scu-dsp/parse10: implement end of chunk and end of input detection

.gitignore

@@ -11,9 +11,12 @@
 *.pcm
 *.su
 *.ss
+*.dsp
+*.exe
 res/mai.data
 tools/ttf-convert
 tools/ttf-bitmap
+tools/ttf_bitmap2
 common/keyboard.cpp
 common/keyboard.hpp
 wordle/word_list.hpp

Makefile

@@ -1,5 +1,5 @@
-CFLAGS = -Isaturn
+CFLAGS = -Isaturn -I. -D__saturn__
-OPT ?= -Og
+OPT ?= -O2
 LIB = ./saturn
 
 ALL =
@@ -28,6 +28,9 @@ include $(LIB)/common.mk
 %.data.o: %.data
 	$(BUILD_BINARY_O)
 
+%.data.h: %.data
+	$(BUILD_BINARY_H)
+
 %.pattern.o: %.pattern
 	$(BUILD_BINARY_O)
 
@@ -37,6 +40,15 @@ include $(LIB)/common.mk
 %.bin.o: %.bin
 	$(BUILD_BINARY_O)
 
+%.bin.h: %.bin
+	$(BUILD_BINARY_H)
+
+%.dsp.o: %.dsp
+	$(BUILD_BINARY_O)
+
+%.dsp.h: %.dsp
+	$(BUILD_BINARY_H)
+
 %.pcm.o: %.pcm
 	$(BUILD_BINARY_O)
 
@@ -52,11 +64,19 @@ vdp2/nbg0.elf: vdp2/nbg0.o res/butterfly.data.o res/butterfly.data.pal.o
 
 vdp2/nbg0_16color.elf: vdp2/nbg0_16color.o res/kirby.data.o res/kirby.data.pal.o
 
+vdp2/nbg0_font.elf: vdp2/nbg0_font.o saturn/start.o font/hp_100lx_4bit.data.o
+
+vdp2/rbg0_font.elf: vdp2/rbg0_font.o saturn/start.o font/hp_100lx_4bit.data.o
+
 vdp2/color_calculation_ratio.elf: vdp2/color_calculation_ratio.o res/mai00.data.o res/mai.data.pal.o res/haohmaru.data.o res/haohmaru.data.pal.o res/forest.data.pal.o res/forest.pattern.o res/forest.tile.o
 
+vdp2/line_color_screen.elf: vdp2/line_color_screen.o $(LIBGCC)
+
 vdp1/polygon.elf: vdp1/polygon.o
 vdp1/cube.elf: vdp1/cube.o $(LIBGCC)
 vdp1/cube2.elf: vdp1/cube2.o
+vdp1/bear.elf: CFLAGS += -DUSE_SH2_DVSR
+vdp1/bear.elf: vdp1/bear.o $(LIBGCC)
 vdp1/normal_sprite.elf: vdp1/normal_sprite.o res/mai00.data.o res/mai.data.pal.o
 
 vdp1/normal_sprite_color_bank.elf: vdp1/normal_sprite_color_bank.o res/mai00.data.o res/mai.data.pal.o
@@ -153,6 +173,15 @@ editor/main_saturn.elf: editor/main_saturn.o res/nec.bitmap.bin.o res/nec_bold.b
 
 cdc/cdc.elf: cdc/cdc.o saturn/start.o memcpy.o cdc/serial.o
 
+scu-dsp/add.elf: scu-dsp/add.o saturn/start.o cdc/serial.o scu-dsp/input.bin.o
+
+scu-dsp/div10.elf: scu-dsp/div10.o saturn/start.o cdc/serial.o scu-dsp/div10.dsp.o
+
+scu-dsp/div10_vdp2.elf: scu-dsp/div10_vdp2.o saturn/start.o cdc/serial.o scu-dsp/div10_vdp2.dsp.o font/hp_100lx_4bit_flip.data.o
+
+%.dsp: %.asm
+	~/scu-dsp-asm/scu-dsp-asm $< $@
+
 # clean
 clean: clean-sh
 clean-sh:

cdc/serial.c

@@ -4,14 +4,14 @@
 
 void serial_char(const char c)
 {
-  while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
+  //while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
   sh2.reg.TDR = c;
 }
 
 void serial_string(const char * s)
 {
   while (*s) {
-    while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
+    //while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
     sh2.reg.TDR = *s++;
   }
 }
@@ -19,7 +19,7 @@ void serial_string(const char * s)
 void serial_bytes(const char * s, uint32_t length)
 {
   while (length > 0) {
-    while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
+    //while ((sh2.reg.SSR & SSR__TDRE) == 0); // wait for transmit data empty
     sh2.reg.TDR = *s++;
     length -= 1;
   }

