QueryTemplateGenerator.cpp File Reference

#include "QueryTemplateGenerator.h"
#include "IRCodegenUtils.h"
#include "Logger/Logger.h"
#include <llvm/IR/Constants.h>
#include <llvm/IR/IRBuilder.h>
#include <llvm/IR/Instructions.h>
#include <llvm/IR/Verifier.h>

Include dependency graph for QueryTemplateGenerator.cpp:

Go to the source code of this file.

Namespaces
	anonymous_namespace{QueryTemplateGenerator.cpp}

Functions
template<class Attributes >
llvm::Function *	anonymous_namespace{QueryTemplateGenerator.cpp}::default_func_builder (llvm::Module *mod, const std::string &name)
template<class Attributes >
llvm::Function *	anonymous_namespace{QueryTemplateGenerator.cpp}::pos_start (llvm::Module *mod)
template<class Attributes >
llvm::Function *	anonymous_namespace{QueryTemplateGenerator.cpp}::group_buff_idx (llvm::Module *mod)
template<class Attributes >
llvm::Function *	anonymous_namespace{QueryTemplateGenerator.cpp}::pos_step (llvm::Module *mod)
template<class Attributes >
llvm::Function *	anonymous_namespace{QueryTemplateGenerator.cpp}::row_process (llvm::Module *mod, const size_t aggr_col_count, const bool hoist_literals)
template<class Attributes >
std::tuple< llvm::Function *, llvm::CallInst * >	query_template_impl (llvm::Module *mod, const size_t aggr_col_count, const bool hoist_literals, const bool is_estimate_query, const GpuSharedMemoryContext &gpu_smem_context)
template<class Attributes >
std::tuple< llvm::Function *, llvm::CallInst * >	query_group_by_template_impl (llvm::Module *mod, const bool hoist_literals, const QueryMemoryDescriptor &query_mem_desc, const ExecutorDeviceType device_type, const bool check_scan_limit, const GpuSharedMemoryContext &gpu_smem_context)
std::tuple< llvm::Function *, llvm::CallInst * >	query_template (llvm::Module *module, const size_t aggr_col_count, const bool hoist_literals, const bool is_estimate_query, const GpuSharedMemoryContext &gpu_smem_context)
std::tuple< llvm::Function *, llvm::CallInst * >	query_group_by_template (llvm::Module *module, const bool hoist_literals, const QueryMemoryDescriptor &query_mem_desc, const ExecutorDeviceType device_type, const bool check_scan_limit, const GpuSharedMemoryContext &gpu_smem_context)

Function Documentation

std::tuple<llvm::Function*, llvm::CallInst*> query_group_by_template	(	llvm::Module *	module,
		const bool	hoist_literals,
		const QueryMemoryDescriptor &	query_mem_desc,
		const ExecutorDeviceType	device_type,
		const bool	check_scan_limit,
		const GpuSharedMemoryContext &	gpu_smem_context
	)

Definition at line 862 of file QueryTemplateGenerator.cpp.

References module(), and query_mem_desc.

                                                    {
  return query_group_by_template_impl<llvm::AttributeSet>(module,
                                                          hoist_literals,
                                                          query_mem_desc,
                                                          device_type,
                                                          check_scan_limit,
                                                          gpu_smem_context);
}

Here is the call graph for this function:

template<class Attributes >

std::tuple<llvm::Function*, llvm::CallInst*> query_group_by_template_impl	(	llvm::Module *	mod,
		const bool	hoist_literals,
		const QueryMemoryDescriptor &	query_mem_desc,
		const ExecutorDeviceType	device_type,
		const bool	check_scan_limit,
		const GpuSharedMemoryContext &	gpu_smem_context
	)

Definition at line 520 of file QueryTemplateGenerator.cpp.

References CHECK(), error_code, logger::FATAL, frag_idx, GpuSharedMemoryContext::getSharedMemorySize(), GPU, anonymous_namespace{QueryTemplateGenerator.cpp}::group_buff_idx(), GpuSharedMemoryContext::isSharedMemoryUsed(), QueryMemoryDescriptor::isWarpSyncRequired(), join_hash_tables, literals, LLVM_ALIGN, LOG, max_matched, anonymous_namespace{QueryTemplateGenerator.cpp}::pos_start(), anonymous_namespace{QueryTemplateGenerator.cpp}::pos_step(), and anonymous_namespace{QueryTemplateGenerator.cpp}::row_process().

