* Makefile.am: dist autogen.sh

* README: add comments
* configure.ac: bump version.  Fix ordering.  failing->broken
* liboil.pc.in: add -lm
* liboil/dct/dct36_f32.c: (dct36_f32_ref): still broken.  Note to
self: only add classes that _have_ a reference implementation.
* liboil/dct/fdct8x8_f64.c: (fdct8x8_f64_1d): fix striding
* liboil/liboilfuncs.h: parameter name change
* liboil/liboilparameter.h: Add guard value.
* liboil/liboiltest.c: (oil_test_check_function), (check_guard),
(oil_test_check_impl), (init_parameter), (check_holes): check
interstitial regions.  Use a guard value instead of 0.
* liboil/simdpack/abs_i386.c: (abs_u16_s16_mmx2),
(abs_u16_s16_sse2): disable broken impls
* liboil/simdpack/scalarmult_i386.c: (scalarmult_f32_sse): fix
impl

12 files changed, 91 insertions, 27 deletions

diff --git a/ChangeLog b/ChangeLog
index 23152b5..f19a2c9 100644
--- a/ChangeLog
+++ b/ChangeLog
@@ -1,3 +1,22 @@
+2005-01-13  David Schleef  <ds@schleef.org>
+
+	* Makefile.am: dist autogen.sh
+	* README: add comments
+	* configure.ac: bump version.  Fix ordering.  failing->broken
+	* liboil.pc.in: add -lm
+	* liboil/dct/dct36_f32.c: (dct36_f32_ref): still broken.  Note to
+	self: only add classes that _have_ a reference implementation.
+	* liboil/dct/fdct8x8_f64.c: (fdct8x8_f64_1d): fix striding
+	* liboil/liboilfuncs.h: parameter name change
+	* liboil/liboilparameter.h: Add guard value.
+	* liboil/liboiltest.c: (oil_test_check_function), (check_guard),
+	(oil_test_check_impl), (init_parameter), (check_holes): check
+	interstitial regions.  Use a guard value instead of 0.
+	* liboil/simdpack/abs_i386.c: (abs_u16_s16_mmx2),
+	(abs_u16_s16_sse2): disable broken impls
+	* liboil/simdpack/scalarmult_i386.c: (scalarmult_f32_sse): fix
+	impl
+
 2005-01-11  David Schleef  <ds@schleef.org>
 
 	* README: additional notes
diff --git a/Makefile.am b/Makefile.am
index 136351a..7c59013 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -1,7 +1,7 @@
 
 SUBDIRS = liboil testsuite examples
 
-EXTRA_DIST = COPYING
+EXTRA_DIST = COPYING autogen.sh
 
 pkgconfig_DATA = liboil-$(LIBOIL_MAJORMINOR).pc
 
diff --git a/README b/README
index eb7445f..ec33f8e 100644
--- a/README
+++ b/README
@@ -72,4 +72,10 @@ Random Ideas
 
 Why is trans8x8_f64 so slow on powerpc (compared to trans8x8_u16)?
 
+oil_restride_XXX() - copy with stride
+oil_scaleconv_s16_f32() - multiply by constant and convert
+oil_scaleclipconv_s16_f32() - multiply by constant, clip, and convert
+oil_splat_XXX()
+
+
 
diff --git a/configure.ac b/configure.ac
index ebdfc8a..6b358fd 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,5 +1,5 @@
 AC_PREREQ([2.58])
-AC_INIT(liboil,0.2.2.1)
+AC_INIT(liboil,0.3.0)
 
 nano=$(echo $VERSION | sed 's/[0-9]\.[0-9]\.[0-9]\.*//')
 if test x"$nano" = x1 ; then
@@ -8,8 +8,8 @@ else
   LIBOIL_CVS="no"
 fi
 
-AM_INIT_AUTOMAKE(1.6)
 AC_CANONICAL_TARGET([])
+AM_INIT_AUTOMAKE(1.6)
 AM_MAINTAINER_MODE
 
 LIBOIL_MAJORMINOR=0.3
@@ -28,7 +28,7 @@ AM_PROG_LIBTOOL
 
 AC_CONFIG_SRCDIR([liboil/liboil.h])
 
-ACLOCAL_AMFLAGS="-I m4 $ACLOCAL_FLAGS"
+ACLOCAL_AMFLAGS="-I m4 $ACLOCAL_AMFLAGS"
 AC_SUBST(ACLOCAL_AMFLAGS)
 
