SPIRV-Tools/source/opt/dead_branch_elim_pass.cpp

// Copyright (c) 2017 The Khronos Group Inc.
// Copyright (c) 2017 Valve Corporation
// Copyright (c) 2017 LunarG Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//     http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.

#include "dead_branch_elim_pass.h"

#include "cfa.h"
#include "iterator.h"

namespace spvtools {
namespace opt {

namespace {

const uint32_t kBranchCondConditionalIdInIdx = 0;
const uint32_t kBranchCondTrueLabIdInIdx = 1;
const uint32_t kBranchCondFalseLabIdInIdx = 2;
const uint32_t kSelectionMergeMergeBlockIdInIdx = 0;
const uint32_t kPhiVal0IdInIdx = 0;
const uint32_t kPhiLab0IdInIdx = 1;
const uint32_t kPhiVal1IdInIdx = 2;
const uint32_t kLoopMergeMergeBlockIdInIdx = 0;
const uint32_t kLoopMergeContinueBlockIdInIdx = 1;

} // anonymous namespace

uint32_t DeadBranchElimPass::MergeBlockIdIfAny(
    const ir::BasicBlock& blk, uint32_t* cbid) const {
  auto merge_ii = blk.cend();
  --merge_ii;
  uint32_t mbid = 0;
  *cbid = 0;
  if (merge_ii != blk.cbegin()) {
    --merge_ii;
    if (merge_ii->opcode() == SpvOpLoopMerge) {
      mbid = merge_ii->GetSingleWordInOperand(kLoopMergeMergeBlockIdInIdx);
      *cbid = merge_ii->GetSingleWordInOperand(kLoopMergeContinueBlockIdInIdx);
    }
    else if (merge_ii->opcode() == SpvOpSelectionMerge) {
      mbid = merge_ii->GetSingleWordInOperand(
          kSelectionMergeMergeBlockIdInIdx);
    }
  }
  return mbid;
}

void DeadBranchElimPass::ComputeStructuredSuccessors(ir::Function* func) {
  // If header, make merge block first successor. If a loop header, make
  // the second successor the continue target.
  for (auto& blk : *func) {
    uint32_t cbid;
    uint32_t mbid = MergeBlockIdIfAny(blk, &cbid);
    if (mbid != 0) {
      block2structured_succs_[&blk].push_back(id2block_[mbid]);
      if (cbid != 0)
        block2structured_succs_[&blk].push_back(id2block_[cbid]);
    }
    // add true successors
    blk.ForEachSuccessorLabel([&blk, this](uint32_t sbid) {
      block2structured_succs_[&blk].push_back(id2block_[sbid]);
    });
  }
}

void DeadBranchElimPass::ComputeStructuredOrder(
    ir::Function* func, std::list<ir::BasicBlock*>* order) {
  // Compute structured successors and do DFS
  ComputeStructuredSuccessors(func);
  auto ignore_block = [](cbb_ptr) {};
  auto ignore_edge = [](cbb_ptr, cbb_ptr) {};
  auto get_structured_successors = [this](const ir::BasicBlock* block) {
      return &(block2structured_succs_[block]); };
  // TODO(greg-lunarg): Get rid of const_cast by making moving const
  // out of the cfa.h prototypes and into the invoking code.
  auto post_order = [&](cbb_ptr b) {
      order->push_front(const_cast<ir::BasicBlock*>(b)); };
  
  spvtools::CFA<ir::BasicBlock>::DepthFirstTraversal(
      &*func->begin(), get_structured_successors, ignore_block, post_order,
      ignore_edge);
}

void DeadBranchElimPass::GetConstCondition(
    uint32_t condId, bool* condVal, bool* condIsConst) {
  ir::Instruction* cInst = def_use_mgr_->GetDef(condId);
  switch (cInst->opcode()) {
    case SpvOpConstantFalse: {
      *condVal = false;
      *condIsConst = true;
    } break;
    case SpvOpConstantTrue: {
      *condVal = true;
      *condIsConst = true;
    } break;
    case SpvOpLogicalNot: {
      bool negVal;
      (void)GetConstCondition(cInst->GetSingleWordInOperand(0),
          &negVal, condIsConst);
      if (*condIsConst)
        *condVal = !negVal;
    } break;
    default: {
      *condIsConst = false;
    } break;
  }
}

void DeadBranchElimPass::AddBranch(uint32_t labelId, ir::BasicBlock* bp) {
  std::unique_ptr<ir::Instruction> newBranch(
    new ir::Instruction(SpvOpBranch, 0, 0,
        {{spv_operand_type_t::SPV_OPERAND_TYPE_ID, {labelId}}}));
  def_use_mgr_->AnalyzeInstDefUse(&*newBranch);
  bp->AddInstruction(std::move(newBranch));
}

void DeadBranchElimPass::AddSelectionMerge(uint32_t labelId,
    ir::BasicBlock* bp) {
  std::unique_ptr<ir::Instruction> newMerge(
    new ir::Instruction(SpvOpSelectionMerge, 0, 0,
        {{spv_operand_type_t::SPV_OPERAND_TYPE_ID, {labelId}},
         {spv_operand_type_t::SPV_OPERAND_TYPE_LITERAL_INTEGER, {0}}}));
  def_use_mgr_->AnalyzeInstDefUse(&*newMerge);
  bp->AddInstruction(std::move(newMerge));
}

