From b2f00ba330250f6697bec20188505cc6dd1c6e81 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 5 Jan 2021 12:10:02 +0100
Subject: [PATCH 01/12] Heavily optimized common copper list routines.

---
 lib/libgfx/CopMoveLong.c | 19 ++++++++++---------
 lib/libgfx/CopMoveWord.c | 11 +++++------
 lib/libgfx/CopWait.c     | 15 ++++++---------
 3 files changed, 21 insertions(+), 24 deletions(-)

diff --git a/lib/libgfx/CopMoveLong.c b/lib/libgfx/CopMoveLong.c
index 831be8ef..9c9de6e7 100644
--- a/lib/libgfx/CopMoveLong.c
+++ b/lib/libgfx/CopMoveLong.c
@@ -1,16 +1,17 @@
 #include <copper.h>
 
-CopInsT *CopMoveLong(CopListT *list, u_short reg, void *data) {
-  CopInsT *ptr = list->curr;
-  u_short *ins = (u_short *)ptr;
+CopInsT *CopMoveLong(CopListT *list, u_short reg, void *ptr) {
+  CopInsT *ins = list->curr;
+  u_int data = (u_int)ptr;
 
   reg &= 0x01fe;
+  reg += 2;
 
-  *ins++ = reg;
-  *ins++ = (u_int)data >> 16;
-  *ins++ = reg + 2;
-  *ins++ = (u_int)data;
+  *((u_short *)ins)++ = reg;
+  *((u_short *)ins)++ = data;
+  *((u_short *)ins)++ = reg - 2;
+  *((u_short *)ins)++ = swap16(data);
 
-  list->curr = (CopInsT *)ins;
-  return ptr;
+  list->curr = ins;
+  return ins - 2;
 }
diff --git a/lib/libgfx/CopMoveWord.c b/lib/libgfx/CopMoveWord.c
index 1ac71bf2..37389328 100644
--- a/lib/libgfx/CopMoveWord.c
+++ b/lib/libgfx/CopMoveWord.c
@@ -1,12 +1,11 @@
 #include <copper.h>
 
 CopInsT *CopMoveWord(CopListT *list, u_short reg, u_short data) {
-  CopInsT *ptr = list->curr;
-  u_short *ins = (u_short *)ptr;
+  CopInsT *ins = list->curr;
 
-  *ins++ = reg & 0x01fe;
-  *ins++ = data;
+  *((u_short *)ins)++ = reg & 0x01fe;
+  *((u_short *)ins)++ = data;
 
-  list->curr = (CopInsT *)ins;
-  return ptr;
+  list->curr = ins;
+  return ins - 1;
 }
diff --git a/lib/libgfx/CopWait.c b/lib/libgfx/CopWait.c
index e7fb03cb..17788627 100644
--- a/lib/libgfx/CopWait.c
+++ b/lib/libgfx/CopWait.c
@@ -1,15 +1,12 @@
 #include <copper.h>
 
 CopInsT *CopWait(CopListT *list, u_short vp, u_short hp) {
-  CopInsT *ptr = list->curr;
-  u_char *bp = (u_char *)ptr;
-  u_short *wp;
+  CopInsT *ins = list->curr;
 
-  *bp++ = vp;
-  *bp++ = hp | 1; 
-  wp = (u_short *)bp;
-  *wp++ = 0xfffe;
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1; 
+  *((u_short *)ins)++ = 0xfffe;
 
-  list->curr = (CopInsT *)wp;
-  return ptr;
+  list->curr = ins;
+  return ins - 1;
 }

From 51473bb19c4c7c7840f1239b5e1b5a3a2381db66 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 5 Jan 2021 12:11:26 +0100
Subject: [PATCH 02/12] Vertical scaling - first attempt.

---
 effects/scaler/Makefile           |  7 +++
 effects/scaler/data/stars-die.png |  3 ++
 effects/scaler/scaler.c           | 80 +++++++++++++++++++++++++++++++
 3 files changed, 90 insertions(+)
 create mode 100644 effects/scaler/Makefile
 create mode 100644 effects/scaler/data/stars-die.png
 create mode 100644 effects/scaler/scaler.c

diff --git a/effects/scaler/Makefile b/effects/scaler/Makefile
new file mode 100644
index 00000000..3eba1ad1
--- /dev/null
+++ b/effects/scaler/Makefile
@@ -0,0 +1,7 @@
+TOPDIR := $(realpath ../..)
+
+CLEAN-FILES := data/stars-die.c
+
+PNG2C.stars-die := --bitmap image,320x256x5,+interleaved --palette image_pal,32
+
+include $(TOPDIR)/build/effect.mk
diff --git a/effects/scaler/data/stars-die.png b/effects/scaler/data/stars-die.png
new file mode 100644
index 00000000..0faf9936
--- /dev/null
+++ b/effects/scaler/data/stars-die.png
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7aecfc44106b2d8168984b948dc78f6be1509d3c6f1db368fac17b2686fb0863
+size 24313
diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
new file mode 100644
index 00000000..5bed85c7
--- /dev/null
+++ b/effects/scaler/scaler.c
@@ -0,0 +1,80 @@
+#include "effect.h"
+#include "custom.h"
+#include "copper.h"
+#include "bitmap.h"
+#include "palette.h"
+#include "profiler.h"
+
+#include "data/stars-die.c"
+
+#define LINES 256
+
+static CopListT *cp0, *cp1;
+
+/* static */ void MakeCopperList(CopListT *cp, short height) {
+  short ys = (LINES - height) / 2;
+  int dy = (LINES << 16) / height;
+
+  CopInit(cp);
+  CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, height);
+  CopSetupBitplanes(cp, NULL, &image, image.depth);
+
+  {
+    short n = (short)(dy >> 16) * (short)image.depth;
+    short mod = (short)image.bytesPerRow * (short)(n - 1);
+    int y = 0;
+    short i;
+
+    for (i = 1; i < height; i++) {
+      short _mod = mod;
+      int ny = y + dy;
+      if ((u_short)ny < (u_short)y)
+        _mod += image.bytesPerRow * image.depth;
+      CopWait(cp, Y(ys + i), 0);
+      CopMove16(cp, bpl1mod, _mod);
+      CopMove16(cp, bpl2mod, _mod);
+      y = ny;
+    }
+  }
+
+  CopEnd(cp);
+}
+
+/* static */ void Init(void) {
+  LoadPalette(&image_pal, 0);
+  SetupPlayfield(MODE_LORES, image.depth,
+                 X(0), Y(0), image.width, image.height);
+
+  cp0 = NewCopList(40 + LINES * 3);
+  cp1 = NewCopList(40 + LINES * 3);
+  MakeCopperList(cp0, LINES);
+  CopListActivate(cp0);
+
+  EnableDMA(DMAF_RASTER);
+}
+
+static void Kill(void) {
+  DisableDMA(DMAF_COPPER | DMAF_RASTER);
+  DeleteCopList(cp0);
+  DeleteCopList(cp1);
+}
+
+PROFILE(Scaler);
+
+static void Render(void) {
+  static short val = 0, dir = 1;
+
+  ProfilerStart(Scaler);
+  MakeCopperList(cp0, LINES - val);
+  ProfilerStop(Scaler);
+
+  CopListRun(cp0);
+  TaskWaitVBlank();
+  { CopListT *tmp = cp0; cp0 = cp1; cp1 = tmp; }
+
+  val += dir;
+  if ((val == 0) || (val == LINES - 2))
+    dir = -dir;
+}
+
+EFFECT(scaler, NULL, NULL, Init, Kill, Render);

From 0c779bd5e34902f3533ba391c47790959fdac70e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 5 Jan 2021 21:33:22 +0100
Subject: [PATCH 03/12] Optimize more copper routines.