                                                    {
  if (gpu_smem_context.isSharedMemoryUsed()) {
    CHECK(device_type == ExecutorDeviceType::GPU);
  }
  using namespace llvm;

  auto func_pos_start = pos_start<Attributes>(mod);
  CHECK(func_pos_start);
  auto func_pos_step = pos_step<Attributes>(mod);
  CHECK(func_pos_step);
  auto func_group_buff_idx = group_buff_idx<Attributes>(mod);
  CHECK(func_group_buff_idx);
  auto func_row_process = row_process<Attributes>(mod, 0, hoist_literals);
  CHECK(func_row_process);
  auto func_init_shared_mem = gpu_smem_context.isSharedMemoryUsed()
                                  ? mod->getFunction("init_shared_mem")
                                  : mod->getFunction("init_shared_mem_nop");
  CHECK(func_init_shared_mem);

  auto func_write_back = mod->getFunction("write_back_nop");
  CHECK(func_write_back);

  auto i32_type = IntegerType::get(mod->getContext(), 32);
  auto i64_type = IntegerType::get(mod->getContext(), 64);
  auto pi8_type = PointerType::get(IntegerType::get(mod->getContext(), 8), 0);
  auto pi32_type = PointerType::get(i32_type, 0);
  auto pi64_type = PointerType::get(i64_type, 0);
  auto ppi64_type = PointerType::get(pi64_type, 0);
  auto ppi8_type = PointerType::get(pi8_type, 0);

  std::vector<Type*> query_args;
  query_args.push_back(ppi8_type);
  if (hoist_literals) {
    query_args.push_back(pi8_type);
  }
  query_args.push_back(pi64_type);
  query_args.push_back(pi64_type);
  query_args.push_back(pi32_type);
  query_args.push_back(pi64_type);

  query_args.push_back(ppi64_type);
  query_args.push_back(i32_type);
  query_args.push_back(pi64_type);
  query_args.push_back(pi32_type);
  query_args.push_back(pi32_type);

  FunctionType* query_func_type = FunctionType::get(
      /*Result=*/Type::getVoidTy(mod->getContext()),
      /*Params=*/query_args,
      /*isVarArg=*/false);

  std::string query_name{"query_group_by_template"};
  auto query_func_ptr = mod->getFunction(query_name);
  CHECK(!query_func_ptr);

  query_func_ptr = Function::Create(
      /*Type=*/query_func_type,
      /*Linkage=*/GlobalValue::ExternalLinkage,
      /*Name=*/"query_group_by_template",
      mod);

  query_func_ptr->setCallingConv(CallingConv::C);

  Attributes query_func_pal;
  {
    SmallVector<Attributes, 4> Attrs;
    Attributes PAS;
    {
      AttrBuilder B;
      B.addAttribute(Attribute::ReadNone);
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 1U, B);
    }

    Attrs.push_back(PAS);
    {
      AttrBuilder B;
      B.addAttribute(Attribute::ReadOnly);
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 2U, B);
    }

    Attrs.push_back(PAS);
    {
      AttrBuilder B;
      B.addAttribute(Attribute::ReadNone);
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 3U, B);
    }

    Attrs.push_back(PAS);
    {
      AttrBuilder B;
      B.addAttribute(Attribute::ReadOnly);
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 4U, B);
    }

    Attrs.push_back(PAS);
    {
      AttrBuilder B;
      B.addAttribute(Attribute::UWTable);
      PAS = Attributes::get(mod->getContext(), ~0U, B);
    }

    Attrs.push_back(PAS);

    query_func_pal = Attributes::get(mod->getContext(), Attrs);
  }
  query_func_ptr->setAttributes(query_func_pal);