font/hp_100lx_4bit.data.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_font_hp_100lx_4bit_data_start __asm("_binary_font_hp_100lx_4bit_data_start");
+extern uint32_t _binary_font_hp_100lx_4bit_data_end __asm("_binary_font_hp_100lx_4bit_data_end");
+extern uint32_t _binary_font_hp_100lx_4bit_data_size __asm("_binary_font_hp_100lx_4bit_data_size");

font/hp_100lx_4bit_flip.data.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_font_hp_100lx_4bit_flip_data_start __asm("_binary_font_hp_100lx_4bit_flip_data_start");
+extern uint32_t _binary_font_hp_100lx_4bit_flip_data_end __asm("_binary_font_hp_100lx_4bit_flip_data_end");
+extern uint32_t _binary_font_hp_100lx_4bit_flip_data_size __asm("_binary_font_hp_100lx_4bit_flip_data_size");

math/mat2x2.hpp

@@ -0,0 +1,105 @@
+#pragma once
+
+#include <utility>
+#include "vec2.hpp"
+
+template <int R, int C, typename T>
+struct mat;
+
+//
+// mat4x4
+//
+
+template <typename T>
+struct mat<2, 2, T>
+{
+  typedef vec<2, T> row_type;
+  typedef vec<2, T> col_type;
+
+private:
+  row_type value[2];
+
+public:
+  inline constexpr mat();
+
+  inline constexpr mat
+  (
+    T const& a00, T const& a01,
+    T const& a10, T const& a11
+  );
+
+  inline static constexpr int length() { return 4; }
+
+  inline constexpr typename mat<2, 2, T>::row_type const &
+  operator[](int i) const;
+
+  void operator=(const mat<2, 2, T>&) = delete;
+
+};
+
+
+template<typename T>
+inline constexpr mat<2, 2, T>::mat()
+  : value{std::move(row_type(1, 0)),
+	  std::move(row_type(0, 1))}
+{ }
+
+template<typename T>
+inline constexpr mat<2, 2, T>::mat
+(
+  T const& a00, T const& a01,
+  T const& a10, T const& a11
+)
+  : value{std::move(row_type(a00, a01)),
+	  std::move(row_type(a10, a11))}
+{ }
+
+template <typename T>
+inline constexpr typename mat<2, 2, T>::row_type const &
+mat<2, 2, T>::operator[](int i) const
+{
+  switch (i)
+  {
+  default: [[fallthrough]];
+  case 0:
+    return value[0];
+  case 1:
+    return value[1];
+  }
+}
+
+template<typename T>
+inline constexpr mat<2, 2, T> operator*(mat<2, 2, T> const& m1, mat<2, 2, T> const& m2)
+{
+#define c(i, j) (                               \
+    m1[i][0] * m2[0][j]                         \
+  + m1[i][1] * m2[1][j])
+
+  return mat<2, 2, T>(c(0,0), c(0,1),
+                      c(1,0), c(1,1));
+#undef c
+}
+
+template<typename T>
+inline constexpr typename mat<2, 2, T>::row_type operator*
+(
+  mat<2, 2, T> const& m,
+  typename mat<2, 2, T>::col_type const& v
+)
+{
+#define c(i) (                                  \
+    m[i][0] * v[0]                              \
+  + m[i][1] * v[1])
+
+  return typename mat<2, 2, T>::row_type(c(0), c(1));
+#undef c
+}
+
+template<typename T>
+inline constexpr mat<2, 2, T> transpose(mat<2, 2, T> const& m)
+{
+  return mat<2, 2, T>(
+    m[0][0], m[1][0],
+    m[0][1], m[1][1]
+  );
+}

math/mat3x3.hpp

@@ -74,6 +74,17 @@ mat<3, 3, T>::operator[](int i) const
   }
 }
 
+template<typename T>
+inline constexpr mat<3, 3, T> operator+(mat<3, 3, T> const& m1, mat<3, 3, T> const& m2)
+{
+#define c(i, j) ( m1[i][j] + m2[i][j] )
+
+  return mat<3, 3, T>(c(0,0), c(0,1), c(0,2),
+                      c(1,0), c(1,1), c(1,2),
+                      c(2,0), c(2,1), c(2,2));
+#undef c
+}
+
 template<typename T>
 inline constexpr mat<3, 3, T> operator*(mat<3, 3, T> const& m1, mat<3, 3, T> const& m2)
 {
@@ -88,6 +99,17 @@ inline constexpr mat<3, 3, T> operator*(mat<3, 3, T> const& m1, mat<3, 3, T> con
 #undef c
 }
 
