[llvm] [ctx_profile] Profile reader and writer (PR #91859)

Snehasish Kumar via llvm-commits llvm-commits at lists.llvm.org
Tue May 14 14:33:16 PDT 2024


================
@@ -0,0 +1,49 @@
+//===- PGOCtxProfWriter.cpp - Contextual Instrumentation profile writer ---===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Write a contextual profile to bitstream.
+//
+//===----------------------------------------------------------------------===//
+
+#include "llvm/ProfileData/PGOCtxProfWriter.h"
+#include "llvm/Bitstream/BitCodeEnums.h"
+
+using namespace llvm;
+using namespace llvm::ctx_profile;
+
+void PGOCtxProfileWriter::writeCounters(const ContextNode &Node) {
+  Writer.EmitCode(bitc::UNABBREV_RECORD);
+  Writer.EmitVBR(PGOCtxProfileRecords::Counters, VBREncodingBits);
+  Writer.EmitVBR(Node.counters_size(), VBREncodingBits);
+  for (auto I = 0U; I < Node.counters_size(); ++I)
+    Writer.EmitVBR64(Node.counters()[I], VBREncodingBits);
+}
+
+// recursively write all the subcontexts. We do need to traverse depth first to
+// model the context->subcontext implicitly, and since this captures call
+// stacks, we don't really need to be worried about stack overflow and we can
+// keep the implementation simple.
+void PGOCtxProfileWriter::writeImpl(std::optional<uint32_t> CallerIndex,
+                                    const ContextNode &Node) {
+  Writer.EnterSubblock(PGOCtxProfileBlockIDs::ContextNodeBlockID, CodeLen);
+  Writer.EmitRecord(PGOCtxProfileRecords::Guid,
+                    SmallVector<uint64_t, 1>{Node.guid()});
+  if (CallerIndex)
+    Writer.EmitRecord(PGOCtxProfileRecords::CalleeIndex,
+                      SmallVector<uint64_t, 1>{*CallerIndex});
+  writeCounters(Node);
+  for (auto I = 0U; I < Node.callsites_size(); ++I)
----------------
snehasish wrote:

nit: spell out the type as uint32_t instead of relying on the reader to notice `0U`.

https://github.com/llvm/llvm-project/pull/91859


More information about the llvm-commits mailing list