void DeadBranchElimPass::AddBranchConditional(uint32_t condId,
    uint32_t trueLabId, uint32_t falseLabId, ir::BasicBlock* bp) {
  std::unique_ptr<ir::Instruction> newBranchCond(
    new ir::Instruction(SpvOpBranchConditional, 0, 0,
        {{spv_operand_type_t::SPV_OPERAND_TYPE_ID, {condId}},
         {spv_operand_type_t::SPV_OPERAND_TYPE_ID, {trueLabId}},
         {spv_operand_type_t::SPV_OPERAND_TYPE_ID, {falseLabId}}}));
  def_use_mgr_->AnalyzeInstDefUse(&*newBranchCond);
  bp->AddInstruction(std::move(newBranchCond));
}

void DeadBranchElimPass::KillAllInsts(ir::BasicBlock* bp) {
  bp->ForEachInst([this](ir::Instruction* ip) {
    KillNamesAndDecorates(ip);
    def_use_mgr_->KillInst(ip);
  });
}

bool DeadBranchElimPass::GetConstConditionalSelectionBranch(ir::BasicBlock* bp,
    ir::Instruction** branchInst, ir::Instruction** mergeInst,
    uint32_t *condId, bool *condVal) {
  auto ii = bp->end();
  --ii;
  *branchInst = &*ii;
  if ((*branchInst)->opcode() != SpvOpBranchConditional)
    return false;
  if (ii == bp->begin())
    return false;
  --ii;
  *mergeInst = &*ii;
  if ((*mergeInst)->opcode() != SpvOpSelectionMerge)
    return false;
  bool condIsConst;
  *condId = (*branchInst)->GetSingleWordInOperand(
      kBranchCondConditionalIdInIdx);
  (void) GetConstCondition(*condId, condVal, &condIsConst);
  return condIsConst;
}

bool DeadBranchElimPass::HasNonPhiRef(uint32_t labelId) {
  analysis::UseList* uses = def_use_mgr_->GetUses(labelId);
  if (uses == nullptr)
    return false;
  for (auto u : *uses)
    if (u.inst->opcode() != SpvOpPhi)
      return true;
  return false;
}

bool DeadBranchElimPass::EliminateDeadBranches(ir::Function* func) {
  // Traverse blocks in structured order
  std::list<ir::BasicBlock*> structuredOrder;
  ComputeStructuredOrder(func, &structuredOrder);
  std::unordered_set<ir::BasicBlock*> elimBlocks;
  bool modified = false;
  for (auto bi = structuredOrder.begin(); bi != structuredOrder.end(); ++bi) {
    // Skip blocks that are already in the elimination set
    if (elimBlocks.find(*bi) != elimBlocks.end())
      continue;
    // Skip blocks that don't have constant conditional branch preceded
    // by OpSelectionMerge
    ir::Instruction* br;
    ir::Instruction* mergeInst;
    uint32_t condId;
    bool condVal;
    if (!GetConstConditionalSelectionBranch(*bi, &br, &mergeInst, &condId,
        &condVal))
      continue;

    // Replace conditional branch with unconditional branch
    const uint32_t trueLabId =
        br->GetSingleWordInOperand(kBranchCondTrueLabIdInIdx);
    const uint32_t falseLabId =
        br->GetSingleWordInOperand(kBranchCondFalseLabIdInIdx);
    const uint32_t mergeLabId =
        mergeInst->GetSingleWordInOperand(kSelectionMergeMergeBlockIdInIdx);
    const uint32_t liveLabId = condVal == true ? trueLabId : falseLabId;
    const uint32_t deadLabId = condVal == true ? falseLabId : trueLabId;
    AddBranch(liveLabId, *bi);
    def_use_mgr_->KillInst(br);
    def_use_mgr_->KillInst(mergeInst);

    // Iterate to merge block adding dead blocks to elimination set
    auto dbi = bi;
    ++dbi;
    uint32_t dLabId = (*dbi)->id();
    while (dLabId != mergeLabId) {
      if (!HasNonPhiRef(dLabId)) {
        // Kill use/def for all instructions and mark block for elimination
        KillAllInsts(*dbi);
        elimBlocks.insert(*dbi);
      }
      ++dbi;
      dLabId = (*dbi)->id();
    }

    // Process phi instructions in merge block.
    // elimBlocks are now blocks which cannot precede merge block. Also,
    // if eliminated branch is to merge label, remember the conditional block
    // also cannot precede merge block.
    uint32_t deadCondLabId = 0;
    if (deadLabId == mergeLabId)
      deadCondLabId = (*bi)->id();
    (*dbi)->ForEachPhiInst([&elimBlocks, &deadCondLabId, this](
        ir::Instruction* phiInst) {
      const uint32_t phiLabId0 =
          phiInst->GetSingleWordInOperand(kPhiLab0IdInIdx);
      const bool useFirst =
          elimBlocks.find(id2block_[phiLabId0]) == elimBlocks.end() &&
          phiLabId0 != deadCondLabId;
      const uint32_t phiValIdx =
          useFirst ? kPhiVal0IdInIdx : kPhiVal1IdInIdx;
      const uint32_t replId = phiInst->GetSingleWordInOperand(phiValIdx);
      const uint32_t phiId = phiInst->result_id();
      KillNamesAndDecorates(phiId);
      (void)def_use_mgr_->ReplaceAllUsesWith(phiId, replId);
      def_use_mgr_->KillInst(phiInst);
    });