+template<typename T>
+inline constexpr mat<3, 3, T> operator*(mat<3, 3, T> const& m1, float s)
+{
+#define c(i, j) ( m1[i][j] * s )
+
+  return mat<3, 3, T>(c(0,0), c(0,1), c(0,2),
+                      c(1,0), c(1,1), c(1,2),
+                      c(2,0), c(2,1), c(2,2));
+#undef c
+}
+
 template<typename T>
 inline constexpr typename mat<3, 3, T>::row_type operator*
 (

math/mat4x4.hpp

@@ -35,7 +35,7 @@ public:
   inline constexpr typename mat<4, 4, T>::row_type const &
   operator[](int i) const;
 
-  void operator=(const mat<4, 4, T>&) = delete;
+  //void operator=(const mat<4, 4, T>&) = delete;
 
 };
 
@@ -80,6 +80,18 @@ mat<4, 4, T>::operator[](int i) const
   }
 }
 
+template<typename T>
+inline constexpr mat<4, 4, T> operator+(mat<4, 4, T> const& m1, mat<4, 4, T> const& m2)
+{
+#define c(i, j) ( m1[i][j] + m2[i][j] )
+
+  return mat<4, 4, T>(c(0,0), c(0,1), c(0,2), c(0,3),
+                      c(1,0), c(1,1), c(1,2), c(1,3),
+                      c(2,0), c(2,1), c(2,2), c(2,3),
+                      c(3,0), c(3,1), c(3,2), c(3,3));
+#undef c
+}
+
 template<typename T>
 inline constexpr mat<4, 4, T> operator*(mat<4, 4, T> const& m1, mat<4, 4, T> const& m2)
 {
@@ -96,6 +108,18 @@ inline constexpr mat<4, 4, T> operator*(mat<4, 4, T> const& m1, mat<4, 4, T> con
 #undef c
 }
 
+template<typename T>
+inline constexpr mat<4, 4, T> operator*(mat<4, 4, T> const& m1, float s)
+{
+#define c(i, j) ( m1[i][j] * s )
+
+  return mat<4, 4, T>(c(0,0), c(0,1), c(0,2), c(0,3),
+                      c(1,0), c(1,1), c(1,2), c(1,3),
+                      c(2,0), c(2,1), c(2,2), c(2,3),
+                      c(3,0), c(3,1), c(3,2), c(3,3));
+#undef c
+}
+
 template<typename T>
 inline constexpr typename mat<4, 4, T>::row_type operator*
 (
@@ -113,6 +137,23 @@ inline constexpr typename mat<4, 4, T>::row_type operator*
 #undef c
 }
 
+template<typename T>
+inline constexpr vec<3, T> operator*
+(
+  mat<4, 4, T> const& m,
+  vec<3, T> const& v
+)
+{
+#define c(i) (                                  \
+    m[i][0] * v[0]                              \
+  + m[i][1] * v[1]                              \
+  + m[i][2] * v[2]                              \
+  + m[i][3] )
+
+  return vec<3, T>(c(0), c(1), c(2));
+#undef c
+}
+
 template<typename T>
 inline constexpr mat<4, 4, T> transpose(mat<4, 4, T> const& m)
 {

