndex = queueFamilyIndex,
+ .queueIndex = queueIndex };
tu_GetDeviceQueue2(_device, &info, pQueue);
}
const VkSubmitInfo *pSubmits,
VkFence _fence)
{
+ TU_FROM_HANDLE(tu_queue, queue, _queue);
+
+ for (uint32_t i = 0; i < submitCount; ++i) {
+ const VkSubmitInfo *submit = pSubmits + i;
+ const bool last_submit = (i == submitCount - 1);
+ struct tu_bo_list bo_list;
+ tu_bo_list_init(&bo_list);
+
+ uint32_t entry_count = 0;
+ for (uint32_t j = 0; j < submit->commandBufferCount; ++j) {
+ TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, submit->pCommandBuffers[j]);
+ entry_count += cmdbuf->cs.entry_count;
+ }
+
+ struct drm_msm_gem_submit_cmd cmds[entry_count];
+ uint32_t entry_idx = 0;
+ for (uint32_t j = 0; j < submit->commandBufferCount; ++j) {
+ TU_FROM_HANDLE(tu_cmd_buffer, cmdbuf, submit->pCommandBuffers[j]);
+ struct tu_cs *cs = &cmdbuf->cs;
+ for (unsigned i = 0; i < cs->entry_count; ++i, ++entry_idx) {
+ cmds[entry_idx].type = MSM_SUBMIT_CMD_BUF;
+ cmds[entry_idx].submit_idx =
+ tu_bo_list_add(&bo_list, cs->entries[i].bo,
+ MSM_SUBMIT_BO_READ | MSM_SUBMIT_BO_DUMP);
+ cmds[entry_idx].submit_offset = cs->entries[i].offset;
+ cmds[entry_idx].size = cs->entries[i].size;
+ cmds[entry_idx].pad = 0;
+ cmds[entry_idx].nr_relocs = 0;
+ cmds[entry_idx].relocs = 0;
+ }
+
+ tu_bo_list_merge(&bo_list, &cmdbuf->bo_list);
+ }
+
+ uint32_t flags = MSM_PIPE_3D0;
+ if (last_submit) {
+ flags |= MSM_SUBMIT_FENCE_FD_OUT;
+ }
+
+ struct drm_msm_gem_submit req = {
+ .flags = flags,
+ .queueid = queue->msm_queue_id,
+ .bos = (uint64_t)(uintptr_t) bo_list.bo_infos,
+ .nr_bos = bo_list.count,
+ .cmds = (uint64_t)(uintptr_t)cmds,
+ .nr_cmds = entry_count,
+ };
+
+ int ret = drmCommandWriteRead(queue->device->physical_device->local_fd,
+ DRM_MSM_GEM_SUBMIT,
+ &req, sizeof(req));
+ if (ret) {
+ fprintf(stderr, "submit failed: %s\n", strerror(errno));
+ abort();
+ }
+
+ tu_bo_list_destroy(&bo_list);
+
+ if (last_submit) {
+ /* no need to merge fences as queue execution is serialized */
+ tu_fence_update_fd(&queue->submit_fence, req.fence_fd);
+ }
+ }
+
+ if (_fence != VK_NULL_HANDLE) {
+ TU_FROM_HANDLE(tu_fence, fence, _fence);
+ tu_fence_copy(fence, &queue->submit_fence);
+ }
+
return VK_SUCCESS;
}
VkResult
tu_QueueWaitIdle(VkQueue _queue)
{
+ TU_FROM_HANDLE(tu_queue, queue, _queue);
+
+ tu_fence_wait_idle(&queue->submit_fence);
+
return VK_SUCCESS;
}
return VK_SUCCESS;
}
+VkResult
+tu_ImportSemaphoreFdKHR(VkDevice _device,
+ const VkImportSemaphoreFdInfoKHR *pImportSemaphoreFdInfo)
+{
+ tu_stub();
+
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_GetSemaphoreFdKHR(VkDevice _device,
+ const VkSemaphoreGetFdInfoKHR *pGetFdInfo,
+ int *pFd)
+{
+ tu_stub();
+
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_ImportFenceFdKHR(VkDevice _device,
+ const VkImportFenceFdInfoKHR *pImportFenceFdInfo)
+{
+ tu_stub();
+
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_GetFenceFdKHR(VkDevice _device,
+ const VkFenceGetFdInfoKHR *pGetFdInfo,
+ int *pFd)
+{
+ tu_stub();
+
+ return VK_SUCCESS;
+}
+
VkResult
tu_EnumerateInstanceExtensionProperties(const char *pLayerName,
uint32_t *pPropertyCount,
{
TU_FROM_HANDLE(tu_instance, instance, _instance);
- return tu_lookup_entrypoint_checked(pName,
