* README: additional notes

* configure.ac: look for mmap()
* examples/md5/md5sum.c: (main): fix for same
* liboil/copy/trans8x8_i386.c: (trans8x8_u16_mmx): Need to be
careful about using %ebx
* liboil/dct/idct8x8_i386.c: (fdct8x8s_s16_mmx): same
* liboil/simdpack/abs_i386.c: (abs_u16_s16_mmx): clobber memory

7 files changed, 71 insertions, 38 deletions

diff --git a/ChangeLog b/ChangeLog
index ac8803d..23152b5 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,13 @@
+2005-01-11  David Schleef  <ds@schleef.org>
+
+	* README: additional notes
+	* configure.ac: look for mmap()
+	* examples/md5/md5sum.c: (main): fix for same
+	* liboil/copy/trans8x8_i386.c: (trans8x8_u16_mmx): Need to be
+	careful about using %ebx
+	* liboil/dct/idct8x8_i386.c: (fdct8x8s_s16_mmx): same
+	* liboil/simdpack/abs_i386.c: (abs_u16_s16_mmx): clobber memory
+
 2005-01-09  David Schleef  <ds@schleef.org>
 
 	* autogen.sh:
diff --git a/README b/README
index 12cfe35..eb7445f 100644
--- a/README
+++ b/README
@@ -1,12 +1,34 @@
 
+
+Requirements
+============
+
+Liboil has no required dependecies (other than a decent C compiler
+and standard C library).  A few extra tests and examples are enabled
+if you have glib-2.0 installed, but these extras are neither required
+nor installed.
+
+Using GCC is strongly recommended, since the most interesting bits
+of liboil are written in GCC-style inline assembly.  GCC versions
+prior to 3.2 are not supported.
+
+
+Problems Compiling?
+===================
+
+Function implementations in liboil are designed to be modular, so
+if your compiler is having difficulty compiling a particular function,
+simply disable it and the corresponding OIL_DEFINE_IMPL() line.
+
+
 ABI warning
 ===========
 
 In general, liboil is not ABI stable.  However, portions of liboil
-are guaranteed to be stable through the 0.3.x series, and using
+are guaranteed to be stable through the 0.3.x series, and by using
 a compatibility library, through the 0.4.x series as well.  This
-policy is designed to provide almost all applications with ABI
-stability for the symbols most used in liboil.
+policy is designed to provide almost all applications the necessary
+ABI stability for the symbols likely to be used in liboil.
 
 Any symbols declared by including <liboil/liboil.h> follow this
 ABI policy.  This includes all liboil function classes as well
@@ -39,9 +61,9 @@ correctly.  Then, at a convenient time, libB and libC can
 independently be recompiled using liboil-0.4.x, and the packages
 will no longer depend on liboil-0.3.so.
 
-The traditional way of dealing with these changes is to
-migrate all packages to the new library as quickly as possible.
-This is unduly cumbersome.
+The traditional way of dealing with these changes is to migrate
+all packages to the new library as quickly as possible.  This is
+unduly cumbersome.
 
 
 
diff --git a/configure.ac b/configure.ac
index 112d29b..ebdfc8a 100644
--- a/configure.ac
+++ b/configure.ac
@@ -101,6 +101,7 @@ AM_CONDITIONAL(HAVE_CPU_IA64, test "x$HAVE_CPU_IA64" = "xyes")
 
 AC_C_BIGENDIAN
 
+AC_FUNC_MMAP()
 AC_CHECK_LIB(m, rintf,
   AC_DEFINE(HAVE_RINTF, 1, [Define if rintf() is available]))
 AC_CHECK_LIB(m, lrint,
diff --git a/examples/md5/md5sum.c b/examples/md5/md5sum.c
index db37330..36a5132 100644
--- a/examples/md5/md5sum.c
+++ b/examples/md5/md5sum.c
@@ -25,8 +25,6 @@
  * POSSIBILITY OF SUCH DAMAGE.
  */
 