math/vec2.hpp

@@ -0,0 +1,158 @@
+#pragma once
+
+#include "math.hpp"
+#include "vec.hpp"
+
+//
+// vec3
+//
+
+template <typename T>
+struct vec<2, T>
+{
+  union {
+    struct { T x, y; };
+    struct { T u, v; };
+  };
+
+  inline constexpr vec();
+  inline constexpr vec(T scalar);
+  inline constexpr vec(T _x, T _y);
+
+  constexpr inline vec<2, T> operator-() const;
+  inline constexpr T const& operator[](int i) const;
+  inline constexpr vec<2, T>& operator=(vec<2, T> const& v);
+  inline constexpr vec<2, T>& operator+=(vec<2, T> const& v);
+  inline constexpr vec<2, T>& operator-=(vec<2, T> const& v);
+};
+
+template <typename T>
+inline constexpr vec<2, T>::vec()
+  : x(0), y(0)
+{}
+
+template <typename T>
+inline constexpr vec<2, T>::vec(T scalar)
+  : x(scalar), y(scalar)
+{}
+
+template <typename T>
+inline constexpr vec<2, T>::vec(T _x, T _y)
+  : x(_x), y(_y)
+{}
+
+template <typename T>
+constexpr inline vec<2, T> vec<2, T>::operator-() const
+{
+  return vec<2, T>(-x, -y);
+}
+
+template <typename T>
+inline constexpr T const& vec<2, T>::operator[](int i) const
+{
+  switch(i)
+  {
+  default: [[fallthrough]];
+  case 0: return x;
+  case 1: return y;
+  }
+}
+
+template <typename T>
+inline constexpr vec<2, T>& vec<2, T>::operator=(vec<2, T> const& v)
+{
+  this->x = static_cast<T>(v.x);
+  this->y = static_cast<T>(v.y);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<2, T>& vec<2, T>::operator+=(vec<2, T> const& v)
+{
+  *this = *this + vec<2, T>(v);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<2, T>& vec<2, T>::operator-=(vec<2, T> const& v)
+{
+  *this = *this - vec<2, T>(v);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator+(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  return vec<2, T>(v1.x + v2.x,
+                   v1.y + v2.y);
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator-(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  return vec<2, T>(v1.x - v2.x,
+                   v1.y - v2.y);
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator*(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  return vec<2, T>(v1.x * v2.x,
+                   v1.y * v2.y);
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator*(vec<2, T> const& v1, T const& scalar)
+{
+  return v1 * vec<2, T>(scalar);
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator*(T const& scalar, vec<2, T> const& v1)
+{
+  return vec<2, T>(scalar) * v1;
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator/(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  return vec<2, T>(v1.x / v2.x,
+                   v1.y / v2.y);
+}
+
+template <typename T>
+inline constexpr vec<2, T> operator/(vec<2, T> const& v1, T const& scalar)
+{
+  return v1 / vec<2, T>(scalar);
+}
+
+template <typename T>
+inline constexpr T dot(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  vec<2, T> tmp(v1 * v2);
+  return tmp.x + tmp.y;
+}
+
+template <typename T>
+inline constexpr T cross(vec<2, T> const& v1, vec<2, T> const& v2)
+{
+  return v1.x * v2.y - v2.x * v1.y;
+}
+
+template <typename T>
+inline constexpr vec<2, T> functor1(T (&func) (T const& x), vec<2, T> const& v)
+{
+  return vec<2, T>(func(v.x), func(v.y));
+}
+
+template <typename T, typename U>
+inline constexpr vec<2, U> functor1(U (&func) (T const& x), vec<2, T> const& v)
+{
+  return vec<2, U>(func(v.x), func(v.y));
+}
+
+template <typename T>
+inline constexpr T magnitude(vec<2, T> const& v)
+{
+  return sqrt(dot(v, v));
+}

math/vec3.hpp

@@ -10,7 +10,10 @@
 template <typename T>
 struct vec<3, T>
 {
-  T x, y, z;
+  union {
+    struct { T x, y, z; };
+    struct { T r, g, b; };
+  };
 
   inline constexpr vec();
   inline constexpr vec(T scalar);
@@ -109,6 +112,12 @@ inline constexpr vec<3, T> operator*(vec<3, T> const& v1, T const& scalar)
   return v1 * vec<3, T>(scalar);
 }
 
+template <typename T>
+inline constexpr vec<3, T> operator*(T const& scalar, vec<3, T> const& v1)
+{
+  return vec<3, T>(scalar) * v1;
+}
+
 template <typename T>
 inline constexpr vec<3, T> operator/(vec<3, T> const& v1, vec<3, T> const& v2)
 {
@@ -130,6 +139,14 @@ inline constexpr T dot(vec<3, T> const& v1, vec<3, T> const& v2)
   return tmp.x + tmp.y + tmp.z;
 }
 
