icd/intel/dev.c

/*
 * Vulkan
 *
 * Copyright (C) 2014 LunarG, Inc.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a
 * copy of this software and associated documentation files (the "Software"),
 * to deal in the Software without restriction, including without limitation
 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
 * and/or sell copies of the Software, and to permit persons to whom the
 * Software is furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included
 * in all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * 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.
 *
 * Authors:
 *   Chia-I Wu <olv@lunarg.com>
 */

#include <stdarg.h>
#include "kmd/winsys.h"
#include "desc.h"
#include "gpu.h"
#include "pipeline.h"
#include "queue.h"
#include "dev.h"

static void dev_destroy_meta_shaders(struct intel_dev *dev)
{
    uint32_t i;

    for (i = 0; i < ARRAY_SIZE(dev->cmd_meta_shaders); i++) {
        if (!dev->cmd_meta_shaders[i])
            break;

        intel_pipeline_shader_destroy(dev, dev->cmd_meta_shaders[i]);
        dev->cmd_meta_shaders[i] = NULL;
    }
}

static bool dev_create_meta_shaders(struct intel_dev *dev)
{
    uint32_t i;

    for (i = 0; i < ARRAY_SIZE(dev->cmd_meta_shaders); i++) {
        struct intel_pipeline_shader *sh;

        sh = intel_pipeline_shader_create_meta(dev, i);
        if (!sh) {
            dev_destroy_meta_shaders(dev);
            return false;
        }

        dev->cmd_meta_shaders[i] = sh;
    }

    return true;
}

static VK_RESULT dev_create_queues(struct intel_dev *dev,
                                    const VkDeviceQueueCreateInfo *queues,
                                    uint32_t count)
{
    uint32_t i;

    if (!count)
        return VK_ERROR_INVALID_POINTER;

    for (i = 0; i < count; i++) {
        const VkDeviceQueueCreateInfo *q = &queues[i];
        VK_RESULT ret = VK_SUCCESS;

        if (q->queueNodeIndex < INTEL_GPU_ENGINE_COUNT &&
            q->queueCount == 1 && !dev->queues[q->queueNodeIndex]) {
            ret = intel_queue_create(dev, q->queueNodeIndex,
                    &dev->queues[q->queueNodeIndex]);
        }
        else {
            ret = VK_ERROR_INVALID_POINTER;
        }

        if (ret != VK_SUCCESS) {
            uint32_t j;
            for (j = 0; j < i; j++)
                intel_queue_destroy(dev->queues[j]);

            return ret;
        }
    }

    return VK_SUCCESS;
}

VK_RESULT intel_dev_create(struct intel_gpu *gpu,
                            const VkDeviceCreateInfo *info,
                            struct intel_dev **dev_ret)
{
    struct intel_dev *dev;
    uint32_t i;
    VK_RESULT ret;

    if (gpu->winsys)
        return VK_ERROR_DEVICE_ALREADY_CREATED;

    dev = (struct intel_dev *) intel_base_create(&gpu->handle,
            sizeof(*dev), info->flags,
            VK_DBG_OBJECT_DEVICE, info, sizeof(struct intel_dev_dbg));
    if (!dev)
        return VK_ERROR_OUT_OF_MEMORY;

    for (i = 0; i < info->extensionCount; i++) {
        const enum intel_ext_type ext = intel_gpu_lookup_extension(gpu,
                info->ppEnabledExtensionNames[i]);

        if (ext != INTEL_EXT_INVALID)
            dev->exts[ext] = true;
    }

    dev->gpu = gpu;

    ret = intel_gpu_init_winsys(gpu);
    if (ret != VK_SUCCESS) {
        intel_dev_destroy(dev);
        return ret;
    }

    dev->winsys = gpu->winsys;

    dev->cmd_scratch_bo = intel_winsys_alloc_bo(dev->winsys,
            "command buffer scratch", 4096, false);
    if (!dev->cmd_scratch_bo) {
        intel_dev_destroy(dev);
        return VK_ERROR_OUT_OF_GPU_MEMORY;
    }

    if (!dev_create_meta_shaders(dev)) {
        intel_dev_destroy(dev);
        return VK_ERROR_OUT_OF_MEMORY;
    }