---
 include/copper.h          | 17 +++++++----------
 lib/libgfx/CopInit.c      |  2 +-
 lib/libgfx/CopLoadColor.c | 12 +++++++-----
 lib/libgfx/CopLoadPal.c   |  8 +++++---
 lib/libgfx/CopMoveWord.c  |  2 +-
 lib/libgfx/CopSkip.c      | 12 ++++++++++++
 lib/libgfx/CopWait.c      |  2 +-
 lib/libgfx/CopWaitSafe.c  |  9 ++++-----
 lib/libgfx/CopWaitSkip.c  | 15 ---------------
 lib/libgfx/Makefile       |  2 +-
 10 files changed, 39 insertions(+), 42 deletions(-)
 create mode 100644 lib/libgfx/CopSkip.c
 delete mode 100644 lib/libgfx/CopWaitSkip.c

diff --git a/include/copper.h b/include/copper.h
index 54a35bfe..12f7ffd2 100644
--- a/include/copper.h
+++ b/include/copper.h
@@ -34,12 +34,10 @@ typedef union {
   } move;
 } CopInsT;
 
-#define CLF_VPOVF 1 /* Vertical Position counter overflowed */
-
 typedef struct {
   CopInsT *curr;
   u_short length;
-  u_short flags;
+  u_char  overflow; /* -1 if Vertical Position counter overflowed */
   CopInsT entry[0]; 
 } CopListT;
 
@@ -57,29 +55,28 @@ static inline void CopListRun(CopListT *list) {
 }
 
 /* Low-level functions */
-CopInsT *CopMoveWord(CopListT *list, u_short reg, u_short data);
+CopInsT *CopMoveWord(CopListT *list, u_short reg, short data);
 CopInsT *CopMoveLong(CopListT *list, u_short reg, void *data);
 
-#define CSREG(reg) (u_short)offsetof(struct Custom, reg)
+#define CSREG(reg) offsetof(struct Custom, reg)
 #define CopMove16(cp, reg, data) CopMoveWord(cp, CSREG(reg), data)
 #define CopMove32(cp, reg, data) CopMoveLong(cp, CSREG(reg), data)
 
 /* Official way to represent no-op copper instruction. */
 #define CopNoOp(cp) CopMoveWord(cp, 0x1FE, 0)
 
-CopInsT *CopWait(CopListT *list, u_short vp, u_short hp);
+CopInsT *CopWait(CopListT *list, short vp, short hp);
 CopInsT *CopWaitMask(CopListT *list, u_short vp, u_short hp, 
                      u_short vpmask asm("d2"), u_short hpmask asm("d3"));
 
 /* Handles Copper Vertical Position counter overflow, by inserting CopWaitEOL
  * at first WAIT instruction with VP >= 256. */
-CopInsT *CopWaitSafe(CopListT *list, u_short vp, u_short hp);
+CopInsT *CopWaitSafe(CopListT *list, short vp, short hp);
 
 /* The most significant bit of vertical position cannot be masked out (overlaps
  * with blitter-finished-disable bit), so we have to pass upper bit as well. */
 #define CopWaitH(cp, vp, hp) CopWaitMask(cp, vp & 128, hp, 0, 255)
 #define CopWaitV(cp, vp) CopWaitMask(cp, vp, 0, 255, 0)
-#define CopWaitEOL(cp, vp) CopWait(cp, vp, LASTHP)
 
 CopInsT *CopSkip(CopListT *list, u_short vp, u_short hp);
 CopInsT *CopSkipMask(CopListT *list, u_short vp, u_short hp, 
@@ -103,8 +100,8 @@ static inline void CopInsSet16(CopInsT *ins, u_short data) {
 }
 
 /* High-level functions */
-CopInsT *CopLoadPal(CopListT *list, const PaletteT *palette, u_short start);
-CopInsT *CopLoadColor(CopListT *list, u_short start, u_short end, u_short color);
+CopInsT *CopLoadPal(CopListT *list, const PaletteT *palette, short start);
+CopInsT *CopLoadColor(CopListT *list, short start, short end, short color);
 
 void CopSetupMode(CopListT *list, u_short mode, u_short depth);
 void CopSetupDisplayWindow(CopListT *list, u_short mode, 
diff --git a/lib/libgfx/CopInit.c b/lib/libgfx/CopInit.c
index 24b3871f..9df16255 100644
--- a/lib/libgfx/CopInit.c
+++ b/lib/libgfx/CopInit.c
@@ -2,5 +2,5 @@
 
 void CopInit(CopListT *list) {
   list->curr = list->entry;
-  list->flags = 0;
+  list->overflow = 0;
 }
diff --git a/lib/libgfx/CopLoadColor.c b/lib/libgfx/CopLoadColor.c
index 4a0c2a1e..72226af3 100644
--- a/lib/libgfx/CopLoadColor.c
+++ b/lib/libgfx/CopLoadColor.c
@@ -1,14 +1,16 @@
 #include <copper.h>
 
-CopInsT *CopLoadColor(CopListT *list, u_short start, u_short end, u_short color)
-{
+CopInsT *CopLoadColor(CopListT *list, short start, short end, short color) {
   CopInsT *ptr = list->curr;
   u_short *ins = (u_short *)ptr;
+  short n = end - start - 1;
+  short reg = CSREG(color[start]);
 
-  while (start <= end) {
-    *ins++ = CSREG(color[start++]);
+  do {
+    *ins++ = reg;
     *ins++ = color;
-  }
+    reg += 2;
+  } while (--n != -1);
 
   list->curr = (CopInsT *)ins;
   return ptr;
diff --git a/lib/libgfx/CopLoadPal.c b/lib/libgfx/CopLoadPal.c
index 68cbbed7..bb974368 100644
--- a/lib/libgfx/CopLoadPal.c
+++ b/lib/libgfx/CopLoadPal.c
@@ -1,14 +1,16 @@
 #include <copper.h>
 
-CopInsT *CopLoadPal(CopListT *list, const PaletteT *palette, u_short start) {
+CopInsT *CopLoadPal(CopListT *list, const PaletteT *palette, short start) {
   CopInsT *ptr = list->curr;
   u_short *ins = (u_short *)ptr;
   u_short *c = palette->colors;
-  short n = min(palette->count, (u_short)(32 - start)) - 1;
+  short n = min((short)palette->count, (short)(32 - start)) - 1;
+  short reg = CSREG(color[start]);
 
   do {
-    *ins++ = CSREG(color[start++]);
+    *ins++ = reg;
     *ins++ = *c++;
+    reg += 2;
   } while (--n != -1);
 
   list->curr = (CopInsT *)ins;
diff --git a/lib/libgfx/CopMoveWord.c b/lib/libgfx/CopMoveWord.c
index 37389328..a3a5108a 100644
--- a/lib/libgfx/CopMoveWord.c
+++ b/lib/libgfx/CopMoveWord.c
@@ -1,6 +1,6 @@
 #include <copper.h>
 