+template <typename T>
+inline constexpr vec<3, T> cross(vec<3, T> const& v1, vec<3, T> const& v2)
+{
+  return vec<3, T>(v1.y * v2.z - v2.y * v1.z,
+                   v1.z * v2.x - v2.z * v1.x,
+                   v1.x * v2.y - v2.x * v1.y);
+}
+
 template <typename T>
 inline constexpr vec<3, T> functor1(T (&func) (T const& x), vec<3, T> const& v)
 {
@@ -143,15 +160,7 @@ inline constexpr vec<3, U> functor1(U (&func) (T const& x), vec<3, T> const& v)
 }
 
 template <typename T>
-inline constexpr T length(vec<3, T> const& v)
+inline constexpr T magnitude(vec<3, T> const& v)
 {
   return sqrt(dot(v, v));
 }
-
-template<typename T>
-inline constexpr vec<3, T> cross(vec<3, T> const& x, vec<3, T> const& y)
-{
-  return vec<3, T>(x.y * y.z - y.y * x.z,
-		   x.z * y.x - y.z * x.x,
-		   x.x * y.y - y.x * x.y);
-}

model/bear/bear.data.h

@@ -0,0 +1,15 @@
+#pragma once
+
+#include <stdint.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+extern uint32_t _binary_model_bear_bear_data_start __asm("_binary_model_bear_bear_data_start");
+extern uint32_t _binary_model_bear_bear_data_end __asm("_binary_model_bear_bear_data_end");
+extern uint32_t _binary_model_bear_bear_data_size __asm("_binary_model_bear_bear_data_size");
+
+#ifdef __cplusplus
+}
+#endif

model/bear/bear.mtl

@@ -0,0 +1,22 @@
+# Blender 4.3.2 MTL File: 'bear.blend'
+# www.blender.org
+
+newmtl matBear
+Ns 250.000000
+Ka 1.000000 1.000000 1.000000
+Kd 0.800000 0.800000 0.800000
+Ks 0.000000 0.000000 0.000000
+Ke 0.000000 0.000000 0.000000
+Ni 1.000000
+d 1.000000
+illum 1
+
+newmtl matOutline
+Ns 250.000000
+Ka 1.000000 1.000000 1.000000
+Kd 0.000000 0.000000 0.000000
+Ks 0.000000 0.000000 0.000000
+Ke 0.000000 0.000000 0.000000
+Ni 1.000000
+d 1.000000
+illum 1

model/bear/material.h

@@ -0,0 +1,13 @@
+#pragma once
+
+#include <stdint.h>
+
+#include "model/material.h"
+
+enum material {
+  bear_matBear,
+  bear_matOutline,
+};
+
+const struct material_descriptor bear_material[] = {
+};

model/material.h

@@ -0,0 +1,25 @@
+#pragma once
+
+#include <stdint.h>
+
+struct pixel_descriptor {
+  uint8_t * start;
+  int32_t size;
+  int32_t vram_offset; // offset into vram texture address space
+
+  int16_t width;
+  int16_t height;
+};
+
+struct palette_descriptor {
+  uint8_t * start;
+  int32_t size;
+  int32_t vram_offset; // offset into vram palette address space
+
+  int32_t palette_size;
+};
+
+struct material_descriptor {
+  struct pixel_descriptor pixel;
+  struct palette_descriptor palette;
+};

model/model.h

@@ -0,0 +1,59 @@
+#pragma once
+
+#include <stdint.h>
+
+#include "math/vec3.hpp"
+#include "math/vec2.hpp"
+
+#ifdef __dreamcast__
+using vertex_position = vec<3, float>;
+using vertex_normal = vec<3, float>;
+using vertex_texture = vec<2, float>;
+#endif
+#ifdef __saturn__
+#include "math/fp.hpp"
+using vertex_position = vec<3, fp16_16>;
+using vertex_normal = vec<3, fp16_16>;
+using vertex_texture = vec<2, fp16_16>;
+#endif
+
+struct index_ptn {
+  uint16_t position;
+  uint16_t texture;
+  uint16_t normal;
+};
+
+union triangle {
+  struct {
+    struct index_ptn a;
+    struct index_ptn b;
+    struct index_ptn c;
+  };
+  struct index_ptn v[3];
+};
+
+union quadrilateral {
+  struct {
+    struct index_ptn a;
+    struct index_ptn b;
+    struct index_ptn c;
+    struct index_ptn d;
+  };
+  struct index_ptn v[4];
+};
+
+struct object {
+  const union triangle * triangle;
+  const union quadrilateral * quadrilateral;
+  const int triangle_count;
+  const int quadrilateral_count;
+  const int material;
+};
+
+struct model {
+  const vertex_position * position;
+  const vertex_texture * texture;
+  const vertex_normal * normal;
+  const struct object ** object;
+  const int object_count;
+};