  Function::arg_iterator query_arg_it = query_func_ptr->arg_begin();
  Value* byte_stream = &*query_arg_it;
  byte_stream->setName("byte_stream");
  Value* literals{nullptr};
  if (hoist_literals) {
    literals = &*(++query_arg_it);
    ;
    literals->setName("literals");
  }
  Value* row_count_ptr = &*(++query_arg_it);
  row_count_ptr->setName("row_count_ptr");
  Value* frag_row_off_ptr = &*(++query_arg_it);
  frag_row_off_ptr->setName("frag_row_off_ptr");
  Value* max_matched_ptr = &*(++query_arg_it);
  max_matched_ptr->setName("max_matched_ptr");
  Value* agg_init_val = &*(++query_arg_it);
  agg_init_val->setName("agg_init_val");
  Value* group_by_buffers = &*(++query_arg_it);
  group_by_buffers->setName("group_by_buffers");
  Value* frag_idx = &*(++query_arg_it);
  frag_idx->setName("frag_idx");
  Value* join_hash_tables = &*(++query_arg_it);
  join_hash_tables->setName("join_hash_tables");
  Value* total_matched = &*(++query_arg_it);
  total_matched->setName("total_matched");
  Value* error_code = &*(++query_arg_it);
  error_code->setName("error_code");

  auto bb_entry = BasicBlock::Create(mod->getContext(), ".entry", query_func_ptr, 0);
  auto bb_preheader =
      BasicBlock::Create(mod->getContext(), ".loop.preheader", query_func_ptr, 0);
  auto bb_forbody = BasicBlock::Create(mod->getContext(), ".forbody", query_func_ptr, 0);
  auto bb_crit_edge =
      BasicBlock::Create(mod->getContext(), "._crit_edge", query_func_ptr, 0);
  auto bb_exit = BasicBlock::Create(mod->getContext(), ".exit", query_func_ptr, 0);

  // Block  .entry
  LoadInst* row_count = new LoadInst(row_count_ptr, "", false, bb_entry);
  row_count->setAlignment(LLVM_ALIGN(8));
  row_count->setName("row_count");

  LoadInst* max_matched = new LoadInst(max_matched_ptr, "", false, bb_entry);
  max_matched->setAlignment(LLVM_ALIGN(8));

  auto crt_matched_ptr = new AllocaInst(i32_type, 0, "crt_matched", bb_entry);
  auto old_total_matched_ptr = new AllocaInst(i32_type, 0, "old_total_matched", bb_entry);
  CallInst* pos_start = CallInst::Create(func_pos_start, "", bb_entry);
  pos_start->setCallingConv(CallingConv::C);
  pos_start->setTailCall(true);
  Attributes pos_start_pal;
  pos_start->setAttributes(pos_start_pal);

  CallInst* pos_step = CallInst::Create(func_pos_step, "", bb_entry);
  pos_step->setCallingConv(CallingConv::C);
  pos_step->setTailCall(true);
  Attributes pos_step_pal;
  pos_step->setAttributes(pos_step_pal);

  CallInst* group_buff_idx = CallInst::Create(func_group_buff_idx, "", bb_entry);
  group_buff_idx->setCallingConv(CallingConv::C);
  group_buff_idx->setTailCall(true);
  Attributes group_buff_idx_pal;
  group_buff_idx->setAttributes(group_buff_idx_pal);

  CastInst* pos_start_i64 = new SExtInst(pos_start, i64_type, "", bb_entry);
  const PointerType* Ty = dyn_cast<PointerType>(group_by_buffers->getType());
  CHECK(Ty);
  GetElementPtrInst* group_by_buffers_gep = GetElementPtrInst::Create(
      Ty->getElementType(), group_by_buffers, group_buff_idx, "", bb_entry);
  LoadInst* col_buffer = new LoadInst(group_by_buffers_gep, "", false, bb_entry);
  col_buffer->setName("col_buffer");
  col_buffer->setAlignment(LLVM_ALIGN(8));

  llvm::ConstantInt* shared_mem_bytes_lv =
      ConstantInt::get(i32_type, gpu_smem_context.getSharedMemorySize());
  llvm::CallInst* result_buffer =
      CallInst::Create(func_init_shared_mem,
                       std::vector<llvm::Value*>{col_buffer, shared_mem_bytes_lv},
                       "result_buffer",
                       bb_entry);
  // TODO(Saman): change this further, normal path should not go through this

  ICmpInst* enter_or_not =
      new ICmpInst(*bb_entry, ICmpInst::ICMP_SLT, pos_start_i64, row_count, "");
  BranchInst::Create(bb_preheader, bb_exit, enter_or_not, bb_entry);