-#define HAVE_MMAN_H
-
 #ifdef HAVE_CONFIG_H
 #include "config.h"
 #endif
@@ -38,7 +36,7 @@
 #include <stdlib.h>
 
 #include <sys/stat.h>
-#ifdef HAVE_MMAN_H
+#ifdef HAVE_MMAP
 #include <sys/mman.h>
 #endif
 #include <errno.h>
@@ -151,7 +149,7 @@ int main(int argc, char *argv[])
   ret = fstat (fd, &st);
 
   n_bytes = st.st_size;
-#ifdef HAVE_MMAN_H
+#ifdef HAVE_MMAP
   while (1) {
     ptr = mmap (NULL, n_bytes, PROT_READ, MAP_SHARED | MAP_POPULATE, fd, 0);
     getpid();
diff --git a/liboil/copy/trans8x8_i386.c b/liboil/copy/trans8x8_i386.c
index d6f299a..4ee5004 100644
--- a/liboil/copy/trans8x8_i386.c
+++ b/liboil/copy/trans8x8_i386.c
@@ -38,99 +38,99 @@ static void
 trans8x8_u16_mmx (uint16_t *dest, int dstr, uint16_t *src, int sstr)
 {
   asm volatile (
-      "  leal (%3,%3,2),%%ebx \n"         // UBER 0:
+      "  leal (%3,%3,2),%%eax \n"         // UBER 0:
       "  movq (%1), %%mm0 \n"             // UBER 1:
       "  movq (%1,%3,2), %%mm2 \n"        // UBER 2: 
       "  movq %%mm0, %%mm4 \n"            // UBER 3: 1
       "  movq %%mm2, %%mm5 \n"            // UBER 4: 2
       "  punpcklwd (%1,%3), %%mm0 \n"     // UBER 5: 1
-      "  punpcklwd (%1,%%ebx), %%mm2 \n"  // UBER 6: 0 2
+      "  punpcklwd (%1,%%eax), %%mm2 \n"  // UBER 6: 0 2
       "  punpckhwd (%1,%3), %%mm4 \n"     // UBER 7: 3
-      "  punpckhwd (%1,%%ebx), %%mm5 \n"  // UBER 8: 4
+      "  punpckhwd (%1,%%eax), %%mm5 \n"  // UBER 8: 4
       "  movq %%mm0, %%mm1 \n"            // UBER 9: 5
       "  movq %%mm4, %%mm3 \n"            // UBER 10: 7
       "  punpckldq %%mm2, %%mm0 \n"       // UBER 11: 5 6
       "  punpckldq %%mm5, %%mm4 \n"       // UBER 12: 7 8
       "  punpckhdq %%mm2, %%mm1 \n"       // UBER 13: 6 9
       "  punpckhdq %%mm5, %%mm3 \n"       // UBER 14: 9 10
-      "  leal (%2,%2,2),%%ebx \n"         // UBER 15: 8
+      "  leal (%2,%2,2),%%eax \n"         // UBER 15: 8
       "  movq %%mm0, 0(%0) \n"            // UBER 16: 11
       "  movq %%mm1, (%0,%2) \n"          // UBER 17: 13
       "  movq %%mm4, (%0,%2,2) \n"        // UBER 18: 12
-      "  movq %%mm3, (%0,%%ebx) \n"       // UBER 19: 14 15
+      "  movq %%mm3, (%0,%%eax) \n"       // UBER 19: 14 15
 