- instance ? instance->api_version : 0,
- instance ? &instance->enabled_extensions
- : NULL,
- NULL);
+ return tu_lookup_entrypoint_checked(
+ pName, instance ? instance->api_version : 0,
+ instance ? &instance->enabled_extensions : NULL, NULL);
}
/* The loader wants us to expose a second GetInstanceProcAddr function
{
TU_FROM_HANDLE(tu_device, device, _device);
- return tu_lookup_entrypoint_checked(pName,
- device->instance->api_version,
- &device->instance->enabled_extensions,
- &device->enabled_extensions);
+ return tu_lookup_entrypoint_checked(pName, device->instance->api_version,
+ &device->instance->enabled_extensions,
+ &device->enabled_extensions);
}
static VkResult
VkDeviceMemory *pMem)
{
struct tu_device_memory *mem;
+ VkResult result;
assert(pAllocateInfo->sType == VK_STRUCTURE_TYPE_MEMORY_ALLOCATE_INFO);
return VK_SUCCESS;
}
- mem = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*mem),
- 8,
+ mem = vk_alloc2(&device->alloc, pAllocator, sizeof(*mem), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (mem == NULL)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
- mem->bo = fd_bo_new(device->physical_device->drm_device, pAllocateInfo->allocationSize,
- DRM_FREEDRENO_GEM_CACHE_WCOMBINE |
- DRM_FREEDRENO_GEM_TYPE_KMEM);
- if (!mem->bo) {
+ const VkImportMemoryFdInfoKHR *fd_info =
+ vk_find_struct_const(pAllocateInfo->pNext, IMPORT_MEMORY_FD_INFO_KHR);
+ if (fd_info && !fd_info->handleType)
+ fd_info = NULL;
+
+ if (fd_info) {
+ assert(fd_info->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT ||
+ fd_info->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT);
+
+ /*
+ * TODO Importing the same fd twice gives us the same handle without
+ * reference counting. We need to maintain a per-instance handle-to-bo
+ * table and add reference count to tu_bo.
+ */
+ result = tu_bo_init_dmabuf(device, &mem->bo,
+ pAllocateInfo->allocationSize, fd_info->fd);
+ if (result == VK_SUCCESS) {
+ /* take ownership and close the fd */
+ close(fd_info->fd);
+ }
+ } else {
+ result =
+ tu_bo_init_new(device, &mem->bo, pAllocateInfo->allocationSize);
+ }
+
+ if (result != VK_SUCCESS) {
vk_free2(&device->alloc, pAllocator, mem);
- return vk_error(device->instance, VK_ERROR_OUT_OF_DEVICE_MEMORY);
+ return result;
}
+
mem->size = pAllocateInfo->allocationSize;
mem->type_index = pAllocateInfo->memoryTypeIndex;
if (mem == NULL)
return;
- if (mem->bo)
- fd_bo_del(mem->bo);
-
+ tu_bo_finish(device, &mem->bo);
vk_free2(&device->alloc, pAllocator, mem);
}
{
TU_FROM_HANDLE(tu_device, device, _device);
TU_FROM_HANDLE(tu_device_memory, mem, _memory);
+ VkResult result;
if (mem == NULL) {
*ppData = NULL;
if (mem->user_ptr) {
*ppData = mem->user_ptr;
- } else if (!mem->map){
- *ppData = mem->map = fd_bo_map(mem->bo);
+ } else if (!mem->map) {
+ result = tu_bo_map(device, &mem->bo);
+ if (result != VK_SUCCESS)
+ return result;
+ *ppData = mem->map = mem->bo.map;
} else
*ppData = mem->map;
TU_FROM_HANDLE(tu_buffer, buffer, _buffer);
pMemoryRequirements->memoryTypeBits = 1;
- pMemoryRequirements->alignment = 16;
+ pMemoryRequirements->alignment = 64;
pMemoryRequirements->size =
- align64(buffer->size, pMemoryRequirements->alignment);