 PKG_CHECK_MODULES(GLIB, glib-2.0, HAVE_GLIB=yes, HAVE_GLIB=no)
@@ -57,11 +57,11 @@ AM_CONDITIONAL(USE_ALT_OPT, test "x$enable_alt_opt" = xyes)
 LIBOIL_OPT_CFLAGS="-O3 -fomit-frame-pointer -funroll-all-loops -Wno-unused-function -D'OIL_OPT_MANGLE(a)=a\#\#_O3' -D'OIL_OPT_SUFFIX=\"_O3\"'"
 AC_SUBST(LIBOIL_OPT_CFLAGS)
 
-AC_ARG_ENABLE(failing-implementations,
-  AC_HELP_STRING([--enable-failing-implementations],[compile with failing implementations]),
-  enable_failing_implementations=$enableval,enable_failing_implementations=no)
-if test "x$enable_failing_implementations" = xyes ; then
-  AC_DEFINE(USE_FAILING_IMPLS, 1, [Define if compiling failing implementations])
+AC_ARG_ENABLE(broken-implementations,
+  AC_HELP_STRING([--enable-broken-implementations],[compile with broken implementations]),
+  enable_broken_implementations=$enableval,enable_broken_implementations=no)
+if test "x$enable_broken_implementations" = xyes ; then
+  AC_DEFINE(ENABLE_BROKEN_IMPLS, 1, [Define if compiling broken implementations])
 fi
 
 dnl Set up conditionals for (target) architecture:
diff --git a/liboil.pc.in b/liboil.pc.in
index 1c8e904..ce115eb 100644
--- a/liboil.pc.in
+++ b/liboil.pc.in
@@ -8,7 +8,7 @@ toolsdir=${exec_prefix}/bin
 Name: liboil-@LIBOIL_MAJORMINOR@
 Description: Libaray of Optimized Inner Loops
 Version: @VERSION@
-Libs: -L${libdir} -loil-@LIBOIL_MAJORMINOR@
+Libs: -L${libdir} -loil-@LIBOIL_MAJORMINOR@ -lm
 Cflags: -I${includedir}
 
 
diff --git a/liboil/dct/dct36_f32.c b/liboil/dct/dct36_f32.c
index c29fb30..f1315c7 100644
--- a/liboil/dct/dct36_f32.c
+++ b/liboil/dct/dct36_f32.c
@@ -37,10 +37,12 @@ OIL_DEFINE_CLASS(dct36_f32, "float *d_36, int dstr, float *s_36, int sstr");
 static void
 dct36_f32_ref(float *dest, int dstr, float *src, int sstr)
 {
+#if 0
   int i;
   for(i=0;i<36;i++){
     dest[i] = src[i];
   }
+#endif
 }
 
 OIL_DEFINE_IMPL_REF (dct36_f32_ref, dct36_f32);
diff --git a/liboil/dct/fdct8x8_f64.c b/liboil/dct/fdct8x8_f64.c
index 0e7c09f..4e4938d 100644
--- a/liboil/dct/fdct8x8_f64.c
+++ b/liboil/dct/fdct8x8_f64.c
@@ -139,7 +139,7 @@ fdct8x8_f64_1d (double *dest, int dstr, double *src, int sstr)
 
 	for(i=0;i<8;i++){
 		oil_fdct8_f64(dest + i, tmp + i,
-                    8*sizeof(double), 8*sizeof(double));
+                    dstr, 8*sizeof(double));
 	}
 }
 