-      "  leal (%3,%3,2),%%ebx \n"
+      "  leal (%3,%3,2),%%eax \n"
       "  movq 8(%1), %%mm0 \n"
       "  movq 8(%1,%3,2), %%mm2 \n"
       "  movq %%mm0, %%mm4 \n"
       "  movq %%mm2, %%mm5 \n"
       "  punpcklwd 8(%1,%3), %%mm0 \n"
-      "  punpcklwd 8(%1,%%ebx), %%mm2 \n"
+      "  punpcklwd 8(%1,%%eax), %%mm2 \n"
       "  punpckhwd 8(%1,%3), %%mm4 \n"
-      "  punpckhwd 8(%1,%%ebx), %%mm5 \n"
+      "  punpckhwd 8(%1,%%eax), %%mm5 \n"
       "  movq %%mm0, %%mm1 \n"
       "  movq %%mm4, %%mm3 \n"
       "  punpckldq %%mm2, %%mm0 \n"
       "  punpckldq %%mm5, %%mm4 \n"
       "  punpckhdq %%mm2, %%mm1 \n"
       "  punpckhdq %%mm5, %%mm3 \n"
-      "  leal (%2,%2,2),%%ebx \n"
+      "  leal (%2,%2,2),%%eax \n"
       "  leal (%0,%2,4),%0 \n"
       "  movq %%mm0, 0(%0) \n"
       "  movq %%mm1, (%0,%2) \n"
       "  movq %%mm4, (%0,%2,2) \n"
-      "  movq %%mm3, (%0,%%ebx) \n"
+      "  movq %%mm3, (%0,%%eax) \n"
 
       "  leal (%1,%3,4),%1 \n"
-      "  leal (%3,%3,2),%%ebx \n"
+      "  leal (%3,%3,2),%%eax \n"
       "  movq 0(%1), %%mm0 \n"
       "  movq 0(%1,%3,2), %%mm2 \n"
       "  movq %%mm0, %%mm4 \n"
       "  movq %%mm2, %%mm5 \n"
       "  punpcklwd 0(%1,%3), %%mm0 \n"
-      "  punpcklwd 0(%1,%%ebx), %%mm2 \n"
+      "  punpcklwd 0(%1,%%eax), %%mm2 \n"
       "  punpckhwd 0(%1,%3), %%mm4 \n"
-      "  punpckhwd 0(%1,%%ebx), %%mm5 \n"
+      "  punpckhwd 0(%1,%%eax), %%mm5 \n"
       "  movq %%mm0, %%mm1 \n"
       "  movq %%mm4, %%mm3 \n"
       "  punpckldq %%mm2, %%mm0 \n"
       "  punpckldq %%mm5, %%mm4 \n"
       "  punpckhdq %%mm2, %%mm1 \n"
       "  punpckhdq %%mm5, %%mm3 \n"
-      "  leal (%2,%2,2),%%ebx \n"
+      "  leal (%2,%2,2),%%eax \n"
       "  neg %2 \n"
       "  leal (%0,%2,4),%0 \n"
       "  neg %2 \n"
       "  movq %%mm0, 8(%0) \n"
       "  movq %%mm1, 8(%0,%2) \n"
       "  movq %%mm4, 8(%0,%2,2) \n"
-      "  movq %%mm3, 8(%0,%%ebx) \n"
+      "  movq %%mm3, 8(%0,%%eax) \n"
 
-      "  leal (%3,%3,2),%%ebx \n"
+      "  leal (%3,%3,2),%%eax \n"
       "  movq 8(%1), %%mm0 \n"
       "  movq 8(%1,%3,2), %%mm2 \n"
       "  movq %%mm0, %%mm4 \n"
       "  movq %%mm2, %%mm5 \n"
       "  punpcklwd 8(%1,%3), %%mm0 \n"
-      "  punpcklwd 8(%1,%%ebx), %%mm2 \n"
+      "  punpcklwd 8(%1,%%eax), %%mm2 \n"
       "  punpckhwd 8(%1,%3), %%mm4 \n"
-      "  punpckhwd 8(%1,%%ebx), %%mm5 \n"
+      "  punpckhwd 8(%1,%%eax), %%mm5 \n"
       "  movq %%mm0, %%mm1 \n"
       "  movq %%mm4, %%mm3 \n"
       "  punpckldq %%mm2, %%mm0 \n"
       "  punpckldq %%mm5, %%mm4 \n"
       "  punpckhdq %%mm2, %%mm1 \n"
       "  punpckhdq %%mm5, %%mm3 \n"
-      "  leal (%2,%2,2),%%ebx \n"
+      "  leal (%2,%2,2),%%eax \n"
       "  leal (%0,%2,4),%0 \n"
       "  movq %%mm0, 8(%0) \n"
       "  movq %%mm1, 8(%0,%2) \n"
       "  movq %%mm4, 8(%0,%2,2) \n"
-      "  movq %%mm3, 8(%0,%%ebx) \n"
+      "  movq %%mm3, 8(%0,%%eax) \n"
       "  emms \n"
       : "+r" (dest), "+r" (src), "+r" (dstr), "+r" (sstr)
       :
-      : "ebx");
+      : "eax");
 
