From 23a543971e6f0549e158e66a08f3631c2ccb617e Mon Sep 17 00:00:00 2001
From: Dan Gohman <dan433584@gmail.com>
Date: Fri, 10 Feb 2017 00:02:58 +0000
Subject: [PATCH] [Support] Extend SLEB128 encoding support.

Add support for padded SLEB128 values, and support for writing SLEB128
values to buffers rather than to ostreams, similar to the existing
ULEB128 support.

llvm-svn: 294675
---
 llvm/include/llvm/Support/LEB128.h    | 40 +++++++++++++++++++++--
 llvm/unittests/Support/LEB128Test.cpp | 47 +++++++++++++++++++--------
 2 files changed, 71 insertions(+), 16 deletions(-)

diff --git a/llvm/include/llvm/Support/LEB128.h b/llvm/include/llvm/Support/LEB128.h
index 6a95432ca2d9..edb923bb6599 100644
--- a/llvm/include/llvm/Support/LEB128.h
+++ b/llvm/include/llvm/Support/LEB128.h
@@ -20,7 +20,8 @@
 namespace llvm {
 
 /// Utility function to encode a SLEB128 value to an output stream.
-inline void encodeSLEB128(int64_t Value, raw_ostream &OS) {
+inline void encodeSLEB128(int64_t Value, raw_ostream &OS,
+                          unsigned Padding = 0) {
   bool More;
   do {
     uint8_t Byte = Value & 0x7f;
@@ -28,10 +29,45 @@ inline void encodeSLEB128(int64_t Value, raw_ostream &OS) {
     Value >>= 7;
     More = !((((Value == 0 ) && ((Byte & 0x40) == 0)) ||
               ((Value == -1) && ((Byte & 0x40) != 0))));
-    if (More)
+    if (More || Padding != 0)
       Byte |= 0x80; // Mark this byte to show that more bytes will follow.
     OS << char(Byte);
   } while (More);
+
+  // Pad with 0x80 and emit a terminating byte at the end.
+  if (Padding != 0) {
+    uint8_t PadValue = Value < 0 ? 0x7f : 0x00;
+    for (; Padding != 1; --Padding)
+      OS << char(PadValue | 0x80);
+    OS << char(PadValue);
+  }
+}
+
+/// Utility function to encode a SLEB128 value to a buffer. Returns
+/// the length in bytes of the encoded value.
+inline unsigned encodeSLEB128(int64_t Value, uint8_t *p,
+                              unsigned Padding = 0) {
+  uint8_t *orig_p = p;
+  bool More;
+  do {
+    uint8_t Byte = Value & 0x7f;
+    // NOTE: this assumes that this signed shift is an arithmetic right shift.
+    Value >>= 7;
+    More = !((((Value == 0 ) && ((Byte & 0x40) == 0)) ||
+              ((Value == -1) && ((Byte & 0x40) != 0))));
+    if (More || Padding != 0)
+      Byte |= 0x80; // Mark this byte to show that more bytes will follow.
+    *p++ = Byte;
+  } while (More);
+
+  // Pad with 0x80 and emit a terminating byte at the end.
+  if (Padding != 0) {
+    uint8_t PadValue = Value < 0 ? 0x7f : 0x00;
+    for (; Padding != 1; --Padding)
+      *p++ = (PadValue | 0x80);
+    *p++ = PadValue;
+  }
+  return (unsigned)(p - orig_p);
 }
 
 /// Utility function to encode a ULEB128 value to an output stream.
diff --git a/llvm/unittests/Support/LEB128Test.cpp b/llvm/unittests/Support/LEB128Test.cpp
index 76b63e5a8381..061936df1d19 100644
--- a/llvm/unittests/Support/LEB128Test.cpp
+++ b/llvm/unittests/Support/LEB128Test.cpp
@@ -17,26 +17,45 @@ using namespace llvm;
 namespace {
 
 TEST(LEB128Test, EncodeSLEB128) {
-#define EXPECT_SLEB128_EQ(EXPECTED, VALUE) \
+#define EXPECT_SLEB128_EQ(EXPECTED, VALUE, PAD) \
   do { \
-    /* encodeSLEB128(uint64_t, raw_ostream &) */ \
     std::string Expected(EXPECTED, sizeof(EXPECTED) - 1); \
-    std::string Actual; \
-    raw_string_ostream Stream(Actual); \
-    encodeSLEB128(VALUE, Stream); \
+    \
+    /* encodeSLEB128(uint64_t, raw_ostream &, unsigned) */ \
+    std::string Actual1; \
+    raw_string_ostream Stream(Actual1); \
+    encodeSLEB128(VALUE, Stream, PAD); \
     Stream.flush(); \
-    EXPECT_EQ(Expected, Actual); \
+    EXPECT_EQ(Expected, Actual1); \
+    \
+    /* encodeSLEB128(uint64_t, uint8_t *, unsigned) */ \
+    uint8_t Buffer[32]; \
+    unsigned Size = encodeSLEB128(VALUE, Buffer, PAD); \
+    std::string Actual2(reinterpret_cast<const char *>(Buffer), Size); \
+    EXPECT_EQ(Expected, Actual2); \
   } while (0)
 
   // Encode SLEB128
-  EXPECT_SLEB128_EQ("\x00", 0);
-  EXPECT_SLEB128_EQ("\x01", 1);
-  EXPECT_SLEB128_EQ("\x7f", -1);
-  EXPECT_SLEB128_EQ("\x3f", 63);
-  EXPECT_SLEB128_EQ("\x41", -63);
-  EXPECT_SLEB128_EQ("\x40", -64);
-  EXPECT_SLEB128_EQ("\xbf\x7f", -65);
-  EXPECT_SLEB128_EQ("\xc0\x00", 64);
+  EXPECT_SLEB128_EQ("\x00", 0, 0);
+  EXPECT_SLEB128_EQ("\x01", 1, 0);
+  EXPECT_SLEB128_EQ("\x7f", -1, 0);
+  EXPECT_SLEB128_EQ("\x3f", 63, 0);
+  EXPECT_SLEB128_EQ("\x41", -63, 0);
+  EXPECT_SLEB128_EQ("\x40", -64, 0);
+  EXPECT_SLEB128_EQ("\xbf\x7f", -65, 0);
+  EXPECT_SLEB128_EQ("\xc0\x00", 64, 0);
+
+  // Encode SLEB128 with some extra padding bytes
+  EXPECT_SLEB128_EQ("\x80\x00", 0, 1);
+  EXPECT_SLEB128_EQ("\x80\x80\x00", 0, 2);
+  EXPECT_SLEB128_EQ("\xff\x80\x00", 0x7f, 1);
+  EXPECT_SLEB128_EQ("\xff\x80\x80\x00", 0x7f, 2);
+  EXPECT_SLEB128_EQ("\x80\x81\x00", 0x80, 1);
+  EXPECT_SLEB128_EQ("\x80\x81\x80\x00", 0x80, 2);
+  EXPECT_SLEB128_EQ("\xc0\x7f", -0x40, 1);
+  EXPECT_SLEB128_EQ("\xc0\xff\x7f", -0x40, 2);
+  EXPECT_SLEB128_EQ("\x80\xff\x7f", -0x80, 1);
+  EXPECT_SLEB128_EQ("\x80\xff\xff\x7f", -0x80, 2);
 
 #undef EXPECT_SLEB128_EQ
 }