  // Block .loop.preheader
  CastInst* pos_step_i64 = new SExtInst(pos_step, i64_type, "", bb_preheader);
  BranchInst::Create(bb_forbody, bb_preheader);

  // Block .forbody
  Argument* pos_pre = new Argument(i64_type);
  PHINode* pos = PHINode::Create(i64_type, check_scan_limit ? 3 : 2, "pos", bb_forbody);

  std::vector<Value*> row_process_params;
  row_process_params.push_back(result_buffer);
  row_process_params.push_back(crt_matched_ptr);
  row_process_params.push_back(total_matched);
  row_process_params.push_back(old_total_matched_ptr);
  row_process_params.push_back(max_matched_ptr);
  row_process_params.push_back(agg_init_val);
  row_process_params.push_back(pos);
  row_process_params.push_back(frag_row_off_ptr);
  row_process_params.push_back(row_count_ptr);
  if (hoist_literals) {
    CHECK(literals);
    row_process_params.push_back(literals);
  }
  if (check_scan_limit) {
    new StoreInst(ConstantInt::get(IntegerType::get(mod->getContext(), 32), 0),
                  crt_matched_ptr,
                  bb_forbody);
  }
  CallInst* row_process =
      CallInst::Create(func_row_process, row_process_params, "", bb_forbody);
  row_process->setCallingConv(CallingConv::C);
  row_process->setTailCall(true);
  Attributes row_process_pal;
  row_process->setAttributes(row_process_pal);

  // Forcing all threads within a warp to be synchronized (Compute >= 7.x)
  if (query_mem_desc.isWarpSyncRequired(device_type)) {
    auto func_sync_warp_protected = mod->getFunction("sync_warp_protected");
    CHECK(func_sync_warp_protected);
    CallInst::Create(func_sync_warp_protected,
                     std::vector<llvm::Value*>{pos, row_count},
                     "",
                     bb_forbody);
  }

  BinaryOperator* pos_inc =
      BinaryOperator::Create(Instruction::Add, pos, pos_step_i64, "", bb_forbody);
  ICmpInst* loop_or_exit =
      new ICmpInst(*bb_forbody, ICmpInst::ICMP_SLT, pos_inc, row_count, "");
  if (check_scan_limit) {
    auto crt_matched = new LoadInst(crt_matched_ptr, "crt_matched", false, bb_forbody);
    auto filter_match = BasicBlock::Create(
        mod->getContext(), "filter_match", query_func_ptr, bb_crit_edge);
    llvm::Value* new_total_matched =
        new LoadInst(old_total_matched_ptr, "", false, filter_match);
    new_total_matched =
        BinaryOperator::CreateAdd(new_total_matched, crt_matched, "", filter_match);
    CHECK(new_total_matched);
    ICmpInst* limit_not_reached = new ICmpInst(*filter_match,
                                               ICmpInst::ICMP_SLT,
                                               new_total_matched,
                                               max_matched,
                                               "limit_not_reached");
    BranchInst::Create(
        bb_forbody,
        bb_crit_edge,
        BinaryOperator::Create(
            BinaryOperator::And, loop_or_exit, limit_not_reached, "", filter_match),
        filter_match);
    auto filter_nomatch = BasicBlock::Create(
        mod->getContext(), "filter_nomatch", query_func_ptr, bb_crit_edge);
    BranchInst::Create(bb_forbody, bb_crit_edge, loop_or_exit, filter_nomatch);
    ICmpInst* crt_matched_nz = new ICmpInst(
        *bb_forbody, ICmpInst::ICMP_NE, crt_matched, ConstantInt::get(i32_type, 0), "");
    BranchInst::Create(filter_match, filter_nomatch, crt_matched_nz, bb_forbody);
    pos->addIncoming(pos_start_i64, bb_preheader);
    pos->addIncoming(pos_pre, filter_match);
    pos->addIncoming(pos_pre, filter_nomatch);
  } else {
    pos->addIncoming(pos_start_i64, bb_preheader);
    pos->addIncoming(pos_pre, bb_forbody);
    BranchInst::Create(bb_forbody, bb_crit_edge, loop_or_exit, bb_forbody);
  }