 }
 OIL_DEFINE_IMPL (trans8x8_u16_mmx, trans8x8_u16);
diff --git a/liboil/dct/idct8x8_i386.c b/liboil/dct/idct8x8_i386.c
index ca16d9d..0f68c72 100644
--- a/liboil/dct/idct8x8_i386.c
+++ b/liboil/dct/idct8x8_i386.c
@@ -96,6 +96,8 @@ fdct8x8s_s16_mmx (uint16_t *dest, int dstr, uint16_t *src, int sstr)
   int32_t tmp[32];
 
   asm volatile (
+      /* Note: this asm is unclean with %ebx, but it's not an issue
+       * in this particular case. */
 #if 1
       /* first half */
       "  movl %1, %%eax \n" // src
@@ -401,7 +403,7 @@ fdct8x8s_s16_mmx (uint16_t *dest, int dstr, uint16_t *src, int sstr)
       "  emms \n"
       :
       : "m" (dest), "m" (src), "m" (dstr), "m" (sstr), "r" (tmp), "r" (dct_mmx_constants)
-      : "eax", "ebx", "ecx", "edx");
+      : "eax", "ecx", "edx");
 
 }
 OIL_DEFINE_IMPL_FULL (fdct8x8s_s16_mmx, fdct8x8s_s16, OIL_IMPL_FLAG_MMX | OIL_IMPL_FLAG_SSE);
diff --git a/liboil/simdpack/abs_i386.c b/liboil/simdpack/abs_i386.c
index 626065e..824c1ce 100644
--- a/liboil/simdpack/abs_i386.c
+++ b/liboil/simdpack/abs_i386.c
@@ -134,7 +134,7 @@ OIL_DEFINE_IMPL_ASM (abs_u16_s16_i386asm3, abs_u16_s16);
 static void
 abs_u16_s16_mmx (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
 {
-  const int16_t p[][4] = {
+  static const int16_t p[][4] = {
     { -32768, -32768, -32768, -32768 },
     { 32767, 32767, 32767, 32767 }
   };
@@ -150,7 +150,7 @@ abs_u16_s16_mmx (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
   __asm__ __volatile__ ("\n"
       "	movq	(%0), %%mm2		\n"
       "	movq	8(%0), %%mm3		\n"
-      :: "c" (p));
+      :: "r" (p));
   while (n--) {
     tmp[0] = *src;
     OIL_INCREMENT (src, sstr);
@@ -161,16 +161,16 @@ abs_u16_s16_mmx (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
     tmp[3] = *src;
     OIL_INCREMENT (src, sstr);
     __asm__ __volatile__ ("\n"
-        "	movq	(%%eax), %%mm1		\n"
+        "	movq	(%0), %%mm1		\n"
         "	movq	%%mm1, %%mm0		\n"
         "	paddsw	%%mm2, %%mm0		\n"
         "	paddsw	%%mm3, %%mm1		\n"
         "	psubsw	%%mm2, %%mm0		\n"
         "	psubsw	%%mm3, %%mm1		\n"
         "	psubw	%%mm1, %%mm0		\n"
-        "	movq	%%mm0, (%%eax)		\n"
-        :: "a" (&tmp)
-    );
+        "	movq	%%mm0, (%0)		\n"
+        : : "r" (tmp)
+        : "memory" );
     *dest = tmp[0];
     OIL_INCREMENT (dest, dstr);
     *dest = tmp[1];