From f4b41a7e8010aad642abc927d7074e9c5f01fb1e Mon Sep 17 00:00:00 2001
From: Alexis Engelke <alexis.engelke@googlemail.com>
Date: Wed, 17 Jun 2020 16:42:07 +0200
Subject: [PATCH] decode: Use uint16_t for trie

---
 decode.c       |  8 ++++----
 parseinstrs.py | 15 ++++++++-------
 2 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/decode.c b/decode.c
index 070aad6..643547c 100644
--- a/decode.c
+++ b/decode.c
@@ -14,7 +14,7 @@
 #define UNLIKELY(x) __builtin_expect((x), 0)
 
 #define FD_DECODE_TABLE_DATA
-static __attribute__((aligned(16))) const uint8_t _decode_table[] = {
+static __attribute__((aligned(16))) const uint16_t _decode_table[] = {
 #include <fadec-decode-table.inc>
 };
 #undef FD_DECODE_TABLE_DATA
@@ -44,7 +44,7 @@ typedef enum DecodeMode DecodeMode;
 
 #define ENTRY_UNPACK(table,kind,entry) do { \
             uint16_t entry_copy = entry; \
-            table = &((uint16_t*) _decode_table)[(entry_copy & ~7) >> 1]; \
+            table = &_decode_table[(entry_copy & ~7) >> 1]; \
             kind = entry_copy & ENTRY_MASK; \
         } while (0)
 
@@ -323,11 +323,11 @@ fd_decode(const uint8_t* buffer, size_t len_sz, int mode_int, uintptr_t address,
     // Ensure that we can actually handle the decode request
 #if defined(ARCH_386)
     if (mode == DECODE_32)
-        table = &((uint16_t*) _decode_table)[FD_TABLE_OFFSET_32 >> 1];
+        table = &_decode_table[FD_TABLE_OFFSET_32];
 #endif
 #if defined(ARCH_X86_64)
     if (mode == DECODE_64)
-        table = &((uint16_t*) _decode_table)[FD_TABLE_OFFSET_64 >> 1];
+        table = &_decode_table[FD_TABLE_OFFSET_64];
 #endif
 
     if (UNLIKELY(table == NULL))
diff --git a/parseinstrs.py b/parseinstrs.py
index 4c955a7..84bf0ae 100644
--- a/parseinstrs.py
+++ b/parseinstrs.py
@@ -336,12 +336,12 @@ class Table:
         for name, entry in self.data.items():
             self.annotations[current] = "%s(%d)" % (name, entry.kind.value)
             self.offsets[name] = current
-            current += (2*entry.encode_length + 7) & ~7
-        if current >= 0x10000:
+            current += (entry.encode_length + 3) & ~3
+        if current >= 0x8000:
             raise Exception("maximum table size exceeded: {:x}".format(current))
 
     def encode_item(self, name):
-        return self.offsets[name] | self.data[name].kind.value
+        return (self.offsets[name] << 1) | self.data[name].kind.value
 
     def compile(self):
         self.calc_offsets()
@@ -349,7 +349,7 @@ class Table:
 
         data = ()
         for off, entry in ordered:
-            data += (0,) * (off//2 - len(data)) + entry.encode(self.encode_item)
+            data += (0,) * (off - len(data)) + entry.encode(self.encode_item)
 
         data = struct.pack("<%dH"%len(data), *data)
 
@@ -358,12 +358,13 @@ class Table:
         return data, self.annotations, [self.offsets[k] for k in self.roots]
 
 def wrap(string):
-    return "\n".join(string[i:i+80] for i in range(0, len(string), 80))
+    return "\n".join(string[i:i+56] for i in range(0, len(string), 56))
 
 def bytes_to_table(data, notes):
-    hexdata = ",".join("0x{:02x}".format(byte) for byte in data)
+    hexdata = ",".join("0x{:04x}".format(int.from_bytes(data[i:i+2], "little"))
+                       for i in range(0, len(data), 2))
     offs = [0] + sorted(notes.keys()) + [len(data)]
-    return "\n".join(wrap(hexdata[p*5:c*5]) + "\n//%04x "%c + notes.get(c, "")
+    return "\n".join(wrap(hexdata[p*7:c*7]) + "\n//%04x "%c + notes.get(c, "")
                      for p, c in zip(offs, offs[1:]))
 
 template = """// Auto-generated file -- do not modify!