  // Block ._crit_edge
  BranchInst::Create(bb_exit, bb_crit_edge);

  // Block .exit
  CallInst::Create(func_write_back,
                   std::vector<Value*>{col_buffer, result_buffer, shared_mem_bytes_lv},
                   "",
                   bb_exit);

  ReturnInst::Create(mod->getContext(), bb_exit);

  // Resolve Forward References
  pos_pre->replaceAllUsesWith(pos_inc);
  delete pos_pre;

  if (verifyFunction(*query_func_ptr, &llvm::errs())) {
    LOG(FATAL) << "Generated invalid code. ";
  }

  return {query_func_ptr, row_process};
}

Here is the call graph for this function:

std::tuple<llvm::Function*, llvm::CallInst*> query_template	(	llvm::Module *	module,
		const size_t	aggr_col_count,
		const bool	hoist_literals,
		const bool	is_estimate_query,
		const GpuSharedMemoryContext &	gpu_smem_context
	)

Definition at line 853 of file QueryTemplateGenerator.cpp.

References module().

                                                    {
  return query_template_impl<llvm::AttributeSet>(
      module, aggr_col_count, hoist_literals, is_estimate_query, gpu_smem_context);
}

Here is the call graph for this function:

template<class Attributes >

std::tuple<llvm::Function*, llvm::CallInst*> query_template_impl	(	llvm::Module *	mod,
		const size_t	aggr_col_count,
		const bool	hoist_literals,
		const bool	is_estimate_query,
		const GpuSharedMemoryContext &	gpu_smem_context
	)

If GPU shared memory optimization is disabled, for each aggregate target, threads copy back their aggregate results (stored in registers) back into memory. This process is performed per processed fragment. In the host the final results are reduced (per target, for all threads and all fragments).

If GPU Shared memory optimization is enabled, we properly (atomically) aggregate all thread's results into memory, which makes the final reduction on host much cheaper. Here, we call a noop dummy write back function which will be properly replaced at runtime depending on the target expressions.

Definition at line 186 of file QueryTemplateGenerator.cpp.

References CHECK(), error_code, logger::FATAL, frag_idx, anonymous_namespace{QueryTemplateGenerator.cpp}::group_buff_idx(), GpuSharedMemoryContext::isSharedMemoryUsed(), join_hash_tables, literals, LLVM_ALIGN, LOG, out, anonymous_namespace{QueryTemplateGenerator.cpp}::pos_start(), anonymous_namespace{QueryTemplateGenerator.cpp}::pos_step(), run_benchmark_import::result, anonymous_namespace{QueryTemplateGenerator.cpp}::row_process(), and to_string().

                                                    {
  using namespace llvm;

  auto func_pos_start = pos_start<Attributes>(mod);
  CHECK(func_pos_start);
  auto func_pos_step = pos_step<Attributes>(mod);
  CHECK(func_pos_step);
  auto func_group_buff_idx = group_buff_idx<Attributes>(mod);
  CHECK(func_group_buff_idx);
  auto func_row_process = row_process<Attributes>(
      mod, is_estimate_query ? 1 : aggr_col_count, hoist_literals);
  CHECK(func_row_process);

  auto i8_type = IntegerType::get(mod->getContext(), 8);
  auto i32_type = IntegerType::get(mod->getContext(), 32);
  auto i64_type = IntegerType::get(mod->getContext(), 64);
  auto pi8_type = PointerType::get(i8_type, 0);
  auto ppi8_type = PointerType::get(pi8_type, 0);
  auto pi32_type = PointerType::get(i32_type, 0);
  auto pi64_type = PointerType::get(i64_type, 0);
  auto ppi64_type = PointerType::get(pi64_type, 0);

  std::vector<Type*> query_args;
  query_args.push_back(ppi8_type);
  if (hoist_literals) {
    query_args.push_back(pi8_type);
  }
  query_args.push_back(pi64_type);
  query_args.push_back(pi64_type);
  query_args.push_back(pi32_type);

  query_args.push_back(pi64_type);
  query_args.push_back(ppi64_type);
  query_args.push_back(i32_type);
  query_args.push_back(pi64_type);
  query_args.push_back(pi32_type);
  query_args.push_back(pi32_type);

