src/intel/vulkan/genX_pipeline.c

   1 /*
   2  * Copyright © 2015 Intel Corporation
   3  *
   4  * Permission is hereby granted, free of charge, to any person obtaining a
   5  * copy of this software and associated documentation files (the "Software"),
   6  * to deal in the Software without restriction, including without limitation
   7  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
   8  * and/or sell copies of the Software, and to permit persons to whom the
   9  * Software is furnished to do so, subject to the following conditions:
  10  *
  11  * The above copyright notice and this permission notice (including the next
  12  * paragraph) shall be included in all copies or substantial portions of the
  13  * Software.
  14  *
  15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  16  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
  18  * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  19  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  20  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  21  * IN THE SOFTWARE.
  22  */
  23
  24 #include "anv_private.h"
  25
  26 #include "genxml/gen_macros.h"
  27 #include "genxml/genX_pack.h"
  28
  29 VkResult
  30 genX(compute_pipeline_create)(
  31     VkDevice                                    _device,
  32     struct anv_pipeline_cache *                 cache,
  33     const VkComputePipelineCreateInfo*          pCreateInfo,
  34     const VkAllocationCallbacks*                pAllocator,
  35     VkPipeline*                                 pPipeline)
  36 {
  37    ANV_FROM_HANDLE(anv_device, device, _device);
  38    struct anv_physical_device *physical_device =
  39       &device->instance->physicalDevice;
  40    struct anv_pipeline *pipeline;
  41    VkResult result;
  42
  43    assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO);
  44
  45    pipeline = anv_alloc2(&device->alloc, pAllocator, sizeof(*pipeline), 8,
  46                          VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
  47    if (pipeline == NULL)
  48       return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
  49
  50    pipeline->device = device;
  51    pipeline->layout = anv_pipeline_layout_from_handle(pCreateInfo->layout);
  52
  53    pipeline->blend_state.map = NULL;
  54
  55    result = anv_reloc_list_init(&pipeline->batch_relocs,
  56                                 pAllocator ? pAllocator : &device->alloc);
  57    if (result != VK_SUCCESS) {
  58       anv_free2(&device->alloc, pAllocator, pipeline);
  59       return result;
  60    }
  61    pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
  62    pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
  63    pipeline->batch.relocs = &pipeline->batch_relocs;
  64
  65    /* When we free the pipeline, we detect stages based on the NULL status
  66     * of various prog_data pointers.  Make them NULL by default.
  67     */
  68    memset(pipeline->shaders, 0, sizeof(pipeline->shaders));
  69
  70    pipeline->vs_simd8 = NO_KERNEL;
  71    pipeline->vs_vec4 = NO_KERNEL;
  72    pipeline->gs_kernel = NO_KERNEL;
  73
  74    pipeline->active_stages = 0;
  75
  76    pipeline->needs_data_cache = false;
  77
  78    assert(pCreateInfo->stage.stage == VK_SHADER_STAGE_COMPUTE_BIT);
  79    ANV_FROM_HANDLE(anv_shader_module, module,  pCreateInfo->stage.module);
  80    result = anv_pipeline_compile_cs(pipeline, cache, pCreateInfo, module,
  81                                     pCreateInfo->stage.pName,
  82                                     pCreateInfo->stage.pSpecializationInfo);
  83    if (result != VK_SUCCESS) {
  84       anv_free2(&device->alloc, pAllocator, pipeline);
  85       return result;
  86    }
  87
  88    pipeline->use_repclear = false;
  89
  90    const struct brw_cs_prog_data *cs_prog_data = get_cs_prog_data(pipeline);
  91
  92    anv_pipeline_setup_l3_config(pipeline, cs_prog_data->base.total_shared > 0);
  93
  94    uint32_t group_size = cs_prog_data->local_size[0] *
  95       cs_prog_data->local_size[1] * cs_prog_data->local_size[2];
  96    uint32_t remainder = group_size & (cs_prog_data->simd_size - 1);
  97
  98    if (remainder > 0)
  99       pipeline->cs_right_mask = ~0u >> (32 - remainder);
 100    else
 101       pipeline->cs_right_mask = ~0u >> (32 - cs_prog_data->simd_size);
 102
 103    const uint32_t vfe_curbe_allocation =
 104       ALIGN(cs_prog_data->push.per_thread.regs * cs_prog_data->threads +
 105             cs_prog_data->push.cross_thread.regs, 2);
 106
 107    anv_batch_emit(&pipeline->batch, GENX(MEDIA_VFE_STATE), vfe) {
 108       vfe.ScratchSpaceBasePointer = (struct anv_address) {
 109          .bo = anv_scratch_pool_alloc(device, &device->scratch_pool,
 110                                       MESA_SHADER_COMPUTE,
 111                                       cs_prog_data->base.total_scratch),
 112          .offset = 0,
 113       };
 114       vfe.PerThreadScratchSpace  = ffs(cs_prog_data->base.total_scratch / 2048);
 115 #if GEN_GEN > 7
 116       vfe.StackSize              = 0;
 117 #else
 118       vfe.GPGPUMode              = true;
 119 #endif
 120       vfe.MaximumNumberofThreads = physical_device->max_cs_threads - 1;
 121       vfe.NumberofURBEntries     = GEN_GEN <= 7 ? 0 : 2;
 122       vfe.ResetGatewayTimer      = true;
 123 #if GEN_GEN <= 8
 124       vfe.BypassGatewayControl   = true;
 125 #endif
 126       vfe.URBEntryAllocationSize = GEN_GEN <= 7 ? 0 : 2;
 127       vfe.CURBEAllocationSize    = vfe_curbe_allocation;
 128    }
 129
 130    *pPipeline = anv_pipeline_to_handle(pipeline);
 131
 132    return VK_SUCCESS;
 133 }