xref: /external/mesa3d/src/gallium/drivers/iris/iris_screen.c

/*
 * Copyright © 2017 Intel Corporation
 *
 * 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.
 */

/**
 * @file iris_screen.c
 *
 * Screen related driver hooks and capability lists.
 *
 * A program may use multiple rendering contexts (iris_context), but
 * they all share a common screen (iris_screen).  Global driver state
 * can be stored in the screen; it may be accessed by multiple threads.
 */

#include <stdio.h>
#include <errno.h>
#include <sys/ioctl.h>
#include "pipe/p_defines.h"
#include "pipe/p_state.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "util/u_debug.h"
#include "util/os_file.h"
#include "util/u_cpu_detect.h"
#include "util/u_inlines.h"
#include "util/format/u_format.h"
#include "util/u_transfer_helper.h"
#include "util/u_upload_mgr.h"
#include "util/ralloc.h"
#include "util/xmlconfig.h"
#include "iris_context.h"
#include "iris_defines.h"
#include "iris_fence.h"
#include "iris_pipe.h"
#include "iris_resource.h"
#include "iris_screen.h"
#include "compiler/glsl_types.h"
#include "intel/common/intel_gem.h"
#include "intel/common/intel_l3_config.h"
#include "intel/common/intel_uuid.h"
#include "iris_monitor.h"

#define genX_call(devinfo, func, ...)             \
   switch ((devinfo)->verx10) {                   \
   case 200:                                      \
      gfx20_##func(__VA_ARGS__);                  \
      break;                                      \
   case 125:                                      \
      gfx125_##func(__VA_ARGS__);                 \
      break;                                      \
   case 120:                                      \
      gfx12_##func(__VA_ARGS__);                  \
      break;                                      \
   case 110:                                      \
      gfx11_##func(__VA_ARGS__);                  \
      break;                                      \
   case 90:                                       \
      gfx9_##func(__VA_ARGS__);                   \
      break;                                      \
   case 80:                                       \
      gfx8_##func(__VA_ARGS__);                   \
      break;                                      \
   default:                                       \
      unreachable("Unknown hardware generation"); \
   }

static const char *
iris_get_vendor(struct pipe_screen *pscreen)
{
   return "Intel";
}

static const char *
iris_get_device_vendor(struct pipe_screen *pscreen)
{
   return "Intel";
}

static void
iris_get_device_uuid(struct pipe_screen *pscreen, char *uuid)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;

   intel_uuid_compute_device_id((uint8_t *)uuid, screen->devinfo, PIPE_UUID_SIZE);
}

static void
iris_get_driver_uuid(struct pipe_screen *pscreen, char *uuid)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct intel_device_info *devinfo = screen->devinfo;

   intel_uuid_compute_driver_id((uint8_t *)uuid, devinfo, PIPE_UUID_SIZE);
}

static bool
iris_enable_clover()
{
   static int enable = -1;
   if (enable < 0)
      enable = debug_get_bool_option("IRIS_ENABLE_CLOVER", false);
   return enable;
}

static void
iris_warn_cl()
{
   static bool warned = false;
   if (warned)
      return;

   warned = true;
   fprintf(stderr, "WARNING: OpenCL support via iris driver is incomplete.\n"
                   "For a complete and conformant OpenCL implementation, use\n"
                   "https://github.com/intel/compute-runtime instead\n");
}

static const char *
iris_get_name(struct pipe_screen *pscreen)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct intel_device_info *devinfo = screen->devinfo;
   static char buf[128];

   snprintf(buf, sizeof(buf), "Mesa %s", devinfo->name);
   return buf;
}

static const char *
iris_get_cl_cts_version(struct pipe_screen *pscreen)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct intel_device_info *devinfo = screen->devinfo;

   /* https://www.khronos.org/conformance/adopters/conformant-products/opencl#submission_405 */
   if (devinfo->verx10 == 120)
      return "v2022-04-22-00";

   return NULL;
}

static int
iris_get_video_memory(struct iris_screen *screen)
{
   uint64_t vram = iris_bufmgr_vram_size(screen->bufmgr);
   uint64_t sram = iris_bufmgr_sram_size(screen->bufmgr);
   if (vram) {
      return vram / (1024 * 1024);
   } else if (sram) {
      return sram / (1024 * 1024);
   } else {
      /* This is the old code path, it get the GGTT size from the kernel
       * (which should always be 4Gb on Gfx8+).
       *
       * We should probably never end up here. This is just a fallback to get
       * some kind of value in case os_get_available_system_memory fails.
       */
      const struct intel_device_info *devinfo = screen->devinfo;
      /* Once a batch uses more than 75% of the maximum mappable size, we
       * assume that there's some fragmentation, and we start doing extra
       * flushing, etc.  That's the big cliff apps will care about.
       */
      const unsigned gpu_mappable_megabytes =
         (devinfo->aperture_bytes * 3 / 4) / (1024 * 1024);