-CopInsT *CopMoveWord(CopListT *list, u_short reg, u_short data) {
+CopInsT *CopMoveWord(CopListT *list, u_short reg, short data) {
   CopInsT *ins = list->curr;
 
   *((u_short *)ins)++ = reg & 0x01fe;
diff --git a/lib/libgfx/CopSkip.c b/lib/libgfx/CopSkip.c
new file mode 100644
index 00000000..7ee1f77c
--- /dev/null
+++ b/lib/libgfx/CopSkip.c
@@ -0,0 +1,12 @@
+#include <copper.h>
+
+CopInsT *CopSkip(CopListT *list, u_short vp, u_short hp) {
+  CopInsT *ins = list->curr;
+
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_short *)ins)++ = 0xffff;
+
+  list->curr = ins;
+  return ins - 1;
+}
diff --git a/lib/libgfx/CopWait.c b/lib/libgfx/CopWait.c
index 17788627..a966c881 100644
--- a/lib/libgfx/CopWait.c
+++ b/lib/libgfx/CopWait.c
@@ -1,6 +1,6 @@
 #include <copper.h>
 
-CopInsT *CopWait(CopListT *list, u_short vp, u_short hp) {
+CopInsT *CopWait(CopListT *list, short vp, short hp) {
   CopInsT *ins = list->curr;
 
   *((u_char *)ins)++ = vp;
diff --git a/lib/libgfx/CopWaitSafe.c b/lib/libgfx/CopWaitSafe.c
index 9900e6ab..6074abbd 100644
--- a/lib/libgfx/CopWaitSafe.c
+++ b/lib/libgfx/CopWaitSafe.c
@@ -1,11 +1,10 @@
 #include <copper.h>
 
-CopInsT *CopWaitSafe(CopListT *list, u_short vp, u_short hp) {
-  if (!(list->flags & CLF_VPOVF) && (vp >= 256)) {
+CopInsT *CopWaitSafe(CopListT *list, short vp, short hp) {
+  if (vp >= 256 && !list->overflow) {
+    list->overflow = -1;
     /* Wait for last waitable position to control when overflow occurs. */
-    CopWaitEOL(list, 255);
-    list->flags |= CLF_VPOVF;
+    *((u_int *)list->curr)++ = 0xffdffffe;
   }
-
   return CopWait(list, vp, hp);
 }
diff --git a/lib/libgfx/CopWaitSkip.c b/lib/libgfx/CopWaitSkip.c
deleted file mode 100644
index 02c4cbef..00000000
--- a/lib/libgfx/CopWaitSkip.c
+++ /dev/null
@@ -1,15 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopSkip(CopListT *list, u_short vp, u_short hp) {
-  CopInsT *ptr = list->curr;
-  u_char *bp = (u_char *)ptr;
-  u_short *wp;
-
-  *bp++ = vp;
-  *bp++ = hp | 1;
-  wp = (u_short *)bp;
-  *wp++ = 0xffff;
-
-  list->curr = (CopInsT *)wp;
-  return ptr;
-}
diff --git a/lib/libgfx/Makefile b/lib/libgfx/Makefile
index d99b2744..db5ff38e 100644
--- a/lib/libgfx/Makefile
+++ b/lib/libgfx/Makefile
@@ -24,12 +24,12 @@ SOURCES := \
 	CopSetupManualSprites.c \
 	CopSetupMode.c \
 	CopSetupSprites.c \
+	CopSkip.c \
 	CopSkipMask.c \
 	CopUpdateBitplanes.c \
 	CopWait.c \
 	CopWaitMask.c \
 	CopWaitSafe.c \
-	CopWaitSkip.c \
 	CpuEdge.c \
 	CpuLine.c \
 	DeleteBitmap.c \

From 2ec6f66a04816d0547158cd9c80b9747cf34f43e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 5 Jan 2021 21:43:32 +0100
Subject: [PATCH 04/12] Move some code to subprocedure.

---
 effects/scaler/scaler.c | 41 +++++++++++++++++++++--------------------
 1 file changed, 21 insertions(+), 20 deletions(-)

diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
index 5bed85c7..baedd1f2 100644
--- a/effects/scaler/scaler.c
+++ b/effects/scaler/scaler.c
@@ -11,32 +11,33 @@
 
 static CopListT *cp0, *cp1;
 
+/* static */ void VerticalScaler(CopListT *cp, short ys, short height) {
+  short rowmod = image.bytesPerRow * image.depth;
+  int dy = (LINES << 16) / height;
+  short n = (short)(dy >> 16) * (short)image.depth;
+  short mod = (short)image.bytesPerRow * (short)(n - 1);
+  int y = 0;
+  short i;
+
+  for (i = 1; i < height; i++) {
+    short _mod = mod;
+    int ny = y + dy;
+    if ((u_short)ny < (u_short)y)
+      _mod += rowmod;
+    CopWaitSafe(cp, Y(ys + i), X(0));
+    CopMove16(cp, bpl1mod, _mod);
+    CopMove16(cp, bpl2mod, _mod);
+    y = ny;
+  }
+}
+
 /* static */ void MakeCopperList(CopListT *cp, short height) {
   short ys = (LINES - height) / 2;
-  int dy = (LINES << 16) / height;
 
   CopInit(cp);
   CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, height);
   CopSetupBitplanes(cp, NULL, &image, image.depth);
-
-  {
-    short n = (short)(dy >> 16) * (short)image.depth;
-    short mod = (short)image.bytesPerRow * (short)(n - 1);
-    int y = 0;
-    short i;
-
-    for (i = 1; i < height; i++) {
-      short _mod = mod;
-      int ny = y + dy;
-      if ((u_short)ny < (u_short)y)
-        _mod += image.bytesPerRow * image.depth;
-      CopWait(cp, Y(ys + i), 0);
-      CopMove16(cp, bpl1mod, _mod);
-      CopMove16(cp, bpl2mod, _mod);
-      y = ny;
-    }
-  }
-
+  VerticalScaler(cp, ys, height);
   CopEnd(cp);
 }
 

From 3f061bc1f1be6a16f78252e734c33cedac41e441 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 5 Jan 2021 23:25:04 +0100
Subject: [PATCH 05/12] Optimize sh*t out of it - now I'm content :)

---
 include/copper.h         | 141 ++++++++++++++++++++++++++++++---------
 lib/libgfx/CopEnd.c      |   9 ---
 lib/libgfx/CopInit.c     |   6 --
 lib/libgfx/CopMoveLong.c |  17 -----
 lib/libgfx/CopMoveWord.c |  11 ---
 lib/libgfx/CopSkip.c     |  12 ----
 lib/libgfx/CopSkipMask.c |  16 -----
 lib/libgfx/CopWait.c     |  12 ----
 lib/libgfx/CopWaitMask.c |  16 -----
 lib/libgfx/CopWaitSafe.c |  10 ---
 lib/libgfx/Makefile      |   9 ---
 11 files changed, 110 insertions(+), 149 deletions(-)
 delete mode 100644 lib/libgfx/CopEnd.c
 delete mode 100644 lib/libgfx/CopInit.c
 delete mode 100644 lib/libgfx/CopMoveLong.c
 delete mode 100644 lib/libgfx/CopMoveWord.c
 delete mode 100644 lib/libgfx/CopSkip.c
 delete mode 100644 lib/libgfx/CopSkipMask.c
 delete mode 100644 lib/libgfx/CopWait.c
 delete mode 100644 lib/libgfx/CopWaitMask.c
 delete mode 100644 lib/libgfx/CopWaitSafe.c

diff --git a/include/copper.h b/include/copper.h
index 12f7ffd2..be2dc9f5 100644
--- a/include/copper.h
+++ b/include/copper.h
@@ -43,7 +43,17 @@ typedef struct {
 