diff --git a/liboil/liboilfuncs.h b/liboil/liboilfuncs.h
index d11cae0..5cd83ca 100644
--- a/liboil/liboilfuncs.h
+++ b/liboil/liboilfuncs.h
@@ -361,7 +361,7 @@ extern OilFunctionClass *oil_function_class_ptr_dequantize8x8_s16;
 typedef void (*_oil_type_dequantize8x8_s16)(int16_t * d_8x8, int dstr, const int16_t * s1_8x8, int sstr1, const int16_t * s2_8x8, int sstr2);
 #define oil_dequantize8x8_s16 ((_oil_type_dequantize8x8_s16)(*(void **)oil_function_class_ptr_dequantize8x8_s16))
 extern OilFunctionClass *oil_function_class_ptr_diffsquaresum_f64;
-typedef void (*_oil_type_diffsquaresum_f64)(double * dest, const double * src1, int sstr1, const double * src2, int sstr2, int n);
+typedef void (*_oil_type_diffsquaresum_f64)(double * d_1, const double * src1, int sstr1, const double * src2, int sstr2, int n);
 #define oil_diffsquaresum_f64 ((_oil_type_diffsquaresum_f64)(*(void **)oil_function_class_ptr_diffsquaresum_f64))
 extern OilFunctionClass *oil_function_class_ptr_fdct8_f64;
 typedef void (*_oil_type_fdct8_f64)(double * d_8, const double * s_8, int dstr, int sstr);
diff --git a/liboil/liboilparameter.h b/liboil/liboilparameter.h
index 8679cd5..f8133a7 100644
--- a/liboil/liboilparameter.h
+++ b/liboil/liboilparameter.h
@@ -104,6 +104,7 @@ struct _OilParameter {
   int post_n;
   int stride;
   int size;
+  int guard;
 };
 
 #endif
diff --git a/liboil/liboiltest.c b/liboil/liboiltest.c
index a862d1b..0c9394c 100644
--- a/liboil/liboiltest.c
+++ b/liboil/liboiltest.c
@@ -42,6 +42,8 @@ static void fill_array (void *ptr, OilType type, int pre_n, int stride,
     int post_n);
 static double check_array (void *data, void *ref, OilType type, int pre_n,
     int stride, int post_n);
+static int check_holes (void *data, OilType type, int pre_n,
+    int stride, int post_n, int guard);
 
 OilTest *
 oil_test_new (OilFunctionClass *klass)
@@ -156,7 +158,7 @@ oil_test_check_function (OilTest *test)
         memcpy (p->test_data, p->src_data, p->size);
         args[i] = (unsigned long)p->test_data + OIL_TEST_HEADER;
       } else if (p->direction == 'd') {
-        memset (p->test_data, 0, p->size);
+        memset (p->test_data, p->guard, p->size);
         args[i] = (unsigned long)p->test_data + OIL_TEST_HEADER;
       } else {
         OIL_ERROR ("not reached");
@@ -204,11 +206,11 @@ oil_test_check_ref (OilTest *test)
 }
 
 static int
-check_zero (uint8_t *data, int n)
+check_guard (uint8_t *data, int n, int guard)
 {
   int i;
   for(i=0;i<n;i++) {
-    if (data[i] != 0) return 0;
+    if (data[i] != guard) return 0;
   }
   return 1;
 }
@@ -240,20 +242,26 @@ oil_test_check_impl (OilTest *test, OilFunctionImpl *impl)
 
     if (p->is_pointer) {
       if (p->direction == 'i' || p->direction == 'd') {
-        x += check_array (p->ref_data, p->test_data, p->type, p->pre_n,
+        x += check_array (p->test_data, p->ref_data, p->type, p->pre_n,
             p->stride, p->post_n);
         n += p->pre_n * p->post_n;
-        if (!check_zero (p->test_data, OIL_TEST_HEADER)) {
+        if (!check_guard (p->test_data, OIL_TEST_HEADER, p->guard)) {
           fail = 1;
           OIL_ERROR("function %s wrote before area for parameter %s",
               test->impl->name, p->parameter_name);
         }
-        if (!check_zero ((uint8_t *)p->test_data + p->size - OIL_TEST_FOOTER,
-              OIL_TEST_FOOTER)) {
+        if (!check_guard ((uint8_t *)p->test_data + p->size - OIL_TEST_FOOTER,
+              OIL_TEST_FOOTER, p->guard)) {
           fail = 1;
           OIL_ERROR("function %s wrote after area for parameter %s",
               test->impl->name, p->parameter_name);
         }
+        if (!check_holes (p->test_data, p->type, p->pre_n, p->stride,
+              p->post_n, p->guard)) {
+          fail = 1;
+          OIL_ERROR("function %s wrote in interstitial area for parameter %s",
+              test->impl->name, p->parameter_name);
+        }
       }
     }
   }
@@ -347,25 +355,26 @@ init_parameter (OilTest *test, OilParameter *p, OilParameter *ps)
   }
 
   p->size = p->stride * p->post_n + OIL_TEST_HEADER + OIL_TEST_FOOTER;
+  p->guard = oil_rand_u8();
 
   if (p->direction == 'i' || p->direction == 's') {
     if (p->src_data) free (p->src_data);
 