scu-dsp/div10.asm

@@ -0,0 +1,76 @@
+        mov 0,ct0               ; num
+        mov 0,ct1               ; vdp2 address
+        mov 0,ct3               ; output
+        mvi 12345,mc0                            ; m0[0] (12345)
+
+        ;; output: m3
+base10_loop:
+                                   mov 0,ct0
+        mvi div10_unsigned,pc
+        ;;   [X     ]   [Y     ]   [D1     ]
+                        mov mc0,y   mov 1,lop     ; mvi imm,pc delay slot (executed twice)
+        ;; after function return:
+                                   mov all,mc0   ; m0[1] (1234)
+                                   mov all,pl    ; ??? why can't this happen in the next instruction?
+        ;; mod10: multiply by 10:
+        sl              mov alu,a
+        sl              mov alu,a
+        sl              mov alu,a
+        add             mov alu,a
+        add             mov alu,a  mov 0,ct0     ; restore ct0 after 'mov all,mc1'
+
+        ;;   a: 12340
+                        mov mc0,a  mov all,pl
+
+        ;;   a: 12345   m0[0]
+        ;;   p: 12340
+
+        ;; mod10: subtract (a - p)
+        sub             mov alu,a
+
+        ;;   a: 5
+
+        ;; store digit in m3
+             mov mc0,p  clr a      mov all,mc3   ; m0[1]
+
+        ;;   p: 1234
+        ;;   a: 0
+        add             mov alu,a  mov 0,ct0
+
+        ;;   a: 1234
+        jmp nz,base10_loop
+                                   mov all,mc0   ; jmp delay slot
+
+        endi
+        nop
+
+        ;; argument: ry
+        ;; return: a ← ry / 10
+        ;; maximum ry is somewhere between 2 ^ 21 and 2 ^ 22
+div10_unsigned:
+        ;; 1 / 10 * (2 ^ 24) ~= 1677722 = 0x19999a
+        mvi 1677722,rx
+             mov mul,p  clr a
+        ad2             mov alu,a
+
+        ;;   ALH: [nmlkjihgfedcba9876543210________]
+        ;;   ALL: [76543210________________________]
+
+                        clr a      mov alh,pl ; alh is (a >> 16)
+        add             mov alu,a
+
+        ;;   ALL: [nmlkjihgfedcba9876543210________]
+
+        ;; rotate left 24 requires fewer cycles than shift right 8
+        rl8             mov alu,a
+        rl8             mov alu,a
+        rl8             mov alu,a
+
+        ;;   ALL: [________nmlkjihgfedcba9876543210]
+
+        ;; mask 24 bit result
+        mvi 0xffffff,pl
+
+        ;; return to caller ; reset ct0
+        btm
+        and             mov alu,a  mov 0,ct0

scu-dsp/div10.cpp

@@ -0,0 +1,35 @@
+#include "cdc/serial.h"
+#include "scu.h"
+#include "vdp2.h"
+
+#include "scu-dsp/div10.dsp.h"
+
+void main()
+{
+  scu.reg.PPAF = 0; // stop execution
+  scu.reg.PPAF = (1 << 15) | 0; // reset PC
+
+  uint32_t * program = (uint32_t *)&_binary_scu_dsp_div10_dsp_start;
+  int program_length = ((int)&_binary_scu_dsp_div10_dsp_size) / 4;
+
+  for (int i = 0; i < program_length; i++) {
+    uint32_t p = program[i];
+    scu.reg.PPD = p;
+    serial_integer(p, 8, '\n');
+  }
+
+  scu.reg.PPAF = (1 << 16); // execute
+
+  int end_flag = 0;
+  while (end_flag == 0) {
+    end_flag = (scu.reg.PPAF >> 18) & 1;
+  }
+
+  scu.reg.PPAF = 0;
+  scu.reg.PDA = 64 * 3; // m3[0]
+
+  serial_string("answer:\n");
+  for (int i = 0; i < 8; i++) {
+    serial_integer(scu.reg.PDD, 8, '\n');
+  }
+}

scu-dsp/div10.dsp.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_scu_dsp_div10_dsp_start __asm("_binary_scu_dsp_div10_dsp_start");
+extern uint32_t _binary_scu_dsp_div10_dsp_end __asm("_binary_scu_dsp_div10_dsp_end");
+extern uint32_t _binary_scu_dsp_div10_dsp_size __asm("_binary_scu_dsp_div10_dsp_size");

