Regenerated.

* app/composite/gimp-composite-{mmx,sse,sse2,altivec,vis}.c: Regenerated.

* app/composite/gimp-composite-regression.[ch]: Ensure all ints
  are now unsigned longs.

* app/composite/gimp-composite-sse2.c (gimp_composite_addition_rgba8_rgba8_rgba8_sse2):
  Enclosed the troublesome (rather dense) asm construct in
  __OPTIMIZE__ in which case the compiler can successfully
  allocate enough registers to load up the asm()

ChangeLog

@@ -1,3 +1,19 @@
+2003-09-16  Helvetix Victorinox  <helvetix@gimp.org>
+
+	* app/composite/gimp-composite-{mmx,sse,sse2,altivec,vis}.c: Regenerated.
+
+	* app/composite/gimp-composite-regression.[ch]: Ensure all ints
+	  are now unsigned longs.
+
+	* app/composite/gimp-composite-sse2.c (gimp_composite_addition_rgba8_rgba8_rgba8_sse2):
+	  Enclosed the troublesome (rather dense) asm construct in
+	  __OPTIMIZE__ in which case the compiler can successfully
+	  allocate enough registers to load up the asm()
+
+	* app/composite/gimp-composite-sse2.c: Removed code that caused
+	  gcc to complain can't find a register in class `GENERAL_REGS' To
+	  be revisited later.
+
 2003-09-17  Seth Burgess <sjburges@gimp.org>
 
 	* plug-ins/common/sparkle.c: use the rowstride of a pixel region
@@ -71,10 +87,21 @@
 
 2003-09-16  Helvetix Victorinox  <helvetix@gimp.org>
 
+	* app/composite/gimp-composite-{mmx,sse,sse2,altivec,vis}.c: Regenerated.
+
+	* app/composite/gimp-composite-regression.[ch]: Ensure all ints
+	  are now unsigned longs.
+
+	* app/composite/gimp-composite-sse2.c (gimp_composite_addition_rgba8_rgba8_rgba8_sse2):
+	  Enclosed the troublesome (rather dense) asm construct in
+	  __OPTIMIZE__ in which case the compiler can successfully
+	  allocate enough registers to load up the asm()
+
 	* app/composite/gimp-composite-sse2.c: Removed code that caused
 	  gcc to complain can't find a register in class `GENERAL_REGS' To
 	  be revisited later.
 
+
 2003-09-16  Sven Neumann  <sven@gimp.org>
 
 	* app/composite/gimp-composite-generic.c

app/composite/gimp-composite-3dnow-test.c

@@ -71,7 +71,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/gimp-composite-altivec-test.c

@@ -71,7 +71,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/gimp-composite-mmx-test.c

@@ -211,7 +211,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/gimp-composite-mmx.c

@@ -978,7 +978,7 @@ gimp_composite_scale_rgba8_rgba8_rgba8_mmx (GimpCompositeContext *_op)
 
   for (; op.n_pixels >= 2; op.n_pixels -= 2)
     {
-      asm volatile ("movq           %0,%%mm2\n"
+      asm volatile ("movq           %1,%%mm2\n"
                     "\tmovq      %%mm2,%%mm1\n"
                     "\tpunpcklbw %%mm0,%%mm1\n"
                     "\tmovq      %%mm3,%%mm5\n"
@@ -993,28 +993,28 @@ gimp_composite_scale_rgba8_rgba8_rgba8_mmx (GimpCompositeContext *_op)
 
                     "\tpackuswb  %%mm4,%%mm1\n"
 
-                    "\tmovq    %%mm1,%1\n"
+                    "\tmovq    %%mm1,%0\n"
-                    : /* empty */
+                    : "=m" (*op.D)
-                    : "m" (*op.A), "m" (*op.D)
+                    : "m" (*op.A)
-                    : "0", "1", "%mm1", "%mm2", "%mm3", "%mm4", "%mm5", "%mm6", "%mm7");
+                    : "%mm1", "%mm2", "%mm3", "%mm4", "%mm5", "%mm6", "%mm7");
       op.A += 8;
       op.D += 8;
     }
 
   if (op.n_pixels)
     {
-    asm volatile ("movd           %0,%%mm2\n"
+      asm volatile ("movd           %1,%%mm2\n"
-                  "\tmovq      %%mm2,%%mm1\n"
+                    "\tmovq      %%mm2,%%mm1\n"
-                  "\tpunpcklbw %%mm0,%%mm1\n"
+                    "\tpunpcklbw %%mm0,%%mm1\n"
-                  "\tmovq      %%mm3,%%mm5\n"
+                    "\tmovq      %%mm3,%%mm5\n"
 
-                  "\t" pmulwX(mm5,mm1,mm7) "\n"
+                    "\t" pmulwX(mm5,mm1,mm7) "\n"
 