 CopListT *NewCopList(u_short length);
 void DeleteCopList(CopListT *list);
-void CopInit(CopListT *list);
+
+static inline void CopInit(CopListT *list) {
+  list->curr = list->entry;
+  list->overflow = 0;
+}
+
+static inline void CopEnd(CopListT *list) {
+  CopInsT *ins = list->curr;
+  *((u_int *)ins)++ = 0xfffffffe;
+  list->curr = ins;
+}
 
 /* @brief Enable copper and activate copper list.
  * @warning This function busy-waits for vertical blank. */
@@ -55,50 +65,119 @@ static inline void CopListRun(CopListT *list) {
 }
 
 /* Low-level functions */
-CopInsT *CopMoveWord(CopListT *list, u_short reg, short data);
-CopInsT *CopMoveLong(CopListT *list, u_short reg, void *data);
-
 #define CSREG(reg) offsetof(struct Custom, reg)
-#define CopMove16(cp, reg, data) CopMoveWord(cp, CSREG(reg), data)
-#define CopMove32(cp, reg, data) CopMoveLong(cp, CSREG(reg), data)
+#define CopMove16(cp, reg, data) CopMoveWord((cp), CSREG(reg), (data))
+#define CopMove32(cp, reg, data) CopMoveLong((cp), CSREG(reg), (int)(data))
+
+static inline CopInsT *CopMoveWord(CopListT *list, short reg, short data) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_short *)ins)++ = reg;
+  *((u_short *)ins)++ = data;
+  list->curr = ins;
+  return pos;
+}
+
+static inline void CopInsSet16(CopInsT *ins, short data) {
+  ins->move.data = data;
+}
+
+static inline CopInsT *CopMoveLong(CopListT *list, short reg, int data) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_short *)ins)++ = reg + 2;
+  *((u_short *)ins)++ = data;
+  *((u_short *)ins)++ = reg;
+  *((u_short *)ins)++ = swap16(data);
+  list->curr = ins;
+  return pos;
+}
+
+static inline void CopInsSet32(CopInsT *ins, void *data) {
+  asm volatile("movew %0,%2\n"
+               "swap  %0\n"
+               "movew %0,%1\n"
+               : "+d" (data)
+               : "m" (ins[1].move.data), "m" (ins[0].move.data));
+}
 
 /* Official way to represent no-op copper instruction. */
 #define CopNoOp(cp) CopMoveWord(cp, 0x1FE, 0)
 
-CopInsT *CopWait(CopListT *list, short vp, short hp);
-CopInsT *CopWaitMask(CopListT *list, u_short vp, u_short hp, 
-                     u_short vpmask asm("d2"), u_short hpmask asm("d3"));
+/* Wait for raster beam position to be greater or equal to (vp, hp). */
+static inline CopInsT *CopWait(CopListT *list, short vp, short hp) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_short *)ins)++ = 0xfffe;
+  list->curr = ins;
+  return pos;
+}
 
 /* Handles Copper Vertical Position counter overflow, by inserting CopWaitEOL
  * at first WAIT instruction with VP >= 256. */
-CopInsT *CopWaitSafe(CopListT *list, short vp, short hp);
+static inline CopInsT *CopWaitSafe(CopListT *list, short vp, short hp) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  if (vp > 255 && !list->overflow) {
+    list->overflow = -1;
+    /* Wait for last waitable position to control when overflow occurs. */
+    *((u_int *)ins)++ = 0xffdffffe;
+  }
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_short *)ins)++ = 0xfffe;
+  list->curr = ins;
+  return pos;
+}
+
+/* Similar to CopWait, but masks bits in beam position counters. */
+static inline CopInsT *CopWaitMask(CopListT *list, short vp, short hp,
+                                   short vpmask, short hpmask) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_char *)ins)++ = 0x80 | vpmask;
+  *((u_char *)ins)++ = hpmask & 0xfe;
+  list->curr = ins;
+  return pos;
+}
 
 /* The most significant bit of vertical position cannot be masked out (overlaps
  * with blitter-finished-disable bit), so we have to pass upper bit as well. */
-#define CopWaitH(cp, vp, hp) CopWaitMask(cp, vp & 128, hp, 0, 255)
-#define CopWaitV(cp, vp) CopWaitMask(cp, vp, 0, 255, 0)
-
-CopInsT *CopSkip(CopListT *list, u_short vp, u_short hp);
-CopInsT *CopSkipMask(CopListT *list, u_short vp, u_short hp, 
-                     u_short vpmask asm("d2"), u_short hpmask asm("d3"));
-
-#define CopSkipH(cp, vp, hp) CopSkipMask(cp, vp & 128, hp, 0, 255)
-#define CopSkipV(cp, vp) CopSkipMask(cp, vp, 0, 255, 0)
-
-void CopEnd(CopListT *list);
-
-static inline void CopInsSet32(CopInsT *ins, void *data) {
-  asm volatile("movew %0,%2\n"
-               "swap  %0\n"
-               "movew %0,%1\n"
-               : "+d" (data)
-               : "m" (ins[0].move.data), "m" (ins[1].move.data));
+#define CopWaitH(cp, vp, hp) CopWaitMask((cp), (vp) & 128, (hp), 0, 255)
+#define CopWaitV(cp, vp) CopWaitMask((cp), (vp), 0, 255, 0)
+
+/* Skip next instruction if the video beam has already reached a specified
+ * (vp, hp) position. */
+static inline CopInsT *CopSkip(CopListT *list, short vp, short hp) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_short *)ins)++ = 0xffff;
+  list->curr = ins;
+  return pos;
 }
 
-static inline void CopInsSet16(CopInsT *ins, u_short data) {
-  ins->move.data = data;
+/* Similar to CopSkip, but masks bits in beam position counters. */
+static inline CopInsT *CopSkipMask(CopListT *list, short vp, short hp, 
+                                   short vpmask, short hpmask) {
+  CopInsT *pos = list->curr;
+  CopInsT *ins = list->curr;
+  *((u_char *)ins)++ = vp;
+  *((u_char *)ins)++ = hp | 1;
+  *((u_char *)ins)++ = 0x80 | vpmask;
+  *((u_char *)ins)++ = hpmask | 1;
+  list->curr = ins;
+  return pos;
 }
 
+#define CopSkipH(cp, vp, hp) CopSkipMask((cp), (vp) & 128, (hp), 0, 255)
+#define CopSkipV(cp, vp) CopSkipMask((cp), (vp), 0, 255, 0)
+
 /* High-level functions */
 CopInsT *CopLoadPal(CopListT *list, const PaletteT *palette, short start);
 CopInsT *CopLoadColor(CopListT *list, short start, short end, short color);
@@ -117,7 +196,7 @@ void CopUpdateBitplanes(CopInsT **bplptr, const BitmapT *bitmap, short n);
 void CopSetupDualPlayfield(CopListT *list, CopInsT **bplptr,
                            const BitmapT *pf1, const BitmapT *pf2);
 