scu-dsp/div10_vdp2.asm

@@ -0,0 +1,104 @@
+        mov 0,ct0               ; num
+        mov 0,ct1               ; vdp2 address
+        mov 0,ct3               ; output
+        mvi 12345,mc0                            ; m0[0] (12345)
+
+        ;; clear m3
+        mov 7,lop
+        lps
+        mov 0,mc3
+        mov 0,ct3
+
+        ;; output: m3
+base10_loop:
+                                   mov 0,ct0
+        mvi div10_unsigned,pc
+        ;;   [X     ]   [Y     ]   [D1     ]
+                        mov mc0,y   mov 1,lop     ; mvi imm,pc delay slot (executed twice)
+        ;; after function return:
+                                   mov all,mc0   ; m0[1] (1234)
+                                   mov all,pl    ; ??? why can't this happen in the next instruction?
+        ;; mod10: multiply by 10:
+        sl              mov alu,a
+        sl              mov alu,a
+        sl              mov alu,a
+        add             mov alu,a
+        add             mov alu,a  mov 0,ct0     ; restore ct0 after 'mov all,mc1'
+
+        ;;   a: 12340
+                        mov mc0,a  mov all,pl
+
+        ;;   a: 12345   m0[0]
+        ;;   p: 12340
+
+        ;; mod10: subtract (a - p)
+        sub             mov alu,a
+
+        ;;   a: 5
+
+        ;; convert to vdp2 character
+        mvi 16,pl
+        add             mov alu,a
+
+        ;; store digit in m3
+             mov mc0,p  clr a      mov all,mc3   ; m0[1]
+
+        ;;   p: 1234
+        ;;   a: 0
+        add             mov alu,a  mov 0,ct0
+
+        ;;   a: 1234
+        jmp nz,base10_loop
+                                   mov all,mc0   ; jmp delay slot
+
+
+        ;;
+        ;; transfer to vdp2
+        ;;
+
+        ;; vdp2 address calculation
+        mvi ((8 * 0x4000 + (64 - 8) * 4 + 0x05e00000) >> 2),wa0
+                                   mov 0,ct3
+        ;;                 clr a      mov 0,ct3
+        ;; add             mov alu,a
+        ;;                            mov all,wa0
+
+        ;; end vdp2 address calculation
+        dma1 mc3,d0,8
+dma_wait:
+        jmp t0,dma_wait
+        nop
+
+        endi
+        nop
+
+        ;; argument: ry
+        ;; return: a ← ry / 10
+        ;; maximum ry is somewhere between 2 ^ 21 and 2 ^ 22
+div10_unsigned:
+        ;; 1 / 10 * (2 ^ 24) ~= 1677722 = 0x19999a
+        mvi 1677722,rx
+             mov mul,p  clr a
+        ad2             mov alu,a
+
+        ;;   ALH: [nmlkjihgfedcba9876543210________]
+        ;;   ALL: [76543210________________________]
+
+                        clr a      mov alh,pl ; alh is (a >> 16)
+        add             mov alu,a
+
+        ;;   ALL: [nmlkjihgfedcba9876543210________]
+
+        ;; rotate left 24 requires fewer cycles than shift right 8
+        rl8             mov alu,a
+        rl8             mov alu,a
+        rl8             mov alu,a
+
+        ;;   ALL: [________nmlkjihgfedcba9876543210]
+
+        ;; mask 24 bit result
+        mvi 0xffffff,pl
+
+        ;; return to caller ; reset ct0
+        btm
+        and             mov alu,a  mov 0,ct0

