/* * Copyright © 2015 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 (including the next * paragraph) 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. */ #include "anv_nir.h" #include "nir/nir_builder.h" #include "compiler/brw_nir.h" #include "util/mesa-sha1.h" #include "util/set.h" #include "vk_enum_to_str.h" #include "genxml/genX_bits.h" /* Sampler tables don't actually have a maximum size but we pick one just so * that we don't end up emitting too much state on-the-fly. */ #define MAX_SAMPLER_TABLE_SIZE 128 #define BINDLESS_OFFSET 255 enum binding_property { BINDING_PROPERTY_NORMAL = BITFIELD_BIT(0), BINDING_PROPERTY_PUSHABLE = BITFIELD_BIT(1), BINDING_PROPERTY_EMBEDDED_SAMPLER = BITFIELD_BIT(2), BINDING_PROPERTY_NO_BINDING_TABLE = BITFIELD_BIT(3), }; struct apply_pipeline_layout_state { const struct anv_physical_device *pdevice; const struct anv_pipeline_sets_layout *layout; nir_address_format desc_addr_format; nir_address_format ssbo_addr_format; nir_address_format ubo_addr_format; /* Place to flag lowered instructions so we don't lower them twice */ struct set *lowered_instrs; bool uses_constants; bool has_dynamic_buffers; bool has_independent_sets; uint8_t constants_offset; struct { bool desc_buffer_used; uint8_t desc_offset; struct anv_binding_apply_layout { uint8_t use_count; /* Binding table offset */ uint8_t surface_offset; /* Sampler table offset */ uint8_t sampler_offset; /* Embedded sampler index */ uint16_t embedded_sampler_index; /* Properties of the binding */ enum binding_property properties; /* For each binding is identified with a unique identifier for push * computation. */ uint32_t push_block; } *binding; } set[MAX_SETS]; }; /* For a given binding, tells us how many binding table entries are needed per * element. */ static uint32_t bti_multiplier(const struct apply_pipeline_layout_state *state, uint32_t set, uint32_t binding) { const struct anv_descriptor_set_layout *set_layout = state->layout->set[set].layout; const struct anv_descriptor_set_binding_layout *bind_layout = &set_layout->binding[binding]; return bind_layout->max_plane_count; } static nir_address_format addr_format_for_desc_type(VkDescriptorType desc_type, struct apply_pipeline_layout_state *state) { switch (desc_type) { case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER: case VK_DESCRIPTOR_TYPE_STORAGE_BUFFER_DYNAMIC: return state->ssbo_addr_format; case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER: case VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC: return state->ubo_addr_format; case VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK: return state->desc_addr_format; default: unreachable("Unsupported descriptor type"); } } static struct anv_binding_apply_layout * add_binding(struct apply_pipeline_layout_state *state, uint32_t set, uint32_t binding) { const struct anv_descriptor_set_layout *set_layout = state->layout->set[set].layout; const struct anv_descriptor_set_binding_layout *bind_layout = &set_layout->binding[binding]; assert(set < state->layout->num_sets); assert(binding < state->layout->set[set].layout->binding_count); if (state->set[set].binding[binding].use_count < UINT8_MAX) state->set[set].binding[binding].use_count++; /* Only flag the descriptor buffer as used if there's actually data for * this binding. This lets us be lazy and call this function constantly * without worrying about unnecessarily enabling the buffer. */ if (bind_layout->descriptor_surface_stride) state->set[set].desc_buffer_used = true; if (bind_layout->dynamic_offset_index >= 0) state->has_dynamic_buffers = true; state->set[set].binding[binding].properties |= BINDING_PROPERTY_NORMAL; if (set_layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_EMBEDDED_IMMUTABLE_SAMPLERS_BIT_EXT) state->set[set].binding[binding].properties |= BINDING_PROPERTY_EMBEDDED_SAMPLER; return &state->set[set].binding[binding]; } const VkDescriptorSetLayoutCreateFlags non_pushable_set_flags = VK_DESCRIPTOR_SET_LAYOUT_CREATE_DESCRIPTOR_BUFFER_BIT_EXT | VK_DESCRIPTOR_SET_LAYOUT_CREATE_EMBEDDED_IMMUTABLE_SAMPLERS_BIT_EXT; const VkDescriptorBindingFlags non_pushable_binding_flags = VK_DESCRIPTOR_BINDING_UPDATE_AFTER_BIND_BIT | VK_DESCRIPTOR_BINDING_UPDATE_UNUSED_WHILE_PENDING_BIT | VK_DESCRIPTOR_BINDING_PARTIALLY_BOUND_BIT; static void add_binding_type(struct apply_pipeline_layout_state *state, uint32_t set, uint32_t binding, VkDescriptorType type) { add_binding(state, set, binding); const struct anv_descriptor_set_layout *set_layout = state->layout->set[set].layout; const struct anv_descriptor_set_binding_layout *bind_layout = &set_layout->binding[binding]; /* We can't push descriptor buffers but we can for push descriptors */ const bool is_set_pushable = (set_layout->flags & non_pushable_set_flags) == 0 || set_layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR; const bool is_binding_pushable = (bind_layout->flags & non_pushable_binding_flags) == 0; if (is_set_pushable && is_binding_pushable && (state->layout->set[set].layout->binding[binding].type == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER || state->layout->set[set].layout->binding[binding].type == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER_DYNAMIC || state->layout->set[set].layout->binding[binding].type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK || state->layout->set[set].layout->binding[binding].type == VK_DESCRIPTOR_TYPE_MUTABLE_EXT) && (type == VK_DESCRIPTOR_TYPE_UNIFORM_BUFFER || type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK)) state->set[set].binding[binding].properties |= BINDING_PROPERTY_PUSHABLE; } static struct anv_binding_apply_layout * add_deref_src_binding(struct apply_pipeline_layout_state *state, nir_src src) { nir_deref_instr *deref = nir_src_as_deref(src); nir_variable *var = nir_deref_instr_get_variable(deref); return add_binding(state, var->data.descriptor_set, var->data.binding); } static void add_tex_src_binding(struct apply_pipeline_layout_state *state, nir_tex_instr *tex, nir_tex_src_type deref_src_type) { int deref_src_idx = nir_tex_instr_src_index(tex, deref_src_type); if (deref_src_idx < 0) return; struct anv_binding_apply_layout *layout = add_deref_src_binding(state, tex->src[deref_src_idx].src); /* This is likely a fallout of Wa_14020375314 but hasn't fully be * understood by HW people yet. * * In HSD-18037984222 we reported that the render target index given * through a descriptor in the address register is broken. I think the same * issue happening here when we use a descriptor given by the address * register for the sampler and when the * RENDER_SURFACE_STATE::EnableSamplerRoutetoLSC bit is enabled. This seems * to affect only texelFetch() operations. * * We probably don't want to loose the performance benefit of the route to * LSC so instead we disable dynamic descriptors by checking if a binding * array is accessed with a non constant value. * * Fixes a bunch of tests in dEQP-VK.binding_model.*.index_push_constant.* */ if (state->pdevice->info.ver >= 20 && tex->op == nir_texop_txf) { nir_deref_instr *deref = nir_src_as_deref(tex->src[deref_src_idx].src); if (deref->deref_type != nir_deref_type_var) { assert(deref->deref_type == nir_deref_type_array); if (!nir_src_is_const(deref->arr.index)) layout->properties |= BINDING_PROPERTY_NO_BINDING_TABLE; } } } static bool get_used_bindings(UNUSED nir_builder *_b, nir_instr *instr, void *_state) { struct apply_pipeline_layout_state *state = _state; switch (instr->type) { case nir_instr_type_intrinsic: { nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); switch (intrin->intrinsic) { case nir_intrinsic_vulkan_resource_index: add_binding_type(state, nir_intrinsic_desc_set(intrin), nir_intrinsic_binding(intrin), nir_intrinsic_desc_type(intrin)); break; case nir_intrinsic_image_deref_load: case nir_intrinsic_image_deref_store: case nir_intrinsic_image_deref_atomic: case nir_intrinsic_image_deref_atomic_swap: case nir_intrinsic_image_deref_size: case nir_intrinsic_image_deref_samples: case nir_intrinsic_image_deref_load_param_intel: case nir_intrinsic_image_deref_load_raw_intel: case nir_intrinsic_image_deref_store_raw_intel: case nir_intrinsic_image_deref_sparse_load: add_deref_src_binding(state, intrin->src[0]); break; case nir_intrinsic_load_constant: state->uses_constants = true; break; default: break; } break; } case nir_instr_type_tex: { nir_tex_instr *tex = nir_instr_as_tex(instr); add_tex_src_binding(state, tex, nir_tex_src_texture_deref); add_tex_src_binding(state, tex, nir_tex_src_sampler_deref); break; } default: break; } return false; } static nir_intrinsic_instr * find_descriptor_for_index_src(nir_src src, struct apply_pipeline_layout_state *state) { nir_intrinsic_instr *intrin = nir_src_as_intrinsic(src); while (intrin && intrin->intrinsic == nir_intrinsic_vulkan_resource_reindex) intrin = nir_src_as_intrinsic(intrin->src[0]); if (!intrin || intrin->intrinsic != nir_intrinsic_vulkan_resource_index) return NULL; return intrin; } static bool descriptor_has_bti(nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { assert(intrin->intrinsic == nir_intrinsic_vulkan_resource_index); uint32_t set = nir_intrinsic_desc_set(intrin); uint32_t binding = nir_intrinsic_binding(intrin); const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; if (state->set[set].binding[binding].properties & BINDING_PROPERTY_EMBEDDED_SAMPLER) return false; uint32_t surface_index; if (bind_layout->data & ANV_DESCRIPTOR_INLINE_UNIFORM) surface_index = state->set[set].desc_offset; else surface_index = state->set[set].binding[binding].surface_offset; /* Only lower to a BTI message if we have a valid binding table index. */ return surface_index < MAX_BINDING_TABLE_SIZE; } static nir_address_format descriptor_address_format(nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { assert(intrin->intrinsic == nir_intrinsic_vulkan_resource_index); return addr_format_for_desc_type(nir_intrinsic_desc_type(intrin), state); } static nir_intrinsic_instr * nir_deref_find_descriptor(nir_deref_instr *deref, struct apply_pipeline_layout_state *state) { while (1) { /* Nothing we will use this on has a variable */ assert(deref->deref_type != nir_deref_type_var); nir_deref_instr *parent = nir_src_as_deref(deref->parent); if (!parent) break; deref = parent; } assert(deref->deref_type == nir_deref_type_cast); nir_intrinsic_instr *intrin = nir_src_as_intrinsic(deref->parent); if (!intrin || intrin->intrinsic != nir_intrinsic_load_vulkan_descriptor) return NULL; return find_descriptor_for_index_src(intrin->src[0], state); } static nir_def * build_load_descriptor_mem(nir_builder *b, nir_def *desc_addr, unsigned desc_offset, unsigned num_components, unsigned bit_size, const struct apply_pipeline_layout_state *state) { switch (state->desc_addr_format) { case nir_address_format_64bit_global_32bit_offset: { nir_def *base_addr = nir_pack_64_2x32(b, nir_trim_vector(b, desc_addr, 2)); nir_def *offset32 = nir_iadd_imm(b, nir_channel(b, desc_addr, 3), desc_offset); return nir_load_global_constant_offset(b, num_components, bit_size, base_addr, offset32, .align_mul = 8, .align_offset = desc_offset % 8); } case nir_address_format_32bit_index_offset: { nir_def *surface_index = nir_channel(b, desc_addr, 0); nir_def *offset32 = nir_iadd_imm(b, nir_channel(b, desc_addr, 1), desc_offset); return nir_load_ubo(b, num_components, bit_size, surface_index, offset32, .align_mul = 8, .align_offset = desc_offset % 8, .range_base = 0, .range = num_components * bit_size / 8); } default: unreachable("Unsupported address format"); } } /* When using direct descriptor, we do not have a structure to read in memory * like anv_address_range_descriptor where all the fields match perfectly the * vec4 address format we need to generate for A64 messages. Instead we need * to build the vec4 from parsing the RENDER_SURFACE_STATE structure. Easy * enough for the surface address, lot less fun for the size where you have to * combine 3 fields scattered over multiple dwords, add one to the total and * do a check against the surface type to deal with the null descriptors. * * Fortunately we can reuse the Auxiliary surface adddress field to stash our * buffer size and just load a vec4. */ static nir_def * build_optimized_load_render_surface_state_address(nir_builder *b, nir_def *desc_addr, struct apply_pipeline_layout_state *state) { const struct intel_device_info *devinfo = &state->pdevice->info; nir_def *surface_addr = build_load_descriptor_mem(b, desc_addr, RENDER_SURFACE_STATE_SurfaceBaseAddress_start(devinfo) / 8, 4, 32, state); nir_def *addr_ldw = nir_channel(b, surface_addr, 0); nir_def *addr_udw = nir_channel(b, surface_addr, 1); nir_def *length = nir_channel(b, surface_addr, 3); return nir_vec4(b, addr_ldw, addr_udw, length, nir_imm_int(b, 0)); } /* When using direct descriptor, we do not have a structure to read in memory * like anv_address_range_descriptor where all the fields match perfectly the * vec4 address format we need to generate for A64 messages. Instead we need * to build the vec4 from parsing the RENDER_SURFACE_STATE structure. Easy * enough for the surface address, lot less fun for the size. */ static nir_def * build_non_optimized_load_render_surface_state_address(nir_builder *b, nir_def *desc_addr, struct apply_pipeline_layout_state *state) { const struct intel_device_info *devinfo = &state->pdevice->info; assert(((RENDER_SURFACE_STATE_SurfaceBaseAddress_start(devinfo) + RENDER_SURFACE_STATE_SurfaceBaseAddress_bits(devinfo) - 1) - RENDER_SURFACE_STATE_Width_start(devinfo)) / 8 <= 32); nir_def *surface_addr = build_load_descriptor_mem(b, desc_addr, RENDER_SURFACE_STATE_SurfaceBaseAddress_start(devinfo) / 8, DIV_ROUND_UP(RENDER_SURFACE_STATE_SurfaceBaseAddress_bits(devinfo), 32), 32, state); nir_def *addr_ldw = nir_channel(b, surface_addr, 0); nir_def *addr_udw = nir_channel(b, surface_addr, 1); /* Take all the RENDER_SURFACE_STATE fields from the beginning of the * structure up to the Depth field. */ const uint32_t type_sizes_dwords = DIV_ROUND_UP(RENDER_SURFACE_STATE_Depth_start(devinfo) + RENDER_SURFACE_STATE_Depth_bits(devinfo), 32); nir_def *type_sizes = build_load_descriptor_mem(b, desc_addr, 0, type_sizes_dwords, 32, state); const unsigned width_start = RENDER_SURFACE_STATE_Width_start(devinfo); /* SKL PRMs, Volume 2d: Command Reference: Structures, RENDER_SURFACE_STATE * * Width: "bits [6:0] of the number of entries in the buffer - 1" * Height: "bits [20:7] of the number of entries in the buffer - 1" * Depth: "bits [31:21] of the number of entries in the buffer - 1" */ const unsigned width_bits = 7; nir_def *width = nir_iand_imm(b, nir_ishr_imm(b, nir_channel(b, type_sizes, width_start / 32), width_start % 32), (1u << width_bits) - 1); const unsigned height_start = RENDER_SURFACE_STATE_Height_start(devinfo); const unsigned height_bits = RENDER_SURFACE_STATE_Height_bits(devinfo); nir_def *height = nir_iand_imm(b, nir_ishr_imm(b, nir_channel(b, type_sizes, height_start / 32), height_start % 32), (1u << height_bits) - 1); const unsigned depth_start = RENDER_SURFACE_STATE_Depth_start(devinfo); const unsigned depth_bits = RENDER_SURFACE_STATE_Depth_bits(devinfo); nir_def *depth = nir_iand_imm(b, nir_ishr_imm(b, nir_channel(b, type_sizes, depth_start / 32), depth_start % 32), (1u << depth_bits) - 1); nir_def *length = width; length = nir_ior(b, length, nir_ishl_imm(b, height, width_bits)); length = nir_ior(b, length, nir_ishl_imm(b, depth, width_bits + height_bits)); length = nir_iadd_imm(b, length, 1); /* Check the surface type, if it's SURFTYPE_NULL, set the length of the * buffer to 0. */ const unsigned type_start = RENDER_SURFACE_STATE_SurfaceType_start(devinfo); const unsigned type_dw = type_start / 32; nir_def *type = nir_iand_imm(b, nir_ishr_imm(b, nir_channel(b, type_sizes, type_dw), type_start % 32), (1u << RENDER_SURFACE_STATE_SurfaceType_bits(devinfo)) - 1); length = nir_bcsel(b, nir_ieq_imm(b, type, 7 /* SURFTYPE_NULL */), nir_imm_int(b, 0), length); return nir_vec4(b, addr_ldw, addr_udw, length, nir_imm_int(b, 0)); } static inline nir_def * build_load_render_surface_state_address(nir_builder *b, nir_def *desc_addr, struct apply_pipeline_layout_state *state) { if (state->pdevice->isl_dev.buffer_length_in_aux_addr) return build_optimized_load_render_surface_state_address(b, desc_addr, state); /* Wa_14019708328 */ return build_non_optimized_load_render_surface_state_address(b, desc_addr, state); } /* Load the depth of a 3D storage image. * * Either by reading the indirect descriptor value, or reading the value from * RENDER_SURFACE_STATE. * * This is necessary for VK_EXT_image_sliced_view_of_3d. */ static nir_def * build_load_storage_3d_image_depth(nir_builder *b, nir_def *desc_addr, nir_def *resinfo_depth, struct apply_pipeline_layout_state *state) { const struct intel_device_info *devinfo = &state->pdevice->info; if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT) { return build_load_descriptor_mem( b, desc_addr, offsetof(struct anv_storage_image_descriptor, image_depth), 1, 32, state); } else { nir_def *data = build_load_descriptor_mem( b, desc_addr, RENDER_SURFACE_STATE_RenderTargetViewExtent_start(devinfo) / 8, 1, 32, state); nir_def *depth = nir_ushr_imm( b, data, RENDER_SURFACE_STATE_RenderTargetViewExtent_start(devinfo) % 32); depth = nir_iand_imm( b, depth, (1u << RENDER_SURFACE_STATE_RenderTargetViewExtent_bits(devinfo)) - 1); depth = nir_iadd_imm(b, depth, 1); /* Return the minimum between the RESINFO value and the * RENDER_SURFACE_STATE::RenderTargetViewExtent value. * * Both are expressed for the current view LOD, but in the case of a * SURFTYPE_NULL, RESINFO will return the right value, while the -1 * value in RENDER_SURFACE_STATE should be ignored. */ return nir_umin(b, resinfo_depth, depth); } } static nir_def * build_load_desc_set_dynamic_index(nir_builder *b, unsigned set_idx) { return nir_iand_imm( b, anv_load_driver_uniform(b, 1, desc_surface_offsets[set_idx]), ANV_DESCRIPTOR_SET_DYNAMIC_INDEX_MASK); } static nir_def * build_load_desc_address(nir_builder *b, nir_def *set_idx, unsigned set_idx_imm, const struct apply_pipeline_layout_state *state) { nir_def *desc_offset = set_idx != NULL ? anv_load_driver_uniform_indexed(b, 1, desc_surface_offsets, set_idx) : anv_load_driver_uniform(b, 1, desc_surface_offsets[set_idx_imm]); desc_offset = nir_iand_imm(b, desc_offset, ANV_DESCRIPTOR_SET_OFFSET_MASK); if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_BUFFER && !state->pdevice->uses_ex_bso) { nir_def *bindless_base_offset = anv_load_driver_uniform(b, 1, surfaces_base_offset); desc_offset = nir_iadd(b, bindless_base_offset, desc_offset); } return nir_pack_64_2x32_split( b, desc_offset, nir_load_reloc_const_intel( b, state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_BUFFER ? BRW_SHADER_RELOC_DESCRIPTORS_BUFFER_ADDR_HIGH : BRW_SHADER_RELOC_DESCRIPTORS_ADDR_HIGH)); } /** Build a Vulkan resource index * * A "resource index" is the term used by our SPIR-V parser and the relevant * NIR intrinsics for a reference into a descriptor set. It acts much like a * deref in NIR except that it accesses opaque descriptors