-static inline CopInsT *CopSetColor(CopListT *list, short i, u_short value) {
+static inline CopInsT *CopSetColor(CopListT *list, short i, short value) {
   return CopMove16(list, color[i], value);
 }
 
diff --git a/lib/libgfx/CopEnd.c b/lib/libgfx/CopEnd.c
deleted file mode 100644
index 85b7be42..00000000
--- a/lib/libgfx/CopEnd.c
+++ /dev/null
@@ -1,9 +0,0 @@
-#include <copper.h>
-
-void CopEnd(CopListT *list) {
-  u_int *ins = (u_int *)list->curr;
-
-  *ins++ = 0xfffffffe;
-
-  list->curr = (CopInsT *)ins;
-}
diff --git a/lib/libgfx/CopInit.c b/lib/libgfx/CopInit.c
deleted file mode 100644
index 9df16255..00000000
--- a/lib/libgfx/CopInit.c
+++ /dev/null
@@ -1,6 +0,0 @@
-#include <copper.h>
-
-void CopInit(CopListT *list) {
-  list->curr = list->entry;
-  list->overflow = 0;
-}
diff --git a/lib/libgfx/CopMoveLong.c b/lib/libgfx/CopMoveLong.c
deleted file mode 100644
index 9c9de6e7..00000000
--- a/lib/libgfx/CopMoveLong.c
+++ /dev/null
@@ -1,17 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopMoveLong(CopListT *list, u_short reg, void *ptr) {
-  CopInsT *ins = list->curr;
-  u_int data = (u_int)ptr;
-
-  reg &= 0x01fe;
-  reg += 2;
-
-  *((u_short *)ins)++ = reg;
-  *((u_short *)ins)++ = data;
-  *((u_short *)ins)++ = reg - 2;
-  *((u_short *)ins)++ = swap16(data);
-
-  list->curr = ins;
-  return ins - 2;
-}
diff --git a/lib/libgfx/CopMoveWord.c b/lib/libgfx/CopMoveWord.c
deleted file mode 100644
index a3a5108a..00000000
--- a/lib/libgfx/CopMoveWord.c
+++ /dev/null
@@ -1,11 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopMoveWord(CopListT *list, u_short reg, short data) {
-  CopInsT *ins = list->curr;
-
-  *((u_short *)ins)++ = reg & 0x01fe;
-  *((u_short *)ins)++ = data;
-
-  list->curr = ins;
-  return ins - 1;
-}
diff --git a/lib/libgfx/CopSkip.c b/lib/libgfx/CopSkip.c
deleted file mode 100644
index 7ee1f77c..00000000
--- a/lib/libgfx/CopSkip.c
+++ /dev/null
@@ -1,12 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopSkip(CopListT *list, u_short vp, u_short hp) {
-  CopInsT *ins = list->curr;
-
-  *((u_char *)ins)++ = vp;
-  *((u_char *)ins)++ = hp | 1;
-  *((u_short *)ins)++ = 0xffff;
-
-  list->curr = ins;
-  return ins - 1;
-}
diff --git a/lib/libgfx/CopSkipMask.c b/lib/libgfx/CopSkipMask.c
deleted file mode 100644
index 24a3e7b3..00000000
--- a/lib/libgfx/CopSkipMask.c
+++ /dev/null
@@ -1,16 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopSkipMask(CopListT *list, u_short vp, u_short hp,
-                     u_short vpmask asm("d2"), u_short hpmask asm("d3"))
-{
-  CopInsT *ptr = (CopInsT *)list->curr;
-  u_char *ins = (u_char *)ptr;
-
-  *ins++ = vp;
-  *ins++ = hp | 1;
-  *ins++ = 0x80 | vpmask;
-  *ins++ = hpmask | 1;
-
-  list->curr = (CopInsT *)ins;
-  return ptr;
-}
diff --git a/lib/libgfx/CopWait.c b/lib/libgfx/CopWait.c
deleted file mode 100644
index a966c881..00000000
--- a/lib/libgfx/CopWait.c
+++ /dev/null
@@ -1,12 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopWait(CopListT *list, short vp, short hp) {
-  CopInsT *ins = list->curr;
-
-  *((u_char *)ins)++ = vp;
-  *((u_char *)ins)++ = hp | 1; 
-  *((u_short *)ins)++ = 0xfffe;
-
-  list->curr = ins;
-  return ins - 1;
-}
diff --git a/lib/libgfx/CopWaitMask.c b/lib/libgfx/CopWaitMask.c
deleted file mode 100644
index 97af0b8d..00000000
--- a/lib/libgfx/CopWaitMask.c
+++ /dev/null
@@ -1,16 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopWaitMask(CopListT *list, u_short vp, u_short hp,
-                     u_short vpmask asm("d2"), u_short hpmask asm("d3"))
-{
-  CopInsT *ptr = list->curr;
-  u_char *ins = (u_char *)ptr;
-
-  *ins++ = vp;
-  *ins++ = hp | 1;
-  *ins++ = 0x80 | vpmask;
-  *ins++ = hpmask & 0xfe;
-
-  list->curr = (CopInsT *)ins;
-  return ptr;
-}
diff --git a/lib/libgfx/CopWaitSafe.c b/lib/libgfx/CopWaitSafe.c
deleted file mode 100644
index 6074abbd..00000000
--- a/lib/libgfx/CopWaitSafe.c
+++ /dev/null
@@ -1,10 +0,0 @@
-#include <copper.h>
-
-CopInsT *CopWaitSafe(CopListT *list, short vp, short hp) {
-  if (vp >= 256 && !list->overflow) {
-    list->overflow = -1;
-    /* Wait for last waitable position to control when overflow occurs. */
-    *((u_int *)list->curr)++ = 0xffdffffe;
-  }
-  return CopWait(list, vp, hp);
-}
diff --git a/lib/libgfx/Makefile b/lib/libgfx/Makefile
index db5ff38e..11422e19 100644
--- a/lib/libgfx/Makefile
+++ b/lib/libgfx/Makefile
@@ -9,13 +9,9 @@ SOURCES := \
 	ClipArea.c \
 	ColorTab.c \
 	ColorTransition.c \
-	CopEnd.c \
-	CopInit.c \
 	CopListActivate.c \
 	CopLoadColor.c \
 	CopLoadPal.c \
-	CopMoveLong.c \
-	CopMoveWord.c \
 	CopSetupBitplaneArea.c \
 	CopSetupBitplaneFetch.c \
 	CopSetupBitplanes.c \
@@ -24,12 +20,7 @@ SOURCES := \
 	CopSetupManualSprites.c \
 	CopSetupMode.c \
 	CopSetupSprites.c \
-	CopSkip.c \
-	CopSkipMask.c \
 	CopUpdateBitplanes.c \
-	CopWait.c \
-	CopWaitMask.c \
-	CopWaitSafe.c \
 	CpuEdge.c \
 	CpuLine.c \
 	DeleteBitmap.c \

From 3354e9dc61ea00691fb845f8864641a28597cad3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Wed, 6 Jan 2021 12:23:14 +0100
Subject: [PATCH 06/12] Display (normal & interleaved) bitmap upside down.