    ret = intel_desc_region_create(dev, &dev->desc_region);
    if (ret != VK_SUCCESS) {
        intel_dev_destroy(dev);
        return ret;
    }

    intel_pipeline_init_default_sample_patterns(dev,
            (uint8_t *) &dev->sample_pattern_1x,
            (uint8_t *) &dev->sample_pattern_2x,
            (uint8_t *) &dev->sample_pattern_4x,
            (uint8_t *) dev->sample_pattern_8x,
            (uint8_t *) dev->sample_pattern_16x);

    ret = dev_create_queues(dev, info->pRequestedQueues,
            info->queueRecordCount);
    if (ret != VK_SUCCESS) {
        intel_dev_destroy(dev);
        return ret;
    }

    *dev_ret = dev;

    return VK_SUCCESS;
}

static void dev_clear_msg_filters(struct intel_dev *dev)
{
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);
    struct intel_dev_dbg_msg_filter *filter;

    filter = dbg->filters;
    while (filter) {
        struct intel_dev_dbg_msg_filter *next = filter->next;
        intel_free(dev, filter);
        filter = next;
    }

    dbg->filters = NULL;
}

void intel_dev_destroy(struct intel_dev *dev)
{
    struct intel_gpu *gpu = dev->gpu;
    uint32_t i;

    if (dev->base.dbg)
        dev_clear_msg_filters(dev);

    for (i = 0; i < ARRAY_SIZE(dev->queues); i++) {
        if (dev->queues[i])
            intel_queue_destroy(dev->queues[i]);
    }

    if (dev->desc_region)
        intel_desc_region_destroy(dev, dev->desc_region);

    dev_destroy_meta_shaders(dev);

    intel_bo_unref(dev->cmd_scratch_bo);

    intel_base_destroy(&dev->base);

    if (gpu->winsys)
        intel_gpu_cleanup_winsys(gpu);
}

VK_RESULT intel_dev_add_msg_filter(struct intel_dev *dev,
                                    int32_t msg_code,
                                    VK_DBG_MSG_FILTER filter)
{
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);
    struct intel_dev_dbg_msg_filter *f = dbg->filters;

    assert(filter != VK_DBG_MSG_FILTER_NONE);

    while (f) {
        if (f->msg_code == msg_code)
            break;
        f = f->next;
    }

    if (f) {
        if (f->filter != filter) {
            f->filter = filter;
            f->triggered = false;
        }
    } else {
        f = intel_alloc(dev, sizeof(*f), 0, VK_SYSTEM_ALLOC_DEBUG);
        if (!f)
            return VK_ERROR_OUT_OF_MEMORY;

        f->msg_code = msg_code;
        f->filter = filter;
        f->triggered = false;

        f->next = dbg->filters;
        dbg->filters = f;
    }

    return VK_SUCCESS;
}

void intel_dev_remove_msg_filter(struct intel_dev *dev,
                                 int32_t msg_code)
{
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);
    struct intel_dev_dbg_msg_filter *f = dbg->filters, *prev = NULL;

    while (f) {
        if (f->msg_code == msg_code) {
            if (prev)
                prev->next = f->next;
            else
                dbg->filters = f->next;

            intel_free(dev, f);
            break;
        }

        prev = f;
        f = f->next;
    }
}

static bool dev_filter_msg(struct intel_dev *dev,
                           int32_t msg_code)
{
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);
    struct intel_dev_dbg_msg_filter *filter;

    if (!dbg)
        return false;

    filter = dbg->filters;
    while (filter) {
        if (filter->msg_code != msg_code) {
            filter = filter->next;
            continue;
        }

        if (filter->filter == VK_DBG_MSG_FILTER_ALL)
            return true;

        if (filter->filter == VK_DBG_MSG_FILTER_REPEATED &&
            filter->triggered)
            return true;

        filter->triggered = true;
        break;
    }

    return false;
}

void intel_dev_log(struct intel_dev *dev,
                   VK_DBG_MSG_TYPE msg_type,
                   VK_VALIDATION_LEVEL validation_level,
                   struct intel_base *src_object,
                   size_t location,
                   int32_t msg_code,
                   const char *format, ...)
{
    va_list ap;

    if (dev_filter_msg(dev, msg_code))
        return;

    va_start(ap, format);
    intel_logv(dev, msg_type, validation_level, (VK_BASE_OBJECT) src_object,
            location, msg_code, format, ap);
    va_end(ap);
}

