[PATCH] D97770: [clangd] Make categorical features 64 bit in DecisionForest Model.

[Utkarsh Saxena via Phabricator via cfe-commits]

usaxena95 created this revision.
Herald added subscribers: kadircet, arphaman.
usaxena95 requested review of this revision.
Herald added subscribers: cfe-commits, MaskRay, ilya-biryukov.
Herald added a project: clang.

CodeCompletionContext::Kind has 36 Kinds. The completion model used to
support categorical features of 32 cardinality.
Due to this clangd tests were failing asan tests due to overflow.

This patch makes the completion model support 64 cardinality of
categorical features by storing ENUM Features as uint64_t instead of
uint32_t.

Verified that this fixes the asan failures.

Latency: 6.7ms (old) VS 6.8ms (new) per 1000 predictions.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D97770

Files:
  clang-tools-extra/clangd/benchmarks/CompletionModel/DecisionForestBenchmark.cpp
  clang-tools-extra/clangd/quality/CompletionModelCodegen.py


Index: clang-tools-extra/clangd/quality/CompletionModelCodegen.py
===================================================================
--- clang-tools-extra/clangd/quality/CompletionModelCodegen.py
+++ clang-tools-extra/clangd/quality/CompletionModelCodegen.py
@@ -131,18 +131,19 @@
                     feature, feature))
         elif f["kind"] == "ENUM":
             setters.append(
-                "void set%s(unsigned V) { %s = 1 << V; }" % (feature, feature))
+                "void set%s(unsigned V) { %s = 1LL << V; }" % (feature, feature))
         else:
             raise ValueError("Unhandled feature type.", f["kind"])
 
     # Class members represent all the features of the Example.
     class_members = [
-        "uint32_t %s = 0;" % f['name']
+        "uint%d_t %s = 0;"
+        % (64 if f["kind"] == "ENUM" else 32, f['name'])
         for f in features_json
     ]
     getters = [
-        "LLVM_ATTRIBUTE_ALWAYS_INLINE uint32_t get%s() const { return %s; }"
-        % (f['name'], f['name'])
+        "LLVM_ATTRIBUTE_ALWAYS_INLINE uint%d_t get%s() const { return %s; }"
+        % (64 if f["kind"] == "ENUM" else 32, f['name'], f['name'])
         for f in features_json
     ]
     nline = "\n  "
@@ -245,7 +246,7 @@
 
 %s
 
-#define BIT(X) (1 << X)
+#define BIT(X) (1LL << X)
 
 %s
 
Index: clang-tools-extra/clangd/benchmarks/CompletionModel/DecisionForestBenchmark.cpp
===================================================================
--- clang-tools-extra/clangd/benchmarks/CompletionModel/DecisionForestBenchmark.cpp
+++ clang-tools-extra/clangd/benchmarks/CompletionModel/DecisionForestBenchmark.cpp
@@ -51,7 +51,7 @@
                       : RandInt(20));
     E.setSemaSaysInScope(FlipCoin(0.5));      // Boolean.
     E.setScope(RandInt(4));                   // 4 Scopes.
-    E.setContextKind(RandInt(32));            // 32 Context kinds.
+    E.setContextKind(RandInt(36));            // 36 Context kinds.
     E.setIsInstanceMember(FlipCoin(0.5));     // Boolean.
     E.setHadContextType(FlipCoin(0.6));       // Boolean.
     E.setHadSymbolType(FlipCoin(0.6));        // Boolean.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D97770.327443.patch
Type: text/x-patch
Size: 2122 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20210302/1d56c356/attachment.bin>