  FunctionType* query_func_type = FunctionType::get(
      /*Result=*/Type::getVoidTy(mod->getContext()),
      /*Params=*/query_args,
      /*isVarArg=*/false);

  std::string query_template_name{"query_template"};
  auto query_func_ptr = mod->getFunction(query_template_name);
  CHECK(!query_func_ptr);

  query_func_ptr = Function::Create(
      /*Type=*/query_func_type,
      /*Linkage=*/GlobalValue::ExternalLinkage,
      /*Name=*/query_template_name,
      mod);
  query_func_ptr->setCallingConv(CallingConv::C);

  Attributes query_func_pal;
  {
    SmallVector<Attributes, 4> Attrs;
    Attributes PAS;
    {
      AttrBuilder B;
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 1U, B);
    }

    Attrs.push_back(PAS);
    {
      AttrBuilder B;
      B.addAttribute(Attribute::NoCapture);
      PAS = Attributes::get(mod->getContext(), 2U, B);
    }

    Attrs.push_back(PAS);

    {
      AttrBuilder B;
      B.addAttribute(Attribute::NoCapture);
      Attrs.push_back(Attributes::get(mod->getContext(), 3U, B));
    }

    {
      AttrBuilder B;
      B.addAttribute(Attribute::NoCapture);
      Attrs.push_back(Attributes::get(mod->getContext(), 4U, B));
    }

    Attrs.push_back(PAS);

    query_func_pal = Attributes::get(mod->getContext(), Attrs);
  }
  query_func_ptr->setAttributes(query_func_pal);

  Function::arg_iterator query_arg_it = query_func_ptr->arg_begin();
  Value* byte_stream = &*query_arg_it;
  byte_stream->setName("byte_stream");
  Value* literals{nullptr};
  if (hoist_literals) {
    literals = &*(++query_arg_it);
    literals->setName("literals");
  }
  Value* row_count_ptr = &*(++query_arg_it);
  row_count_ptr->setName("row_count_ptr");
  Value* frag_row_off_ptr = &*(++query_arg_it);
  frag_row_off_ptr->setName("frag_row_off_ptr");
  Value* max_matched_ptr = &*(++query_arg_it);
  max_matched_ptr->setName("max_matched_ptr");
  Value* agg_init_val = &*(++query_arg_it);
  agg_init_val->setName("agg_init_val");
  Value* out = &*(++query_arg_it);
  out->setName("out");
  Value* frag_idx = &*(++query_arg_it);
  frag_idx->setName("frag_idx");
  Value* join_hash_tables = &*(++query_arg_it);
  join_hash_tables->setName("join_hash_tables");
  Value* total_matched = &*(++query_arg_it);
  total_matched->setName("total_matched");
  Value* error_code = &*(++query_arg_it);
  error_code->setName("error_code");

  auto bb_entry = BasicBlock::Create(mod->getContext(), ".entry", query_func_ptr, 0);
  auto bb_preheader =
      BasicBlock::Create(mod->getContext(), ".loop.preheader", query_func_ptr, 0);
  auto bb_forbody = BasicBlock::Create(mod->getContext(), ".for.body", query_func_ptr, 0);
  auto bb_crit_edge =
      BasicBlock::Create(mod->getContext(), "._crit_edge", query_func_ptr, 0);
  auto bb_exit = BasicBlock::Create(mod->getContext(), ".exit", query_func_ptr, 0);

  // Block  (.entry)
  std::vector<Value*> result_ptr_vec;
  llvm::CallInst* smem_output_buffer{nullptr};
  if (!is_estimate_query) {
    for (size_t i = 0; i < aggr_col_count; ++i) {
      auto result_ptr = new AllocaInst(i64_type, 0, "result", bb_entry);
      result_ptr->setAlignment(LLVM_ALIGN(8));
      result_ptr_vec.push_back(result_ptr);
    }
    if (gpu_smem_context.isSharedMemoryUsed()) {
      auto init_smem_func = mod->getFunction("init_shared_mem");
      CHECK(init_smem_func);
      // only one slot per aggregate column is needed, and so we can initialize shared
      // memory buffer for intermediate results to be exactly like the agg_init_val array
      smem_output_buffer = CallInst::Create(
          init_smem_func,
          std::vector<llvm::Value*>{
              agg_init_val,
              llvm::ConstantInt::get(i32_type, aggr_col_count * sizeof(int64_t))},
          "smem_buffer",
          bb_entry);
    }
  }