+ align64(buffer->size, pMemoryRequirements->alignment);
}
void
tu_GetBufferMemoryRequirements2(
VkDevice device,
- const VkBufferMemoryRequirementsInfo2KHR *pInfo,
- VkMemoryRequirements2KHR *pMemoryRequirements)
+ const VkBufferMemoryRequirementsInfo2 *pInfo,
+ VkMemoryRequirements2 *pMemoryRequirements)
{
- tu_GetBufferMemoryRequirements(
- device, pInfo->buffer, &pMemoryRequirements->memoryRequirements);
+ tu_GetBufferMemoryRequirements(device, pInfo->buffer,
+ &pMemoryRequirements->memoryRequirements);
}
void
TU_FROM_HANDLE(tu_image, image, _image);
pMemoryRequirements->memoryTypeBits = 1;
- pMemoryRequirements->size = image->size;
- pMemoryRequirements->alignment = image->alignment;
+ pMemoryRequirements->size = image->layout.size;
+ pMemoryRequirements->alignment = image->layout.base_align;
}
void
tu_GetImageMemoryRequirements2(VkDevice device,
- const VkImageMemoryRequirementsInfo2KHR *pInfo,
- VkMemoryRequirements2KHR *pMemoryRequirements)
+ const VkImageMemoryRequirementsInfo2 *pInfo,
+ VkMemoryRequirements2 *pMemoryRequirements)
{
- tu_GetImageMemoryRequirements(
- device, pInfo->image, &pMemoryRequirements->memoryRequirements);
+ tu_GetImageMemoryRequirements(device, pInfo->image,
+ &pMemoryRequirements->memoryRequirements);
}
void
uint32_t *pSparseMemoryRequirementCount,
VkSparseImageMemoryRequirements *pSparseMemoryRequirements)
{
- stub();
+ tu_stub();
}
void
tu_GetImageSparseMemoryRequirements2(
VkDevice device,
- const VkImageSparseMemoryRequirementsInfo2KHR *pInfo,
+ const VkImageSparseMemoryRequirementsInfo2 *pInfo,
uint32_t *pSparseMemoryRequirementCount,
- VkSparseImageMemoryRequirements2KHR *pSparseMemoryRequirements)
+ VkSparseImageMemoryRequirements2 *pSparseMemoryRequirements)
{
- stub();
+ tu_stub();
}
void
VkResult
tu_BindBufferMemory2(VkDevice device,
uint32_t bindInfoCount,
- const VkBindBufferMemoryInfoKHR *pBindInfos)
+ const VkBindBufferMemoryInfo *pBindInfos)
{
+ for (uint32_t i = 0; i < bindInfoCount; ++i) {
+ TU_FROM_HANDLE(tu_device_memory, mem, pBindInfos[i].memory);
+ TU_FROM_HANDLE(tu_buffer, buffer, pBindInfos[i].buffer);
+
+ if (mem) {
+ buffer->bo = &mem->bo;
+ buffer->bo_offset = pBindInfos[i].memoryOffset;
+ } else {
+ buffer->bo = NULL;
+ }
+ }
return VK_SUCCESS;
}
VkDeviceMemory memory,
VkDeviceSize memoryOffset)
{
- const VkBindBufferMemoryInfoKHR info = {
- .sType = VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO_KHR,
+ const VkBindBufferMemoryInfo info = {
+ .sType = VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO,
.buffer = buffer,
.memory = memory,
.memoryOffset = memoryOffset
VkResult
tu_BindImageMemory2(VkDevice device,
uint32_t bindInfoCount,
- const VkBindImageMemoryInfoKHR *pBindInfos)
-{
+ const VkBindImageMemoryInfo *pBindInfos)
+{
+ for (uint32_t i = 0; i < bindInfoCount; ++i) {
+ TU_FROM_HANDLE(tu_image, image, pBindInfos[i].image);
+ TU_FROM_HANDLE(tu_device_memory, mem, pBindInfos[i].memory);
+
+ if (mem) {
+ image->bo = &mem->bo;
+ image->bo_offset = pBindInfos[i].memoryOffset;
+ } else {
+ image->bo = NULL;
+ image->bo_offset = 0;
+ }
+ }
+
return VK_SUCCESS;
}
VkDeviceMemory memory,
VkDeviceSize memoryOffset)
{
- const VkBindImageMemoryInfoKHR info = {
- .sType = VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO_KHR,
