From a23bab96cb9eb7a869eb260336e5837e9c63d69b Mon Sep 17 00:00:00 2001
From: Mark Wooding <mdw@distorted.org.uk>
Date: Sat, 26 May 2018 23:29:09 +0100
Subject: [PATCH] utils/bits.h: Support compiler magic for unaligned loads and
 stores.

Introduce `RAWw' macros to access the raw memory, and implement `LOADwe'
and `STOREwe' in terms of these and `eTOHw'/`HTOew'.

These are remarkably tricky because GCC (I think mistakenly) thinks that
type-based aliasing is applicable when in fact the other type in
question is `unsigned char', which is known to be able to alias
anything.  Hit things with the `may_alias' hammer and hope they quieten
down.
---
 utils/bits.3 | 21 +++++++++++++++++
 utils/bits.h | 77 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 98 insertions(+)

diff --git a/utils/bits.3 b/utils/bits.3
index 7b1c715..a002e24 100644
--- a/utils/bits.3
+++ b/utils/bits.3
@@ -112,6 +112,11 @@ bits \- portable bit manipulation macros
 .\" HTOB64
 .\" HTOL64
 .\"
+.\" RAW8
+.\" RAW16
+.\" RAW32
+.\" RAW64
+.\"
 .\" @GETBYTE
 .\" @PUTBYTE
 .\"
@@ -485,6 +490,22 @@ These macros always operate on byte offsets regardless of the type of
 the pointer
 .IR p .
 .PP
+For each size suffix
+.IR w ,
+there may be a macro such that the invocation
+.BI RAW w ( p )
+is an lvalue designating the
+.IR w /8
+octets starting at address
+.IR p ,
+interpreted according to the environment's preferred representation,
+except that
+.I p
+need not be aligned in any particular fashion.  There are many reasons
+why this might not be possible; programmers are not normally expected to
+use these macros directly, and they are documented in case they are
+useful for special effects.
+.PP
 For each size-and-endian suffix
 .IR we ,
 the macro invocation
diff --git a/utils/bits.h b/utils/bits.h
index 3b12f96..656eee5 100644
--- a/utils/bits.h
+++ b/utils/bits.h
@@ -40,6 +40,10 @@
 #  include <stdint.h>
 #endif
 
+#ifndef MLIB_COMPILER_H
+#  include "compiler.h"
+#endif
+
 /*----- Decide on some types ----------------------------------------------*/
 
 /* --- Make GNU C shut up --- */
@@ -316,6 +320,16 @@ typedef unsigned char octet, uint8;
 
 /* --- Endianness swapping --- */
 
+#if GCC_VERSION_P(4, 8)
+#  define ENDSWAP16(x) ((uint16)__builtin_bswap16(x))
+#endif
+#if GCC_VERSION_P(4, 3)
+#  define ENDSWAP32(x) ((uint32)__builtin_bswap32(x))
+#endif
+#if GCC_VERSION_P(4, 3) && defined(HAVE_UINT64)
+#  define ENDSWAP64(x) ((uint64)__builtin_bswap64(x))
+#endif
+
 #ifndef ENDSWAP8
 #  define ENDSWAP8(x) U8(x)
 #endif
@@ -411,8 +425,71 @@ typedef unsigned char octet, uint8;
 #  define BTOH64_(z, x) ASSIGN64(z, x)
 #endif
 
+/* --- Unaligned access (GCC-specific) --- */
+
+#if GCC_VERSION_P(3, 3) && CHAR_BIT == 8
+#  define MLIB_MISALIGNED __attribute__((aligned(1), may_alias))
+#  if __SIZEOF_SHORT__ == 2
+     typedef MLIB_MISALIGNED unsigned short misaligned_uint16;
+#    define RAW16(p) (*(misaligned_uint16 *)(p))
+#  endif
+#  if __SIZEOF_INT__ == 4
+     typedef MLIB_MISALIGNED unsigned int misaligned_uint32;
+#    define RAW32(p) (*(misaligned_uint32 *)(p))
+#  elif __SIZEOF_LONG__ == 4
+     typedef MLIB_MISALIGNED unsigned long misaligned_uint32;
+#    define RAW32(p) (*(misaligned_uint32 *)(p))
+#  endif
+#  if __SIZEOF_LONG__ == 8
+     typedef MLIB_MISALIGNED unsigned long misaligned_uint64;
+#    define RAW64(p) (*(misaligned_uint64 *)(p))
+#  elif __SIZEOF_LONG_LONG__ == 8
+     typedef MLIB_MISALIGNED unsigned long long misaligned_uint64;
+#    define RAW64(p) (*(misaligned_uint64 *)(p))
+#  endif
+#endif
+
 /* --- Storage and retrieval --- */
 
+#if defined(RAW16) && defined(LTOH16)
+#  define LOAD16_L(p) LTOH16(RAW16(p))
+#endif
+#if defined(RAW16) && defined(HTOL16)
+#  define STORE16_L(p, x) (RAW16(p) = HTOL16(x))
+#endif
+#if defined(RAW16) && defined(BTOH16)
+#  define LOAD16_B(p) BTOH16(RAW16(p))
+#endif
+#if defined(RAW16) && defined(HTOB16)
+#  define STORE16_B(p, x) (RAW16(p) = HTOB16(x))
+#endif
+
+#if defined(RAW32) && defined(LTOH32)
+#  define LOAD32_L(p) LTOH32(RAW32(p))
+#endif
+#if defined(RAW32) && defined(HTOL32)
+#  define STORE32_L(p, x) (RAW32(p) = HTOL32(x))
+#endif
+#if defined(RAW32) && defined(BTOH32)
+#  define LOAD32_B(p) BTOH32(RAW32(p))
+#endif
+#if defined(RAW32) && defined(HTOB32)
+#  define STORE32_B(p, x) (RAW32(p) = HTOB32(x))
+#endif
+
+#if defined(RAW64) && defined(LTOH64)
+#  define LOAD64_L(p) LTOH64(RAW64(p))
+#endif
+#if defined(RAW64) && defined(HTOL64)
+#  define STORE64_L(p, x) (RAW64(p) = HTOL64(x))
+#endif
+#if defined(RAW64) && defined(BTOH64)
+#  define LOAD64_B(p) BTOH64(RAW64(p))
+#endif
+#if defined(RAW64) && defined(HTOB64)
+#  define STORE64_B(p, x) (RAW64(p) = HTOB64(x))
+#endif
+
 #define GETBYTE(p, o) (((octet *)(p))[o] & MASK8)
 #define PUTBYTE(p, o, v) (((octet *)(p))[o] = U8((v)))
 
-- 
2.11.0