  LoadInst* row_count = new LoadInst(row_count_ptr, "row_count", false, bb_entry);
  row_count->setAlignment(LLVM_ALIGN(8));
  row_count->setName("row_count");
  std::vector<Value*> agg_init_val_vec;
  if (!is_estimate_query) {
    for (size_t i = 0; i < aggr_col_count; ++i) {
      auto idx_lv = ConstantInt::get(i32_type, i);
      auto agg_init_gep =
          GetElementPtrInst::CreateInBounds(agg_init_val, idx_lv, "", bb_entry);
      auto agg_init_val = new LoadInst(agg_init_gep, "", false, bb_entry);
      agg_init_val->setAlignment(LLVM_ALIGN(8));
      agg_init_val_vec.push_back(agg_init_val);
      auto init_val_st = new StoreInst(agg_init_val, result_ptr_vec[i], false, bb_entry);
      init_val_st->setAlignment(LLVM_ALIGN(8));
    }
  }

  CallInst* pos_start = CallInst::Create(func_pos_start, "pos_start", bb_entry);
  pos_start->setCallingConv(CallingConv::C);
  pos_start->setTailCall(true);
  Attributes pos_start_pal;
  pos_start->setAttributes(pos_start_pal);

  CallInst* pos_step = CallInst::Create(func_pos_step, "pos_step", bb_entry);
  pos_step->setCallingConv(CallingConv::C);
  pos_step->setTailCall(true);
  Attributes pos_step_pal;
  pos_step->setAttributes(pos_step_pal);

  CallInst* group_buff_idx = nullptr;
  if (!is_estimate_query) {
    group_buff_idx = CallInst::Create(func_group_buff_idx, "group_buff_idx", bb_entry);
    group_buff_idx->setCallingConv(CallingConv::C);
    group_buff_idx->setTailCall(true);
    Attributes group_buff_idx_pal;
    group_buff_idx->setAttributes(group_buff_idx_pal);
  }

  CastInst* pos_start_i64 = new SExtInst(pos_start, i64_type, "", bb_entry);
  ICmpInst* enter_or_not =
      new ICmpInst(*bb_entry, ICmpInst::ICMP_SLT, pos_start_i64, row_count, "");
  BranchInst::Create(bb_preheader, bb_exit, enter_or_not, bb_entry);

  // Block .loop.preheader
  CastInst* pos_step_i64 = new SExtInst(pos_step, i64_type, "", bb_preheader);
  BranchInst::Create(bb_forbody, bb_preheader);

  // Block  .forbody
  Argument* pos_inc_pre = new Argument(i64_type);
  PHINode* pos = PHINode::Create(i64_type, 2, "pos", bb_forbody);
  pos->addIncoming(pos_start_i64, bb_preheader);
  pos->addIncoming(pos_inc_pre, bb_forbody);

  std::vector<Value*> row_process_params;
  row_process_params.insert(
      row_process_params.end(), result_ptr_vec.begin(), result_ptr_vec.end());
  if (is_estimate_query) {
    row_process_params.push_back(new LoadInst(out, "", false, bb_forbody));
  }
  row_process_params.push_back(agg_init_val);
  row_process_params.push_back(pos);
  row_process_params.push_back(frag_row_off_ptr);
  row_process_params.push_back(row_count_ptr);
  if (hoist_literals) {
    CHECK(literals);
    row_process_params.push_back(literals);
  }
  CallInst* row_process =
      CallInst::Create(func_row_process, row_process_params, "", bb_forbody);
  row_process->setCallingConv(CallingConv::C);
  row_process->setTailCall(false);
  Attributes row_process_pal;
  row_process->setAttributes(row_process_pal);

  BinaryOperator* pos_inc =
      BinaryOperator::CreateNSW(Instruction::Add, pos, pos_step_i64, "", bb_forbody);
  ICmpInst* loop_or_exit =
      new ICmpInst(*bb_forbody, ICmpInst::ICMP_SLT, pos_inc, row_count, "");
  BranchInst::Create(bb_forbody, bb_crit_edge, loop_or_exit, bb_forbody);