+ const VkBindImageMemoryInfo info = {
+ .sType = VK_STRUCTURE_TYPE_BIND_BUFFER_MEMORY_INFO,
.image = image,
.memory = memory,
.memoryOffset = memoryOffset
return VK_SUCCESS;
}
-VkResult
-tu_CreateFence(VkDevice _device,
- const VkFenceCreateInfo *pCreateInfo,
- const VkAllocationCallbacks *pAllocator,
- VkFence *pFence)
-{
- TU_FROM_HANDLE(tu_device, device, _device);
-
- struct tu_fence *fence = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*fence),
- 8,
- VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
-
- if (!fence)
- return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
-
- *pFence = tu_fence_to_handle(fence);
-
- return VK_SUCCESS;
-}
-
-void
-tu_DestroyFence(VkDevice _device,
- VkFence _fence,
- const VkAllocationCallbacks *pAllocator)
-{
- TU_FROM_HANDLE(tu_device, device, _device);
- TU_FROM_HANDLE(tu_fence, fence, _fence);
-
- if (!fence)
- return;
-
- vk_free2(&device->alloc, pAllocator, fence);
-}
-
-VkResult
-tu_WaitForFences(VkDevice _device,
- uint32_t fenceCount,
- const VkFence *pFences,
- VkBool32 waitAll,
- uint64_t timeout)
-{
- return VK_SUCCESS;
-}
-
-VkResult
-tu_ResetFences(VkDevice _device, uint32_t fenceCount, const VkFence *pFences)
-{
- return VK_SUCCESS;
-}
-
-VkResult
-tu_GetFenceStatus(VkDevice _device, VkFence _fence)
-{
- return VK_SUCCESS;
-}
-
// Queue semaphore functions
VkResult
{
TU_FROM_HANDLE(tu_device, device, _device);
- struct tu_semaphore *sem = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*sem),
- 8,
- VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ struct tu_semaphore *sem =
+ vk_alloc2(&device->alloc, pAllocator, sizeof(*sem), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (!sem)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
VkEvent *pEvent)
{
TU_FROM_HANDLE(tu_device, device, _device);
- struct tu_event *event = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*event),
- 8,
- VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ struct tu_event *event =
+ vk_alloc2(&device->alloc, pAllocator, sizeof(*event), 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (!event)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
+ VkResult result = tu_bo_init_new(device, &event->bo, 0x1000);
+ if (result != VK_SUCCESS)
+ goto fail_alloc;
+
+ result = tu_bo_map(device, &event->bo);
+ if (result != VK_SUCCESS)
+ goto fail_map;
+
*pEvent = tu_event_to_handle(event);
return VK_SUCCESS;
+
+fail_map:
+ tu_bo_finish(device, &event->bo);
+fail_alloc:
+ vk_free2(&device->alloc, pAllocator, event);
+ return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
}
void
if (!event)
return;
+
+ tu_bo_finish(device, &event->bo);
vk_free2(&device->alloc, pAllocator, event);
}
{
TU_FROM_HANDLE(tu_event, event, _event);
- if (*event->map == 1)
+ if (*(uint64_t*) event->bo.map == 1)
return VK_EVENT_SET;
return VK_EVENT_RESET;
}
tu_SetEvent(VkDevice _device, VkEvent _event)
{
TU_FROM_HANDLE(tu_event, event, _event);
- *event->map = 1;
+ *(uint64_t*) event->bo.map = 1;
return VK_SUCCESS;
}
tu_ResetEvent(VkDevice _device, VkEvent _event)
{
TU_FROM_HANDLE(tu_event, event, _event);
- *event->map = 0;
+ *(uint64_t*) event->bo.map = 0;
return VK_SUCCESS;
}
