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 #if (ANV_GEN == 9)
  27 #  include "genxml/gen9_pack.h"
  28 #elif (ANV_GEN == 8)
  29 #  include "genxml/gen8_pack.h"
  30 #elif (ANV_IS_HASWELL)
  31 #  include "genxml/gen75_pack.h"
  32 #elif (ANV_GEN == 7)
  33 #  include "genxml/gen7_pack.h"
  34 #endif
  35
  36 VkResult
  37 genX(compute_pipeline_create)(
  38     VkDevice                                    _device,
  39     struct anv_pipeline_cache *                 cache,
  40     const VkComputePipelineCreateInfo*          pCreateInfo,
  41     const VkAllocationCallbacks*                pAllocator,
  42     VkPipeline*                                 pPipeline)
  43 {
  44    ANV_FROM_HANDLE(anv_device, device, _device);
  45    struct anv_pipeline *pipeline;
  46    VkResult result;
  47
  48    assert(pCreateInfo->sType == VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO);
  49
  50    pipeline = anv_alloc2(&device->alloc, pAllocator, sizeof(*pipeline), 8,
  51                          VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
  52    if (pipeline == NULL)
  53       return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
  54
  55    pipeline->device = device;
  56    pipeline->layout = anv_pipeline_layout_from_handle(pCreateInfo->layout);
  57
  58    pipeline->blend_state.map = NULL;
  59
  60    result = anv_reloc_list_init(&pipeline->batch_relocs,
  61                                 pAllocator ? pAllocator : &device->alloc);
  62    if (result != VK_SUCCESS) {
  63       anv_free2(&device->alloc, pAllocator, pipeline);
  64       return result;
  65    }
  66    pipeline->batch.next = pipeline->batch.start = pipeline->batch_data;
  67    pipeline->batch.end = pipeline->batch.start + sizeof(pipeline->batch_data);
  68    pipeline->batch.relocs = &pipeline->batch_relocs;
  69
  70    /* When we free the pipeline, we detect stages based on the NULL status
  71     * of various prog_data pointers.  Make them NULL by default.
  72     */
  73    memset(pipeline->prog_data, 0, sizeof(pipeline->prog_data));
  74    memset(pipeline->scratch_start, 0, sizeof(pipeline->scratch_start));
  75
  76    pipeline->vs_simd8 = NO_KERNEL;
  77    pipeline->vs_vec4 = NO_KERNEL;
  78    pipeline->gs_kernel = NO_KERNEL;
  79
  80    pipeline->active_stages = 0;
  81    pipeline->total_scratch = 0;
  82
  83    assert(pCreateInfo->stage.stage == VK_SHADER_STAGE_COMPUTE_BIT);
  84    ANV_FROM_HANDLE(anv_shader_module, module,  pCreateInfo->stage.module);
  85    anv_pipeline_compile_cs(pipeline, cache, pCreateInfo, module,
  86                            pCreateInfo->stage.pName,
  87                            pCreateInfo->stage.pSpecializationInfo);
  88
  89    pipeline->use_repclear = false;
  90
  91    const struct brw_cs_prog_data *cs_prog_data = &pipeline->cs_prog_data;
  92
  93    anv_batch_emit(&pipeline->batch, GENX(MEDIA_VFE_STATE),
  94                   .ScratchSpaceBasePointer = pipeline->scratch_start[MESA_SHADER_COMPUTE],
  95                   .PerThreadScratchSpace = ffs(cs_prog_data->base.total_scratch / 2048),
  96 #if ANV_GEN > 7
  97                   .ScratchSpaceBasePointerHigh = 0,
  98                   .StackSize = 0,
  99 #else
 100                   .GPGPUMode = true,
 101 #endif
 102                   .MaximumNumberofThreads = device->info.max_cs_threads - 1,
 103                   .NumberofURBEntries = ANV_GEN <= 7 ? 0 : 2,
 104                   .ResetGatewayTimer = true,
 105 #if ANV_GEN <= 8
 106                   .BypassGatewayControl = true,
 107 #endif
 108                   .URBEntryAllocationSize = ANV_GEN <= 7 ? 0 : 2,
 109                   .CURBEAllocationSize = 0);
 110
 111    struct brw_cs_prog_data *prog_data = &pipeline->cs_prog_data;
 112    uint32_t group_size = prog_data->local_size[0] *
 113       prog_data->local_size[1] * prog_data->local_size[2];
 114    pipeline->cs_thread_width_max = DIV_ROUND_UP(group_size, prog_data->simd_size);
 115    uint32_t remainder = group_size & (prog_data->simd_size - 1);
 116
 117    if (remainder > 0)
 118       pipeline->cs_right_mask = ~0u >> (32 - remainder);
 119    else
 120       pipeline->cs_right_mask = ~0u >> (32 - prog_data->simd_size);
 121
 122
 123    *pPipeline = anv_pipeline_to_handle(pipeline);
 124
 125    return VK_SUCCESS;
 126 }