---
 include/copper.h                         |  2 ++
 lib/libgfx/CopSetupBitplanes.c           |  6 ++--
 lib/libgfx/CopSetupBitplanesUpsideDown.c | 40 ++++++++++++++++++++++++
 lib/libgfx/Makefile                      |  1 +
 4 files changed, 47 insertions(+), 2 deletions(-)
 create mode 100644 lib/libgfx/CopSetupBitplanesUpsideDown.c

diff --git a/include/copper.h b/include/copper.h
index be2dc9f5..ad2cea40 100644
--- a/include/copper.h
+++ b/include/copper.h
@@ -189,6 +189,8 @@ void CopSetupBitplaneFetch(CopListT *list, u_short mode,
                            u_short xs, u_short w);
 void CopSetupBitplanes(CopListT *list, CopInsT **bplptr,
                        const BitmapT *bitmap, u_short depth);
+void CopSetupBitplanesUpsideDown(CopListT *list, CopInsT **bplptr,
+                                 const BitmapT *bitmap, u_short depth);
 void CopSetupBitplaneArea(CopListT *list, u_short mode, u_short depth,
                           const BitmapT *bitmap, short x, short y,
                           const Area2D *area);
diff --git a/lib/libgfx/CopSetupBitplanes.c b/lib/libgfx/CopSetupBitplanes.c
index eafdcb6c..b6b67d60 100644
--- a/lib/libgfx/CopSetupBitplanes.c
+++ b/lib/libgfx/CopSetupBitplanes.c
@@ -6,13 +6,15 @@ void CopSetupBitplanes(CopListT *list, CopInsT **bplptr,
   {
     void **planes = bitmap->planes;
     short n = depth - 1;
-    short i = 0;
+    short reg = CSREG(bplpt);
 
     do {
-      CopInsT *ins = CopMove32(list, bplpt[i++], *planes++);
+      CopInsT *ins = CopMoveLong(list, reg, (int)(*planes++));
 
       if (bplptr)
         *bplptr++ = ins;
+
+      reg += 4;
     } while (--n != -1);
   }
 
diff --git a/lib/libgfx/CopSetupBitplanesUpsideDown.c b/lib/libgfx/CopSetupBitplanesUpsideDown.c
new file mode 100644
index 00000000..6b2633b4
--- /dev/null
+++ b/lib/libgfx/CopSetupBitplanesUpsideDown.c
@@ -0,0 +1,40 @@
+#include <copper.h>
+
+void CopSetupBitplanesUpsideDown(CopListT *list, CopInsT **bplptr,
+                                 const BitmapT *bitmap, u_short depth) 
+{
+  short bytesPerLine = bitmap->bytesPerRow;
+  int start;
+
+  if (bitmap->flags & BM_INTERLEAVED)
+    bytesPerLine *= (short)bitmap->depth;
+
+  start = bytesPerLine * (short)(bitmap->height - 1);
+
+  {
+    void **planes = bitmap->planes;
+    short n = depth - 1;
+    short reg = CSREG(bplpt);
+
+    do {
+      CopInsT *ins = CopMoveLong(list, reg, (int)(*planes++) + start);
+
+      if (bplptr)
+        *bplptr++ = ins;
+
+      reg += 4;
+    } while (--n != -1);
+  }
+
+  {
+    short modulo;
+
+    if (bitmap->flags & BM_INTERLEAVED)
+      modulo = (short)bitmap->bytesPerRow * (short)(depth + 1);
+    else
+      modulo = 2 * bitmap->bytesPerRow;
+
+    CopMove16(list, bpl1mod, -modulo);
+    CopMove16(list, bpl2mod, -modulo);
+  }
+}
diff --git a/lib/libgfx/Makefile b/lib/libgfx/Makefile
index 11422e19..657698cb 100644
--- a/lib/libgfx/Makefile
+++ b/lib/libgfx/Makefile
@@ -15,6 +15,7 @@ SOURCES := \
 	CopSetupBitplaneArea.c \
 	CopSetupBitplaneFetch.c \
 	CopSetupBitplanes.c \
+	CopSetupBitplanesUpsideDown.c \
 	CopSetupDisplayWindow.c \
 	CopSetupDualPlayfield.c \
 	CopSetupManualSprites.c \

From a99997d2cde4306deeded9ae33ba795fc4b93095 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 19 Jan 2021 08:22:48 +0100
Subject: [PATCH 07/12] This procedure is not generic enough to be a part of
 library.

---
 include/copper.h                         |  2 --
 lib/libgfx/CopSetupBitplanesUpsideDown.c | 40 ------------------------
 lib/libgfx/Makefile                      |  1 -
 3 files changed, 43 deletions(-)
 delete mode 100644 lib/libgfx/CopSetupBitplanesUpsideDown.c

diff --git a/include/copper.h b/include/copper.h
index ad2cea40..be2dc9f5 100644
--- a/include/copper.h
+++ b/include/copper.h
@@ -189,8 +189,6 @@ void CopSetupBitplaneFetch(CopListT *list, u_short mode,
                            u_short xs, u_short w);
 void CopSetupBitplanes(CopListT *list, CopInsT **bplptr,
                        const BitmapT *bitmap, u_short depth);
-void CopSetupBitplanesUpsideDown(CopListT *list, CopInsT **bplptr,
-                                 const BitmapT *bitmap, u_short depth);
 void CopSetupBitplaneArea(CopListT *list, u_short mode, u_short depth,
                           const BitmapT *bitmap, short x, short y,
                           const Area2D *area);
diff --git a/lib/libgfx/CopSetupBitplanesUpsideDown.c b/lib/libgfx/CopSetupBitplanesUpsideDown.c
deleted file mode 100644
index 6b2633b4..00000000
--- a/lib/libgfx/CopSetupBitplanesUpsideDown.c
+++ /dev/null
@@ -1,40 +0,0 @@
-#include <copper.h>
-
-void CopSetupBitplanesUpsideDown(CopListT *list, CopInsT **bplptr,
-                                 const BitmapT *bitmap, u_short depth) 
-{
-  short bytesPerLine = bitmap->bytesPerRow;
-  int start;
-
-  if (bitmap->flags & BM_INTERLEAVED)
-    bytesPerLine *= (short)bitmap->depth;
-
-  start = bytesPerLine * (short)(bitmap->height - 1);
-
-  {
-    void **planes = bitmap->planes;
-    short n = depth - 1;
-    short reg = CSREG(bplpt);
-
-    do {
-      CopInsT *ins = CopMoveLong(list, reg, (int)(*planes++) + start);
-
-      if (bplptr)
-        *bplptr++ = ins;
-
-      reg += 4;
-    } while (--n != -1);
-  }
-
-  {
-    short modulo;
-
-    if (bitmap->flags & BM_INTERLEAVED)
-      modulo = (short)bitmap->bytesPerRow * (short)(depth + 1);
-    else
-      modulo = 2 * bitmap->bytesPerRow;
-
-    CopMove16(list, bpl1mod, -modulo);
-    CopMove16(list, bpl2mod, -modulo);
-  }
-}
diff --git a/lib/libgfx/Makefile b/lib/libgfx/Makefile
index 657698cb..11422e19 100644
--- a/lib/libgfx/Makefile
+++ b/lib/libgfx/Makefile
@@ -15,7 +15,6 @@ SOURCES := \
 	CopSetupBitplaneArea.c \
 	CopSetupBitplaneFetch.c \
 	CopSetupBitplanes.c \
-	CopSetupBitplanesUpsideDown.c \
 	CopSetupDisplayWindow.c \
 	CopSetupDualPlayfield.c \
 	CopSetupManualSprites.c \

From 6cbbd5bf2c6d8f484b5c2fb70adadab45bbf4a0a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 19 Jan 2021 09:18:08 +0100
Subject: [PATCH 08/12] Proper double buffering.