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_BUFFER_CREATE_INFO);
- buffer = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*buffer),
- 8,
+ buffer = vk_alloc2(&device->alloc, pAllocator, sizeof(*buffer), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (buffer == NULL)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
vk_free2(&device->alloc, pAllocator, buffer);
}
-static uint32_t
-tu_surface_max_layer_count(struct tu_image_view *iview)
-{
- return iview->type == VK_IMAGE_VIEW_TYPE_3D
- ? iview->extent.depth
- : (iview->base_layer + iview->layer_count);
-}
-
VkResult
tu_CreateFramebuffer(VkDevice _device,
const VkFramebufferCreateInfo *pCreateInfo,
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_FRAMEBUFFER_CREATE_INFO);
- size_t size =
- sizeof(*framebuffer) +
- sizeof(struct tu_attachment_info) * pCreateInfo->attachmentCount;
- framebuffer = vk_alloc2(
- &device->alloc, pAllocator, size, 8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
+ size_t size = sizeof(*framebuffer) + sizeof(struct tu_attachment_info) *
+ pCreateInfo->attachmentCount;
+ framebuffer = vk_alloc2(&device->alloc, pAllocator, size, 8,
+ VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (framebuffer == NULL)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
VkImageView _iview = pCreateInfo->pAttachments[i];
struct tu_image_view *iview = tu_image_view_from_handle(_iview);
framebuffer->attachments[i].attachment = iview;
-
- framebuffer->width = MIN2(framebuffer->width, iview->extent.width);
- framebuffer->height = MIN2(framebuffer->height, iview->extent.height);
- framebuffer->layers =
- MIN2(framebuffer->layers, tu_surface_max_layer_count(iview));
}
*pFramebuffer = tu_framebuffer_to_handle(framebuffer);
vk_free2(&device->alloc, pAllocator, fb);
}
+static enum a6xx_tex_clamp
+tu6_tex_wrap(VkSamplerAddressMode address_mode)
+{
+ switch (address_mode) {
+ case VK_SAMPLER_ADDRESS_MODE_REPEAT:
+ return A6XX_TEX_REPEAT;
+ case VK_SAMPLER_ADDRESS_MODE_MIRRORED_REPEAT:
+ return A6XX_TEX_MIRROR_REPEAT;
+ case VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_EDGE:
+ return A6XX_TEX_CLAMP_TO_EDGE;
+ case VK_SAMPLER_ADDRESS_MODE_CLAMP_TO_BORDER:
+ return A6XX_TEX_CLAMP_TO_BORDER;
+ case VK_SAMPLER_ADDRESS_MODE_MIRROR_CLAMP_TO_EDGE:
+ /* only works for PoT.. need to emulate otherwise! */
+ return A6XX_TEX_MIRROR_CLAMP;
+ default:
+ unreachable("illegal tex wrap mode");
+ break;
+ }
+}
+
+static enum a6xx_tex_filter
+tu6_tex_filter(VkFilter filter, unsigned aniso)
+{
+ switch (filter) {
+ case VK_FILTER_NEAREST:
+ return A6XX_TEX_NEAREST;
+ case VK_FILTER_LINEAR:
+ return aniso ? A6XX_TEX_ANISO : A6XX_TEX_LINEAR;
+ case VK_FILTER_CUBIC_EXT:
+ return A6XX_TEX_CUBIC;
+ default:
+ unreachable("illegal texture filter");
+ break;
+ }
+}
+
+static inline enum adreno_compare_func
+tu6_compare_func(VkCompareOp op)
+{
+ return (enum adreno_compare_func) op;
+}
+
static void
tu_init_sampler(struct tu_device *device,
struct tu_sampler *sampler,
const VkSamplerCreateInfo *pCreateInfo)
{
+ const struct VkSamplerReductionModeCreateInfo *reduction =
+ vk_find_struct_const(pCreateInfo->pNext, SAMPLER_REDUCTION_MODE_CREATE_INFO);
+
+ unsigned aniso = pCreateInfo->anisotropyEnable ?