     OIL_DEBUG("allocating %d bytes for src_data for %s", p->size, p->parameter_name);
     p->src_data = malloc (p->size);
-    memset (p->src_data, 0, p->size);
+    memset (p->src_data, p->guard, p->size);
     fill_array (p->src_data + OIL_TEST_HEADER, p->type, p->pre_n, p->stride, p->post_n);
   }
 
   if (p->direction == 'i' || p->direction == 'd') {
     if (p->ref_data) free (p->ref_data);
     p->ref_data = malloc (p->size);
-    memset (p->ref_data, 0, p->size);
+    memset (p->ref_data, p->guard, p->size);
     OIL_DEBUG("allocating %d bytes for ref_data and test_data for %s", p->size, p->parameter_name);
 
     if (p->test_data) free (p->test_data);
     p->test_data = malloc (p->size);
-    memset (p->test_data, 0, p->size);
+    memset (p->test_data, p->guard, p->size);
   }
 }
 
@@ -490,3 +499,28 @@ check_array (void *data, void *ref, OilType type, int pre_n, int stride, int pos
   }
   return x;
 }
+
+static int
+check_holes (void *data, OilType type, int pre_n, int stride, int post_n,
+    int guard)
+{
+  int i;
+  int j;
+  int chunk_size;
+  int hole_size;
+
+  chunk_size = pre_n * oil_type_sizeof (type);
+  hole_size = stride - chunk_size;
+  if (hole_size == 0) {
+    return 1;
+  }
+
+  for(i=0;i<post_n;i++){
+    if (!check_guard (data + stride * i + chunk_size, hole_size, guard)) {
+      return 0;
+    }
+  }
+
+  return 1;
+}
+
diff --git a/liboil/simdpack/abs_i386.c b/liboil/simdpack/abs_i386.c
index 824c1ce..ea0a9f2 100644
--- a/liboil/simdpack/abs_i386.c
+++ b/liboil/simdpack/abs_i386.c
@@ -231,6 +231,7 @@ abs_u16_s16_mmxx (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
 OIL_DEFINE_IMPL_FULL (abs_u16_s16_mmxx, abs_u16_s16, OIL_IMPL_FLAG_MMX);
 #endif
 
+#ifdef ENABLE_BROKEN_IMPLS
 static void
 abs_u16_s16_mmx2 (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
 {
@@ -264,9 +265,10 @@ abs_u16_s16_mmx2 (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
       ::"ecx", "edx");
   asm volatile ("emms");
 }
-
 OIL_DEFINE_IMPL_FULL (abs_u16_s16_mmx2, abs_u16_s16, OIL_IMPL_FLAG_MMXEXT);
+#endif
 
+#ifdef ENABLE_BROKEN_IMPLS
 static void
 abs_u16_s16_sse2 (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
 {
@@ -294,6 +296,6 @@ abs_u16_s16_sse2 (uint16_t * dest, int dstr, int16_t * src, int sstr, int n)
       "	popl	%%ebp			\n":"+D" (src), "+a" (dest), "+S" (n)
       ::"ecx", "edx");
 }
-
 OIL_DEFINE_IMPL_FULL (abs_u16_s16_sse2, abs_u16_s16, OIL_IMPL_FLAG_SSE2);
+#endif
 
diff --git a/liboil/simdpack/scalarmult_i386.c b/liboil/simdpack/scalarmult_i386.c
index 11d8e9a..232dc83 100644
--- a/liboil/simdpack/scalarmult_i386.c
+++ b/liboil/simdpack/scalarmult_i386.c
@@ -60,9 +60,9 @@ scalarmult_f32_sse (float *dest, int dstr, float *src, int sstr,
         : 
         : "r" (t));
     OIL_GET(dest,dstr*(i + 0), float) = t[0];
-    OIL_GET(dest,dstr*(i + 0), float) = t[1];
-    OIL_GET(dest,dstr*(i + 0), float) = t[2];
-    OIL_GET(dest,dstr*(i + 0), float) = t[3];
+    OIL_GET(dest,dstr*(i + 1), float) = t[1];
+    OIL_GET(dest,dstr*(i + 2), float) = t[2];
+    OIL_GET(dest,dstr*(i + 3), float) = t[3];
   }
 
 }