      const long system_memory_pages = sysconf(_SC_PHYS_PAGES);
      const long system_page_size = sysconf(_SC_PAGE_SIZE);

      if (system_memory_pages <= 0 || system_page_size <= 0)
         return -1;

      const uint64_t system_memory_bytes =
         (uint64_t) system_memory_pages * (uint64_t) system_page_size;

      const unsigned system_memory_megabytes =
         (unsigned) (system_memory_bytes / (1024 * 1024));

      return MIN2(system_memory_megabytes, gpu_mappable_megabytes);
   }
}

static int
iris_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct intel_device_info *devinfo = screen->devinfo;

   switch (param) {
   case PIPE_CAP_NPOT_TEXTURES:
   case PIPE_CAP_ANISOTROPIC_FILTER:
   case PIPE_CAP_OCCLUSION_QUERY:
   case PIPE_CAP_QUERY_TIME_ELAPSED:
   case PIPE_CAP_TEXTURE_SWIZZLE:
   case PIPE_CAP_TEXTURE_MIRROR_CLAMP_TO_EDGE:
   case PIPE_CAP_BLEND_EQUATION_SEPARATE:
   case PIPE_CAP_FRAGMENT_SHADER_TEXTURE_LOD:
   case PIPE_CAP_FRAGMENT_SHADER_DERIVATIVES:
   case PIPE_CAP_PRIMITIVE_RESTART:
   case PIPE_CAP_PRIMITIVE_RESTART_FIXED_INDEX:
   case PIPE_CAP_INDEP_BLEND_ENABLE:
   case PIPE_CAP_INDEP_BLEND_FUNC:
   case PIPE_CAP_FS_COORD_ORIGIN_UPPER_LEFT:
   case PIPE_CAP_FS_COORD_PIXEL_CENTER_INTEGER:
   case PIPE_CAP_DEPTH_CLIP_DISABLE:
   case PIPE_CAP_VS_INSTANCEID:
   case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR:
   case PIPE_CAP_SEAMLESS_CUBE_MAP:
   case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
   case PIPE_CAP_CONDITIONAL_RENDER:
   case PIPE_CAP_TEXTURE_BARRIER:
   case PIPE_CAP_STREAM_OUTPUT_PAUSE_RESUME:
   case PIPE_CAP_VERTEX_COLOR_UNCLAMPED:
   case PIPE_CAP_COMPUTE:
   case PIPE_CAP_START_INSTANCE:
   case PIPE_CAP_QUERY_TIMESTAMP:
   case PIPE_CAP_TEXTURE_MULTISAMPLE:
   case PIPE_CAP_CUBE_MAP_ARRAY:
   case PIPE_CAP_TEXTURE_BUFFER_OBJECTS:
   case PIPE_CAP_QUERY_PIPELINE_STATISTICS_SINGLE:
   case PIPE_CAP_TEXTURE_QUERY_LOD:
   case PIPE_CAP_SAMPLE_SHADING:
   case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
   case PIPE_CAP_DRAW_INDIRECT:
   case PIPE_CAP_MULTI_DRAW_INDIRECT:
   case PIPE_CAP_MULTI_DRAW_INDIRECT_PARAMS:
   case PIPE_CAP_MIXED_FRAMEBUFFER_SIZES:
   case PIPE_CAP_VS_LAYER_VIEWPORT:
   case PIPE_CAP_TES_LAYER_VIEWPORT:
   case PIPE_CAP_FS_FINE_DERIVATIVE:
   case PIPE_CAP_SHADER_PACK_HALF_FLOAT:
   case PIPE_CAP_ACCELERATED:
   case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
   case PIPE_CAP_CLIP_HALFZ:
   case PIPE_CAP_TGSI_TEXCOORD:
   case PIPE_CAP_STREAM_OUTPUT_INTERLEAVE_BUFFERS:
   case PIPE_CAP_DOUBLES:
   case PIPE_CAP_INT64:
   case PIPE_CAP_SAMPLER_VIEW_TARGET:
   case PIPE_CAP_ROBUST_BUFFER_ACCESS_BEHAVIOR:
   case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
   case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
   case PIPE_CAP_FRAMEBUFFER_NO_ATTACHMENT:
   case PIPE_CAP_CULL_DISTANCE:
   case PIPE_CAP_PACKED_UNIFORMS:
   case PIPE_CAP_SIGNED_VERTEX_BUFFER_OFFSET:
   case PIPE_CAP_TEXTURE_FLOAT_LINEAR:
   case PIPE_CAP_TEXTURE_HALF_FLOAT_LINEAR:
   case PIPE_CAP_POLYGON_OFFSET_CLAMP:
   case PIPE_CAP_QUERY_SO_OVERFLOW:
   case PIPE_CAP_QUERY_BUFFER_OBJECT:
   case PIPE_CAP_TGSI_TEX_TXF_LZ:
   case PIPE_CAP_TEXTURE_QUERY_SAMPLES:
   case PIPE_CAP_SHADER_CLOCK:
   case PIPE_CAP_SHADER_BALLOT:
   case PIPE_CAP_MULTISAMPLE_Z_RESOLVE:
   case PIPE_CAP_CLEAR_SCISSORED:
   case PIPE_CAP_SHADER_GROUP_VOTE:
   case PIPE_CAP_VS_WINDOW_SPACE_POSITION:
   case PIPE_CAP_TEXTURE_GATHER_SM5:
   case PIPE_CAP_SHADER_ARRAY_COMPONENTS:
   case PIPE_CAP_GLSL_TESS_LEVELS_AS_INPUTS:
   case PIPE_CAP_LOAD_CONSTBUF:
   case PIPE_CAP_NIR_COMPACT_ARRAYS:
   case PIPE_CAP_DRAW_PARAMETERS:
   case PIPE_CAP_FS_POSITION_IS_SYSVAL:
   case PIPE_CAP_FS_FACE_IS_INTEGER_SYSVAL:
   case PIPE_CAP_COMPUTE_SHADER_DERIVATIVES:
   case PIPE_CAP_INVALIDATE_BUFFER:
   case PIPE_CAP_SURFACE_REINTERPRET_BLOCKS:
   case PIPE_CAP_TEXTURE_SHADOW_LOD:
   case PIPE_CAP_SHADER_SAMPLES_IDENTICAL:
   case PIPE_CAP_GL_SPIRV:
   case PIPE_CAP_GL_SPIRV_VARIABLE_POINTERS:
   case PIPE_CAP_DEMOTE_TO_HELPER_INVOCATION:
   case PIPE_CAP_NATIVE_FENCE_FD:
   case PIPE_CAP_MEMOBJ:
   case PIPE_CAP_MIXED_COLOR_DEPTH_BITS:
   case PIPE_CAP_FENCE_SIGNAL:
   case PIPE_CAP_IMAGE_STORE_FORMATTED:
   case PIPE_CAP_LEGACY_MATH_RULES:
   case PIPE_CAP_ALPHA_TO_COVERAGE_DITHER_CONTROL:
   case PIPE_CAP_MAP_UNSYNCHRONIZED_THREAD_SAFE:
   case PIPE_CAP_HAS_CONST_BW:
   case PIPE_CAP_CL_GL_SHARING:
      return true;
   case PIPE_CAP_UMA:
      return iris_bufmgr_vram_size(screen->bufmgr) == 0;
   case PIPE_CAP_QUERY_MEMORY_INFO:
      return iris_bufmgr_vram_size(screen->bufmgr) != 0;
   case PIPE_CAP_PREFER_BACK_BUFFER_REUSE:
      return false;
   case PIPE_CAP_FBFETCH:
      return IRIS_MAX_DRAW_BUFFERS;
   case PIPE_CAP_FBFETCH_COHERENT:
   case PIPE_CAP_CONSERVATIVE_RASTER_INNER_COVERAGE:
   case PIPE_CAP_POST_DEPTH_COVERAGE:
   case PIPE_CAP_SHADER_STENCIL_EXPORT:
   case PIPE_CAP_DEPTH_CLIP_DISABLE_SEPARATE:
   case PIPE_CAP_FRAGMENT_SHADER_INTERLOCK:
   case PIPE_CAP_ATOMIC_FLOAT_MINMAX:
      return devinfo->ver >= 9;
   case PIPE_CAP_DEPTH_BOUNDS_TEST:
      return devinfo->ver >= 12;
   case PIPE_CAP_MAX_DUAL_SOURCE_RENDER_TARGETS:
      return 1;
   case PIPE_CAP_MAX_RENDER_TARGETS:
      return IRIS_MAX_DRAW_BUFFERS;
   case PIPE_CAP_MAX_TEXTURE_2D_SIZE:
      return 16384;
   case PIPE_CAP_MAX_TEXTURE_CUBE_LEVELS:
      return IRIS_MAX_MIPLEVELS; /* 16384x16384 */
   case PIPE_CAP_MAX_TEXTURE_3D_LEVELS:
      return 12; /* 2048x2048 */
   case PIPE_CAP_MAX_STREAM_OUTPUT_BUFFERS:
      return 4;
   case PIPE_CAP_MAX_TEXTURE_ARRAY_LAYERS:
      return 2048;
   case PIPE_CAP_MAX_STREAM_OUTPUT_SEPARATE_COMPONENTS:
      return IRIS_MAX_SOL_BINDINGS / IRIS_MAX_SOL_BUFFERS;
   case PIPE_CAP_MAX_STREAM_OUTPUT_INTERLEAVED_COMPONENTS:
      return IRIS_MAX_SOL_BINDINGS;
   case PIPE_CAP_GLSL_FEATURE_LEVEL:
   case PIPE_CAP_GLSL_FEATURE_LEVEL_COMPATIBILITY:
      return 460;
   case PIPE_CAP_CONSTANT_BUFFER_OFFSET_ALIGNMENT:
      /* 3DSTATE_CONSTANT_XS requires the start of UBOs to be 32B aligned */
      return 32;
   case PIPE_CAP_MIN_MAP_BUFFER_ALIGNMENT:
      return IRIS_MAP_BUFFER_ALIGNMENT;
   case PIPE_CAP_SHADER_BUFFER_OFFSET_ALIGNMENT:
      return 4;
   case PIPE_CAP_MAX_SHADER_BUFFER_SIZE_UINT:
      return 1 << 27;
   case PIPE_CAP_TEXTURE_BUFFER_OFFSET_ALIGNMENT:
      return 16; // XXX: u_screen says 256 is the minimum value...
   case PIPE_CAP_LINEAR_IMAGE_PITCH_ALIGNMENT:
      return 1;
   case PIPE_CAP_LINEAR_IMAGE_BASE_ADDRESS_ALIGNMENT:
      return 1;
   case PIPE_CAP_TEXTURE_TRANSFER_MODES:
      return PIPE_TEXTURE_TRANSFER_BLIT;
   case PIPE_CAP_MAX_TEXEL_BUFFER_ELEMENTS_UINT:
      return IRIS_MAX_TEXTURE_BUFFER_SIZE;
   case PIPE_CAP_MAX_VIEWPORTS:
      return 16;
   case PIPE_CAP_MAX_GEOMETRY_OUTPUT_VERTICES:
      return 256;
   case PIPE_CAP_MAX_GEOMETRY_TOTAL_OUTPUT_COMPONENTS:
      return 1024;
   case PIPE_CAP_MAX_GS_INVOCATIONS:
      return 32;
   case PIPE_CAP_MAX_TEXTURE_GATHER_COMPONENTS:
      return 4;
   case PIPE_CAP_MIN_TEXTURE_GATHER_OFFSET:
      return -32;
   case PIPE_CAP_MAX_TEXTURE_GATHER_OFFSET:
      return 31;
   case PIPE_CAP_MAX_VERTEX_STREAMS:
      return 4;
   case PIPE_CAP_VENDOR_ID:
      return 0x8086;
   case PIPE_CAP_DEVICE_ID:
      return screen->devinfo->pci_device_id;
   case PIPE_CAP_VIDEO_MEMORY:
      return iris_get_video_memory(screen);
   case PIPE_CAP_MAX_SHADER_PATCH_VARYINGS:
   case PIPE_CAP_MAX_VARYINGS:
      return 32;
   case PIPE_CAP_PREFER_IMM_ARRAYS_AS_CONSTBUF:
      /* We want immediate arrays to go get uploaded as nir->constant_data by
       * nir_opt_large_constants() instead.
       */
      return 0;
   case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
      /* AMD_pinned_memory assumes the flexibility of using client memory
       * for any buffer (incl. vertex buffers) which rules out the prospect
       * of using snooped buffers, as using snooped buffers without
       * cogniscience is likely to be detrimental to performance and require
       * extensive checking in the driver for correctness, e.g. to prevent
       * illegal snoop <-> snoop transfers.
       */
      return devinfo->has_llc;
   case PIPE_CAP_THROTTLE:
      return screen->driconf.disable_throttling ? 0 : 1;

   case PIPE_CAP_CONTEXT_PRIORITY_MASK:
      return PIPE_CONTEXT_PRIORITY_LOW |
             PIPE_CONTEXT_PRIORITY_MEDIUM |
             PIPE_CONTEXT_PRIORITY_HIGH;

   case PIPE_CAP_FRONTEND_NOOP:
      return true;

   // XXX: don't hardcode 00:00:02.0 PCI here
   case PIPE_CAP_PCI_GROUP:
      return 0;
   case PIPE_CAP_PCI_BUS:
      return 0;
   case PIPE_CAP_PCI_DEVICE:
      return 2;
   case PIPE_CAP_PCI_FUNCTION:
      return 0;

   case PIPE_CAP_OPENCL_INTEGER_FUNCTIONS:
   case PIPE_CAP_INTEGER_MULTIPLY_32X16:
      return true;

   case PIPE_CAP_ALLOW_DYNAMIC_VAO_FASTPATH:
      /* Internal details of VF cache make this optimization harmful on GFX
       * version 8 and 9, because generated VERTEX_BUFFER_STATEs are cached
       * separately.
       */
      return devinfo->ver >= 11;

   case PIPE_CAP_TIMER_RESOLUTION:
      return DIV_ROUND_UP(1000000000ull, devinfo->timestamp_frequency);

   case PIPE_CAP_DEVICE_PROTECTED_CONTEXT:
      return screen->kernel_features & KERNEL_HAS_PROTECTED_CONTEXT;

   case PIPE_CAP_ASTC_VOID_EXTENTS_NEED_DENORM_FLUSH:
      return devinfo->ver == 9 && !intel_device_info_is_9lp(devinfo);

   default:
      return u_pipe_screen_get_param_defaults(pscreen, param);
   }
   return 0;
}

static float
iris_get_paramf(struct pipe_screen *pscreen, enum pipe_capf param)
{
   switch (param) {
   case PIPE_CAPF_MIN_LINE_WIDTH:
   case PIPE_CAPF_MIN_LINE_WIDTH_AA:
   case PIPE_CAPF_MIN_POINT_SIZE:
   case PIPE_CAPF_MIN_POINT_SIZE_AA:
      return 1;

   case PIPE_CAPF_POINT_SIZE_GRANULARITY:
   case PIPE_CAPF_LINE_WIDTH_GRANULARITY:
      return 0.1;

   case PIPE_CAPF_MAX_LINE_WIDTH:
   case PIPE_CAPF_MAX_LINE_WIDTH_AA:
      return 7.375f;

   case PIPE_CAPF_MAX_POINT_SIZE:
   case PIPE_CAPF_MAX_POINT_SIZE_AA:
      return 255.0f;

   case PIPE_CAPF_MAX_TEXTURE_ANISOTROPY:
      return 16.0f;
   case PIPE_CAPF_MAX_TEXTURE_LOD_BIAS:
      return 15.0f;
   case PIPE_CAPF_MIN_CONSERVATIVE_RASTER_DILATE:
   case PIPE_CAPF_MAX_CONSERVATIVE_RASTER_DILATE:
   case PIPE_CAPF_CONSERVATIVE_RASTER_DILATE_GRANULARITY:
      return 0.0f;
   default:
      unreachable("unknown param");
   }
}

static int
iris_get_shader_param(struct pipe_screen *pscreen,
                      enum pipe_shader_type p_stage,
                      enum pipe_shader_cap param)
{
   gl_shader_stage stage = stage_from_pipe(p_stage);

   if (p_stage == PIPE_SHADER_MESH ||
       p_stage == PIPE_SHADER_TASK)
      return 0;

   /* this is probably not totally correct.. but it's a start: */
   switch (param) {
   case PIPE_SHADER_CAP_MAX_INSTRUCTIONS:
      return stage == MESA_SHADER_FRAGMENT ? 1024 : 16384;
   case PIPE_SHADER_CAP_MAX_ALU_INSTRUCTIONS:
   case PIPE_SHADER_CAP_MAX_TEX_INSTRUCTIONS:
   case PIPE_SHADER_CAP_MAX_TEX_INDIRECTIONS:
      return stage == MESA_SHADER_FRAGMENT ? 1024 : 0;

   case PIPE_SHADER_CAP_MAX_CONTROL_FLOW_DEPTH:
      return UINT_MAX;

   case PIPE_SHADER_CAP_MAX_INPUTS:
      return stage == MESA_SHADER_VERTEX ? 16 : 32;
   case PIPE_SHADER_CAP_MAX_OUTPUTS:
      return 32;
   case PIPE_SHADER_CAP_MAX_CONST_BUFFER0_SIZE:
      return 16 * 1024 * sizeof(float);
   case PIPE_SHADER_CAP_MAX_CONST_BUFFERS:
      return 16;
   case PIPE_SHADER_CAP_MAX_TEMPS:
      return 256; /* GL_MAX_PROGRAM_TEMPORARIES_ARB */
   case PIPE_SHADER_CAP_CONT_SUPPORTED:
      return 0;
   case PIPE_SHADER_CAP_INDIRECT_INPUT_ADDR:
   case PIPE_SHADER_CAP_INDIRECT_OUTPUT_ADDR:
   case PIPE_SHADER_CAP_INDIRECT_TEMP_ADDR:
   case PIPE_SHADER_CAP_INDIRECT_CONST_ADDR:
      /* Lie about these to avoid st/mesa's GLSL IR lowering of indirects,
       * which we don't want.  Our compiler backend will check brw_compiler's
       * options and call nir_lower_indirect_derefs appropriately anyway.
       */
      return true;
   case PIPE_SHADER_CAP_SUBROUTINES:
      return 0;
   case PIPE_SHADER_CAP_INTEGERS:
      return 1;
   case PIPE_SHADER_CAP_INT64_ATOMICS:
   case PIPE_SHADER_CAP_FP16:
   case PIPE_SHADER_CAP_FP16_DERIVATIVES:
   case PIPE_SHADER_CAP_FP16_CONST_BUFFERS:
   case PIPE_SHADER_CAP_INT16:
   case PIPE_SHADER_CAP_GLSL_16BIT_CONSTS:
      return 0;
   case PIPE_SHADER_CAP_MAX_TEXTURE_SAMPLERS:
      return IRIS_MAX_SAMPLERS;
   case PIPE_SHADER_CAP_MAX_SAMPLER_VIEWS:
      return IRIS_MAX_TEXTURES;
   case PIPE_SHADER_CAP_MAX_SHADER_IMAGES:
      return IRIS_MAX_IMAGES;
   case PIPE_SHADER_CAP_MAX_SHADER_BUFFERS:
      return IRIS_MAX_ABOS + IRIS_MAX_SSBOS;
   case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS:
   case PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTER_BUFFERS:
      return 0;
   case PIPE_SHADER_CAP_SUPPORTED_IRS: {
      int irs = 1 << PIPE_SHADER_IR_NIR;
      if (iris_enable_clover())
         irs |= 1 << PIPE_SHADER_IR_NIR_SERIALIZED;
      return irs;
   }
   case PIPE_SHADER_CAP_TGSI_ANY_INOUT_DECL_RANGE:
   case PIPE_SHADER_CAP_TGSI_SQRT_SUPPORTED:
      return 0;
   default:
      unreachable("unknown shader param");
   }
}

static int
iris_get_compute_param(struct pipe_screen *pscreen,
                       enum pipe_shader_ir ir_type,
                       enum pipe_compute_cap param,
                       void *ret)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   const struct intel_device_info *devinfo = screen->devinfo;

   const uint32_t max_invocations =
      MIN2(1024, 32 * devinfo->max_cs_workgroup_threads);

#define RET(x) do {                  \
   if (ret)                          \
      memcpy(ret, x, sizeof(x));     \
   return sizeof(x);                 \
} while (0)

   switch (param) {
   case PIPE_COMPUTE_CAP_ADDRESS_BITS:
      /* This gets queried on OpenCL device init and is never queried by the
       * OpenGL state tracker.
       */
      iris_warn_cl();
      RET((uint32_t []){ 64 });

   case PIPE_COMPUTE_CAP_IR_TARGET:
      if (ret)
         strcpy(ret, "gen");
      return 4;

   case PIPE_COMPUTE_CAP_GRID_DIMENSION:
      RET((uint64_t []) { 3 });

   case PIPE_COMPUTE_CAP_MAX_GRID_SIZE:
      RET(((uint64_t []) { UINT32_MAX, UINT32_MAX, UINT32_MAX }));

   case PIPE_COMPUTE_CAP_MAX_BLOCK_SIZE:
      /* MaxComputeWorkGroupSize[0..2] */
      RET(((uint64_t []) {max_invocations, max_invocations, max_invocations}));

   case PIPE_COMPUTE_CAP_MAX_THREADS_PER_BLOCK:
      /* MaxComputeWorkGroupInvocations */
   case PIPE_COMPUTE_CAP_MAX_VARIABLE_THREADS_PER_BLOCK:
      /* MaxComputeVariableGroupInvocations */
      RET((uint64_t []) { max_invocations });

   case PIPE_COMPUTE_CAP_MAX_LOCAL_SIZE:
      /* MaxComputeSharedMemorySize */
      RET((uint64_t []) { 64 * 1024 });

   case PIPE_COMPUTE_CAP_IMAGES_SUPPORTED:
      RET((uint32_t []) { 1 });

   case PIPE_COMPUTE_CAP_SUBGROUP_SIZES:
      RET((uint32_t []) { 32 | 16 | 8 });

   case PIPE_COMPUTE_CAP_MAX_SUBGROUPS:
      RET((uint32_t []) { devinfo->max_cs_workgroup_threads });

   case PIPE_COMPUTE_CAP_MAX_MEM_ALLOC_SIZE:
   case PIPE_COMPUTE_CAP_MAX_GLOBAL_SIZE:
      RET((uint64_t []) { 1 << 30 }); /* TODO */

   case PIPE_COMPUTE_CAP_MAX_CLOCK_FREQUENCY:
      RET((uint32_t []) { 400 }); /* TODO */

   case PIPE_COMPUTE_CAP_MAX_COMPUTE_UNITS: {
      RET((uint32_t []) { intel_device_info_subslice_total(devinfo) });
   }

   case PIPE_COMPUTE_CAP_MAX_PRIVATE_SIZE:
      /* MaxComputeSharedMemorySize */
      RET((uint64_t []) { 64 * 1024 });

   case PIPE_COMPUTE_CAP_MAX_INPUT_SIZE:
      /* We could probably allow more; this is the OpenCL minimum */
      RET((uint64_t []) { 1024 });

   default:
      unreachable("unknown compute param");
   }
}

static uint64_t
iris_get_timestamp(struct pipe_screen *pscreen)
{
   struct iris_screen *screen = (struct iris_screen *) pscreen;
   uint64_t result;

   if (!intel_gem_read_render_timestamp(iris_bufmgr_get_fd(screen->bufmgr),
                                        screen->devinfo->kmd_type, &result))
      return 0;

   result = intel_device_info_timebase_scale(screen->devinfo, result);

   return result;
}

void
iris_screen_destroy(struct iris_screen *screen)
{
   iris_destroy_screen_measure(screen);
   util_queue_destroy(&screen->shader_compiler_queue);
   glsl_type_singleton_decref();
   iris_bo_unreference(screen->workaround_bo);
   iris_bo_unreference(screen->breakpoint_bo);
   u_transfer_helper_destroy(screen->base.transfer_helper);
   iris_bufmgr_unref(screen->bufmgr);
   disk_cache_destroy(screen->disk_cache);
   close(screen->winsys_fd);
   ralloc_free(screen);
}

static void
iris_screen_unref(struct pipe_screen *pscreen)
{
   iris_pscreen_unref(pscreen);
}

static void
iris_query_memory_info(struct pipe_screen *pscreen,
                       struct pipe_memory_info *info)
{
   struct iris_screen *screen = (struct iris_screen *)pscreen;
   struct intel_device_info di;
   memcpy(&di, screen->devinfo, sizeof(di));

   if (!intel_device_info_update_memory_info(&di, screen->fd))
      return;

   info->total_device_memory =
      (di.mem.vram.mappable.size + di.mem.vram.unmappable.size) / 1024;
   info->avail_device_memory =
      (di.mem.vram.mappable.free + di.mem.vram.unmappable.free) / 1024;
   info->total_staging_memory = di.mem.sram.mappable.size / 1024;
   info->avail_staging_memory = di.mem.sram.mappable.free / 1024;

   /* Neither kernel gives us any way to calculate this information */
   info->device_memory_evicted = 0;
   info->nr_device_memory_evictions = 0;
}

static struct disk_cache *
iris_get_disk_shader_cache(struct pipe_screen *pscreen)
{
   struct iris_screen *screen = (struct iris_screen *) pscreen;
   return screen->disk_cache;
}

static const struct intel_l3_config *
iris_get_default_l3_config(const struct intel_device_info *devinfo,
                           bool compute)
{
   bool wants_dc_cache = true;
   bool has_slm = compute;
   const struct intel_l3_weights w =
      intel_get_default_l3_weights(devinfo, wants_dc_cache, has_slm);
   return intel_get_l3_config(devinfo, w);
}

static void
iris_detect_kernel_features(struct iris_screen *screen)
{
   const struct intel_device_info *devinfo = screen->devinfo;
   /* Kernel 5.2+ */
   if (intel_gem_supports_syncobj_wait(screen->fd))
      screen->kernel_features |= KERNEL_HAS_WAIT_FOR_SUBMIT;
   if (intel_gem_supports_protected_context(screen->fd, devinfo->kmd_type))
      screen->kernel_features |= KERNEL_HAS_PROTECTED_CONTEXT;
}

static bool
iris_init_identifier_bo(struct iris_screen *screen)
{
   void *bo_map;

   bo_map = iris_bo_map(NULL, screen->workaround_bo, MAP_READ | MAP_WRITE);
   if (!bo_map)
      return false;

   assert(iris_bo_is_real(screen->workaround_bo));

   screen->workaround_address = (struct iris_address) {
      .bo = screen->workaround_bo,
      .offset = ALIGN(
         intel_debug_write_identifiers(bo_map, 4096, "Iris"), 32),
   };

   iris_bo_unmap(screen->workaround_bo);

   return true;
}

static int
iris_screen_get_fd(struct pipe_screen *pscreen)
{
   struct iris_screen *screen = (struct iris_screen *) pscreen;

   return screen->winsys_fd;
}

struct pipe_screen *
iris_screen_create(int fd, const struct pipe_screen_config *config)
{
   struct iris_screen *screen = rzalloc(NULL, struct iris_screen);
   if (!screen)
      return NULL;

   driParseConfigFiles(config->options, config->options_info, 0, "iris",
                       NULL, NULL, NULL, 0, NULL, 0);

   bool bo_reuse = false;
   int bo_reuse_mode = driQueryOptioni(config->options, "bo_reuse");
   switch (bo_reuse_mode) {
   case DRI_CONF_BO_REUSE_DISABLED:
      break;
   case DRI_CONF_BO_REUSE_ALL:
      bo_reuse = true;
      break;
   }

   process_intel_debug_variable();

   screen->bufmgr = iris_bufmgr_get_for_fd(fd, bo_reuse);
   if (!screen->bufmgr)
      return NULL;

   screen->devinfo = iris_bufmgr_get_device_info(screen->bufmgr);
   p_atomic_set(&screen->refcount, 1);

   /* Here are the i915 features we need for Iris (in chronological order) :
    *    - I915_PARAM_HAS_EXEC_NO_RELOC     (3.10)
    *    - I915_PARAM_HAS_EXEC_HANDLE_LUT   (3.10)
    *    - I915_PARAM_HAS_EXEC_BATCH_FIRST  (4.13)
    *    - I915_PARAM_HAS_EXEC_FENCE_ARRAY  (4.14)
    *    - I915_PARAM_HAS_CONTEXT_ISOLATION (4.16)
    *
    * Checking the last feature availability will include all previous ones.
    */
   if (!screen->devinfo->has_context_isolation) {
      debug_error("Kernel is too old (4.16+ required) or unusable for Iris.\n"
                  "Check your dmesg logs for loading failures.\n");
      return NULL;
   }

   screen->fd = iris_bufmgr_get_fd(screen->bufmgr);
   screen->winsys_fd = os_dupfd_cloexec(fd);

   screen->id = iris_bufmgr_create_screen_id(screen->bufmgr);