-                  "\tpackuswb  %%mm0,%%mm1\n"
+                    "\tpackuswb  %%mm0,%%mm1\n"
-                  "\tmovd      %%mm1,%1\n"
+                    "\tmovd      %%mm1,%0\n"
-                  : /* empty */
+                    : "=m" (*op.D)
-                  : "m" (*op.A), "m" (*op.D)
+                    : "m" (*op.A)
-                  : "0", "1", "%mm1", "%mm2", "%mm3", "%mm4", "%mm5", "%mm6", "%mm7");
+                    : "%mm1", "%mm2", "%mm3", "%mm4", "%mm5", "%mm6", "%mm7");
   }
 
   asm("emms");
@@ -1149,8 +1149,8 @@ gimp_composite_subtract_rgba8_rgba8_rgba8_mmx (GimpCompositeContext *_op)
 
   for (; op.n_pixels >= 2; op.n_pixels -= 2)
     {
-      asm volatile ("  movq       %0,%%mm2\n"
+      asm volatile ("  movq       %1,%%mm2\n"
-                    "\tmovq       %1,%%mm3\n"
+                    "\tmovq       %2,%%mm3\n"
 
                     "\tmovq    %%mm2,%%mm4\n"
                     "\tpsubusb %%mm3,%%mm4\n"
@@ -1162,9 +1162,9 @@ gimp_composite_subtract_rgba8_rgba8_rgba8_mmx (GimpCompositeContext *_op)
 
                     "\tpand    %%mm0,%%mm2\n"
                     "\tpor     %%mm2,%%mm1\n"
-                    "\tmovq    %%mm1,%2\n"
+                    "\tmovq    %%mm1,%0\n"
-                    : /* empty */
+                    : "=m" (*op.D)
-                    : "m" (*op.A), "m" (*op.B), "m" (*op.D)
+                    : "m" (*op.A), "m" (*op.B)
                     : "0", "1", "2", "%mm1", "%mm2", "%mm3", "%mm4", "%mm5");
       op.A += 8;
       op.B += 8;

app/composite/gimp-composite-regression.c

@@ -134,7 +134,7 @@ gimp_composite_regression_compare_contexts (char *operation, GimpCompositeContex
 int
 gimp_composite_regression_comp_rgba8 (char *str, gimp_rgba8_t *rgba8A, gimp_rgba8_t *rgba8B, gimp_rgba8_t *expected, gimp_rgba8_t *got, u_long length)
 {
-  int i;
+  u_long i;
   int failed;
   int fail_count;
 
@@ -223,12 +223,12 @@ gimp_composite_regression_timer_report (char *name, double t1, double t2)
 }
 
 double
-gimp_composite_regression_time_function (int iterations, void (*func)(), GimpCompositeContext *ctx)
+gimp_composite_regression_time_function (u_long iterations, void (*func)(), GimpCompositeContext *ctx)
 {
   struct timeval t0;
   struct timeval t1;
   struct timeval tv_elapsed;
-  int i;
+  u_long i;
 
   gettimeofday(&t0, NULL);
   for (i = 0; i < iterations; i++) { (*func)(ctx); }
@@ -239,10 +239,10 @@ gimp_composite_regression_time_function (int iterations, void (*func)(), GimpCom
 }
 
 gimp_rgba8_t *
-gimp_composite_regression_random_rgba8 (unsigned long n_pixels)
+gimp_composite_regression_random_rgba8 (u_long n_pixels)
 {
   gimp_rgba8_t *rgba8;
-  int i;
+  u_long i;
 
   if ((rgba8 = (gimp_rgba8_t *) calloc (sizeof(gimp_rgba8_t), n_pixels))) {
     for (i = 0; i < n_pixels; i++) {
@@ -257,11 +257,11 @@ gimp_composite_regression_random_rgba8 (unsigned long n_pixels)
 }
 
 gimp_rgba8_t *
-gimp_composite_regression_fixed_rgba8 (unsigned long n_pixels)
+gimp_composite_regression_fixed_rgba8 (u_long n_pixels)
 {
   gimp_rgba8_t *rgba8;
-  int i;
+  u_long i;
-  int v;
+  u_long v;
 
   if ((rgba8 = (gimp_rgba8_t *) calloc(sizeof(gimp_rgba8_t), n_pixels))) {
     for (i = 0; i < n_pixels; i++) {

app/composite/gimp-composite-regression.h

@@ -85,7 +85,7 @@ typedef struct
 } gimp_rgba32_t;
 #endif
 
-extern double gimp_composite_regression_time_function (int, void (*)(), GimpCompositeContext *);
+extern double gimp_composite_regression_time_function (u_long, void (*)(), GimpCompositeContext *);
 extern int gimp_composite_regression_comp_rgba8 (char *, gimp_rgba8_t *, gimp_rgba8_t *, gimp_rgba8_t *, gimp_rgba8_t *, u_long);
 extern int gimp_composite_regression_comp_va8 (char *, gimp_va8_t *, gimp_va8_t *, gimp_va8_t *, gimp_va8_t *, u_long);
 extern int gimp_composite_regression_compare_contexts (char *, GimpCompositeContext *, GimpCompositeContext *);
@@ -94,8 +94,8 @@ extern void gimp_composite_regression_print_rgba8 (gimp_rgba8_t *);
 extern void gimp_composite_regression_print_va8 (gimp_va8_t *);
 extern void gimp_composite_regression_timer_report (char *, double, double);
 