  // Block ._crit_edge
  std::vector<Instruction*> result_vec_pre;
  if (!is_estimate_query) {
    for (size_t i = 0; i < aggr_col_count; ++i) {
      auto result = new LoadInst(result_ptr_vec[i], ".pre.result", false, bb_crit_edge);
      result->setAlignment(LLVM_ALIGN(8));
      result_vec_pre.push_back(result);
    }
  }

  BranchInst::Create(bb_exit, bb_crit_edge);

  // Block  .exit
  if (!is_estimate_query) {
    std::vector<PHINode*> result_vec;
    for (int64_t i = aggr_col_count - 1; i >= 0; --i) {
      auto result =
          PHINode::Create(IntegerType::get(mod->getContext(), 64), 2, "", bb_exit);
      result->addIncoming(result_vec_pre[i], bb_crit_edge);
      result->addIncoming(agg_init_val_vec[i], bb_entry);
      result_vec.insert(result_vec.begin(), result);
    }

    for (size_t i = 0; i < aggr_col_count; ++i) {
      auto col_idx = ConstantInt::get(i32_type, i);
      if (gpu_smem_context.isSharedMemoryUsed()) {
        auto target_addr =
            GetElementPtrInst::CreateInBounds(smem_output_buffer, col_idx, "", bb_exit);
        // TODO: generalize this once we want to support other types of aggregate
        // functions besides COUNT.
        auto agg_func = mod->getFunction("agg_sum_shared");
        CHECK(agg_func);
        CallInst::Create(
            agg_func, std::vector<llvm::Value*>{target_addr, result_vec[i]}, "", bb_exit);
      } else {
        auto out_gep = GetElementPtrInst::CreateInBounds(out, col_idx, "", bb_exit);
        auto col_buffer = new LoadInst(out_gep, "", false, bb_exit);
        col_buffer->setAlignment(LLVM_ALIGN(8));
        auto slot_idx = BinaryOperator::CreateAdd(
            group_buff_idx,
            BinaryOperator::CreateMul(frag_idx, pos_step, "", bb_exit),
            "",
            bb_exit);
        auto target_addr =
            GetElementPtrInst::CreateInBounds(col_buffer, slot_idx, "", bb_exit);
        StoreInst* result_st = new StoreInst(result_vec[i], target_addr, false, bb_exit);
        result_st->setAlignment(LLVM_ALIGN(8));
      }
    }
    if (gpu_smem_context.isSharedMemoryUsed()) {
      // final reduction of results from shared memory buffer back into global memory.
      auto sync_thread_func = mod->getFunction("sync_threadblock");
      CHECK(sync_thread_func);
      CallInst::Create(sync_thread_func, std::vector<llvm::Value*>{}, "", bb_exit);
      auto reduce_smem_to_gmem_func = mod->getFunction("write_back_non_grouped_agg");
      CHECK(reduce_smem_to_gmem_func);
      // each thread reduce the aggregate target corresponding to its own thread ID.
      // If there are more targets than threads we do not currently use shared memory
      // optimization. This can be relaxed if necessary
      for (size_t i = 0; i < aggr_col_count; i++) {
        auto out_gep = GetElementPtrInst::CreateInBounds(
            out, ConstantInt::get(i32_type, i), "", bb_exit);
        auto gmem_output_buffer = new LoadInst(
            out_gep, "gmem_output_buffer_" + std::to_string(i), false, bb_exit);
        CallInst::Create(
            reduce_smem_to_gmem_func,
            std::vector<llvm::Value*>{
                smem_output_buffer, gmem_output_buffer, ConstantInt::get(i32_type, i)},
            "",
            bb_exit);
      }
    }
  }

  ReturnInst::Create(mod->getContext(), bb_exit);

  // Resolve Forward References
  pos_inc_pre->replaceAllUsesWith(pos_inc);
  delete pos_inc_pre;

  if (verifyFunction(*query_func_ptr)) {
    LOG(FATAL) << "Generated invalid code. ";
  }

  return {query_func_ptr, row_process};
}

Here is the call graph for this function: