X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Ffreedreno%2Fvulkan%2Ftu_device.c;h=91e1a846daf5bd9d6da809753c0630f165372109;hb=e43fc003e0ed9ad5ba6e19a1132457233edba6eb;hp=3abaf0ff729a881d06bc5abb73c9c9a5bf578edc;hpb=6cb5fd0d7163a2430f14686faad037912f18b557;p=mesa.git diff --git a/src/freedreno/vulkan/tu_device.c b/src/freedreno/vulkan/tu_device.c index 3abaf0ff729..91e1a846daf 100644 --- a/src/freedreno/vulkan/tu_device.c +++ b/src/freedreno/vulkan/tu_device.c @@ -36,13 +36,13 @@ #include #include +#include "compiler/glsl_types.h" #include "util/debug.h" #include "util/disk_cache.h" -#include "util/strtod.h" #include "vk_format.h" #include "vk_util.h" -#include "drm/msm_drm.h" +#include "drm-uapi/msm_drm.h" static int tu_device_get_cache_uuid(uint16_t family, void *uuid) @@ -198,8 +198,6 @@ tu_physical_device_init(struct tu_physical_device *device, 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); @@ -254,13 +252,32 @@ tu_physical_device_init(struct tu_physical_device *device, goto fail; } + 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; + } + memset(device->name, 0, sizeof(device->name)); sprintf(device->name, "FD%d", device->gpu_id); 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: - device->tile_align_w = 32; - device->tile_align_h = 32; + 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: result = vk_errorf(instance, VK_ERROR_INITIALIZATION_FAILED, @@ -293,6 +310,12 @@ tu_physical_device_init(struct tu_physical_device *device, goto fail; } + result = tu_wsi_init(device); + if (result != VK_SUCCESS) { + vk_error(instance, result); + goto fail; + } + return VK_SUCCESS; fail: @@ -305,13 +328,15 @@ fail: 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, @@ -320,7 +345,7 @@ default_alloc_func(void *pUserData, return malloc(size); } -static void * +static VKAPI_ATTR void * default_realloc_func(void *pUserData, void *pOriginal, size_t size, @@ -330,7 +355,7 @@ default_realloc_func(void *pUserData, return realloc(pOriginal, size); } -static void +static VKAPI_ATTR void default_free_func(void *pUserData, void *pMemory) { free(pMemory); @@ -347,6 +372,9 @@ 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 } }; @@ -424,7 +452,7 @@ tu_CreateInstance(const VkInstanceCreateInfo *pCreateInfo, return vk_error(instance, result); } - _mesa_locale_init(); + glsl_type_singleton_init_or_ref(); VG(VALGRIND_CREATE_MEMPOOL(instance, 0, false)); @@ -448,7 +476,7 @@ tu_DestroyInstance(VkInstance _instance, VG(VALGRIND_DESTROY_MEMPOOL(instance)); - _mesa_locale_fini(); + glsl_type_singleton_decref(); vk_debug_report_instance_destroy(&instance->debug_report_callbacks); @@ -555,17 +583,17 @@ tu_GetPhysicalDeviceFeatures(VkPhysicalDevice physicalDevice, *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, @@ -573,11 +601,11 @@ tu_GetPhysicalDeviceFeatures(VkPhysicalDevice physicalDevice, .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, @@ -609,8 +637,8 @@ tu_GetPhysicalDeviceFeatures2(VkPhysicalDevice physicalDevice, vk_foreach_struct(ext, pFeatures->pNext) { switch (ext->sType) { - case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTER_FEATURES: { - VkPhysicalDeviceVariablePointerFeatures *features = (void *) ext; + case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_VARIABLE_POINTERS_FEATURES: { + VkPhysicalDeviceVariablePointersFeatures *features = (void *) ext; features->variablePointersStorageBuffer = false; features->variablePointers = false; break; @@ -623,9 +651,9 @@ tu_GetPhysicalDeviceFeatures2(VkPhysicalDevice physicalDevice, features->multiviewTessellationShader = false; break; } - case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETER_FEATURES: { - VkPhysicalDeviceShaderDrawParameterFeatures *features = - (VkPhysicalDeviceShaderDrawParameterFeatures *) ext; + case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_DRAW_PARAMETERS_FEATURES: { + VkPhysicalDeviceShaderDrawParametersFeatures *features = + (VkPhysicalDeviceShaderDrawParametersFeatures *) ext; features->shaderDrawParameters = false; break; } @@ -682,6 +710,13 @@ tu_GetPhysicalDeviceFeatures2(VkPhysicalDevice physicalDevice, 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; } @@ -694,20 +729,16 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, 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), @@ -716,8 +747,8 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -729,7 +760,7 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -738,10 +769,10 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -752,12 +783,12 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -765,7 +796,7 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .maxComputeWorkGroupCount = { 65535, 65535, 65535 }, .maxComputeWorkGroupInvocations = 2048, .maxComputeWorkGroupSize = { 2048, 2048, 2048 }, - .subPixelPrecisionBits = 4 /* FIXME */, + .subPixelPrecisionBits = 8, .subTexelPrecisionBits = 4 /* FIXME */, .mipmapPrecisionBits = 4 /* FIXME */, .maxDrawIndexedIndexValue = UINT32_MAX, @@ -777,9 +808,9 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -802,7 +833,7 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice, .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, @@ -880,6 +911,45 @@ tu_GetPhysicalDeviceProperties2(VkPhysicalDevice physicalDevice, 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; + } + 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; } @@ -890,8 +960,8 @@ 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 @@ -1004,6 +1074,61 @@ tu_get_device_extension_index(const char *name) 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, @@ -1065,7 +1190,7 @@ tu_CreateDevice(VkPhysicalDevice physicalDevice, 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, @@ -1077,13 +1202,38 @@ tu_CreateDevice(VkPhysicalDevice physicalDevice, 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; + 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; @@ -1095,14 +1245,27 @@ tu_CreateDevice(VkPhysicalDevice physicalDevice, result = 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]); @@ -1110,9 +1273,6 @@ fail: vk_free(&device->alloc, device->queues[i]); } - if (device->compiler) - ralloc_free(device->compiler); - vk_free(&device->alloc, device); return result; } @@ -1125,6 +1285,9 @@ tu_DestroyDevice(VkDevice _device, const VkAllocationCallbacks *pAllocator) 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]); @@ -1299,6 +1462,44 @@ tu_DeviceWaitIdle(VkDevice _device) 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, @@ -1531,7 +1732,7 @@ tu_GetBufferMemoryRequirements(VkDevice _device, TU_FROM_HANDLE(tu_buffer, buffer, _buffer); pMemoryRequirements->memoryTypeBits = 1; - pMemoryRequirements->alignment = 16; + pMemoryRequirements->alignment = 64; pMemoryRequirements->size = align64(buffer->size, pMemoryRequirements->alignment); } @@ -1554,8 +1755,8 @@ tu_GetImageMemoryRequirements(VkDevice _device, 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 @@ -1723,9 +1924,23 @@ tu_CreateEvent(VkDevice _device, 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 @@ -1738,6 +1953,8 @@ tu_DestroyEvent(VkDevice _device, if (!event) return; + + tu_bo_finish(device, &event->bo); vk_free2(&device->alloc, pAllocator, event); } @@ -1746,7 +1963,7 @@ tu_GetEventStatus(VkDevice _device, VkEvent _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; } @@ -1755,7 +1972,7 @@ VkResult 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; } @@ -1764,7 +1981,7 @@ VkResult 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; } @@ -1808,14 +2025,6 @@ tu_DestroyBuffer(VkDevice _device, 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, @@ -1842,11 +2051,6 @@ tu_CreateFramebuffer(VkDevice _device, 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); @@ -1866,11 +2070,94 @@ tu_DestroyFramebuffer(VkDevice _device, 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 @@ -2063,3 +2350,16 @@ tu_GetDeviceGroupPeerMemoryFeatures( 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 }; +}