---
 effects/scaler/scaler.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)

diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
index baedd1f2..2f2c5ecc 100644
--- a/effects/scaler/scaler.c
+++ b/effects/scaler/scaler.c
@@ -8,10 +8,13 @@
 #include "data/stars-die.c"
 
 #define LINES 256
+#define DEPTH 5
 
-static CopListT *cp0, *cp1;
+static CopListT *cp[2];
+static CopInsT *bplptr[2][DEPTH];
+static short active = 0;
 
-/* static */ void VerticalScaler(CopListT *cp, short ys, short height) {
+/* static */ void VerticalScalerForward(CopListT *cp, short ys, short height) {
   short rowmod = image.bytesPerRow * image.depth;
   int dy = (LINES << 16) / height;
   short n = (short)(dy >> 16) * (short)image.depth;
@@ -32,12 +35,13 @@ static CopListT *cp0, *cp1;
 }
 
 /* static */ void MakeCopperList(CopListT *cp, short height) {
+/* static */ void MakeCopperList(CopListT *cp, CopInsT **bplptr, short height) {
   short ys = (LINES - height) / 2;
 
   CopInit(cp);
   CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, height);
-  CopSetupBitplanes(cp, NULL, &image, image.depth);
-  VerticalScaler(cp, ys, height);
+  CopSetupBitplanes(cp, bplptr, &image, image.depth);
+  VerticalScalerForward(cp, ys, height);
   CopEnd(cp);
 }
 
@@ -46,18 +50,18 @@ static CopListT *cp0, *cp1;
   SetupPlayfield(MODE_LORES, image.depth,
                  X(0), Y(0), image.width, image.height);
 
-  cp0 = NewCopList(40 + LINES * 3);
-  cp1 = NewCopList(40 + LINES * 3);
-  MakeCopperList(cp0, LINES);
-  CopListActivate(cp0);
+  cp[0] = NewCopList(40 + LINES * 3);
+  cp[1] = NewCopList(40 + LINES * 3);
+  MakeCopperList(cp[active], bplptr[active], LINES);
+  CopListActivate(cp[active]);
 
   EnableDMA(DMAF_RASTER);
 }
 
 static void Kill(void) {
   DisableDMA(DMAF_COPPER | DMAF_RASTER);
-  DeleteCopList(cp0);
-  DeleteCopList(cp1);
+  DeleteCopList(cp[0]);
+  DeleteCopList(cp[1]);
 }
 
 PROFILE(Scaler);
@@ -66,16 +70,17 @@ static void Render(void) {
   static short val = 0, dir = 1;
 
   ProfilerStart(Scaler);
-  MakeCopperList(cp0, LINES - val);
+  MakeCopperList(cp[active], bplptr[active], LINES - val);
   ProfilerStop(Scaler);
 
-  CopListRun(cp0);
+  CopListRun(cp[active]);
   TaskWaitVBlank();
-  { CopListT *tmp = cp0; cp0 = cp1; cp1 = tmp; }
 
   val += dir;
   if ((val == 0) || (val == LINES - 2))
     dir = -dir;
+
+  active ^= 1;
 }
 
 EFFECT(scaler, NULL, NULL, Init, Kill, Render);

From 3193e3d897309f760523588563aa5312759cd32a Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 19 Jan 2021 09:30:26 +0100
Subject: [PATCH 09/12] Flip image upside down when height is negative.

---
 effects/scaler/scaler.c | 63 ++++++++++++++++++++++++++++++++++-------
 1 file changed, 52 insertions(+), 11 deletions(-)

diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
index 2f2c5ecc..8d4e5f3e 100644
--- a/effects/scaler/scaler.c
+++ b/effects/scaler/scaler.c
@@ -11,10 +11,9 @@
 #define DEPTH 5
 
 static CopListT *cp[2];
-static CopInsT *bplptr[2][DEPTH];
 static short active = 0;
 
-/* static */ void VerticalScalerForward(CopListT *cp, short ys, short height) {
+static void VerticalScalerForward(CopListT *cp, short ys, short height) {
   short rowmod = image.bytesPerRow * image.depth;
   int dy = (LINES << 16) / height;
   short n = (short)(dy >> 16) * (short)image.depth;
@@ -34,14 +33,56 @@ static short active = 0;
   }
 }
 
-/* static */ void MakeCopperList(CopListT *cp, short height) {
-/* static */ void MakeCopperList(CopListT *cp, CopInsT **bplptr, short height) {
-  short ys = (LINES - height) / 2;
+static void CopSetupBitplanesReverse(CopListT *list, CopInsT **bplptr,
+                                     const BitmapT *bitmap, u_short depth)
+{
+  short bytesPerLine = bitmap->bytesPerRow;
+  int start;
+
+  if (bitmap->flags & BM_INTERLEAVED)
+    bytesPerLine *= (short)bitmap->depth;
+
+  start = bytesPerLine * (short)(bitmap->height - 1);
+
+  {
+    void **planes = bitmap->planes;
+    short n = depth - 1;
+    short reg = CSREG(bplpt);
+
+    do {
+      CopInsT *ins = CopMoveLong(list, reg, (int)(*planes++) + start);
+
+      if (bplptr)
+        *bplptr++ = ins;
+
+      reg += 4;
+    } while (--n != -1);
+  }
+
+  {
+    short modulo;
+
+    if (bitmap->flags & BM_INTERLEAVED)
+      modulo = (short)bitmap->bytesPerRow * (short)(depth + 1);
+    else
+      modulo = 2 * bitmap->bytesPerRow;
+
+    CopMove16(list, bpl1mod, -modulo);
+    CopMove16(list, bpl2mod, -modulo);
+  }
+}
+
+static void MakeCopperList(CopListT *cp, short height) {
+  short ys = (LINES - abs(height)) / 2;
 
   CopInit(cp);
   CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, height);
-  CopSetupBitplanes(cp, bplptr, &image, image.depth);
-  VerticalScalerForward(cp, ys, height);
+  if (height > 0) {
+    CopSetupBitplanes(cp, NULL, &image, image.depth);
+    VerticalScalerForward(cp, ys, height);
+  } else if (height < 0) {
+    CopSetupBitplanesReverse(cp, NULL, &image, image.depth);
+  }
   CopEnd(cp);
 }
 
@@ -52,7 +93,7 @@ static short active = 0;
 
   cp[0] = NewCopList(40 + LINES * 3);
   cp[1] = NewCopList(40 + LINES * 3);
-  MakeCopperList(cp[active], bplptr[active], LINES);
+  MakeCopperList(cp[active], LINES);
   CopListActivate(cp[active]);
 
   EnableDMA(DMAF_RASTER);
@@ -67,17 +108,17 @@ static void Kill(void) {
 PROFILE(Scaler);
 
 static void Render(void) {
-  static short val = 0, dir = 1;
+  static short val = LINES, dir = -1;
 
   ProfilerStart(Scaler);
-  MakeCopperList(cp[active], bplptr[active], LINES - val);
+  MakeCopperList(cp[active], val);
   ProfilerStop(Scaler);
 