ICD_EXPORT VK_RESULT VKAPI vkCreateDevice(
    VK_PHYSICAL_GPU                            gpu_,
    const VkDeviceCreateInfo*               pCreateInfo,
    VK_DEVICE*                                 pDevice)
{
    struct intel_gpu *gpu = intel_gpu(gpu_);

    return intel_dev_create(gpu, pCreateInfo, (struct intel_dev **) pDevice);
}

ICD_EXPORT VK_RESULT VKAPI vkDestroyDevice(
    VK_DEVICE                                  device)
{
    struct intel_dev *dev = intel_dev(device);

    intel_dev_destroy(dev);

    return VK_SUCCESS;
}

ICD_EXPORT VK_RESULT VKAPI vkGetDeviceQueue(
    VK_DEVICE                                  device,
    uint32_t                                    queueNodeIndex,
    uint32_t                                    queueIndex,
    VK_QUEUE*                                  pQueue)
{
    struct intel_dev *dev = intel_dev(device);

    if (queueNodeIndex >= INTEL_GPU_ENGINE_COUNT) {
        return VK_ERROR_UNAVAILABLE;
    }

    if (queueIndex > 0)
        return VK_ERROR_UNAVAILABLE;

    *pQueue = dev->queues[queueNodeIndex];
    return VK_SUCCESS;
}

ICD_EXPORT VK_RESULT VKAPI vkDeviceWaitIdle(
    VK_DEVICE                                  device)
{
    struct intel_dev *dev = intel_dev(device);
    VK_RESULT ret = VK_SUCCESS;
    uint32_t i;

    for (i = 0; i < ARRAY_SIZE(dev->queues); i++) {
        if (dev->queues[i]) {
            const VK_RESULT r = intel_queue_wait(dev->queues[i], -1);
            if (r != VK_SUCCESS)
                ret = r;
        }
    }

    return ret;
}

ICD_EXPORT VK_RESULT VKAPI vkDbgSetValidationLevel(
    VK_DEVICE                                  device,
    VK_VALIDATION_LEVEL                        validationLevel)
{
    struct intel_dev *dev = intel_dev(device);
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);

    if (dbg)
        dbg->validation_level = validationLevel;

    return VK_SUCCESS;
}

ICD_EXPORT VK_RESULT VKAPI vkDbgSetMessageFilter(
    VK_DEVICE                                  device,
    int32_t                                     msgCode,
    VK_DBG_MSG_FILTER                          filter)
{
    struct intel_dev *dev = intel_dev(device);

    if (!dev->base.dbg)
        return VK_SUCCESS;

    if (filter == VK_DBG_MSG_FILTER_NONE) {
        intel_dev_remove_msg_filter(dev, msgCode);
        return VK_SUCCESS;
    }

    return intel_dev_add_msg_filter(dev, msgCode, filter);
}

ICD_EXPORT VK_RESULT VKAPI vkDbgSetDeviceOption(
    VK_DEVICE                                  device,
    VK_DBG_DEVICE_OPTION                       dbgOption,
    size_t                                      dataSize,
    const void*                                 pData)
{
    struct intel_dev *dev = intel_dev(device);
    struct intel_dev_dbg *dbg = intel_dev_dbg(dev);
    VK_RESULT ret = VK_SUCCESS;

    if (dataSize == 0)
        return VK_ERROR_INVALID_VALUE;

    switch (dbgOption) {
    case VK_DBG_OPTION_DISABLE_PIPELINE_LOADS:
        if (dbg)
            dbg->disable_pipeline_loads = *((const bool *) pData);
        break;
    case VK_DBG_OPTION_FORCE_OBJECT_MEMORY_REQS:
        if (dbg)
            dbg->force_object_memory_reqs = *((const bool *) pData);
        break;
    case VK_DBG_OPTION_FORCE_LARGE_IMAGE_ALIGNMENT:
        if (dbg)
            dbg->force_large_image_alignment = *((const bool *) pData);
        break;
    default:
        ret = VK_ERROR_INVALID_VALUE;
        break;
    }

    return ret;
}