From 211a9f2057321e6a3500d9b7873085621604e336 Mon Sep 17 00:00:00 2001 From: Rhys Perry Date: Wed, 15 Apr 2020 15:39:44 +0100 Subject: [PATCH] aco: use emit_load helper for VMEM/SMEM loads MIME-Version: 1.0 Content-Type: text/plain; charset=utf8 Content-Transfer-Encoding: 8bit Also implements 8/16-bit loads for scratch/global. Signed-off-by: Rhys Perry Reviewed-by: Daniel Schürmann Part-of: --- .../compiler/aco_instruction_selection.cpp | 720 ++++++------------ 1 file changed, 226 insertions(+), 494 deletions(-) diff --git a/src/amd/compiler/aco_instruction_selection.cpp b/src/amd/compiler/aco_instruction_selection.cpp index f2d40b484a7..ca376e5052b 100644 --- a/src/amd/compiler/aco_instruction_selection.cpp +++ b/src/amd/compiler/aco_instruction_selection.cpp @@ -3410,6 +3410,183 @@ Temp lds_load_callback(Builder& bld, const LoadEmitInfo *info, static auto emit_lds_load = emit_load; +Temp smem_load_callback(Builder& bld, const LoadEmitInfo *info, + Temp offset, unsigned bytes_needed, + unsigned align, unsigned const_offset, + Temp dst_hint) +{ + unsigned size = 0; + aco_opcode op; + if (bytes_needed <= 4) { + size = 1; + op = info->resource.id() ? aco_opcode::s_buffer_load_dword : aco_opcode::s_load_dword; + } else if (bytes_needed <= 8) { + size = 2; + op = info->resource.id() ? aco_opcode::s_buffer_load_dwordx2 : aco_opcode::s_load_dwordx2; + } else if (bytes_needed <= 16) { + size = 4; + op = info->resource.id() ? aco_opcode::s_buffer_load_dwordx4 : aco_opcode::s_load_dwordx4; + } else if (bytes_needed <= 32) { + size = 8; + op = info->resource.id() ? aco_opcode::s_buffer_load_dwordx8 : aco_opcode::s_load_dwordx8; + } else { + size = 16; + op = info->resource.id() ? aco_opcode::s_buffer_load_dwordx16 : aco_opcode::s_load_dwordx16; + } + aco_ptr load{create_instruction(op, Format::SMEM, 2, 1)}; + if (info->resource.id()) { + load->operands[0] = Operand(info->resource); + load->operands[1] = Operand(offset); + } else { + load->operands[0] = Operand(offset); + load->operands[1] = Operand(0u); + } + RegClass rc(RegType::sgpr, size); + Temp val = dst_hint.id() && dst_hint.regClass() == rc ? dst_hint : bld.tmp(rc); + load->definitions[0] = Definition(val); + load->glc = info->glc; + load->dlc = info->glc && bld.program->chip_class >= GFX10; + load->barrier = info->barrier; + load->can_reorder = false; // FIXME: currently, it doesn't seem beneficial due to how our scheduler works + bld.insert(std::move(load)); + return val; +} + +static auto emit_smem_load = emit_load; + +Temp mubuf_load_callback(Builder& bld, const LoadEmitInfo *info, + Temp offset, unsigned bytes_needed, + unsigned align_, unsigned const_offset, + Temp dst_hint) +{ + Operand vaddr = offset.type() == RegType::vgpr ? Operand(offset) : Operand(v1); + Operand soffset = offset.type() == RegType::sgpr ? Operand(offset) : Operand((uint32_t) 0); + + if (info->soffset.id()) { + if (soffset.isTemp()) + vaddr = bld.copy(bld.def(v1), soffset); + soffset = Operand(info->soffset); + } + + unsigned bytes_size = 0; + aco_opcode op; + if (bytes_needed == 1) { + bytes_size = 1; + op = aco_opcode::buffer_load_ubyte; + } else if (bytes_needed == 2) { + bytes_size = 2; + op = aco_opcode::buffer_load_ushort; + } else if (bytes_needed <= 4) { + bytes_size = 4; + op = aco_opcode::buffer_load_dword; + } else if (bytes_needed <= 8) { + bytes_size = 8; + op = aco_opcode::buffer_load_dwordx2; + } else if (bytes_needed <= 12 && bld.program->chip_class > GFX6) { + bytes_size = 12; + op = aco_opcode::buffer_load_dwordx3; + } else { + bytes_size = 16; + op = aco_opcode::buffer_load_dwordx4; + } + aco_ptr mubuf{create_instruction(op, Format::MUBUF, 3, 1)}; + mubuf->operands[0] = Operand(info->resource); + mubuf->operands[1] = vaddr; + mubuf->operands[2] = soffset; + mubuf->offen = (offset.type() == RegType::vgpr); + mubuf->glc = info->glc; + mubuf->dlc = info->glc && bld.program->chip_class >= GFX10; + mubuf->barrier = info->barrier; + mubuf->can_reorder = info->can_reorder; + mubuf->offset = const_offset; + RegClass rc = RegClass::get(RegType::vgpr, align(bytes_size, 4)); + Temp val = dst_hint.id() && rc == dst_hint.regClass() ? dst_hint : bld.tmp(rc); + mubuf->definitions[0] = Definition(val); + bld.insert(std::move(mubuf)); + + if (bytes_size < 4) + val = bld.pseudo(aco_opcode::p_extract_vector, bld.def(RegClass::get(RegType::vgpr, bytes_size)), val, Operand(0u)); + + return val; +} + +static auto emit_mubuf_load = emit_load; + +Temp get_gfx6_global_rsrc(Builder& bld, Temp addr) +{ + uint32_t rsrc_conf = S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) | + S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32); + + if (addr.type() == RegType::vgpr) + return bld.pseudo(aco_opcode::p_create_vector, bld.def(s4), Operand(0u), Operand(0u), Operand(-1u), Operand(rsrc_conf)); + return bld.pseudo(aco_opcode::p_create_vector, bld.def(s4), addr, Operand(-1u), Operand(rsrc_conf)); +} + +Temp global_load_callback(Builder& bld, const LoadEmitInfo *info, + Temp offset, unsigned bytes_needed, + unsigned align_, unsigned const_offset, + Temp dst_hint) +{ + unsigned bytes_size = 0; + bool mubuf = bld.program->chip_class == GFX6; + bool global = bld.program->chip_class >= GFX9; + aco_opcode op; + if (bytes_needed == 1) { + bytes_size = 1; + op = mubuf ? aco_opcode::buffer_load_ubyte : global ? aco_opcode::global_load_ubyte : aco_opcode::flat_load_ubyte; + } else if (bytes_needed == 2) { + bytes_size = 2; + op = mubuf ? aco_opcode::buffer_load_ushort : global ? aco_opcode::global_load_ushort : aco_opcode::flat_load_ushort; + } else if (bytes_needed <= 4) { + bytes_size = 4; + op = mubuf ? aco_opcode::buffer_load_dword : global ? aco_opcode::global_load_dword : aco_opcode::flat_load_dword; + } else if (bytes_needed <= 8) { + bytes_size = 8; + op = mubuf ? aco_opcode::buffer_load_dwordx2 : global ? aco_opcode::global_load_dwordx2 : aco_opcode::flat_load_dwordx2; + } else if (bytes_needed <= 12 && !mubuf) { + bytes_size = 12; + op = global ? aco_opcode::global_load_dwordx3 : aco_opcode::flat_load_dwordx3; + } else { + bytes_size = 16; + op = mubuf ? aco_opcode::buffer_load_dwordx4 : global ? aco_opcode::global_load_dwordx4 : aco_opcode::flat_load_dwordx4; + } + RegClass rc = RegClass::get(RegType::vgpr, align(bytes_size, 4)); + Temp val = dst_hint.id() && rc == dst_hint.regClass() ? dst_hint : bld.tmp(rc); + if (mubuf) { + aco_ptr mubuf{create_instruction(op, Format::MUBUF, 3, 1)}; + mubuf->operands[0] = Operand(get_gfx6_global_rsrc(bld, offset)); + mubuf->operands[1] = offset.type() == RegType::vgpr ? Operand(offset) : Operand(v1); + mubuf->operands[2] = Operand(0u); + mubuf->glc = info->glc; + mubuf->dlc = false; + mubuf->offset = 0; + mubuf->addr64 = offset.type() == RegType::vgpr; + mubuf->disable_wqm = false; + mubuf->barrier = info->barrier; + mubuf->definitions[0] = Definition(val); + bld.insert(std::move(mubuf)); + } else { + offset = offset.regClass() == s2 ? bld.copy(bld.def(v2), offset) : offset; + + aco_ptr flat{create_instruction(op, global ? Format::GLOBAL : Format::FLAT, 2, 1)}; + flat->operands[0] = Operand(offset); + flat->operands[1] = Operand(s1); + flat->glc = info->glc; + flat->dlc = info->glc && bld.program->chip_class >= GFX10; + flat->barrier = info->barrier; + flat->offset = 0u; + flat->definitions[0] = Definition(val); + bld.insert(std::move(flat)); + } + + if (bytes_size < 4) + val = bld.pseudo(aco_opcode::p_extract_vector, bld.def(RegClass::get(RegType::vgpr, bytes_size)), val, Operand(0u)); + + return val; +} + +static auto emit_global_load = emit_load; + Temp load_lds(isel_context *ctx, unsigned elem_size_bytes, Temp dst, Temp address, unsigned base_offset, unsigned align) { @@ -3691,29 +3868,6 @@ void store_vmem_mubuf(isel_context *ctx, Temp src, Temp descriptor, Temp voffset } } -Temp emit_single_mubuf_load(isel_context *ctx, Temp descriptor, Temp voffset, Temp soffset, - unsigned const_offset, unsigned size_dwords, bool allow_reorder = true) -{ - assert(size_dwords != 3 || ctx->program->chip_class != GFX6); - assert(size_dwords >= 1 && size_dwords <= 4); - - Builder bld(ctx->program, ctx->block); - Temp vdata = bld.tmp(RegClass(RegType::vgpr, size_dwords)); - aco_opcode op = (aco_opcode) ((unsigned) aco_opcode::buffer_load_dword + size_dwords - 1); - const_offset = resolve_excess_vmem_const_offset(bld, voffset, const_offset); - - Operand voffset_op = voffset.id() ? Operand(as_vgpr(ctx, voffset)) : Operand(v1); - Operand soffset_op = soffset.id() ? Operand(soffset) : Operand(0u); - Builder::Result r = bld.mubuf(op, Definition(vdata), Operand(descriptor), voffset_op, soffset_op, const_offset, - /* offen */ !voffset_op.isUndefined(), /* idxen*/ false, /* addr64 */ false, - /* disable_wqm */ false, /* glc */ true, - /* dlc*/ ctx->program->chip_class >= GFX10, /* slc */ false); - - static_cast(r.instr)->can_reorder = allow_reorder; - - return vdata; -} - void load_vmem_mubuf(isel_context *ctx, Temp dst, Temp descriptor, Temp voffset, Temp soffset, unsigned base_const_offset, unsigned elem_size_bytes, unsigned num_components, unsigned stride = 0u, bool allow_combining = true, bool allow_reorder = true) @@ -3723,35 +3877,16 @@ void load_vmem_mubuf(isel_context *ctx, Temp dst, Temp descriptor, Temp voffset, assert(!!stride != allow_combining); Builder bld(ctx->program, ctx->block); - unsigned split_cnt = num_components; - - if (elem_size_bytes == 8) { - elem_size_bytes = 4; - num_components *= 2; - } - if (!stride) - stride = elem_size_bytes; - - unsigned load_size = 1; - if (allow_combining) { - if ((num_components % 4) == 0) - load_size = 4; - else if ((num_components % 3) == 0 && ctx->program->chip_class != GFX6) - load_size = 3; - else if ((num_components % 2) == 0) - load_size = 2; - } - - unsigned num_loads = num_components / load_size; - std::array elems; - - for (unsigned i = 0; i < num_loads; ++i) { - unsigned const_offset = i * stride * load_size + base_const_offset; - elems[i] = emit_single_mubuf_load(ctx, descriptor, voffset, soffset, const_offset, load_size, allow_reorder); - } - - create_vec_from_array(ctx, elems.data(), num_loads, RegType::vgpr, load_size * 4u, split_cnt, dst); + LoadEmitInfo info = {Operand(voffset), dst, num_components, elem_size_bytes, descriptor}; + info.component_stride = allow_combining ? 0 : stride; + info.glc = true; + info.swizzle_component_size = allow_combining ? 0 : 4; + info.align_mul = MIN2(elem_size_bytes, 4); + info.align_offset = 0; + info.soffset = soffset; + info.const_offset = base_const_offset; + emit_mubuf_load(ctx, bld, &info); } std::pair offset_add_from_nir(isel_context *ctx, const std::pair &base_offset, nir_src *off_src, unsigned stride = 1u) @@ -4855,236 +4990,25 @@ void visit_load_resource(isel_context *ctx, nir_intrinsic_instr *instr) } void load_buffer(isel_context *ctx, unsigned num_components, unsigned component_size, - Temp dst, Temp rsrc, Temp offset, int byte_align, + Temp dst, Temp rsrc, Temp offset, unsigned align_mul, unsigned align_offset, bool glc=false, bool readonly=true) { Builder bld(ctx->program, ctx->block); - bool dlc = glc && ctx->options->chip_class >= GFX10; - unsigned num_bytes = num_components * component_size; - - aco_opcode op; - if (dst.type() == RegType::vgpr || ((ctx->options->chip_class < GFX8 || component_size < 4) && !readonly)) { - Operand vaddr = offset.type() == RegType::vgpr ? Operand(offset) : Operand(v1); - Operand soffset = offset.type() == RegType::sgpr ? Operand(offset) : Operand((uint32_t) 0); - unsigned const_offset = 0; - - /* for small bit sizes add buffer for unaligned loads */ - if (byte_align) { - if (num_bytes > 2) - num_bytes += byte_align == -1 ? 4 - component_size : byte_align; - else - byte_align = 0; - } - - Temp lower = Temp(); - if (num_bytes > 16) { - assert(num_components == 3 || num_components == 4); - op = aco_opcode::buffer_load_dwordx4; - lower = bld.tmp(v4); - aco_ptr mubuf{create_instruction(op, Format::MUBUF, 3, 1)}; - mubuf->definitions[0] = Definition(lower); - mubuf->operands[0] = Operand(rsrc); - mubuf->operands[1] = vaddr; - mubuf->operands[2] = soffset; - mubuf->offen = (offset.type() == RegType::vgpr); - mubuf->glc = glc; - mubuf->dlc = dlc; - mubuf->barrier = readonly ? barrier_none : barrier_buffer; - mubuf->can_reorder = readonly; - bld.insert(std::move(mubuf)); - emit_split_vector(ctx, lower, 2); - num_bytes -= 16; - const_offset = 16; - } else if (num_bytes == 12 && ctx->options->chip_class == GFX6) { - /* GFX6 doesn't support loading vec3, expand to vec4. */ - num_bytes = 16; - } - - switch (num_bytes) { - case 1: - op = aco_opcode::buffer_load_ubyte; - break; - case 2: - op = aco_opcode::buffer_load_ushort; - break; - case 3: - case 4: - op = aco_opcode::buffer_load_dword; - break; - case 5: - case 6: - case 7: - case 8: - op = aco_opcode::buffer_load_dwordx2; - break; - case 10: - case 12: - assert(ctx->options->chip_class > GFX6); - op = aco_opcode::buffer_load_dwordx3; - break; - case 16: - op = aco_opcode::buffer_load_dwordx4; - break; - default: - unreachable("Load SSBO not implemented for this size."); - } - aco_ptr mubuf{create_instruction(op, Format::MUBUF, 3, 1)}; - mubuf->operands[0] = Operand(rsrc); - mubuf->operands[1] = vaddr; - mubuf->operands[2] = soffset; - mubuf->offen = (offset.type() == RegType::vgpr); - mubuf->glc = glc; - mubuf->dlc = dlc; - mubuf->barrier = readonly ? barrier_none : barrier_buffer; - mubuf->can_reorder = readonly; - mubuf->offset = const_offset; - aco_ptr instr = std::move(mubuf); - - if (component_size < 4) { - Temp vec = num_bytes <= 4 ? bld.tmp(v1) : num_bytes <= 8 ? bld.tmp(v2) : bld.tmp(v3); - instr->definitions[0] = Definition(vec); - bld.insert(std::move(instr)); - - if (byte_align == -1 || (byte_align && dst.type() == RegType::sgpr)) { - Operand align = byte_align == -1 ? Operand(offset) : Operand((uint32_t)byte_align); - Temp tmp[3] = {vec, vec, vec}; - - if (vec.size() == 3) { - tmp[0] = bld.tmp(v1), tmp[1] = bld.tmp(v1), tmp[2] = bld.tmp(v1); - bld.pseudo(aco_opcode::p_split_vector, Definition(tmp[0]), Definition(tmp[1]), Definition(tmp[2]), vec); - } else if (vec.size() == 2) { - tmp[0] = bld.tmp(v1), tmp[1] = bld.tmp(v1), tmp[2] = tmp[1]; - bld.pseudo(aco_opcode::p_split_vector, Definition(tmp[0]), Definition(tmp[1]), vec); - } - for (unsigned i = 0; i < dst.size(); i++) - tmp[i] = bld.vop3(aco_opcode::v_alignbyte_b32, bld.def(v1), tmp[i + 1], tmp[i], align); - - vec = tmp[0]; - if (dst.size() == 2) - vec = bld.pseudo(aco_opcode::p_create_vector, bld.def(v2), tmp[0], tmp[1]); - byte_align = 0; - } - - if (dst.type() == RegType::vgpr && num_components == 1) { - bld.pseudo(aco_opcode::p_extract_vector, Definition(dst), vec, Operand(byte_align / component_size)); - } else { - trim_subdword_vector(ctx, vec, dst, 4 * vec.size() / component_size, ((1 << num_components) - 1) << byte_align / component_size); - } - - return; - - } else if (dst.size() > 4) { - assert(lower != Temp()); - Temp upper = bld.tmp(RegType::vgpr, dst.size() - lower.size()); - instr->definitions[0] = Definition(upper); - bld.insert(std::move(instr)); - if (dst.size() == 8) - emit_split_vector(ctx, upper, 2); - instr.reset(create_instruction(aco_opcode::p_create_vector, Format::PSEUDO, dst.size() / 2, 1)); - instr->operands[0] = Operand(emit_extract_vector(ctx, lower, 0, v2)); - instr->operands[1] = Operand(emit_extract_vector(ctx, lower, 1, v2)); - instr->operands[2] = Operand(emit_extract_vector(ctx, upper, 0, v2)); - if (dst.size() == 8) - instr->operands[3] = Operand(emit_extract_vector(ctx, upper, 1, v2)); - } else if (dst.size() == 3 && ctx->options->chip_class == GFX6) { - Temp vec = bld.tmp(v4); - instr->definitions[0] = Definition(vec); - bld.insert(std::move(instr)); - emit_split_vector(ctx, vec, 4); - - instr.reset(create_instruction(aco_opcode::p_create_vector, Format::PSEUDO, 3, 1)); - instr->operands[0] = Operand(emit_extract_vector(ctx, vec, 0, v1)); - instr->operands[1] = Operand(emit_extract_vector(ctx, vec, 1, v1)); - instr->operands[2] = Operand(emit_extract_vector(ctx, vec, 2, v1)); - } - - if (dst.type() == RegType::sgpr) { - Temp vec = bld.tmp(RegType::vgpr, dst.size()); - instr->definitions[0] = Definition(vec); - bld.insert(std::move(instr)); - expand_vector(ctx, vec, dst, num_components, (1 << num_components) - 1); - } else { - instr->definitions[0] = Definition(dst); - bld.insert(std::move(instr)); - emit_split_vector(ctx, dst, num_components); - } - } else { - /* for small bit sizes add buffer for unaligned loads */ - if (byte_align) - num_bytes += byte_align == -1 ? 4 - component_size : byte_align; - - switch (num_bytes) { - case 1: - case 2: - case 3: - case 4: - op = aco_opcode::s_buffer_load_dword; - break; - case 5: - case 6: - case 7: - case 8: - op = aco_opcode::s_buffer_load_dwordx2; - break; - case 10: - case 12: - case 16: - op = aco_opcode::s_buffer_load_dwordx4; - break; - case 24: - case 32: - op = aco_opcode::s_buffer_load_dwordx8; - break; - default: - unreachable("Load SSBO not implemented for this size."); - } + bool use_smem = dst.type() != RegType::vgpr && ((ctx->options->chip_class >= GFX8 && component_size >= 4) || readonly); + if (use_smem) offset = bld.as_uniform(offset); - aco_ptr load{create_instruction(op, Format::SMEM, 2, 1)}; - load->operands[0] = Operand(rsrc); - load->operands[1] = Operand(offset); - assert(load->operands[1].getTemp().type() == RegType::sgpr); - load->definitions[0] = Definition(dst); - load->glc = glc; - load->dlc = dlc; - load->barrier = readonly ? barrier_none : barrier_buffer; - load->can_reorder = false; // FIXME: currently, it doesn't seem beneficial due to how our scheduler works - assert(ctx->options->chip_class >= GFX8 || !glc); - - /* adjust misaligned small bit size loads */ - if (byte_align) { - Temp vec = num_bytes <= 4 ? bld.tmp(s1) : num_bytes <= 8 ? bld.tmp(s2) : bld.tmp(s4); - load->definitions[0] = Definition(vec); - bld.insert(std::move(load)); - Operand byte_offset = byte_align > 0 ? Operand(uint32_t(byte_align)) : Operand(offset); - byte_align_scalar(ctx, vec, byte_offset, dst); - - /* trim vector */ - } else if (dst.size() == 3) { - Temp vec = bld.tmp(s4); - load->definitions[0] = Definition(vec); - bld.insert(std::move(load)); - emit_split_vector(ctx, vec, 4); - bld.pseudo(aco_opcode::p_create_vector, Definition(dst), - emit_extract_vector(ctx, vec, 0, s1), - emit_extract_vector(ctx, vec, 1, s1), - emit_extract_vector(ctx, vec, 2, s1)); - } else if (dst.size() == 6) { - Temp vec = bld.tmp(s8); - load->definitions[0] = Definition(vec); - bld.insert(std::move(load)); - emit_split_vector(ctx, vec, 4); - - bld.pseudo(aco_opcode::p_create_vector, Definition(dst), - emit_extract_vector(ctx, vec, 0, s2), - emit_extract_vector(ctx, vec, 1, s2), - emit_extract_vector(ctx, vec, 