+ util_last_bit(MIN2((uint32_t)pCreateInfo->maxAnisotropy >> 1, 8)) : 0;
+ bool miplinear = (pCreateInfo->mipmapMode == VK_SAMPLER_MIPMAP_MODE_LINEAR);
+
+ sampler->descriptor[0] =
+ COND(miplinear, A6XX_TEX_SAMP_0_MIPFILTER_LINEAR_NEAR) |
+ A6XX_TEX_SAMP_0_XY_MAG(tu6_tex_filter(pCreateInfo->magFilter, aniso)) |
+ A6XX_TEX_SAMP_0_XY_MIN(tu6_tex_filter(pCreateInfo->minFilter, aniso)) |
+ A6XX_TEX_SAMP_0_ANISO(aniso) |
+ A6XX_TEX_SAMP_0_WRAP_S(tu6_tex_wrap(pCreateInfo->addressModeU)) |
+ A6XX_TEX_SAMP_0_WRAP_T(tu6_tex_wrap(pCreateInfo->addressModeV)) |
+ A6XX_TEX_SAMP_0_WRAP_R(tu6_tex_wrap(pCreateInfo->addressModeW)) |
+ A6XX_TEX_SAMP_0_LOD_BIAS(pCreateInfo->mipLodBias);
+ sampler->descriptor[1] =
+ /* COND(!cso->seamless_cube_map, A6XX_TEX_SAMP_1_CUBEMAPSEAMLESSFILTOFF) | */
+ COND(pCreateInfo->unnormalizedCoordinates, A6XX_TEX_SAMP_1_UNNORM_COORDS) |
+ A6XX_TEX_SAMP_1_MIN_LOD(pCreateInfo->minLod) |
+ A6XX_TEX_SAMP_1_MAX_LOD(pCreateInfo->maxLod) |
+ COND(pCreateInfo->compareEnable,
+ A6XX_TEX_SAMP_1_COMPARE_FUNC(tu6_compare_func(pCreateInfo->compareOp)));
+ /* This is an offset into the border_color BO, which we fill with all the
+ * possible Vulkan border colors in the correct order, so we can just use
+ * the Vulkan enum with no translation necessary.
+ */
+ sampler->descriptor[2] =
+ A6XX_TEX_SAMP_2_BCOLOR_OFFSET((unsigned) pCreateInfo->borderColor *
+ sizeof(struct bcolor_entry));
+ sampler->descriptor[3] = 0;
+
+ if (reduction) {
+ /* note: vulkan enum matches hw */
+ sampler->descriptor[2] |= A6XX_TEX_SAMP_2_REDUCTION_MODE(reduction->reductionMode);
+ }
+
+ /* TODO:
+ * A6XX_TEX_SAMP_1_MIPFILTER_LINEAR_FAR disables mipmapping, but vk has no NONE mipfilter?
+ */
}
VkResult
assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_SAMPLER_CREATE_INFO);
- sampler = vk_alloc2(&device->alloc,
- pAllocator,
- sizeof(*sampler),
- 8,
+ sampler = vk_alloc2(&device->alloc, pAllocator, sizeof(*sampler), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (!sampler)
return vk_error(device->instance, VK_ERROR_OUT_OF_HOST_MEMORY);
vk_icdNegotiateLoaderICDInterfaceVersion(uint32_t *pSupportedVersion)
{
/* For the full details on loader interface versioning, see
- * <https://github.com/KhronosGroup/Vulkan-LoaderAndValidationLayers/blob/master/loader/LoaderAndLayerInterface.md>.