    // If merge block has no predecessors, replace the new branch with
    // a MergeSelection/BranchCondition using the original constant condition
    // and the mergeblock as the false branch. This is done so the merge block
    // is not orphaned, which could cause invalid control flow in certain case.
    // TODO(greg-lunarg): Do this only in cases where invalid code is caused.
    if (!HasNonPhiRef(mergeLabId)) {
      auto eii = (*bi)->end();
      --eii;
      ir::Instruction* nbr = &*eii;
      AddSelectionMerge(mergeLabId, *bi);
      if (condVal == true)
        AddBranchConditional(condId, liveLabId, mergeLabId, *bi);
      else
        AddBranchConditional(condId, mergeLabId, liveLabId, *bi);
      def_use_mgr_->KillInst(nbr);
    }
    modified = true;
  }

  // Erase dead blocks
  for (auto ebi = func->begin(); ebi != func->end(); )
    if (elimBlocks.find(&*ebi) != elimBlocks.end())
      ebi = ebi.Erase();
    else
      ++ebi;
  return modified;
}

void DeadBranchElimPass::Initialize(ir::Module* module) {

  module_ = module;

  // Initialize function and block maps
  id2block_.clear();
  block2structured_succs_.clear();

  // Initialize block map
  for (auto& fn : *module_)
    for (auto& blk : fn)
      id2block_[blk.id()] = &blk;

  // TODO(greg-lunarg): Reuse def/use from previous passes
  def_use_mgr_.reset(new analysis::DefUseManager(consumer(), module_));

  // Initialize extension whitelist
  InitExtensions();
};

bool DeadBranchElimPass::AllExtensionsSupported() const {
  // If any extension not in whitelist, return false
  for (auto& ei : module_->extensions()) {
    const char* extName = reinterpret_cast<const char*>(
        &ei.GetInOperand(0).words[0]);
    if (extensions_whitelist_.find(extName) == extensions_whitelist_.end())
      return false;
  }
  return true;
}

Pass::Status DeadBranchElimPass::ProcessImpl() {
  // Current functionality assumes structured control flow. 
  // TODO(greg-lunarg): Handle non-structured control-flow.
  if (!module_->HasCapability(SpvCapabilityShader))
    return Status::SuccessWithoutChange;
  // Do not process if module contains OpGroupDecorate. Additional
  // support required in KillNamesAndDecorates().
  // TODO(greg-lunarg): Add support for OpGroupDecorate
  for (auto& ai : module_->annotations())
    if (ai.opcode() == SpvOpGroupDecorate)
      return Status::SuccessWithoutChange;
  // Do not process if any disallowed extensions are enabled
  if (!AllExtensionsSupported())
    return Status::SuccessWithoutChange;
  // Collect all named and decorated ids
  FindNamedOrDecoratedIds();
  // Process all entry point functions
  ProcessFunction pfn = [this](ir::Function* fp) {
    return EliminateDeadBranches(fp);
  };
  bool modified = ProcessEntryPointCallTree(pfn, module_);
  return modified ? Status::SuccessWithChange : Status::SuccessWithoutChange;
}

DeadBranchElimPass::DeadBranchElimPass() {}

Pass::Status DeadBranchElimPass::Process(ir::Module* module) {
  Initialize(module);
  return ProcessImpl();
}

void DeadBranchElimPass::InitExtensions() {
  extensions_whitelist_.clear();
  extensions_whitelist_.insert({
    "SPV_AMD_shader_explicit_vertex_parameter",
    "SPV_AMD_shader_trinary_minmax",
    "SPV_AMD_gcn_shader",
    "SPV_KHR_shader_ballot",
    "SPV_AMD_shader_ballot",
    "SPV_AMD_gpu_shader_half_float",
    "SPV_KHR_shader_draw_parameters",
    "SPV_KHR_subgroup_vote",
    "SPV_KHR_16bit_storage",
    "SPV_KHR_device_group",
    "SPV_KHR_multiview",
    "SPV_NVX_multiview_per_view_attributes",
    "SPV_NV_viewport_array2",
    "SPV_NV_stereo_view_rendering",
    "SPV_NV_sample_mask_override_coverage",
    "SPV_NV_geometry_shader_passthrough",
    "SPV_AMD_texture_gather_bias_lod",
    "SPV_KHR_storage_buffer_storage_class",
    "SPV_KHR_variable_pointers",
    "SPV_AMD_gpu_shader_int16",
    "SPV_KHR_post_depth_coverage",
    "SPV_KHR_shader_atomic_counter_ops",
  });
}

}  // namespace opt
}  // namespace spvtools