   CopListRun(cp[active]);
   TaskWaitVBlank();
 
   val += dir;
-  if ((val == 0) || (val == LINES - 2))
+  if (val == -LINES)
     dir = -dir;
 
   active ^= 1;

From ffc3ffed58541a583936f29eaa7507d2545506c2 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 19 Jan 2021 21:45:16 +0100
Subject: [PATCH 10/12] Works correctly! :D

---
 effects/scaler/scaler.c | 25 +++++++++++++++++++++++--
 1 file changed, 23 insertions(+), 2 deletions(-)

diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
index 8d4e5f3e..1684c8eb 100644
--- a/effects/scaler/scaler.c
+++ b/effects/scaler/scaler.c
@@ -72,21 +72,42 @@ static void CopSetupBitplanesReverse(CopListT *list, CopInsT **bplptr,
   }
 }
 
+static void VerticalScalerReverse(CopListT *cp, short ys, short height) {
+  short rowmod = image.bytesPerRow * image.depth;
+  int dy = (LINES << 16) / height;
+  short n = (short)(dy >> 16) * (short)image.depth;
+  short mod = (short)image.bytesPerRow * (short)(n + 1);
+  int y = 0;
+  short i;
+
+  for (i = 1; i < height; i++) {
+    short _mod = mod;
+    int ny = y + dy;
+    if ((u_short)ny < (u_short)y)
+      _mod += rowmod;
+    CopWaitSafe(cp, Y(ys + i), X(0));
+    CopMove16(cp, bpl1mod, -_mod);
+    CopMove16(cp, bpl2mod, -_mod);
+    y = ny;
+  }
+}
+
 static void MakeCopperList(CopListT *cp, short height) {
   short ys = (LINES - abs(height)) / 2;
 
   CopInit(cp);
-  CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, height);
+  CopSetupDisplayWindow(cp, MODE_LORES, X(0), Y(ys), image.width, abs(height));
   if (height > 0) {
     CopSetupBitplanes(cp, NULL, &image, image.depth);
     VerticalScalerForward(cp, ys, height);
   } else if (height < 0) {
     CopSetupBitplanesReverse(cp, NULL, &image, image.depth);
+    VerticalScalerReverse(cp, ys, -height);
   }
   CopEnd(cp);
 }
 
-/* static */ void Init(void) {
+static void Init(void) {
   LoadPalette(&image_pal, 0);
   SetupPlayfield(MODE_LORES, image.depth,
                  X(0), Y(0), image.width, image.height);

From 26d165703041f69944a6ecdf3ee2303d6c7aaab3 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Tue, 19 Jan 2021 22:05:50 +0100
Subject: [PATCH 11/12] Fix second bounce.

---
 effects/scaler/scaler.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/effects/scaler/scaler.c b/effects/scaler/scaler.c
index 1684c8eb..aee139f6 100644
--- a/effects/scaler/scaler.c
+++ b/effects/scaler/scaler.c
@@ -139,7 +139,7 @@ static void Render(void) {
   TaskWaitVBlank();
 
   val += dir;
-  if (val == -LINES)
+  if (abs(val) == LINES)
     dir = -dir;
 
   active ^= 1;

From 0f6b9fd7686c8d84d707ce96f36a1128ab052246 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Krystian=20Bac=C5=82awski?= <krystian.baclawski@gmail.com>
Date: Mon, 25 Jan 2021 09:12:33 +0100
Subject: [PATCH 12/12] Add tool for precalculating scaler table.

---
 effects/scaler/Makefile             |  1 +
 effects/scaler/scalertab/.gitignore |  1 +
 effects/scaler/scalertab/Makefile   |  3 ++
 effects/scaler/scalertab/main.go    | 75 +++++++++++++++++++++++++++++
 4 files changed, 80 insertions(+)
 create mode 100644 effects/scaler/scalertab/.gitignore
 create mode 100644 effects/scaler/scalertab/Makefile
 create mode 100644 effects/scaler/scalertab/main.go

diff --git a/effects/scaler/Makefile b/effects/scaler/Makefile
index 3eba1ad1..86a04c5a 100644
--- a/effects/scaler/Makefile
+++ b/effects/scaler/Makefile
@@ -1,6 +1,7 @@
 TOPDIR := $(realpath ../..)
 
 CLEAN-FILES := data/stars-die.c
+SUBDIRS := scalertab
 
 PNG2C.stars-die := --bitmap image,320x256x5,+interleaved --palette image_pal,32
 
diff --git a/effects/scaler/scalertab/.gitignore b/effects/scaler/scalertab/.gitignore
new file mode 100644
index 00000000..d675eb8e
--- /dev/null
+++ b/effects/scaler/scalertab/.gitignore
@@ -0,0 +1 @@
+scalertab
diff --git a/effects/scaler/scalertab/Makefile b/effects/scaler/scalertab/Makefile
new file mode 100644
index 00000000..339e6e1c
--- /dev/null
+++ b/effects/scaler/scalertab/Makefile
@@ -0,0 +1,3 @@
+TOPDIR := $(realpath ../../..)
+
+include $(TOPDIR)/build/go.mk
diff --git a/effects/scaler/scalertab/main.go b/effects/scaler/scalertab/main.go
new file mode 100644
index 00000000..969ef43a
--- /dev/null
+++ b/effects/scaler/scalertab/main.go
@@ -0,0 +1,75 @@
+package main
+
+import (
+	"fmt"
+	"math"
+	"sort"
+)
+
+type Row struct {
+	err      float64 // average error
+	columns  []int   // current set of columns
+	removed  int     // which column was removed from working image
+	original int     // the column from original picture that was removed
+}
+
+func estimate(S []int, n int, N int) []Row {
+	if n == 1 {
+		R := make([]Row, 1)
+		R[0] = Row{0.0, []int{S[0]}, 1, S[1]}
+		return R
+	}
+
+	di := float64(N-1) / float64(n-1)
+	I := make([]float64, n)
+	for i := 0; i < n; i++ {
+		I[i] = float64(i) * di
+	}
+
+	rows := make([]Row, n+1)
+
+	for i := 0; i <= n; i++ {
+		C := make([]int, n)
+		sum := 0.0
+		for j := 0; j < n; j++ {
+			if j >= i {
+				C[j] = S[j+1]
+			} else {
+				C[j] = S[j]
+			}
+			delta := I[j] - float64(C[j])
+			sum += delta * delta
+		}
+		rows[i] = Row{math.Sqrt(sum) / float64(n), C, i, S[i]}
+	}
+
+	sort.Slice(rows, func(i, j int) bool {
+		return rows[i].err < rows[j].err
+	})
+
+	R := estimate(rows[0].columns, n-1, N)
+	for _, row := range rows[1:1] {
+		C := estimate(row.columns, n-1, N)
+		if R[len(R)-1].err > C[len(C)-1].err {
+			R = C
+		}
+	}
+
+	return append(R, rows[0])
+}
+
+func main() {
+	N := 40
+	S := make([]int, N)
+	for i := 0; i < N; i++ {
+		S[i] = i
+	}
+	R := estimate(S, N-1, N)
+	R = append(R, Row{0.0, S, -1, -1})
+	err := 0.0
+	for i, row := range R {
+		fmt.Printf("%3d: [%d] = %d\n", i+1, row.removed, row.original)
+		err += row.err
+	}
+	fmt.Printf("\naverage error: %.3f\n", err/float64(N-1))
+}