Pass configuration parser (#152)

Author: ahueck

demo/02_broken_struct_example.c

@@ -6,7 +6,7 @@
 #define COUNT   5
 
 // Comment out for heap allocation
-//#define USE_STACK
+// #define USE_STACK
 
 struct particle {
   int id;

lib/passes/Commandline.cpp

@@ -238,7 +238,7 @@ CommandLineOptions::CommandLineOptions() {
 
 std::optional<typeart::config::OptionValue> CommandLineOptions::getValue(std::string_view opt_path) const {
   auto key = llvm::StringRef(opt_path.data());
-  if (mapping_.count(key) != 0U) {
+  if (occurence_mapping_.lookup(key)) {
     return mapping_.lookup(key);
   }
   return {};

lib/passes/TypeARTConfiguration.cpp

@@ -33,58 +33,46 @@ TypeARTConfiguration::TypeARTConfiguration(std::unique_ptr<file::FileOptions> co
 }
 
 std::optional<OptionValue> TypeARTConfiguration::getValue(std::string_view opt_path) const {
-  auto get_value = [&](const auto& options, const char* source) -> std::optional<OptionValue> {
-    LOG_DEBUG("Query " << source << " " << opt_path.data())
-    if (prioritize_commandline && options.valueSpecified(opt_path)) {
-      LOG_DEBUG("Take " << source << " arg for " << opt_path.data());
-      return options.getValue(opt_path);
+  if (prioritize_commandline) {
+    if (auto value = env_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take ENV " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
+    }
+    if (auto value = commandline_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take CL " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
     }
-    return std::nullopt;
-  };
-  if (auto value = get_value(*env_options_, "ENV")) {
-    return value;
-  }
-  if (auto value = get_value(*commandline_options_, "CL")) {
-    return value;
   }
   return configuration_options_->getValue(opt_path);
 }
 
 OptionValue TypeARTConfiguration::getValueOr(std::string_view opt_path, OptionValue alt) const {
-  auto get_value = [&](const auto& options, const char* source) -> std::optional<OptionValue> {
-    LOG_DEBUG("Query " << source << " " << opt_path.data())
-    if (prioritize_commandline && options.valueSpecified(opt_path)) {
-      LOG_DEBUG("Take " << source << " arg for " << opt_path.data());
-      return options.getValueOr(opt_path, alt);
+  if (prioritize_commandline) {
+    if (auto value = env_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take ENV " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
+    }
+    if (auto value = commandline_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take CL " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
     }
-    return std::nullopt;
-  };
-  if (auto value = get_value(*env_options_, "ENV")) {
-    return value.value();
-  }
-  if (auto value = get_value(*commandline_options_, "CL")) {
-    return value.value();
   }
   return configuration_options_->getValueOr(opt_path, alt);
 }
 
 OptionValue TypeARTConfiguration::operator[](std::string_view opt_path) const {
-  auto get_value = [&](const auto& options, const char* source) -> std::optional<OptionValue> {
-    LOG_DEBUG("Query " << source << " " << opt_path.data())
-    if (prioritize_commandline && options.valueSpecified(opt_path)) {
-      LOG_DEBUG("Take " << source << " arg for " << opt_path.data());
-      return options.operator[](opt_path);
+  if (prioritize_commandline) {
+    if (auto value = env_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take ENV " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
+    }
+    if (auto value = commandline_options_->getValue(opt_path)) {
+      LOG_DEBUG("Take CL " << opt_path << "=" << (std::string(value.value())));
+      return value.value();
     }
-    return std::nullopt;
-  };
-  if (auto value = get_value(*env_options_, "ENV")) {
-    return value.value();
-  }
-  if (auto value = get_value(*commandline_options_, "CL")) {
-    return value.value();
   }
   auto result = configuration_options_->operator[](opt_path);
-  LOG_DEBUG("Query file " << static_cast<std::string>(result))
+  LOG_DEBUG("Take File " << opt_path << "=" << (std::string(result)));
   return result;
 }
 

lib/passes/TypeARTPass.cpp

@@ -16,6 +16,9 @@
 #include "configuration/Configuration.h"
 #include "configuration/EnvironmentConfiguration.h"
 #include "configuration/FileConfiguration.h"
+#include "configuration/PassBuilderUtil.h"
+#include "configuration/PassConfiguration.h"
+#include "configuration/TypeARTOptions.h"
 #include "instrumentation/MemOpArgCollector.h"
 #include "instrumentation/MemOpInstrumentation.h"
 #include "instrumentation/TypeARTFunctions.h"
@@ -41,6 +44,7 @@
 
 #include <cassert>
 #include <cstddef>
+#include <llvm/Support/Error.h>
 #include <optional>
 #include <sstream>
 #include <string>
@@ -84,6 +88,7 @@ std::optional<std::string> get_configuration_file_path() {
 }
 
 class TypeArtPass : public llvm::PassInfoMixin<TypeArtPass> {
+  std::optional<config::TypeARTConfigOptions> pass_opts{std::nullopt};
   std::unique_ptr<config::Configuration> pass_config;
 
   struct TypeArtFunc {
@@ -109,20 +114,26 @@ class TypeArtPass : public llvm::PassInfoMixin<TypeArtPass> {
   std::unique_ptr<InstrumentationContext> instrumentation_context;
 
  public:
+  TypeArtPass() = default;
+  explicit TypeArtPass(config::TypeARTConfigOptions opts) : pass_opts(opts) {
+    // LOG_INFO("Created with \n" << opts)
+  }
+
   bool doInitialization(Module& m) {
     auto config_file_path = get_configuration_file_path();
 
-    const auto init     = config_file_path.has_value()
-                              ? config::TypeARTConfigInit{config_file_path.value()}
-                              : config::TypeARTConfigInit{{}, config::TypeARTConfigInit::FileConfigurationMode::Empty};
-    auto typeart_config = config::make_typeart_configuration(init);
+    const auto init = config_file_path.has_value()
+                          ? config::TypeARTConfigInit{config_file_path.value()}
+                          : config::TypeARTConfigInit{{}, config::TypeARTConfigInit::FileConfigurationMode::Empty};
+
+    auto typeart_config = [&](const auto& init) {
+      if (init.mode == config::TypeARTConfigInit::FileConfigurationMode::Empty) {
+        return config::make_typeart_configuration_from_opts(pass_opts.value_or(config::TypeARTConfigOptions{}));
+      }
+      return config::make_typeart_configuration(init);
+    }(init);
+
     if (typeart_config) {
-      // {
-      //   std::string typeart_conf_str;
-      //   llvm::raw_string_ostream conf_out_stream{typeart_conf_str};
-      //   typeart_config->get()->emitTypeartFileConfiguration(conf_out_stream);
-      //   LOG_INFO("Emitting TypeART file content\n" << conf_out_stream.str())
-      // }
       LOG_INFO("Emitting TypeART configuration content\n" << typeart_config.get()->getOptions())
       pass_config = std::move(*typeart_config);
     } else {
@@ -214,8 +225,9 @@ class TypeArtPass : public llvm::PassInfoMixin<TypeArtPass> {
     meminst_finder->printStats(out);
 
     const auto get_ta_mode = [&]() {
-      const bool heap  = (*pass_config)[config::ConfigStdArgs::heap];
-      const bool stack = (*pass_config)[config::ConfigStdArgs::stack];
+      const bool heap   = (*pass_config)[config::ConfigStdArgs::heap];
+      const bool stack  = (*pass_config)[config::ConfigStdArgs::stack];
+      const bool global = (*pass_config)[config::ConfigStdArgs::global];
 
       if (heap) {
         if (stack) {
@@ -228,7 +240,13 @@ class TypeArtPass : public llvm::PassInfoMixin<TypeArtPass> {
         return " [Stack]";
       }
 
-      llvm_unreachable("Did not find heap or stack, or combination thereof!");
+      if (global) {
+        return " [Global]";
+      }
+
+      LOG_ERROR("Did not find heap or stack, or combination thereof!");
+      assert((heap || stack || global) && "Needs stack, heap, global or combination thereof");
+      return " [Unknown]";
     };
 
     Table stats("TypeArtPass");
@@ -363,13 +381,21 @@ bool LegacyTypeArtPass::doFinalization(llvm::Module&) {
 // New PM
 //.....................
 llvm::PassPluginLibraryInfo getTypeartPassPluginInfo() {
+  using namespace llvm;
   return {LLVM_PLUGIN_API_VERSION, "TypeART", LLVM_VERSION_STRING, [](PassBuilder& pass_builder) {
             pass_builder.registerPipelineParsingCallback(
                 [](StringRef name, ModulePassManager& module_pm, ArrayRef<PassBuilder::PipelineElement>) {
-                  if (name == "typeart") {
-                    module_pm.addPass(typeart::pass::TypeArtPass());
+                  if (typeart::util::pass::checkParametrizedPassName(name, "typeart")) {
+                    auto parameters = typeart::util::pass::parsePassParameters(
+                        typeart::config::pass::parse_typeart_config, name, "typeart");
+                    if (!parameters) {
+                      LOG_FATAL("Error parsing params: " << parameters.takeError())
+                      return false;
+                    }
+                    module_pm.addPass(typeart::pass::TypeArtPass(parameters.get()));
                     return true;
                   }
+                  LOG_FATAL("Not a valid parametrized pass name: " << name)
                   return false;
                 });
           }};

lib/passes/analysis/MemInstFinder.cpp

@@ -22,6 +22,7 @@
 #include "filter/Filter.h"
 #include "filter/Matcher.h"
 #include "filter/StdForwardFilter.h"
+#include "support/ConfigurationBase.h"
 #include "support/Logger.h"
 #include "support/Table.h"
 #include "support/TypeUtil.h"
@@ -43,6 +44,7 @@
 #include <algorithm>
 #include <cstdlib>
 #include <sstream>
+#include <string>
 #include <utility>
 
 using namespace llvm;
@@ -61,7 +63,7 @@ ALWAYS_ENABLED_STATISTIC(NumCallFilteredGlobals, "Number of filtered globals");
 
 namespace typeart::analysis {
 
-using MemInstFinderConfig = config::TypeARTConfigOptions;
+using MemInstFinderConfig = config::Configuration;
 
 namespace filter {
 class CallFilter {
@@ -85,25 +87,27 @@ namespace filter {
 namespace detail {
 static std::unique_ptr<typeart::filter::Filter> make_filter(const MemInstFinderConfig& config) {
   using namespace typeart::filter;
-  const auto filter_id   = config.filter_config.implementation;
-  const std::string glob = config.filter_config.glob;
+  const bool filter                    = config[config::ConfigStdArgs::filter];
+  const FilterImplementation filter_id = config[config::ConfigStdArgs::filter_impl];
+  const std::string glob               = config[config::ConfigStdArgs::filter_glob];
 
-  if (filter_id == FilterImplementation::none || !config.filter) {
+  if (filter_id == FilterImplementation::none || !filter) {
     LOG_DEBUG("Return no-op filter")
     return std::make_unique<NoOpFilter>();
   } else if (filter_id == FilterImplementation::cg) {
-    if (config.filter_config.cg_file.empty()) {
+    const std::string cg_file = config[config::ConfigStdArgs::filter_cg_file];
+    if (cg_file.empty()) {
       LOG_FATAL("CG File not set!");
       std::exit(1);
     }
-    LOG_DEBUG("Return CG filter with CG file @ " << config.filter_config.cg_file)
-    auto json_cg = JSONCG::getJSON(config.filter_config.cg_file);
+    LOG_DEBUG("Return CG filter with CG file @ " << cg_file)
+    auto json_cg = JSONCG::getJSON(cg_file);
     auto matcher = std::make_unique<DefaultStringMatcher>(util::glob2regex(glob));
     return std::make_unique<CGForwardFilter>(glob, std::move(json_cg), std::move(matcher));
   } else {
     LOG_DEBUG("Return default filter")
     auto matcher         = std::make_unique<DefaultStringMatcher>(util::glob2regex(glob));
-    const auto deep_glob = config.filter_config.glob_deep;
+    const auto deep_glob = config[config::ConfigStdArgs::filter_glob_deep];
     auto deep_matcher    = std::make_unique<DefaultStringMatcher>(util::glob2regex(deep_glob));
     return std::make_unique<StandardForwardFilter>(std::move(matcher), std::move(deep_matcher));
   }
@@ -150,7 +154,7 @@ class MemInstFinderPass : public MemInstFinder {
   MemOpVisitor mOpsCollector;
   filter::CallFilter filter;
   llvm::DenseMap<const llvm::Function*, FunctionData> functionMap;
-  MemInstFinderConfig config;
+  const MemInstFinderConfig& config;
 
  public:
   explicit MemInstFinderPass(const MemInstFinderConfig&);
@@ -167,14 +171,14 @@ class MemInstFinderPass : public MemInstFinder {
 };
 
 MemInstFinderPass::MemInstFinderPass(const MemInstFinderConfig& conf_)
-    : mOpsCollector(conf_.stack, conf_.heap), filter(conf_), config(conf_) {
+    : mOpsCollector(conf_), filter(conf_), config(conf_) {
 }
 
 bool MemInstFinderPass::runOnModule(Module& module) {
   mOpsCollector.collectGlobals(module);
   auto& globals = mOpsCollector.globals;
   NumDetectedGlobals += globals.size();
-  if (config.analysis_config.filter_global) {
+  if (config[config::ConfigStdArgs::analysis_filter_global]) {
     globals.erase(llvm::remove_if(
                       globals,
                       [&](const auto gdata) {  // NOLINT
@@ -295,7 +299,7 @@ bool MemInstFinderPass::runOnFunction(llvm::Function& function) {
 
   NumDetectedAllocs += mOpsCollector.allocas.size();
 
-  if (config.analysis_config.filter_alloca_non_array) {
+  if (config[config::ConfigStdArgs::analysis_filter_alloca_non_array]) {
     auto& allocs = mOpsCollector.allocas;
     allocs.erase(llvm::remove_if(allocs,
                                  [&](const auto& data) {
@@ -308,7 +312,7 @@ bool MemInstFinderPass::runOnFunction(llvm::Function& function) {
                  allocs.end());
   }
 
-  if (config.analysis_config.filter_heap_alloc) {
+  if (config[config::ConfigStdArgs::analysis_filter_heap_alloc]) {
     auto& allocs  = mOpsCollector.allocas;
     auto& mallocs = mOpsCollector.mallocs;
 
@@ -344,7 +348,7 @@ bool MemInstFinderPass::runOnFunction(llvm::Function& function) {
                  allocs.end());
   }
 
-  if (config.analysis_config.filter_pointer_alloc) {
+  if (config[config::ConfigStdArgs::analysis_filter_pointer_alloc]) {
     auto& allocs = mOpsCollector.allocas;
     allocs.erase(llvm::remove_if(allocs,
                                  [&](const auto& data) {
@@ -359,7 +363,7 @@ bool MemInstFinderPass::runOnFunction(llvm::Function& function) {
   }
 
   // if (config.filter.useCallFilter) {
-  if (config.filter) {
+  if (config[config::ConfigStdArgs::filter]) {
     auto& allocs = mOpsCollector.allocas;
     allocs.erase(llvm::remove_if(allocs,
                                  [&](const auto& data) {
@@ -414,7 +418,8 @@ void MemInstFinderPass::printStats(llvm::raw_ostream& out) const {
   Table stats("MemInstFinderPass");
   stats.wrap_header = true;
   stats.wrap_length = true;
-  stats.put(Row::make("Filter string", config.filter_config.glob));
+  std::string glob  = config[config::ConfigStdArgs::filter_glob];
+  stats.put(Row::make("Filter string", glob));
   stats.put(Row::make_row("> Heap Memory"));
   stats.put(Row::make("Heap alloc", NumDetectedHeap.getValue()));
   stats.put(Row::make("Heap call filtered %", call_filter_heap_p));
@@ -449,8 +454,8 @@ const GlobalDataList& MemInstFinderPass::getModuleGlobals() const {
 
 std::unique_ptr<MemInstFinder> create_finder(const config::Configuration& config) {
   LOG_DEBUG("Constructing MemInstFinder")
-  const auto meminst_conf = config::helper::config_to_options(config);
-  return std::make_unique<MemInstFinderPass>(meminst_conf);
+  // const auto meminst_conf = config::helper::config_to_options(config);
+  return std::make_unique<MemInstFinderPass>(config);
 }
 
 }  // namespace typeart::analysis

lib/passes/analysis/MemOpVisitor.cpp

@@ -14,6 +14,8 @@
 
 #include "analysis/MemOpData.h"
 #include "compat/CallSite.h"
+#include "configuration/Configuration.h"
+#include "support/ConfigurationBase.h"
 #include "support/Error.h"
 #include "support/Logger.h"
 #include "support/TypeUtil.h"
@@ -48,6 +50,9 @@ using namespace llvm;
 MemOpVisitor::MemOpVisitor() : MemOpVisitor(true, true) {
 }
 
+MemOpVisitor::MemOpVisitor(const config::Configuration& config)
+    : MemOpVisitor(config[config::ConfigStdArgs::stack], config[config::ConfigStdArgs::heap]) {
+}
 MemOpVisitor::MemOpVisitor(bool collect_allocas, bool collect_heap)
     : collect_allocas(collect_allocas), collect_heap(collect_heap) {
 }

lib/passes/analysis/MemOpVisitor.h

@@ -14,6 +14,7 @@
 #define LIB_MEMOPVISITOR_H_
 
 #include "MemOpData.h"
+#include "configuration/Configuration.h"
 
 #include "llvm/IR/InstVisitor.h"
 
@@ -41,6 +42,7 @@ struct MemOpVisitor : public llvm::InstVisitor<MemOpVisitor> {
 
  public:
   MemOpVisitor();
+  explicit MemOpVisitor(const config::Configuration& config);
   MemOpVisitor(bool collect_allocas, bool collect_heap);
   void collect(llvm::Function& function);
   void collectGlobals(llvm::Module& module);