   screen->workaround_bo =
      iris_bo_alloc(screen->bufmgr, "workaround", 4096, 4096,
                    IRIS_MEMZONE_OTHER, BO_ALLOC_NO_SUBALLOC | BO_ALLOC_CAPTURE);
   if (!screen->workaround_bo)
      return NULL;

   screen->breakpoint_bo = iris_bo_alloc(screen->bufmgr, "breakpoint", 4, 4,
                                         IRIS_MEMZONE_OTHER, BO_ALLOC_ZEROED);
   if (!screen->breakpoint_bo)
      return NULL;

   if (!iris_init_identifier_bo(screen))
      return NULL;

   screen->driconf.dual_color_blend_by_location =
      driQueryOptionb(config->options, "dual_color_blend_by_location");
   screen->driconf.disable_throttling =
      driQueryOptionb(config->options, "disable_throttling");
   screen->driconf.always_flush_cache = INTEL_DEBUG(DEBUG_STALL) ||
      driQueryOptionb(config->options, "always_flush_cache");
   screen->driconf.sync_compile =
      driQueryOptionb(config->options, "sync_compile");
   screen->driconf.limit_trig_input_range =
      driQueryOptionb(config->options, "limit_trig_input_range");
   screen->driconf.lower_depth_range_rate =
      driQueryOptionf(config->options, "lower_depth_range_rate");
   screen->driconf.intel_enable_wa_14018912822 =
      driQueryOptionb(config->options, "intel_enable_wa_14018912822");
   screen->driconf.enable_tbimr =
      driQueryOptionb(config->options, "intel_tbimr");
   screen->driconf.generated_indirect_threshold =
      driQueryOptioni(config->options, "generated_indirect_threshold");

   screen->precompile = debug_get_bool_option("shader_precompile", true);

   isl_device_init(&screen->isl_dev, screen->devinfo);

   iris_compiler_init(screen);

   screen->l3_config_3d = iris_get_default_l3_config(screen->devinfo, false);
   screen->l3_config_cs = iris_get_default_l3_config(screen->devinfo, true);

   iris_disk_cache_init(screen);

   slab_create_parent(&screen->transfer_pool,
                      sizeof(struct iris_transfer), 64);

   iris_detect_kernel_features(screen);

   struct pipe_screen *pscreen = &screen->base;

   iris_init_screen_fence_functions(pscreen);
   iris_init_screen_resource_functions(pscreen);
   iris_init_screen_measure(screen);

   pscreen->destroy = iris_screen_unref;
   pscreen->get_name = iris_get_name;
   pscreen->get_vendor = iris_get_vendor;
   pscreen->get_device_vendor = iris_get_device_vendor;
   pscreen->get_cl_cts_version = iris_get_cl_cts_version;
   pscreen->get_screen_fd = iris_screen_get_fd;
   pscreen->get_param = iris_get_param;
   pscreen->get_shader_param = iris_get_shader_param;
   pscreen->get_compute_param = iris_get_compute_param;
   pscreen->get_paramf = iris_get_paramf;
   pscreen->get_compiler_options = iris_get_compiler_options;
   pscreen->get_device_uuid = iris_get_device_uuid;
   pscreen->get_driver_uuid = iris_get_driver_uuid;
   pscreen->get_disk_shader_cache = iris_get_disk_shader_cache;
   pscreen->is_format_supported = iris_is_format_supported;
   pscreen->context_create = iris_create_context;
   pscreen->get_timestamp = iris_get_timestamp;
   pscreen->query_memory_info = iris_query_memory_info;
   pscreen->get_driver_query_group_info = iris_get_monitor_group_info;
   pscreen->get_driver_query_info = iris_get_monitor_info;
   iris_init_screen_program_functions(pscreen);

   genX_call(screen->devinfo, init_screen_state, screen);
   genX_call(screen->devinfo, init_screen_gen_state, screen);

   genX_call(screen->devinfo, init_screen_state, screen);

   glsl_type_singleton_init_or_ref();

   intel_driver_ds_init();

   /* FINISHME: Big core vs little core (for CPUs that have both kinds of
    * cores) and, possibly, thread vs core should be considered here too.
    */
   unsigned compiler_threads = 1;
   const struct util_cpu_caps_t *caps = util_get_cpu_caps();
   unsigned hw_threads = caps->nr_cpus;

   if (hw_threads >= 12) {
      compiler_threads = hw_threads * 3 / 4;
   } else if (hw_threads >= 6) {
      compiler_threads = hw_threads - 2;
   } else if (hw_threads >= 2) {
      compiler_threads = hw_threads - 1;
   }

   if (!util_queue_init(&screen->shader_compiler_queue,
                        "sh", 64, compiler_threads,
                        UTIL_QUEUE_INIT_RESIZE_IF_FULL |
                        UTIL_QUEUE_INIT_SET_FULL_THREAD_AFFINITY,
                        NULL)) {
      iris_screen_destroy(screen);
      return NULL;
   }

   return pscreen;
}