instead of memory. * * Coming out of SPIR-V, both the resource indices (in the form of * vulkan_resource_[re]index intrinsics) and the memory derefs (in the form * of nir_deref_instr) use the same vector component/bit size. The meaning * of those values for memory derefs (nir_deref_instr) is given by the * nir_address_format associated with the descriptor type. For resource * indices, it's an entirely internal to ANV encoding which describes, in some * sense, the address of the descriptor. Thanks to the NIR/SPIR-V rules, it * must be packed into the same size SSA values as a memory address. For this * reason, the actual encoding may depend both on the address format for * memory derefs and the descriptor address format. * * The load_vulkan_descriptor intrinsic exists to provide a transition point * between these two forms of derefs: descriptor and memory. */ static nir_def * build_res_index(nir_builder *b, uint32_t set, uint32_t binding, nir_def *array_index, struct apply_pipeline_layout_state *state) { const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; uint32_t array_size = bind_layout->array_size; uint32_t set_idx; switch (state->desc_addr_format) { case nir_address_format_64bit_global_32bit_offset: /* Descriptor set buffer accesses will go through A64 messages, so the * index to get the descriptor set buffer address is located in the * anv_push_constants::desc_surface_offsets and it's indexed by the set * number. */ set_idx = set; break; case nir_address_format_32bit_index_offset: /* Descriptor set buffer accesses will go through the binding table. The * offset is the entry in the binding table. */ assert(state->set[set].desc_offset < MAX_BINDING_TABLE_SIZE); set_idx = state->set[set].desc_offset; break; default: unreachable("Unsupported address format"); } assert(bind_layout->dynamic_offset_index < MAX_DYNAMIC_BUFFERS); nir_def *dynamic_offset_index; if (bind_layout->dynamic_offset_index >= 0) { if (state->has_independent_sets) { nir_def *dynamic_offset_start = build_load_desc_set_dynamic_index(b, set); dynamic_offset_index = nir_iadd_imm(b, dynamic_offset_start, bind_layout->dynamic_offset_index); } else { dynamic_offset_index = nir_imm_int(b, state->layout->set[set].dynamic_offset_start + bind_layout->dynamic_offset_index); } } else { dynamic_offset_index = nir_imm_int(b, 0xff); /* No dynamic offset */ } const uint32_t desc_bti = state->set[set].binding[binding].surface_offset; /* We don't care about the stride field for inline uniforms (see * build_desc_addr_for_res_index), but for anything else we should be * aligned to 8 bytes because we store a multiple of 8 in the packed info * to be able to encode a stride up to 2040 (8 * 255). */ assert(bind_layout->type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK || bind_layout->descriptor_surface_stride % 8 == 0); const uint32_t desc_stride = bind_layout->type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK ? 0 : bind_layout->descriptor_surface_stride / 8; nir_def *packed = nir_ior_imm(b, dynamic_offset_index, (desc_stride << 24) | (desc_bti << 16) | (set_idx << 8)); return nir_vec4(b, packed, nir_imm_int(b, bind_layout->descriptor_surface_offset), nir_imm_int(b, array_size - 1), array_index); } struct res_index_defs { nir_def *bti_idx; nir_def *set_idx; nir_def *dyn_offset_base; nir_def *desc_offset_base; nir_def *array_index; nir_def *desc_stride; }; static struct res_index_defs unpack_res_index(nir_builder *b, nir_def *index) { struct res_index_defs defs; nir_def *packed = nir_channel(b, index, 0); defs.desc_stride = nir_imul_imm(b, nir_extract_u8(b, packed, nir_imm_int(b, 3)), 8); defs.bti_idx = nir_extract_u8(b, packed, nir_imm_int(b, 2)); defs.set_idx = nir_extract_u8(b, packed, nir_imm_int(b, 1)); defs.dyn_offset_base = nir_extract_u8(b, packed, nir_imm_int(b, 0)); defs.desc_offset_base = nir_channel(b, index, 1); defs.array_index = nir_channel(b, index, 3); return defs; } /** Whether a surface is accessed through the bindless surface state heap */ static bool is_binding_bindless(unsigned set, unsigned binding, bool sampler, const struct apply_pipeline_layout_state *state) { /* Has binding table entry has been allocated for this binding? */ if (sampler && state->set[set].binding[binding].sampler_offset != BINDLESS_OFFSET) return false; if (!sampler && state->set[set].binding[binding].surface_offset != BINDLESS_OFFSET) return false; return true; } /** Adjust a Vulkan resource index * * This is the equivalent of nir_deref_type_ptr_as_array for resource indices. * For array descriptors, it allows us to adjust the array index. Thanks to * variable pointers, we cannot always fold this re-index operation into the * vulkan_resource_index intrinsic and we have to do it based on nothing but * the address format. */ static nir_def * build_res_reindex(nir_builder *b, nir_def *orig, nir_def *delta) { return nir_vec4(b, nir_channel(b, orig, 0), nir_channel(b, orig, 1), nir_channel(b, orig, 2), nir_iadd(b, nir_channel(b, orig, 3), delta)); } /** Get the address for a descriptor given its resource index * * Because of the re-indexing operations, we can't bounds check descriptor * array access until we have the final index. That means we end up doing the * bounds check here, if needed. See unpack_res_index() for more details. * * This function takes both a bind_layout and a desc_type which are used to * determine the descriptor stride for array descriptors. The bind_layout is * optional for buffer descriptor types. */ static nir_def * build_desc_addr_for_res_index(nir_builder *b, const VkDescriptorType desc_type, nir_def *index, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { struct res_index_defs res = unpack_res_index(b, index); nir_def *desc_offset = res.desc_offset_base; if (desc_type != VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) { /* Compute the actual descriptor offset. For inline uniform blocks, * the array index is ignored as they are only allowed to be a single * descriptor (not an array) and there is no concept of a "stride". * */ desc_offset = nir_iadd(b, desc_offset, nir_imul(b, res.array_index, res.desc_stride)); } switch (addr_format) { case nir_address_format_64bit_global_32bit_offset: case nir_address_format_64bit_bounded_global: { switch (state->desc_addr_format) { case nir_address_format_64bit_global_32bit_offset: { nir_def *base_addr = build_load_desc_address(b, res.set_idx, 0, state); return nir_vec4(b, nir_unpack_64_2x32_split_x(b, base_addr), nir_unpack_64_2x32_split_y(b, base_addr), nir_imm_int(b, UINT32_MAX), desc_offset); } case nir_address_format_32bit_index_offset: return nir_vec2(b, res.set_idx, desc_offset); default: unreachable("Unhandled address format"); } } case nir_address_format_32bit_index_offset: assert(desc_type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK); assert(state->desc_addr_format == nir_address_format_32bit_index_offset); return nir_vec2(b, res.set_idx, desc_offset); default: unreachable("Unhandled address format"); } } static nir_def * build_desc_addr_for_binding(nir_builder *b, unsigned set, unsigned binding, nir_def *array_index, unsigned plane, const struct apply_pipeline_layout_state *state) { const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; switch (state->desc_addr_format) { case nir_address_format_64bit_global_32bit_offset: case nir_address_format_64bit_bounded_global: { nir_def *set_addr = build_load_desc_address(b, NULL, set, state); nir_def *desc_offset = nir_iadd_imm(b, nir_imul_imm(b, array_index, bind_layout->descriptor_surface_stride), bind_layout->descriptor_surface_offset); if (plane != 0) { desc_offset = nir_iadd_imm( b, desc_offset, plane * bind_layout->descriptor_data_surface_size); } return nir_vec4(b, nir_unpack_64_2x32_split_x(b, set_addr), nir_unpack_64_2x32_split_y(b, set_addr), nir_imm_int(b, UINT32_MAX), desc_offset); } case nir_address_format_32bit_index_offset: { nir_def *desc_offset = nir_iadd_imm(b, nir_imul_imm(b, array_index, bind_layout->descriptor_surface_stride), bind_layout->descriptor_surface_offset); if (plane != 0) { desc_offset = nir_iadd_imm( b, desc_offset, plane * bind_layout->descriptor_data_surface_size); } return nir_vec2(b, nir_imm_int(b, state->set[set].desc_offset), desc_offset); } default: unreachable("Unhandled address format"); } } static unsigned binding_descriptor_offset(const struct apply_pipeline_layout_state *state, const struct anv_descriptor_set_binding_layout *bind_layout, bool sampler) { if (sampler && state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_DIRECT) return bind_layout->descriptor_sampler_offset; return bind_layout->descriptor_surface_offset; } static unsigned binding_descriptor_stride(const struct apply_pipeline_layout_state *state, const struct anv_descriptor_set_binding_layout *bind_layout, bool sampler) { if (sampler && state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_DIRECT) return bind_layout->descriptor_sampler_stride; return bind_layout->descriptor_surface_stride; } static nir_def * build_surface_index_for_binding(nir_builder *b, unsigned set, unsigned binding, nir_def *array_index, unsigned plane, bool non_uniform, const struct apply_pipeline_layout_state *state) { const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; const unsigned descriptor_offset = binding_descriptor_offset(state, bind_layout, false /* sampler */); const unsigned descriptor_stride = binding_descriptor_stride(state, bind_layout, false /* sampler */); const bool is_bindless = is_binding_bindless(set, binding, false /* sampler */, state); nir_def *set_offset, *surface_index; if (is_bindless) { if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT) { set_offset = nir_imm_int(b, 0xdeaddead); nir_def *desc_addr = build_desc_addr_for_binding(b, set, binding, array_index, plane, state); surface_index = build_load_descriptor_mem(b, desc_addr, 0, 1, 32, state); } else { set_offset = anv_load_driver_uniform(b, 1, desc_surface_offsets[set]); /* With bindless indexes are offsets in the descriptor buffer */ surface_index = nir_iadd_imm(b, nir_imul_imm(b, array_index, descriptor_stride), descriptor_offset); if (plane != 0) { assert(plane < bind_layout->max_plane_count); surface_index = nir_iadd_imm(b, surface_index, plane * (descriptor_stride / bind_layout->max_plane_count)); } assert(descriptor_offset % 64 == 0); assert(descriptor_stride % 64 == 0); } } else { /* Unused */ set_offset = nir_imm_int(b, 0xdeaddead); unsigned bti_stride = bti_multiplier(state, set, binding); assert(bti_stride >= 1); /* For Ycbcr descriptors, add the plane offset */ unsigned element_index = plane; /* With the binding table, it's an index in the table */ surface_index = nir_iadd_imm(b, nir_imul_imm(b, array_index, bti_stride), state->set[set].binding[binding].surface_offset + element_index); assert(state->set[set].binding[binding].surface_offset < MAX_BINDING_TABLE_SIZE); } return nir_resource_intel(b, set_offset, surface_index, array_index, nir_imm_int(b, 0) /* bindless_base_offset */, .desc_set = set, .binding = binding, .resource_block_intel = state->set[set].binding[binding].push_block, .resource_access_intel = (is_bindless ? nir_resource_intel_bindless : 0) | (non_uniform ? nir_resource_intel_non_uniform : 0) | ((state->set[set].binding[binding].properties & BINDING_PROPERTY_PUSHABLE) ? nir_resource_intel_pushable : 0)); } static nir_def * build_sampler_handle_for_binding(nir_builder *b, unsigned set, unsigned binding, nir_def *array_index, unsigned plane, bool non_uniform, const struct apply_pipeline_layout_state *state) { const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; const unsigned descriptor_offset = binding_descriptor_offset(state, bind_layout, true /* sampler */); const unsigned descriptor_stride = binding_descriptor_stride(state, bind_layout, true /* sampler */); const bool is_embedded = state->set[set].binding[binding].properties & BINDING_PROPERTY_EMBEDDED_SAMPLER; const bool is_bindless = is_binding_bindless(set, binding, true /* sampler */, state); nir_def *set_offset, *sampler_index, *sampler_base_offset = nir_imm_int(b, 0); if (is_embedded) { set_offset = nir_imm_int(b, 0xdeaddead); sampler_index = nir_load_reloc_const_intel( b, BRW_SHADER_RELOC_EMBEDDED_SAMPLER_HANDLE + state->set[set].binding[binding].embedded_sampler_index); } else if (is_bindless) { if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT) { set_offset = nir_imm_int(b, 0xdeaddead); nir_def *desc_addr = build_desc_addr_for_binding(b, set, binding, array_index, plane, state); /* This is anv_sampled_image_descriptor, the sampler handle is always * in component 1. */ nir_def *desc_data = build_load_descriptor_mem(b, desc_addr, 0, 2, 32, state); sampler_index = nir_channel(b, desc_data, 1); } else { set_offset = anv_load_driver_uniform(b, 1, desc_sampler_offsets[set]); uint32_t base_offset = descriptor_offset; /* The SAMPLER_STATE can only be located at a 64 byte in the combined * image/sampler case. Combined image/sampler is not supported to be * used with mutable descriptor types. */ if (bind_layout->data & ANV_DESCRIPTOR_SURFACE_SAMPLER) base_offset += ANV_SURFACE_STATE_SIZE; if (plane != 0) { assert(plane < bind_layout->max_plane_count); base_offset += plane * (descriptor_stride / bind_layout->max_plane_count); } sampler_index = nir_iadd_imm(b, nir_imul_imm(b, array_index, descriptor_stride), base_offset); } } else { /* Unused */ set_offset = nir_imm_int(b, 0xdeaddead); sampler_index = nir_iadd_imm(b, array_index, state->set[set].binding[binding].sampler_offset + plane); } nir_resource_data_intel sampler_resource = nir_resource_intel_sampler; if (is_bindless) sampler_resource |= nir_resource_intel_bindless; if (is_embedded) sampler_resource |= nir_resource_intel_sampler_embedded; if (non_uniform) sampler_resource |= nir_resource_intel_non_uniform; return nir_resource_intel(b, set_offset, sampler_index, array_index, sampler_base_offset, .desc_set = set, .binding = binding, .resource_access_intel = sampler_resource); } static nir_def * build_buffer_dynamic_offset_for_res_index(nir_builder *b, nir_def *dyn_offset_base, nir_def *array_index, struct apply_pipeline_layout_state *state) { nir_def *dyn_offset_idx = nir_iadd(b, dyn_offset_base, array_index); nir_def *dyn_load = anv_load_driver_uniform_indexed(b, 1, dynamic_offsets, dyn_offset_idx); return nir_bcsel(b, nir_ieq_imm(b, dyn_offset_base, 0xff), nir_imm_int(b, 0), dyn_load); } /** Convert a Vulkan resource index into a buffer address * * In some cases, this does a memory load from the descriptor set and, in * others, it simply converts from one form to another. * * See build_res_index for details about each resource index format. */ static nir_def * build_indirect_buffer_addr_for_res_index(nir_builder *b, const VkDescriptorType desc_type, nir_def *res_index, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { struct res_index_defs res = unpack_res_index(b, res_index); if (desc_type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) { assert(addr_format == state->desc_addr_format); return build_desc_addr_for_res_index(b, desc_type, res_index, addr_format, state); } else if (addr_format == nir_address_format_32bit_index_offset) { return nir_vec2(b, nir_iadd(b, res.bti_idx, res.array_index), nir_imm_int(b, 0)); } nir_def *desc_addr = build_desc_addr_for_res_index(b, desc_type, res_index, addr_format, state); nir_def *desc = build_load_descriptor_mem(b, desc_addr, 0, 4, 32, state); if (state->has_dynamic_buffers) { /* This shader has dynamic offsets and we have no way of knowing * (save from the dynamic offset base index) if this buffer has a * dynamic offset. */ nir_def *dyn_offset_idx = nir_iadd(b, res.dyn_offset_base, res.array_index); nir_def *dyn_load = anv_load_driver_uniform_indexed(b, 1, dynamic_offsets, dyn_offset_idx); nir_def *dynamic_offset = nir_bcsel(b, nir_ieq_imm(b, res.dyn_offset_base, 0xff), nir_imm_int(b, 0), dyn_load); /* The dynamic offset gets added to the base pointer so that we * have a sliding window range. */ nir_def *base_ptr = nir_pack_64_2x32(b, nir_trim_vector(b, desc, 2)); base_ptr = nir_iadd(b, base_ptr, nir_u2u64(b, dynamic_offset)); desc = nir_vec4(b, nir_unpack_64_2x32_split_x(b, base_ptr), nir_unpack_64_2x32_split_y(b, base_ptr), nir_channel(b, desc, 2), nir_channel(b, desc, 3)); } /* The last element of the vec4 is always zero. * * See also struct anv_address_range_descriptor */ return nir_vec4(b, nir_channel(b, desc, 0), nir_channel(b, desc, 1), nir_channel(b, desc, 2), nir_imm_int(b, 0)); } static nir_def * build_direct_buffer_addr_for_res_index(nir_builder *b, const VkDescriptorType desc_type, nir_def *res_index, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { if (desc_type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) { assert(addr_format == state->desc_addr_format); return build_desc_addr_for_res_index(b, desc_type, res_index, addr_format, state); } else if (addr_format == nir_address_format_32bit_index_offset) { struct res_index_defs res = unpack_res_index(b, res_index); return nir_vec2(b, nir_iadd(b, res.desc_offset_base, nir_imul(b, res.array_index, res.desc_stride)), nir_imm_int(b, 0)); } nir_def *desc_addr = build_desc_addr_for_res_index(b, desc_type, res_index, addr_format, state); nir_def *addr = build_load_render_surface_state_address(b, desc_addr, state); if (state->has_dynamic_buffers) { struct res_index_defs res = unpack_res_index(b, res_index); /* This shader has dynamic offsets and we have no way of knowing (save * from the dynamic offset base index) if this buffer has a dynamic * offset. */ nir_def *dynamic_offset = build_buffer_dynamic_offset_for_res_index( b, res.dyn_offset_base, res.array_index, state); /* The dynamic offset gets added to the base pointer so that we * have a sliding window range. */ nir_def *base_ptr = nir_pack_64_2x32(b, nir_trim_vector(b, addr, 2)); base_ptr = nir_iadd(b, base_ptr, nir_u2u64(b, dynamic_offset)); addr = nir_vec4(b, nir_unpack_64_2x32_split_x(b, base_ptr), nir_unpack_64_2x32_split_y(b, base_ptr), nir_channel(b, addr, 2), nir_channel(b, addr, 3)); } /* The last element of the vec4 is always zero. * * See also struct anv_address_range_descriptor */ return nir_vec4(b, nir_channel(b, addr, 0), nir_channel(b, addr, 1), nir_channel(b, addr, 2), nir_imm_int(b, 0)); } static nir_def * build_buffer_addr_for_res_index(nir_builder *b, const VkDescriptorType desc_type, nir_def *res_index, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT) return build_indirect_buffer_addr_for_res_index(b, desc_type, res_index, addr_format, state); else return build_direct_buffer_addr_for_res_index(b, desc_type, res_index, addr_format, state); } static nir_def * build_buffer_addr_for_binding(nir_builder *b, const VkDescriptorType desc_type, unsigned set, unsigned binding, nir_def *res_index, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { if (addr_format != nir_address_format_32bit_index_offset) return build_buffer_addr_for_res_index(b, desc_type, res_index, addr_format, state); if (desc_type == VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) { const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; return