tu: ir3: Emit push constants directly

[mesa.git] / src / freedreno / vulkan / tu_device.c

diff --git a/src/freedreno/vulkan/tu_device.c b/src/freedreno/vulkan/tu_device.c
index 97838ece23da520c611ee126300d887def4cf254..394c3ce18009d18f06352dfb703ed0c9a39e4d43 100644 (file)
--- a/src/freedreno/vulkan/tu_device.c
+++ b/src/freedreno/vulkan/tu_device.c
@@ -265,19 +265,17 @@ tu_physical_device_init(struct tu_physical_device *device,
 
    switch (device->gpu_id) {
    case 618:
-      device->tile_align_w = 64;
-      device->tile_align_h = 16;
       device->magic.RB_UNKNOWN_8E04_blit = 0x00100000;
-      device->magic.RB_CCU_CNTL_gmem     = 0x3e400004;
+      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->tile_align_w = 64;
-      device->tile_align_h = 16;
       device->magic.RB_UNKNOWN_8E04_blit = 0x01000000;
-      device->magic.RB_CCU_CNTL_gmem     = 0x7c400004;
+      device->ccu_offset_gmem = 0xf8000;
+      device->ccu_offset_bypass = 0x20000;
       device->magic.PC_UNKNOWN_9805 = 0x1;
       device->magic.SP_UNKNOWN_A0F8 = 0x1;
       break;
@@ -338,7 +336,7 @@ tu_physical_device_finish(struct tu_physical_device *device)
       close(device->master_fd);
 }
 
-static void *
+static VKAPI_ATTR void *
 default_alloc_func(void *pUserData,
                    size_t size,
                    size_t align,
@@ -347,7 +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,
@@ -357,7 +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);
@@ -588,14 +586,14 @@ tu_GetPhysicalDeviceFeatures(VkPhysicalDevice physicalDevice,
       .fullDrawIndexUint32 = true,
       .imageCubeArray = false,
       .independentBlend = true,
-      .geometryShader = false,
+      .geometryShader = true,
       .tessellationShader = false,
       .sampleRateShading = true,
       .dualSrcBlend = true,
       .logicOp = true,
       .multiDrawIndirect = false,
       .drawIndirectFirstInstance = false,
-      .depthClamp = false,
+      .depthClamp = true,
       .depthBiasClamp = false,
       .fillModeNonSolid = false,
       .depthBounds = false,
@@ -731,8 +729,8 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice,
                                VkPhysicalDeviceProperties *pProperties)
 {
    TU_FROM_HANDLE(tu_physical_device, pdevice, physicalDevice);
-   VkSampleCountFlags sample_counts = VK_SAMPLE_COUNT_1_BIT |
-      VK_SAMPLE_COUNT_2_BIT | VK_SAMPLE_COUNT_4_BIT | VK_SAMPLE_COUNT_8_BIT;
+   VkSampleCountFlags sample_counts =
+      VK_SAMPLE_COUNT_1_BIT | VK_SAMPLE_COUNT_2_BIT | VK_SAMPLE_COUNT_4_BIT;
 
    /* 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
@@ -779,7 +777,7 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice,
       .maxDescriptorSetInputAttachments = max_descriptor_set_size,
       .maxVertexInputAttributes = 32,
       .maxVertexInputBindings = 32,
-      .maxVertexInputAttributeOffset = 2047,
+      .maxVertexInputAttributeOffset = 4095,
       .maxVertexInputBindingStride = 2048,
       .maxVertexOutputComponents = 128,
       .maxTessellationGenerationLevel = 64,
@@ -790,7 +788,7 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice,
       .maxTessellationControlTotalOutputComponents = 4096,
       .maxTessellationEvaluationInputComponents = 128,
       .maxTessellationEvaluationOutputComponents = 128,
-      .maxGeometryShaderInvocations = 127,
+      .maxGeometryShaderInvocations = 32,
       .maxGeometryInputComponents = 64,
       .maxGeometryOutputComponents = 128,
       .maxGeometryOutputVertices = 256,
@@ -803,7 +801,7 @@ 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,
@@ -839,8 +837,8 @@ tu_GetPhysicalDeviceProperties(VkPhysicalDevice physicalDevice,
       .sampledImageStencilSampleCounts = sample_counts,
       .storageImageSampleCounts = VK_SAMPLE_COUNT_1_BIT,
       .maxSampleMaskWords = 1,
-      .timestampComputeAndGraphics = false, /* FINISHME */
-      .timestampPeriod = 1,
+      .timestampComputeAndGraphics = true,
+      .timestampPeriod = 1000000000.0 / 19200000.0, /* CP_ALWAYS_ON_COUNTER is fixed 19.2MHz */
       .maxClipDistances = 8,
       .maxCullDistances = 8,
       .maxCombinedClipAndCullDistances = 8,
@@ -945,7 +943,7 @@ static const VkQueueFamilyProperties tu_queue_family_properties = {
    .queueFlags =
       VK_QUEUE_GRAPHICS_BIT | VK_QUEUE_COMPUTE_BIT | VK_QUEUE_TRANSFER_BIT,
    .queueCount = 1,
-   .timestampValidBits = 0, /* FINISHME */
+   .timestampValidBits = 48,
    .minImageTransferGranularity = { 1, 1, 1 },
 };
 
@@ -1447,6 +1445,44 @@ 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,
@@ -1703,7 +1739,7 @@ tu_GetImageMemoryRequirements(VkDevice _device,
 
    pMemoryRequirements->memoryTypeBits = 1;
    pMemoryRequirements->size = image->layout.size;
-   pMemoryRequirements->alignment = image->alignment;
+   pMemoryRequirements->alignment = image->layout.base_align;
 }
 
 void