- * What follows is a condensed summary, to help you navigate the large and
- * confusing official doc.
- *
- * - Loader interface v0 is incompatible with later versions. We don't
- * support it.
- *
- * - In loader interface v1:
- * - The first ICD entrypoint called by the loader is
- * vk_icdGetInstanceProcAddr(). The ICD must statically expose this
- * entrypoint.
- * - The ICD must statically expose no other Vulkan symbol unless it is
- * linked with -Bsymbolic.
- * - Each dispatchable Vulkan handle created by the ICD must be
- * a pointer to a struct whose first member is VK_LOADER_DATA. The
- * ICD must initialize VK_LOADER_DATA.loadMagic to ICD_LOADER_MAGIC.
- * - The loader implements vkCreate{PLATFORM}SurfaceKHR() and
- * vkDestroySurfaceKHR(). The ICD must be capable of working with
- * such loader-managed surfaces.
- *
- * - Loader interface v2 differs from v1 in:
- * - The first ICD entrypoint called by the loader is
- * vk_icdNegotiateLoaderICDInterfaceVersion(). The ICD must
- * statically expose this entrypoint.
- *
- * - Loader interface v3 differs from v2 in:
- * - The ICD must implement vkCreate{PLATFORM}SurfaceKHR(),
- * vkDestroySurfaceKHR(), and other API which uses VKSurfaceKHR,
- * because the loader no longer does so.
- */
+ * <https://github.com/KhronosGroup/Vulkan-LoaderAndValidationLayers/blob/master/loader/LoaderAndLayerInterface.md>.
+ * What follows is a condensed summary, to help you navigate the large and
+ * confusing official doc.
+ *
+ * - Loader interface v0 is incompatible with later versions. We don't
+ * support it.
+ *
+ * - In loader interface v1:
+ * - The first ICD entrypoint called by the loader is
+ * vk_icdGetInstanceProcAddr(). The ICD must statically expose this
+ * entrypoint.
+ * - The ICD must statically expose no other Vulkan symbol unless it
+ * is linked with -Bsymbolic.
+ * - Each dispatchable Vulkan handle created by the ICD must be
+ * a pointer to a struct whose first member is VK_LOADER_DATA. The
+ * ICD must initialize VK_LOADER_DATA.loadMagic to
+ * ICD_LOADER_MAGIC.
+ * - The loader implements vkCreate{PLATFORM}SurfaceKHR() and
+ * vkDestroySurfaceKHR(). The ICD must be capable of working with
+ * such loader-managed surfaces.
+ *
+ * - Loader interface v2 differs from v1 in:
+ * - The first ICD entrypoint called by the loader is
+ * vk_icdNegotiateLoaderICDInterfaceVersion(). The ICD must
+ * statically expose this entrypoint.
+ *
+ * - Loader interface v3 differs from v2 in:
+ * - The ICD must implement vkCreate{PLATFORM}SurfaceKHR(),
+ * vkDestroySurfaceKHR(), and other API which uses VKSurfaceKHR,
+ * because the loader no longer does so.
+ */
*pSupportedVersion = MIN2(*pSupportedVersion, 3u);
return VK_SUCCESS;
}
+VkResult
+tu_GetMemoryFdKHR(VkDevice _device,
+ const VkMemoryGetFdInfoKHR *pGetFdInfo,
+ int *pFd)
+{
+ TU_FROM_HANDLE(tu_device, device, _device);
+ TU_FROM_HANDLE(tu_device_memory, memory, pGetFdInfo->memory);
+
+ assert(pGetFdInfo->sType == VK_STRUCTURE_TYPE_MEMORY_GET_FD_INFO_KHR);
+
+ /* At the moment, we support only the below handle types. */
+ assert(pGetFdInfo->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_OPAQUE_FD_BIT ||
+ pGetFdInfo->handleType ==
+ VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT);
+
+ int prime_fd = tu_bo_export_dmabuf(device, &memory->bo);
+ if (prime_fd < 0)
+ return vk_error(device->instance, VK_ERROR_OUT_OF_DEVICE_MEMORY);
+
+ *pFd = prime_fd;
+ return VK_SUCCESS;
+}
+
+VkResult
+tu_GetMemoryFdPropertiesKHR(VkDevice _device,
+ VkExternalMemoryHandleTypeFlagBits handleType,
+ int fd,
+ VkMemoryFdPropertiesKHR *pMemoryFdProperties)
+{
+ assert(handleType == VK_EXTERNAL_MEMORY_HANDLE_TYPE_DMA_BUF_BIT_EXT);
+ pMemoryFdProperties->memoryTypeBits = 1;
+ return VK_SUCCESS;
+}
+
void
tu_GetPhysicalDeviceExternalSemaphoreProperties(
VkPhysicalDevice physicalDevice,
- const VkPhysicalDeviceExternalSemaphoreInfoKHR *pExternalSemaphoreInfo,
- VkExternalSemaphorePropertiesKHR *pExternalSemaphoreProperties)
+ const VkPhysicalDeviceExternalSemaphoreInfo *pExternalSemaphoreInfo,
+ VkExternalSemaphoreProperties *pExternalSemaphoreProperties)
{
pExternalSemaphoreProperties->exportFromImportedHandleTypes = 0;
pExternalSemaphoreProperties->compatibleHandleTypes = 0;
void
tu_GetPhysicalDeviceExternalFenceProperties(
VkPhysicalDevice physicalDevice,
- const VkPhysicalDeviceExternalFenceInfoKHR *pExternalFenceInfo,
- VkExternalFencePropertiesKHR *pExternalFenceProperties)
+ const VkPhysicalDeviceExternalFenceInfo *pExternalFenceInfo,
+ VkExternalFenceProperties *pExternalFenceProperties)
{
pExternalFenceProperties->exportFromImportedHandleTypes = 0;
pExternalFenceProperties->compatibleHandleTypes = 0;
{
TU_FROM_HANDLE(tu_instance, instance, _instance);
return vk_create_debug_report_callback(&instance->debug_report_callbacks,
- pCreateInfo,
- pAllocator,
- &instance->alloc,
- pCallback);
+ pCreateInfo, pAllocator,
+ &instance->alloc, pCallback);
}
void
{
TU_FROM_HANDLE(tu_instance, instance, _instance);
vk_destroy_debug_report_callback(&instance->debug_report_callbacks,
- _callback,
- pAllocator,
- &instance->alloc);
+ _callback, pAllocator, &instance->alloc);
}
void
const char *pMessage)
{
TU_FROM_HANDLE(tu_instance, instance, _instance);
- vk_debug_report(&instance->debug_report_callbacks,
- flags,
- objectType,
- object,
- location,
- messageCode,
- pLayerPrefix,
- pMessage);
+ vk_debug_report(&instance->debug_report_callbacks, flags, objectType,
+ object, location, messageCode, pLayerPrefix, pMessage);
}
void
VK_PEER_MEMORY_FEATURE_GENERIC_SRC_BIT |
VK_PEER_MEMORY_FEATURE_GENERIC_DST_BIT;
}
+
+void tu_GetPhysicalDeviceMultisamplePropertiesEXT(
+ VkPhysicalDevice physicalDevice,
+ VkSampleCountFlagBits samples,
+ VkMultisamplePropertiesEXT* pMultisampleProperties)
+{
+ TU_FROM_HANDLE(tu_physical_device, pdevice, physicalDevice);
+
+ if (samples <= VK_SAMPLE_COUNT_4_BIT && pdevice->supported_extensions.EXT_sample_locations)
+ pMultisampleProperties->maxSampleLocationGridSize = (VkExtent2D){ 1, 1 };
+ else
+ pMultisampleProperties->maxSampleLocationGridSize = (VkExtent2D){ 0, 0 };
+}
projects / mesa.git / blobdiff