nir_vec2(b, nir_imm_int(b, state->set[set].desc_offset), nir_imm_int(b, bind_layout->descriptor_surface_offset)); } struct res_index_defs res = unpack_res_index(b, res_index); return nir_vec2(b, build_surface_index_for_binding(b, set, binding, res.array_index, 0 /* plane */, false /* non_uniform */, state), nir_imm_int(b, 0)); } /** Loads descriptor memory for a variable-based deref chain * * The deref chain has to terminate at a variable with a descriptor_set and * binding set. This is used for images, textures, and samplers. */ static nir_def * build_load_var_deref_surface_handle(nir_builder *b, nir_deref_instr *deref, bool non_uniform, bool *out_is_bindless, struct apply_pipeline_layout_state *state) { nir_variable *var = nir_deref_instr_get_variable(deref); const uint32_t set = var->data.descriptor_set; const uint32_t binding = var->data.binding; *out_is_bindless = is_binding_bindless(set, binding, false /* sampler */, state); nir_def *array_index; if (deref->deref_type != nir_deref_type_var) { assert(deref->deref_type == nir_deref_type_array); assert(nir_deref_instr_parent(deref)->deref_type == nir_deref_type_var); array_index = deref->arr.index.ssa; } else { array_index = nir_imm_int(b, 0); } return build_surface_index_for_binding(b, set, binding, array_index, 0 /* plane */, non_uniform, state); } /** A recursive form of build_res_index() * * This recursively walks a resource [re]index chain and builds the resource * index. It places the new code with the resource [re]index operation in the * hopes of better CSE. This means the cursor is not where you left it when * this function returns. */ static nir_def * build_res_index_for_chain(nir_builder *b, nir_intrinsic_instr *intrin, nir_address_format addr_format, uint32_t *set, uint32_t *binding, struct apply_pipeline_layout_state *state) { if (intrin->intrinsic == nir_intrinsic_vulkan_resource_index) { b->cursor = nir_before_instr(&intrin->instr); *set = nir_intrinsic_desc_set(intrin); *binding = nir_intrinsic_binding(intrin); return build_res_index(b, *set, *binding, intrin->src[0].ssa, state); } else { assert(intrin->intrinsic == nir_intrinsic_vulkan_resource_reindex); nir_intrinsic_instr *parent = nir_src_as_intrinsic(intrin->src[0]); nir_def *index = build_res_index_for_chain(b, parent, addr_format, set, binding, state); b->cursor = nir_before_instr(&intrin->instr); return build_res_reindex(b, index, intrin->src[1].ssa); } } /** Builds a buffer address for a given vulkan [re]index intrinsic * * The cursor is not where you left it when this function returns. */ static nir_def * build_buffer_addr_for_idx_intrin(nir_builder *b, nir_intrinsic_instr *idx_intrin, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { uint32_t set = UINT32_MAX, binding = UINT32_MAX; nir_def *res_index = build_res_index_for_chain(b, idx_intrin, addr_format, &set, &binding, state); const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; return build_buffer_addr_for_binding(b, bind_layout->type, set, binding, res_index, addr_format, state); } /** Builds a buffer address for deref chain * * This assumes that you can chase the chain all the way back to the original * vulkan_resource_index intrinsic. * * The cursor is not where you left it when this function returns. */ static nir_def * build_buffer_addr_for_deref(nir_builder *b, nir_deref_instr *deref, nir_address_format addr_format, struct apply_pipeline_layout_state *state) { nir_deref_instr *parent = nir_deref_instr_parent(deref); if (parent) { nir_def *addr = build_buffer_addr_for_deref(b, parent, addr_format, state); b->cursor = nir_before_instr(&deref->instr); return nir_explicit_io_address_from_deref(b, deref, addr, addr_format); } nir_intrinsic_instr *load_desc = nir_src_as_intrinsic(deref->parent); assert(load_desc->intrinsic == nir_intrinsic_load_vulkan_descriptor); nir_intrinsic_instr *idx_intrin = nir_src_as_intrinsic(load_desc->src[0]); b->cursor = nir_before_instr(&deref->instr); return build_buffer_addr_for_idx_intrin(b, idx_intrin, addr_format, state); } static bool try_lower_direct_buffer_intrinsic(nir_builder *b, nir_intrinsic_instr *intrin, bool is_atomic, struct apply_pipeline_layout_state *state) { nir_deref_instr *deref = nir_src_as_deref(intrin->src[0]); if (!nir_deref_mode_is_one_of(deref, nir_var_mem_ubo | nir_var_mem_ssbo)) return false; nir_intrinsic_instr *desc = nir_deref_find_descriptor(deref, state); if (desc == NULL) { /* We should always be able to find the descriptor for UBO access. */ assert(nir_deref_mode_is_one_of(deref, nir_var_mem_ssbo)); return false; } const unsigned set = nir_intrinsic_desc_set(desc); const unsigned binding = nir_intrinsic_binding(desc); const struct anv_descriptor_set_binding_layout *bind_layout = &state->layout->set[set].layout->binding[binding]; nir_address_format addr_format = descriptor_address_format(desc, state); /* Although we could lower non uniform binding table accesses with * nir_opt_non_uniform_access, we might as well use an A64 message and * avoid the loops inserted by that lowering pass. */ if (nir_intrinsic_access(intrin) & ACCESS_NON_UNIFORM) return false; if (nir_deref_mode_is(deref, nir_var_mem_ssbo)) { /* 64-bit atomics only support A64 messages so we can't lower them to * the index+offset model. */ if (is_atomic && intrin->def.bit_size == 64 && !state->pdevice->info.has_lsc) return false; /* If we don't have a BTI for this binding and we're using indirect * descriptors, we'll use A64 messages. This is handled in the main * lowering path. */ if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT && !descriptor_has_bti(desc, state)) return false; /* Rewrite to 32bit_index_offset whenever we can */ addr_format = nir_address_format_32bit_index_offset; } else { assert(nir_deref_mode_is(deref, nir_var_mem_ubo)); /* If we don't have a BTI for this binding and we're using indirect * descriptors, we'll use A64 messages. This is handled in the main * lowering path. * * We make an exception for uniform blocks which are built from the * descriptor set base address + offset. There is no indirect data to * fetch. */ if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT && bind_layout->type != VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK && !descriptor_has_bti(desc, state)) return false; /* If this is an inline uniform and the shader stage is bindless, we * can't switch to 32bit_index_offset. */ if (bind_layout->type != VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK || !brw_shader_stage_requires_bindless_resources(b->shader->info.stage)) addr_format = nir_address_format_32bit_index_offset; } /* If a dynamic has not been assigned a binding table entry, we need to * bail here. */ if (vk_descriptor_type_is_dynamic(bind_layout->type) && !descriptor_has_bti(desc, state)) return false; nir_def *addr = build_buffer_addr_for_deref(b, deref, addr_format, state); b->cursor = nir_before_instr(&intrin->instr); nir_lower_explicit_io_instr(b, intrin, addr, addr_format); return true; } static bool lower_load_accel_struct_desc(nir_builder *b, nir_intrinsic_instr *load_desc, struct apply_pipeline_layout_state *state) { assert(load_desc->intrinsic == nir_intrinsic_load_vulkan_descriptor); nir_intrinsic_instr *idx_intrin = nir_src_as_intrinsic(load_desc->src[0]); /* It doesn't really matter what address format we choose as * everything will constant-fold nicely. Choose one that uses the * actual descriptor buffer. */ const nir_address_format addr_format = nir_address_format_64bit_bounded_global; uint32_t set = UINT32_MAX, binding = UINT32_MAX; nir_def *res_index = build_res_index_for_chain(b, idx_intrin, addr_format, &set, &binding, state); b->cursor = nir_before_instr(&load_desc->instr); struct res_index_defs res = unpack_res_index(b, res_index); nir_def *desc_addr = build_desc_addr_for_binding(b, set, binding, res.array_index, 0 /* plane */, state); /* Acceleration structure descriptors are always uint64_t */ nir_def *desc = build_load_descriptor_mem(b, desc_addr, 0, 1, 64, state); assert(load_desc->def.bit_size == 64); assert(load_desc->def.num_components == 1); nir_def_replace(&load_desc->def, desc); return true; } static bool lower_direct_buffer_instr(nir_builder *b, nir_instr *instr, void *_state) { struct apply_pipeline_layout_state *state = _state; if (instr->type != nir_instr_type_intrinsic) return false; nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); switch (intrin->intrinsic) { case nir_intrinsic_load_deref: case nir_intrinsic_store_deref: return try_lower_direct_buffer_intrinsic(b, intrin, false, state); case nir_intrinsic_deref_atomic: case nir_intrinsic_deref_atomic_swap: return try_lower_direct_buffer_intrinsic(b, intrin, true, state); case nir_intrinsic_get_ssbo_size: { /* The get_ssbo_size intrinsic always just takes a * index/reindex intrinsic. */ nir_intrinsic_instr *idx_intrin = find_descriptor_for_index_src(intrin->src[0], state); if (idx_intrin == NULL) return false; /* We just checked that this is a BTI descriptor */ const nir_address_format addr_format = nir_address_format_32bit_index_offset; b->cursor = nir_before_instr(&intrin->instr); uint32_t set = UINT32_MAX, binding = UINT32_MAX; nir_def *res_index = build_res_index_for_chain(b, idx_intrin, addr_format, &set, &binding, state); bool non_uniform = nir_intrinsic_access(intrin) & ACCESS_NON_UNIFORM; nir_def *surface_index = build_surface_index_for_binding(b, set, binding, nir_channel(b, res_index, 3), 0 /* plane */, non_uniform, state); nir_src_rewrite(&intrin->src[0], surface_index); _mesa_set_add(state->lowered_instrs, intrin); return true; } case