2, s2)); - } else { - bld.insert(std::move(load)); - } - emit_split_vector(ctx, dst, num_components); - } + LoadEmitInfo info = {Operand(offset), dst, num_components, component_size, rsrc}; + info.glc = glc; + info.barrier = readonly ? barrier_none : barrier_buffer; + info.can_reorder = readonly; + info.align_mul = align_mul; + info.align_offset = align_offset; + if (use_smem) + emit_smem_load(ctx, bld, &info); + else + emit_mubuf_load(ctx, bld, &info); } void visit_load_ubo(isel_context *ctx, nir_intrinsic_instr *instr) @@ -5123,13 +5047,8 @@ void visit_load_ubo(isel_context *ctx, nir_intrinsic_instr *instr) rsrc = bld.smem(aco_opcode::s_load_dwordx4, bld.def(s4), rsrc, Operand(0u)); } unsigned size = instr->dest.ssa.bit_size / 8; - int byte_align = 0; - if (size < 4) { - unsigned align_mul = nir_intrinsic_align_mul(instr); - unsigned align_offset = nir_intrinsic_align_offset(instr); - byte_align = align_mul % 4 == 0 ? align_offset : -1; - } - load_buffer(ctx, instr->num_components, size, dst, rsrc, get_ssa_temp(ctx, instr->src[1].ssa), byte_align); + load_buffer(ctx, instr->num_components, size, dst, rsrc, get_ssa_temp(ctx, instr->src[1].ssa), + nir_intrinsic_align_mul(instr), nir_intrinsic_align_offset(instr)); } void visit_load_push_constant(isel_context *ctx, nir_intrinsic_instr *instr) @@ -5255,8 +5174,7 @@ void visit_load_constant(isel_context *ctx, nir_intrinsic_instr *instr) Operand(desc_type)); unsigned size = instr->dest.ssa.bit_size / 8; // TODO: get alignment information for subdword constants - unsigned byte_align = size < 4 ? -1 : 0; - load_buffer(ctx, instr->num_components, size, dst, rsrc, offset, byte_align); + load_buffer(ctx, instr->num_components, size, dst, rsrc, offset, size, 0); } void visit_discard_if(isel_context *ctx, nir_intrinsic_instr *instr) @@ -6070,13 +5988,8 @@ void visit_load_ssbo(isel_context *ctx, nir_intrinsic_instr *instr) bool glc = nir_intrinsic_access(instr) & (ACCESS_VOLATILE | ACCESS_COHERENT); unsigned size = instr->dest.ssa.bit_size / 8; - int byte_align = 0; - if (size < 4) { - unsigned align_mul = nir_intrinsic_align_mul(instr); - unsigned align_offset = nir_intrinsic_align_offset(instr); - byte_align = align_mul % 4 == 0 ? align_offset : -1; - } - load_buffer(ctx, num_components, size, dst, rsrc, get_ssa_temp(ctx, instr->src[1].ssa), byte_align, glc, false); + load_buffer(ctx, num_components, size, dst, rsrc, get_ssa_temp(ctx, instr->src[1].ssa), + nir_intrinsic_align_mul(instr), nir_intrinsic_align_offset(instr), glc, false); } void visit_store_ssbo(isel_context *ctx, nir_intrinsic_instr *instr) @@ -6323,166 +6236,28 @@ void visit_get_buffer_size(isel_context *ctx, nir_intrinsic_instr *instr) { get_buffer_size(ctx, desc, get_ssa_temp(ctx, &instr->dest.ssa), false); } -Temp get_gfx6_global_rsrc(Builder& bld, Temp addr) -{ - uint32_t rsrc_conf = S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) | - S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32); - - if (addr.type() == RegType::vgpr) - return bld.pseudo(aco_opcode::p_create_vector, bld.def(s4), Operand(0u), Operand(0u), Operand(-1u), Operand(rsrc_conf)); - return bld.pseudo(aco_opcode::p_create_vector, bld.def(s4), addr, Operand(-1u), Operand(rsrc_conf)); -} - void visit_load_global(isel_context *ctx, nir_intrinsic_instr *instr) { Builder bld(ctx->program, ctx->block); unsigned num_components = instr->num_components; - unsigned num_bytes = num_components * instr->dest.ssa.bit_size / 8; - - Temp dst = get_ssa_temp(ctx, &instr->dest.ssa); - Temp addr = get_ssa_temp(ctx, instr->src[0].ssa); - - bool glc = nir_intrinsic_access(instr) & (ACCESS_VOLATILE | ACCESS_COHERENT); - bool dlc = glc && ctx->options->chip_class >= GFX10; + unsigned component_size = instr->dest.ssa.bit_size / 8; + + LoadEmitInfo info = {Operand(get_ssa_temp(ctx, instr->src[0].ssa)), + get_ssa_temp(ctx, &instr->dest.ssa), + num_components, component_size}; + info.glc = nir_intrinsic_access(instr) & (ACCESS_VOLATILE | ACCESS_COHERENT); + info.align_mul = nir_intrinsic_align_mul(instr); + info.align_offset = nir_intrinsic_align_offset(instr); + info.barrier = barrier_buffer; + info.can_reorder = false; /* VMEM stores don't update the SMEM cache and it's difficult to prove that * it's safe to use SMEM */ bool can_use_smem = nir_intrinsic_access(instr) & ACCESS_NON_WRITEABLE; - aco_opcode op; - if (dst.type() == RegType::vgpr || (glc && ctx->options->chip_class < GFX8) || !can_use_smem) { - bool global = ctx->options->chip_class >= GFX9; - - if (ctx->options->chip_class >= GFX7) { - switch (num_bytes) { - case 4: - op = global ? aco_opcode::global_load_dword : aco_opcode::flat_load_dword; - break; - case 8: - op = global ? aco_opcode::global_load_dwordx2 : aco_opcode::flat_load_dwordx2; - break; - case 12: - op = global ? aco_opcode::global_load_dwordx3 : aco_opcode::flat_load_dwordx3; - break; - case 16: - op = global ? aco_opcode::global_load_dwordx4 : aco_opcode::flat_load_dwordx4; - break; - default: - unreachable("load_global not implemented for this size."); - } - - aco_ptr flat{create_instruction(op, global ? Format::GLOBAL : Format::FLAT, 2, 1)}; - flat->operands[0] = Operand(addr); - flat->operands[1] = Operand(s1); - flat->glc = glc; - flat->dlc = dlc; - flat->barrier = barrier_buffer; - - if (dst.type() == RegType::sgpr) { - Temp vec = bld.tmp(RegType::vgpr, dst.size()); - flat->definitions[0] = Definition(vec); - ctx->block->instructions.emplace_back(std::move(flat)); - bld.pseudo(aco_opcode::p_as_uniform, Definition(dst), vec); - } else { - flat->definitions[0] = Definition(dst); - ctx->block->instructions.emplace_back(std::move(flat)); - } - emit_split_vector(ctx, dst, num_components); - } else { - assert(ctx->options->chip_class == GFX6); - - /* GFX6 doesn't support loading vec3, expand to vec4. */ - num_bytes = num_bytes == 12 ? 16 : num_bytes; - - switch (num_bytes) { - case 4: - op = aco_opcode::buffer_load_dword; - break; - case 8: - op = aco_opcode::buffer_load_dwordx2; - break; - case 16: - op = aco_opcode::buffer_load_dwordx4; - break; - default: - unreachable("load_global not implemented for this size."); - } - - Temp rsrc = get_gfx6_global_rsrc(bld, addr); - - aco_ptr mubuf{create_instruction(op, Format::MUBUF, 3, 1)}; - mubuf->operands[0] = Operand(rsrc); - mubuf->operands[1] = addr.type() == RegType::vgpr ? Operand(addr) : Operand(v1); - mubuf->operands[2] = Operand(0u); - mubuf->glc = glc; - mubuf->dlc = false; - mubuf->offset = 0; - mubuf->addr64 = addr.type() == RegType::vgpr; - mubuf->disable_wqm = false; - mubuf->barrier = barrier_buffer; - aco_ptr instr = std::move(mubuf); - - /* expand vector */ - if (dst.size() == 3) { - Temp vec = bld.tmp(v4); - instr->definitions[0] = Definition(vec); - bld.insert(std::move(instr)); - emit_split_vector(ctx, vec, 4); - - instr.reset(create_instruction(aco_opcode::p_create_vector, Format::PSEUDO, 3, 1)); - instr->operands[0] = Operand(emit_extract_vector(ctx, vec, 0, v1)); - instr->operands[1] = Operand(emit_extract_vector(ctx, vec, 1, v1)); - instr->operands[2] = Operand(emit_extract_vector(ctx, vec, 2, v1)); - } - - if (dst.type() == RegType::sgpr) { - Temp vec = bld.tmp(RegType::vgpr, dst.size()); - instr->definitions[0] = Definition(vec); - bld.insert(std::move(instr)); - expand_vector(ctx, vec, dst, num_components, (1 << num_components) - 1); - bld.pseudo(aco_opcode::p_as_uniform, Definition(dst), vec); - } else { - instr->definitions[0] = Definition(dst); - bld.insert(std::move(instr)); - emit_split_vector(ctx, dst, num_components); - } - } + if (info.dst.type() == RegType::vgpr || (info.glc && ctx->options->chip_class < GFX8) || !can_use_smem) { + emit_global_load(ctx, bld, &info); } else { - switch (num_bytes) { - case 4: - op = aco_opcode::s_load_dword; - break; - case 8: - op = aco_opcode::s_load_dwordx2; - break; - case 12: - case 16: - op = aco_opcode::s_load_dwordx4; - break; - default: - unreachable("load_global not implemented for this size."); - } - aco_ptr load{create_instruction(op, Format::SMEM, 2, 1)}; - load->operands[0] = Operand(addr); - load->operands[1] = Operand(0u); - load->definitions[0] = Definition(dst); - load->glc = glc; - load->dlc = dlc; - load->barrier = barrier_buffer; - assert(ctx->options->chip_class >= GFX8 || !glc); - - if (dst.size() == 3) { - /* trim vector */ - Temp vec = bld.tmp(s4); - load->definitions[0] = Definition(vec); - ctx->block->instructions.emplace_back(std::move(load)); - emit_split_vector(ctx, vec, 4); - - bld.pseudo(aco_opcode::p_create_vector, Definition(dst), - emit_extract_vector(ctx, vec, 0, s1), - emit_extract_vector(ctx, vec, 1, s1), - emit_extract_vector(ctx, vec, 2, s1)); - } else { - ctx->block->instructions.emplace_back(std::move(load)); - } + info.offset = Operand(bld.as_uniform(info.offset)); + emit_smem_load(ctx, bld, &info); } } @@ -6954,62 +6729,19 @@ Temp get_scratch_resource(isel_context *ctx) } void visit_load_scratch(isel_context *ctx, nir_intrinsic_instr *instr) { - assert(instr->dest.ssa.bit_size == 32 || instr->dest.ssa.bit_size == 64); Builder bld(ctx->program, ctx->block); Temp rsrc = get_scratch_resource(ctx); Temp offset = as_vgpr(ctx, get_ssa_temp(ctx, instr->src[0].ssa)); Temp dst = get_ssa_temp(ctx, &instr->dest.ssa); - aco_opcode op; - switch (dst.size()) { - case 1: - op = aco_opcode::buffer_load_dword; - break; - case 2: - op = aco_opcode::buffer_load_dwordx2; - break; - case 3: - op = aco_opcode::buffer_load_dwordx3; - break; - case 4: - op = aco_opcode::buffer_load_dwordx4; - break; - case 6: - case 8: { - std::array elems; - Temp lower = bld.mubuf(aco_opcode::buffer_load_dwordx4, - bld.def(v4), rsrc, offset, - ctx->program->scratch_offset, 0, true); - Temp upper = bld.mubuf(dst.size() == 6 ? aco_opcode::buffer_load_dwordx2 : - aco_opcode::buffer_load_dwordx4, - dst.size() == 6 ? bld.def(v2) : bld.def(v4), - rsrc, offset, ctx->program->scratch_offset, 16, true); - emit_split_vector(ctx, lower, 2); - elems[0] = emit_extract_vector(ctx, lower, 0, v2); - elems[1] = emit_extract_vector(ctx, lower, 1, v2); - if (dst.size() == 8) { - emit_split_vector(ctx, upper, 2); - elems[2] = emit_extract_vector(ctx, upper, 0, v2); - elems[3] = emit_extract_vector(ctx, upper, 1, v2); - } else { - elems[2] = upper; - } - - aco_ptr vec{create_instruction(aco_opcode::p_create_vector, - Format::PSEUDO, dst.size() / 2, 1)}; - for (unsigned i = 0; i < dst.size() / 2; i++) - vec->operands[i] = Operand(elems[i]); - vec->definitions[0] = Definition(dst); - bld.insert(std::move(vec)); - ctx->allocated_vec.emplace(dst.id(), elems); - return; - } - default: - unreachable("Wrong dst size for nir_intrinsic_load_scratch"); - } - - bld.mubuf(op, Definition(dst), rsrc, offset, ctx->program->scratch_offset, 0, true); - emit_split_vector(ctx, dst, instr->num_components); + LoadEmitInfo info = {Operand(offset), dst, instr->dest.ssa.num_components, + instr->dest.ssa.bit_size / 8u, rsrc}; + info.align_mul = nir_intrinsic_align_mul(instr); + info.align_offset = nir_intrinsic_align_offset(instr); + info.swizzle_component_size = 16; + info.can_reorder = false; + info.soffset = ctx->program->scratch_offset; + emit_mubuf_load(ctx, bld, &info); } void visit_store_scratch(isel_context *ctx, nir_intrinsic_instr *instr) { -- 2.30.2