-extern gimp_rgba8_t *gimp_composite_regression_random_rgba8 (unsigned long);
+extern gimp_rgba8_t *gimp_composite_regression_random_rgba8 (u_long);
-extern gimp_rgba8_t *gimp_composite_regression_fixed_rgba8 (unsigned long);
+extern gimp_rgba8_t *gimp_composite_regression_fixed_rgba8 (u_long);
 extern GimpCompositeContext *gimp_composite_context_init (GimpCompositeContext *,
 																																																										GimpCompositeOperation,
 																																																										GimpPixelFormat,

app/composite/gimp-composite-sse-test.c

@@ -211,7 +211,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/gimp-composite-sse2-test.c

@@ -151,7 +151,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/gimp-composite-sse2.c

@@ -685,12 +685,13 @@ gimp_composite_swap_rgba8_rgba8_rgba8_sse2 (GimpCompositeContext *_op)
 {
   GimpCompositeContext op = *_op;
 
-#if 0
   /*
-   * Inhale one whole i686 cache line at once. 64 bytes, 16 rgba8 pixels, 4 128 bit xmm registers.
+   * Inhale one whole i686 cache line at once. 64 bytes, 16 rgba8
+   * pixels, 4 128 bit xmm registers.
    */
   for (; op.n_pixels >= 16; op.n_pixels -= 16)
     {
+#ifdef __OPTIMIZE__
       asm volatile ("  movdqu      %0,%%xmm0\n"
                     "\tmovdqu      %1,%%xmm1\n"
                     "\tmovdqu      %2,%%xmm2\n"
@@ -715,10 +716,51 @@ gimp_composite_swap_rgba8_rgba8_rgba8_sse2 (GimpCompositeContext *_op)
                     : /* empty */
                     : "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
                     );
+#else
+      asm volatile ("  movdqu      %0,%%xmm0\n"
+                    "\tmovdqu      %1,%%xmm1\n"
+                    "\tmovdqu      %2,%%xmm2\n"
+                    "\tmovdqu      %3,%%xmm3\n"
+                    : "+m" (op.A[0]), "+m" (op.B[0]),
+                      "+m" (op.A[1]), "+m" (op.B[1])
+                    : /* empty */
+                    : "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
+                    );
+
+      asm volatile ("\tmovdqu      %4,%%xmm4\n"
+                    "\tmovdqu      %5,%%xmm5\n"
+                    "\tmovdqu      %6,%%xmm6\n"
+                    "\tmovdqu      %7,%%xmm7\n"
+                    : "+m" (op.A[2]), "+m" (op.B[2]),
+                      "+m" (op.A[3]), "+m" (op.B[3])
+                    : /* empty */
+                    : "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
+                    );
+
+      asm volatile ("\tmovdqu      %%xmm0,%1\n"
+                    "\tmovdqu      %%xmm1,%0\n"
+                    "\tmovdqu      %%xmm2,%3\n"
+                    "\tmovdqu      %%xmm3,%2\n"
+                    : "+m" (op.A[0]), "+m" (op.B[0]),
+                      "+m" (op.A[1]), "+m" (op.B[1])
+                    : /* empty */
+                    : "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
+                    );
+
+      asm volatile ("\tmovdqu      %%xmm4,%5\n"
+                    "\tmovdqu      %%xmm5,%4\n"
+                    "\tmovdqu      %%xmm6,%7\n"
+                    "\tmovdqu      %%xmm7,%6\n"
+                    : "+m" (op.A[2]), "+m" (op.B[2]),
+                      "+m" (op.A[3]), "+m" (op.B[3])
+                    : /* empty */
+                    : "%xmm0", "%xmm1", "%xmm2", "%xmm3", "%xmm4", "%xmm5", "%xmm6", "%xmm7"
+                    );
+#endif
       op.A += 64;
       op.B += 64;
     }
-#endif
+
 
   for (; op.n_pixels >= 4; op.n_pixels -= 4)
     {

app/composite/gimp-composite-vis-test.c

@@ -71,7 +71,7 @@ main (int argc, char *argv[])
   putenv ("GIMP_COMPOSITE=0x1");
 
   iterations = 1;
-  n_pixels = 163921;
+  n_pixels = 1048593;
 
   argv++, argc--;
   while (argc >= 2) {

app/composite/make-installer.py

@@ -522,7 +522,7 @@ op.add_option('-t', '--test', action='store_true',                 dest='test',
               help='generate regression testing code')
 op.add_option('-i', '--iterations', action='store', type='int',    dest='iterations', default=1,
               help='number of iterations in regression tests')
-op.add_option('-n', '--n-pixels', action='store',   type="int",    dest='n_pixels',   default=64*2049+16*2049+1,
+op.add_option('-n', '--n-pixels', action='store',   type="int",    dest='n_pixels',   default=128*8192+16+1,
               help='number of pixels in each regression test iteration')
 op.add_option('-r', '--requires', action='append',  type='string', dest='requires',   default=[],
               help='cpp #if conditionals')