nir_intrinsic_load_vulkan_descriptor: if (nir_intrinsic_desc_type(intrin) == VK_DESCRIPTOR_TYPE_ACCELERATION_STRUCTURE_KHR) return lower_load_accel_struct_desc(b, intrin, state); return false; default: return false; } } static bool lower_res_index_intrinsic(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_before_instr(&intrin->instr); nir_def *index = build_res_index(b, nir_intrinsic_desc_set(intrin), nir_intrinsic_binding(intrin), intrin->src[0].ssa, state); assert(intrin->def.bit_size == index->bit_size); assert(intrin->def.num_components == index->num_components); nir_def_replace(&intrin->def, index); return true; } static bool lower_res_reindex_intrinsic(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_before_instr(&intrin->instr); nir_def *index = build_res_reindex(b, intrin->src[0].ssa, intrin->src[1].ssa); assert(intrin->def.bit_size == index->bit_size); assert(intrin->def.num_components == index->num_components); nir_def_replace(&intrin->def, index); return true; } static bool lower_load_vulkan_descriptor(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_before_instr(&intrin->instr); const VkDescriptorType desc_type = nir_intrinsic_desc_type(intrin); nir_address_format addr_format = addr_format_for_desc_type(desc_type, state); nir_def *desc = build_buffer_addr_for_res_index(b, desc_type, intrin->src[0].ssa, addr_format, state); assert(intrin->def.bit_size == desc->bit_size); assert(intrin->def.num_components == desc->num_components); nir_def_replace(&intrin->def, desc); return true; } static bool lower_get_ssbo_size(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { if (_mesa_set_search(state->lowered_instrs, intrin)) return false; b->cursor = nir_before_instr(&intrin->instr); const nir_address_format addr_format = nir_address_format_64bit_bounded_global; nir_def *desc_addr = nir_build_addr_iadd_imm( b, build_desc_addr_for_res_index(b, VK_DESCRIPTOR_TYPE_STORAGE_BUFFER, intrin->src[0].ssa, addr_format, state), addr_format, nir_var_mem_ssbo, state->pdevice->isl_dev.ss.size); nir_def *desc_range; if (state->layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_INDIRECT) { /* Load the anv_address_range_descriptor */ desc_range = build_load_descriptor_mem(b, desc_addr, 0, 4, 32, state); } else { /* Build a vec4 similar to anv_address_range_descriptor using the * RENDER_SURFACE_STATE. */ desc_range = build_load_render_surface_state_address(b, desc_addr, state); } nir_def *size = nir_channel(b, desc_range, 2); nir_def_replace(&intrin->def, size); return true; } static bool lower_image_intrinsic(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { nir_deref_instr *deref = nir_src_as_deref(intrin->src[0]); b->cursor = nir_before_instr(&intrin->instr); bool non_uniform = nir_intrinsic_access(intrin) & ACCESS_NON_UNIFORM; bool is_bindless; nir_def *handle = build_load_var_deref_surface_handle(b, deref, non_uniform, &is_bindless, state); nir_rewrite_image_intrinsic(intrin, handle, is_bindless); return true; } static bool lower_image_size_intrinsic(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { if (nir_intrinsic_image_dim(intrin) != GLSL_SAMPLER_DIM_3D) return lower_image_intrinsic(b, intrin, state); nir_deref_instr *deref = nir_src_as_deref(intrin->src[0]); b->cursor = nir_before_instr(&intrin->instr); bool non_uniform = nir_intrinsic_access(intrin) & ACCESS_NON_UNIFORM; bool is_bindless; nir_def *handle = build_load_var_deref_surface_handle(b, deref, non_uniform, &is_bindless, state); nir_rewrite_image_intrinsic(intrin, handle, is_bindless); nir_variable *var = nir_deref_instr_get_variable(deref); const uint32_t set = var->data.descriptor_set; const uint32_t binding = var->data.binding; nir_def *array_index; if (deref->deref_type != nir_deref_type_var) { assert(deref->deref_type == nir_deref_type_array); assert(nir_deref_instr_parent(deref)->deref_type == nir_deref_type_var); array_index = deref->arr.index.ssa; } else { array_index = nir_imm_int(b, 0); } nir_def *desc_addr = build_desc_addr_for_binding( b, set, binding, array_index, 0 /* plane */, state); b->cursor = nir_after_instr(&intrin->instr); nir_def *image_depth = build_load_storage_3d_image_depth(b, desc_addr, nir_channel(b, &intrin->def, 2), state); nir_def *comps[4] = {}; for (unsigned c = 0; c < intrin->def.num_components; c++) comps[c] = c == 2 ? image_depth : nir_channel(b, &intrin->def, c); nir_def *vec = nir_vec(b, comps, intrin->def.num_components); nir_def_rewrite_uses_after(&intrin->def, vec, vec->parent_instr); return true; } static bool lower_load_constant(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_instr_remove(&intrin->instr); /* Any constant-offset load_constant instructions should have been removed * by constant folding. */ assert(!nir_src_is_const(intrin->src[0])); nir_def *offset = nir_iadd_imm(b, intrin->src[0].ssa, nir_intrinsic_base(intrin)); unsigned load_size = intrin->def.num_components * intrin->def.bit_size / 8; unsigned load_align = intrin->def.bit_size / 8; assert(load_size < b->shader->constant_data_size); unsigned max_offset = b->shader->constant_data_size - load_size; offset = nir_umin(b, offset, nir_imm_int(b, max_offset)); nir_def *const_data_addr = nir_pack_64_2x32_split(b, nir_iadd(b, nir_load_reloc_const_intel(b, BRW_SHADER_RELOC_CONST_DATA_ADDR_LOW), offset), nir_load_reloc_const_intel(b, BRW_SHADER_RELOC_CONST_DATA_ADDR_HIGH)); nir_def *data = nir_load_global_constant(b, const_data_addr, load_align, intrin->def.num_components, intrin->def.bit_size); nir_def_rewrite_uses(&intrin->def, data); return true; } static bool lower_base_workgroup_id(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_instr_remove(&intrin->instr); nir_def *base_workgroup_id = anv_load_driver_uniform(b, 3, cs.base_work_group_id[0]); nir_def_rewrite_uses(&intrin->def, base_workgroup_id); return true; } static void lower_tex_deref(nir_builder *b, nir_tex_instr *tex, nir_tex_src_type deref_src_type, unsigned base_index, unsigned plane, struct apply_pipeline_layout_state *state) { int deref_src_idx = nir_tex_instr_src_index(tex, deref_src_type); if (deref_src_idx < 0) return; nir_deref_instr *deref = nir_src_as_deref(tex->src[deref_src_idx].src); nir_variable *var = nir_deref_instr_get_variable(deref); const bool is_sampler = deref_src_type == nir_tex_src_sampler_deref; const unsigned set = var->data.descriptor_set; const unsigned binding = var->data.binding; const bool bindless = is_binding_bindless(set, binding, is_sampler, state); nir_def *array_index = NULL; if (deref->deref_type != nir_deref_type_var) { assert(deref->deref_type == nir_deref_type_array); array_index = deref->arr.index.ssa; } else { array_index = nir_imm_int(b, 0); } nir_tex_src_type offset_src_type; nir_def *index; if (deref_src_type == nir_tex_src_texture_deref) { index = build_surface_index_for_binding(b, set, binding, array_index, plane, tex->texture_non_uniform, state); offset_src_type = bindless ? nir_tex_src_texture_handle : nir_tex_src_texture_offset; } else { assert(deref_src_type == nir_tex_src_sampler_deref); index = build_sampler_handle_for_binding(b, set, binding, array_index, plane, tex->sampler_non_uniform, state); offset_src_type = bindless ? nir_tex_src_sampler_handle : nir_tex_src_sampler_offset; } nir_src_rewrite(&tex->src[deref_src_idx].src, index); tex->src[deref_src_idx].src_type = offset_src_type; } static uint32_t tex_instr_get_and_remove_plane_src(nir_tex_instr *tex) { int plane_src_idx = nir_tex_instr_src_index(tex, nir_tex_src_plane); if (plane_src_idx < 0) return 0; unsigned plane = nir_src_as_uint(tex->src[plane_src_idx].src); nir_tex_instr_remove_src(tex, plane_src_idx); return plane; } static nir_def * build_def_array_select(nir_builder *b, nir_def **srcs, nir_def *idx, unsigned start, unsigned end) { if (start == end - 1) { return srcs[start]; } else { unsigned mid = start + (end - start) / 2; return nir_bcsel(b, nir_ilt_imm(b, idx, mid), build_def_array_select(b, srcs, idx, start, mid), build_def_array_select(b, srcs, idx, mid, end)); } } static bool lower_tex(nir_builder *b, nir_tex_instr *tex, struct apply_pipeline_layout_state *state) { unsigned plane = tex_instr_get_and_remove_plane_src(tex); b->cursor = nir_before_instr(&tex->instr); lower_tex_deref(b, tex, nir_tex_src_texture_deref, tex->texture_index, plane, state); lower_tex_deref(b, tex, nir_tex_src_sampler_deref, tex->sampler_index, plane, state); /* The whole lot will be embedded in the offset/handle source */ tex->texture_index = 0; tex->sampler_index = 0; return true; } static bool lower_ray_query_globals(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { b->cursor = nir_instr_remove(&intrin->instr); nir_def *rq_globals = anv_load_driver_uniform(b, 1, ray_query_globals); nir_def_rewrite_uses(&intrin->def, rq_globals); return true; } static bool lower_num_workgroups(nir_builder *b, nir_intrinsic_instr *intrin, struct apply_pipeline_layout_state *state) { /* For those stages, HW will generate values through payload registers. */ if (gl_shader_stage_is_mesh(b->shader->info.stage)) return false; b->cursor = nir_instr_remove(&intrin->instr); nir_def *num_workgroups; /* On Gfx12.5+ we use the inline register to push the values, on prior * generation we use push constants. */ if (state->pdevice->info.verx10 >= 125) { num_workgroups = nir_load_inline_data_intel( b, 3, 32, .base = ANV_INLINE_PARAM_NUM_WORKGROUPS_OFFSET); } else { num_workgroups = anv_load_driver_uniform(b, 3, cs.num_work_groups[0]); } nir_def *num_workgroups_indirect; nir_push_if(b, nir_ieq_imm(b, nir_channel(b, num_workgroups, 0), UINT32_MAX)); { nir_def *addr = nir_pack_64_2x32_split(b, nir_channel(b, num_workgroups, 1), nir_channel(b, num_workgroups, 2)); num_workgroups_indirect = nir_load_global_constant(b, addr, 4, 3, 32); } nir_pop_if(b, NULL); num_workgroups = nir_if_phi(b, num_workgroups_indirect, num_workgroups); nir_def_rewrite_uses(&intrin->def, num_workgroups); return true; } static bool apply_pipeline_layout(nir_builder *b, nir_instr *instr, void *_state) { struct apply_pipeline_layout_state *state = _state; switch (instr->type) { case nir_instr_type_intrinsic: { nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); switch (intrin->intrinsic) { case nir_intrinsic_vulkan_resource_index: return lower_res_index_intrinsic(b, intrin, state); case nir_intrinsic_vulkan_resource_reindex: return lower_res_reindex_intrinsic(b, intrin, state); case nir_intrinsic_load_vulkan_descriptor: return lower_load_vulkan_descriptor(b, intrin, state); case nir_intrinsic_get_ssbo_size: return lower_get_ssbo_size(b, intrin, state); case nir_intrinsic_image_deref_load: case nir_intrinsic_image_deref_store: case nir_intrinsic_image_deref_atomic: case nir_intrinsic_image_deref_atomic_swap: case nir_intrinsic_image_deref_samples: case nir_intrinsic_image_deref_load_param_intel: case nir_intrinsic_image_deref_load_raw_intel: case nir_intrinsic_image_deref_store_raw_intel: case nir_intrinsic_image_deref_sparse_load: return lower_image_intrinsic(b, intrin, state); case nir_intrinsic_image_deref_size: return lower_image_size_intrinsic(b, intrin, state); case nir_intrinsic_load_constant: return lower_load_constant(b, intrin, state); case nir_intrinsic_load_base_workgroup_id: return lower_base_workgroup_id(b, intrin, state); case nir_intrinsic_load_ray_query_global_intel: return lower_ray_query_globals(b, intrin, state); case nir_intrinsic_load_num_workgroups: return lower_num_workgroups(b, intrin, state); default: return false; } break; } case nir_instr_type_tex: return lower_tex(b, nir_instr_as_tex(instr), state); default: return false; } } struct binding_info { uint32_t binding; uint8_t set; uint16_t score; }; static int compare_binding_infos(const void *_a, const void *_b) { const struct binding_info *a = _a, *b = _b; if (a->score != b->score) return b->score - a->score; if (a->set != b->set) return a->set - b->set; return a->binding - b->binding; } #ifndef NDEBUG static void anv_validate_pipeline_layout(const struct anv_pipeline_sets_layout *layout, nir_shader *shader) { nir_foreach_function_impl(impl, shader) { nir_foreach_block(block, impl) { nir_foreach_instr(instr, block) { if (instr->type != nir_instr_type_intrinsic) continue; nir_intrinsic_instr *intrin = nir_instr_as_intrinsic(instr); if (intrin->intrinsic != nir_intrinsic_vulkan_resource_index) continue; unsigned set = nir_intrinsic_desc_set(intrin); assert(layout->set[set].layout); } } } } #endif static bool binding_is_promotable_to_push(const struct anv_descriptor_set_layout *set_layout, const struct anv_descriptor_set_binding_layout *bind_layout) { if (set_layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR) return true; if (set_layout->flags & (VK_DESCRIPTOR_SET_LAYOUT_CREATE_DESCRIPTOR_BUFFER_BIT_EXT | VK_DESCRIPTOR_SET_LAYOUT_CREATE_EMBEDDED_IMMUTABLE_SAMPLERS_BIT_EXT)) return false; return (bind_layout->flags & non_pushable_binding_flags) == 0; } static void add_null_bti_entry(struct anv_pipeline_bind_map *map) { map->surface_to_descriptor[map->surface_count++] = (struct anv_pipeline_binding) { .set = ANV_DESCRIPTOR_SET_NULL, }; assert(map->surface_count <= MAX_BINDING_TABLE_SIZE); } static void add_bti_entry(struct anv_pipeline_bind_map *map, uint32_t set, uint32_t binding, uint32_t element, uint32_t plane, const struct anv_descriptor_set_binding_layout *bind_layout) { map->surface_to_descriptor[map->surface_count++] = (struct anv_pipeline_binding) { .set = set, .binding = binding, .index = bind_layout->descriptor_index + element, .set_offset = bind_layout->descriptor_surface_offset + element * bind_layout->descriptor_surface_stride + plane * bind_layout->descriptor_data_surface_size, .plane = plane, }; assert(map->surface_count <= MAX_BINDING_TABLE_SIZE); } static void add_dynamic_bti_entry(struct anv_pipeline_bind_map *map, uint32_t set, uint32_t binding, uint32_t element, const struct anv_pipeline_sets_layout *layout, const struct anv_descriptor_set_binding_layout *bind_layout) { map->surface_to_descriptor[map->surface_count++] = (struct anv_pipeline_binding) { .set = set, .binding = binding, .index = bind_layout->descriptor_index + element, .set_offset = bind_layout->descriptor_surface_offset + element * bind_layout->descriptor_surface_stride, .dynamic_offset_index = bind_layout->dynamic_offset_index + element, }; assert(map->surface_count <= MAX_BINDING_TABLE_SIZE); } static void add_sampler_entry(struct anv_pipeline_bind_map *map, uint32_t set, uint32_t binding, uint32_t element, uint32_t plane, const struct anv_pipeline_sets_layout *layout, const struct anv_descriptor_set_binding_layout *bind_layout) { assert((bind_layout->descriptor_index + element) < layout->set[set].layout->descriptor_count); map->sampler_to_descriptor[map->sampler_count++] = (struct anv_pipeline_binding) { .set = set, .binding = binding, .index = bind_layout->descriptor_index + element, .plane = plane, }; } static void add_push_entry(struct anv_pipeline_push_map *push_map, uint32_t set, uint32_t binding, uint32_t element, const struct anv_pipeline_sets_layout *layout, const struct anv_descriptor_set_binding_layout *bind_layout) { push_map->block_to_descriptor[push_map->block_count++] = (struct anv_pipeline_binding) { .set = set, .binding = binding, .index = bind_layout->descriptor_index + element, .dynamic_offset_index = bind_layout->dynamic_offset_index + element, }; } static void add_embedded_sampler_entry(struct apply_pipeline_layout_state *state, struct anv_pipeline_bind_map *map, uint32_t set, uint32_t binding) { state->set[set].binding[binding].embedded_sampler_index = map->embedded_sampler_count; struct anv_pipeline_embedded_sampler_binding *sampler = &map->embedded_sampler_to_binding[map->embedded_sampler_count++]; const struct anv_descriptor_set_layout *set_layout = state->layout->set[set].layout; const struct anv_descriptor_set_binding_layout *bind_layout = &set_layout->binding[binding]; *sampler = (struct anv_pipeline_embedded_sampler_binding) { .set = set, .binding = binding, }; assert(sizeof(sampler->key.sampler) == sizeof(bind_layout->immutable_samplers[0]->state_no_bc[0])); memcpy(sampler->key.sampler, bind_layout->immutable_samplers[0]->state_no_bc[0], sizeof(sampler->key.sampler)); assert(sizeof(sampler->key.color) == sizeof(bind_layout->immutable_samplers[0]->vk.border_color_value.uint32)); memcpy(sampler->key.color, bind_layout->immutable_samplers[0]->vk.border_color_value.uint32, sizeof(sampler->key.color)); } static bool binding_should_use_surface_binding_table(const struct apply_pipeline_layout_state *state, const struct anv_descriptor_set_binding_layout *bind_layout, uint32_t set, uint32_t binding) { if ((bind_layout->data & ANV_DESCRIPTOR_BTI_SURFACE_STATE) == 0) return false; if (state->pdevice->always_use_bindless && (bind_layout->data & ANV_DESCRIPTOR_SURFACE)) return false; if (state->set[set].binding[binding].properties & BINDING_PROPERTY_NO_BINDING_TABLE) return false; return true; } static bool binding_should_use_sampler_binding_table(const struct apply_pipeline_layout_state *state, const struct anv_descriptor_set_binding_layout *binding) { if ((binding->data & ANV_DESCRIPTOR_BTI_SAMPLER_STATE) == 0) return false; if (state->pdevice->always_use_bindless && (binding->data & ANV_DESCRIPTOR_SAMPLER)) return false; return true; } void anv_nir_apply_pipeline_layout(nir_shader *shader, const struct anv_physical_device *pdevice, enum brw_robustness_flags robust_flags, bool independent_sets, const struct anv_pipeline_sets_layout *layout, struct anv_pipeline_bind_map *map, struct anv_pipeline_push_map *push_map, void *push_map_mem_ctx) { void *mem_ctx = ralloc_context(NULL); #ifndef NDEBUG /* We should not have have any reference to a descriptor set that is not * given through the pipeline layout (layout->set[set].layout = NULL). */ anv_validate_pipeline_layout(layout, shader); #endif const bool bindless_stage = brw_shader_stage_requires_bindless_resources(shader->info.stage); struct apply_pipeline_layout_state state = { .pdevice = pdevice, .layout = layout, .desc_addr_format = bindless_stage ? nir_address_format_64bit_global_32bit_offset : nir_address_format_32bit_index_offset, .ssbo_addr_format = anv_nir_ssbo_addr_format(pdevice, robust_flags), .ubo_addr_format = anv_nir_ubo_addr_format(pdevice, robust_flags), .lowered_instrs = _mesa_pointer_set_create(mem_ctx), .has_independent_sets = independent_sets, }; /* Compute the amount of push block items required. */ unsigned push_block_count = 0; for (unsigned s = 0; s < layout->num_sets; s++) { if (!layout->set[s].layout) continue; const unsigned count = layout->set[s].layout->binding_count; state.set[s].binding = rzalloc_array_size(mem_ctx, sizeof(state.set[s].binding[0]), count); const struct anv_descriptor_set_layout *set_layout = layout->set[s].layout; for (unsigned b = 0; b < set_layout->binding_count; b++) { if (set_layout->binding[b].type != VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) push_block_count += set_layout->binding[b].array_size; } } /* Find all use sets/bindings */ nir_shader_instructions_pass(shader, get_used_bindings, nir_metadata_all, &state); /* Assign a BTI to each used descriptor set */ for (unsigned s = 0; s < layout->num_sets; s++) { if (state.desc_addr_format != nir_address_format_32bit_index_offset) { state.set[s].desc_offset = BINDLESS_OFFSET; } else if (state.set[s].desc_buffer_used) { map->surface_to_descriptor[map->surface_count] = (struct anv_pipeline_binding) { .set = (layout->type == ANV_PIPELINE_DESCRIPTOR_SET_LAYOUT_TYPE_BUFFER) ? ANV_DESCRIPTOR_SET_DESCRIPTORS_BUFFER : ANV_DESCRIPTOR_SET_DESCRIPTORS, .binding = UINT32_MAX, .index = s, }; state.set[s].desc_offset = map->surface_count++; } } /* Assign a block index for each surface */ push_map->block_to_descriptor = rzalloc_array(push_map_mem_ctx, struct anv_pipeline_binding, map->surface_count + push_block_count); memcpy(push_map->block_to_descriptor, map->surface_to_descriptor, sizeof(push_map->block_to_descriptor[0]) * map->surface_count); push_map->block_count = map->surface_count; /* Count used bindings, assign embedded sampler indices & add push blocks * for promotion to push constants */ unsigned used_binding_count = 0; for (uint32_t set = 0; set < layout->num_sets; set++) { struct anv_descriptor_set_layout *set_layout = layout->set[set].layout; if (!set_layout) continue; for (unsigned b = 0; b < set_layout->binding_count; b++) { if (state.set[set].binding[b].use_count == 0) continue; used_binding_count++; const struct anv_descriptor_set_binding_layout *bind_layout = &set_layout->binding[b]; if (state.set[set].binding[b].properties & BINDING_PROPERTY_EMBEDDED_SAMPLER) add_embedded_sampler_entry(&state, map, set, b); if (binding_is_promotable_to_push(set_layout, bind_layout)) { if (bind_layout->type != VK_DESCRIPTOR_TYPE_INLINE_UNIFORM_BLOCK) { state.set[set].binding[b].push_block = push_map->block_count; for (unsigned i = 0; i < bind_layout->array_size; i++) add_push_entry(push_map, set, b, i, layout, bind_layout); } else { state.set[set].binding[b].push_block = state.set[set].desc_offset; } } } } struct binding_info *infos = rzalloc_array(mem_ctx, struct binding_info, used_binding_count); used_binding_count = 0; for (uint32_t set = 0; set < layout->num_sets; set++) { const struct anv_descriptor_set_layout *set_layout = layout->set[set].layout; if (!set_layout) continue; for (unsigned b = 0; b < set_layout->binding_count; b++) { if (state.set[set].binding[b].use_count == 0) continue; const struct anv_descriptor_set_binding_layout *binding = &layout->set[set].layout->binding[b]; /* Do a fixed-point calculation to generate a score based on the * number of uses and the binding array size. We shift by 7 instead * of 8 because we're going to use the top bit below to make * everything which does not support bindless super higher priority * than things which do. */ uint16_t score = ((uint16_t)state.set[set].binding[b].use_count << 7) / binding->array_size; /* If the descriptor type doesn't support bindless then put it at the * beginning so we guarantee it gets a slot. */ if (!anv_descriptor_supports_bindless(pdevice, set_layout, binding)) score |= 1 << 15; infos[used_binding_count++] = (struct binding_info) { .set = set, .binding = b, .score = score, }; } } /* Order the binding infos based on score with highest scores first. If * scores are equal we then order by set and binding. */ qsort(infos, used_binding_count, sizeof(struct binding_info), compare_binding_infos); for (unsigned i = 0; i < used_binding_count; i++) { unsigned set = infos[i].set, b = infos[i].binding; assert(layout->set[set].layout); const struct anv_descriptor_set_layout *set_layout = layout->set[set].layout; const struct anv_descriptor_set_binding_layout *binding = &set_layout->binding[b]; const uint32_t array_size = binding->array_size; if (binding->dynamic_offset_index >= 0) state.has_dynamic_buffers = true; const unsigned array_multiplier = bti_multiplier(&state, set, b); assert(array_multiplier >= 1); /* Assume bindless by default */ state.set[set].binding[b].surface_offset = BINDLESS_OFFSET; state.set[set].binding[b].sampler_offset = BINDLESS_OFFSET; if (binding_should_use_surface_binding_table(&state, binding, set, b)) { if (map->surface_count + array_size * array_multiplier > MAX_BINDING_TABLE_SIZE || anv_descriptor_requires_bindless(pdevice, set_layout, binding) || brw_shader_stage_requires_bindless_resources(shader->info.stage)) { /* If this descriptor doesn't fit in the binding table or if it * requires bindless for some reason, flag it as bindless. */ assert(anv_descriptor_supports_bindless(pdevice, set_layout, binding)); } else { state.set[set].binding[b].surface_offset = map->surface_count; if (binding->dynamic_offset_index < 0) { struct anv_sampler **samplers = binding->immutable_samplers; uint8_t max_planes = bti_multiplier(&state, set, b); for (unsigned i = 0; i < binding->array_size; i++) { uint8_t planes = samplers ? samplers[i]->n_planes : 1; for (uint8_t p = 0; p < max_planes; p++) { if (p < planes) { add_bti_entry(map, set, b, i, p, binding); } else { add_null_bti_entry(map); } } } } else { for (unsigned i = 0; i < binding->array_size; i++) add_dynamic_bti_entry(map, set, b, i, layout, binding); } } assert(map->surface_count <= MAX_BINDING_TABLE_SIZE); } if (binding_should_use_sampler_binding_table(&state, binding)) { if (map->sampler_count + array_size * array_multiplier > MAX_SAMPLER_TABLE_SIZE || anv_descriptor_requires_bindless(pdevice, set_layout, binding) || brw_shader_stage_requires_bindless_resources(shader->info.stage)) { /* If this descriptor doesn't fit in the binding table or if it * requires bindless for some reason, flag it as bindless. * * We also make large sampler arrays bindless because we can avoid * using indirect sends thanks to bindless samplers being packed * less tightly than the sampler table. */ assert(anv_descriptor_supports_bindless(pdevice, set_layout, binding)); } else { state.set[set].binding[b].sampler_offset = map->sampler_count; uint8_t max_planes = bti_multiplier(&state, set, b); for (unsigned i = 0; i < binding->array_size; i++) { for (uint8_t p = 0; p < max_planes; p++) { add_sampler_entry(map, set, b, i, p, layout, binding); } } } } if (binding->data & ANV_DESCRIPTOR_INLINE_UNIFORM) { state.set[set].binding[b].surface_offset = state.set[set].desc_offset; } #if 0 fprintf(stderr, "set=%u binding=%u surface_offset=0x%08x require_bindless=%u type=%s\n", set, b, state.set[set].binding[b].surface_offset, anv_descriptor_requires_bindless(pdevice, set_layout, binding), vk_DescriptorType_to_str(binding->type)); #endif } /* Before we do the normal lowering, we look for any SSBO operations * that we can lower to the BTI model and lower them up-front. The BTI * model can perform better than the A64 model for a couple reasons: * * 1. 48-bit address calculations are potentially expensive and using * the BTI model lets us simply compute 32-bit offsets and the * hardware adds the 64-bit surface base address. * * 2. The BTI messages, because they use surface states, do bounds * checking for us. With the A64 model, we have to do our own * bounds checking and this means wider pointers and extra * calculations and branching in the shader. * * The solution to both of these is to convert things to the BTI model * opportunistically. The reason why we need to do this as a pre-pass * is for two reasons: * * 1. The BTI model requires nir_address_format_32bit_index_offset * pointers which are not the same type as the pointers needed for * the A64 model. Because all our derefs are set up for the A64 * model (in case we have variable pointers), we have to crawl all * the way back to the vulkan_resource_index intrinsic and build a * completely fresh index+offset calculation. * * 2. Because the variable-pointers-capable lowering that we do as part * of apply_pipeline_layout_block is destructive (It really has to * be to handle variable pointers properly), we've lost the deref * information by the time we get to the load/store/atomic * intrinsics in that pass. */ nir_shader_instructions_pass(shader, lower_direct_buffer_instr, nir_metadata_control_flow, &state); /* We just got rid of all the direct access. Delete it so it's not in the * way when we do our indirect lowering. */ nir_opt_dce(shader); nir_shader_instructions_pass(shader, apply_pipeline_layout, nir_metadata_none, &state); ralloc_free(mem_ctx); if (brw_shader_stage_is_bindless(shader->info.stage)) { assert(map->surface_count == 0); assert(map->sampler_count == 0); } #if 0 fprintf(stderr, "bti:\n"); for (unsigned i = 0; i < map->surface_count; i++) { fprintf(stderr, " %03i: set=%03u binding=%06i index=%u plane=%u set_offset=0x%08x dyn_offset=0x%08x\n", i, map->surface_to_descriptor[i].set, map->surface_to_descriptor[i].binding, map->surface_to_descriptor[i].index, map->surface_to_descriptor[i].plane, map->surface_to_descriptor[i].set_offset, map->surface_to_descriptor[i].dynamic_offset_index); } fprintf(stderr, "sti:\n"); for (unsigned i = 0; i < map->sampler_count; i++) { fprintf(stderr, " %03i: set=%03u binding=%06i index=%u plane=%u\n", i, map->sampler_to_descriptor[i].set, map->sampler_to_descriptor[i].binding, map->sampler_to_descriptor[i].index, map->sampler_to_descriptor[i].plane); } #endif /* Now that we're done computing the surface and sampler portions of the * bind map, hash them. This lets us quickly determine if the actual * mapping has changed and not just a no-op pipeline change. */ _mesa_sha1_compute(map->surface_to_descriptor, map->surface_count * sizeof(struct anv_pipeline_binding), map->surface_sha1); _mesa_sha1_compute(map->sampler_to_descriptor, map->sampler_count * sizeof(struct anv_pipeline_binding), map->sampler_sha1); }