scu-dsp/div10_vdp2.cpp

@@ -0,0 +1,171 @@
+#include "cdc/serial.h"
+#include "scu.h"
+#include "vdp2.h"
+#include "../common/vdp2_func.hpp"
+
+#include "scu-dsp/div10_vdp2.dsp.h"
+
+#include "font/hp_100lx_4bit_flip.data.h"
+
+void cell_data()
+{
+  const uint32_t * start = reinterpret_cast<uint32_t *>(&_binary_font_hp_100lx_4bit_flip_data_start);
+  const int size = reinterpret_cast<uint32_t>(&_binary_font_hp_100lx_4bit_flip_data_size);
+
+  // the start of VRAM-A0
+  for (int i = 0; i < (size / 4); i++) {
+    vdp2.vram.u32[i] = start[i];
+  }
+}
+
+void palette_data()
+{
+  vdp2.cram.u16[0] = 0x0000;
+  vdp2.cram.u16[1] = 0xffff;
+}
+
+void vdp2_init()
+{
+  v_blank_in();
+
+  // DISP: Please make sure to change this bit from 0 to 1 during V blank.
+  vdp2.reg.TVMD = ( TVMD__DISP | TVMD__LSMD__NON_INTERLACE
+                  | TVMD__VRESO__240 | TVMD__HRESO__NORMAL_320);
+
+  /* set the color mode to 5bits per channel, 1024 colors */
+  vdp2.reg.RAMCTL = RAMCTL__CRMD__RGB_5BIT_1024
+                  | RAMCTL__VRAMD
+                  | RAMCTL__VRBMD
+                  | RAMCTL__RDBSA0__CHARACTER_PATTERN_TABLE // VRAM-A0 0x000000
+                  | RAMCTL__RDBSA1__PATTERN_NAME_TABLE;     // VRAM-A1 0x020000
+
+  vdp2.reg.VRSIZE = 0;
+
+  /* enable display of NBG0 */
+  vdp2.reg.BGON = BGON__R0ON | BGON__R0TPON;
+
+  /* set character format for NBG0 to palettized 16 color
+     set enable "cell format" for NBG0
+     set character size for NBG0 to 1x1 cell */
+  vdp2.reg.CHCTLB = CHCTLB__R0CHCN__16_COLOR
+                  | CHCTLB__R0BMEN__CELL_FORMAT
+                  | CHCTLB__R0CHSZ__1x1_CELL;
+
+  /* plane size */
+  vdp2.reg.PLSZ = PLSZ__RAPLSZ__1x1
+                | PLSZ__RBPLSZ__1x1;
+
+  /* map plane offset
+     1-word: value of bit 6-0 * 0x2000
+     2-word: value of bit 5-0 * 0x4000
+  */
+  // plane_a_offset is at the start of VRAM-A1
+  constexpr int plane_a = 8;
+  constexpr int plane_a_offset = plane_a * 0x4000;
+
+  constexpr int page_size = 64 * 64 * 2; // N0PNB__1WORD (16-bit)
+  constexpr int plane_size = page_size * 1;
+
+  /* cycle pattern table not used for RBG0 ? */
+  vdp2.reg.CYCA0 = 0x0F44F99F;
+  vdp2.reg.CYCA1 = 0x0F44F99F;
+  vdp2.reg.CYCB0 = 0x0F44F99F;
+  vdp2.reg.CYCB1 = 0x0F44F99F;
+
+  vdp2.reg.MPOFR = MPOFR__RAMP(0); // bits 8~6
+  vdp2.reg.MPABRA = MPABRA__RAMPB(plane_a) | MPABRA__RAMPA(plane_a); // bits 5~0
+  vdp2.reg.MPCDRA = MPCDRA__RAMPD(plane_a) | MPCDRA__RAMPC(plane_a); // bits 5~0
+  vdp2.reg.MPEFRA = MPEFRA__RAMPF(plane_a) | MPEFRA__RAMPE(plane_a); // bits 5~0
+  vdp2.reg.MPGHRA = MPGHRA__RAMPH(plane_a) | MPGHRA__RAMPG(plane_a); // bits 5~0
+  vdp2.reg.MPIJRA = MPIJRA__RAMPJ(plane_a) | MPIJRA__RAMPI(plane_a); // bits 5~0
+  vdp2.reg.MPKLRA = MPKLRA__RAMPL(plane_a) | MPKLRA__RAMPK(plane_a); // bits 5~0
+  vdp2.reg.MPMNRA = MPMNRA__RAMPN(plane_a) | MPMNRA__RAMPM(plane_a); // bits 5~0
+  vdp2.reg.MPOPRA = MPOPRA__RAMPP(plane_a) | MPOPRA__RAMPO(plane_a); // bits 5~0
+
+  vdp2.reg.PNCR = PNCR__R0PNB__2WORD;
+
+  vdp2.reg.RPMD = RPMD__ROTATION_PARAMETER_A;
+
+  //vdp2.reg.RPRCTL = 0;
+
+  vdp2.reg.KTCTL = 0;
+
+  vdp2.reg.KTAOF = 0;
+
+  vdp2.reg.PRIR = 3;
+
+  palette_data();
+  cell_data();
+
+  volatile struct vdp2_rotation_parameter_table * table = (struct vdp2_rotation_parameter_table *)&vdp2.vram.u32[0x4000 / 4];
+  table->screen_start_coordinate_xst = 0;
+  table->screen_start_coordinate_yst = 0;
+  table->screen_start_coordinate_zst = 0;
+  table->screen_vertical_coordinate_increment_dxst = 0;
+  table->screen_vertical_coordinate_increment_dyst = (1 << 16);
+  table->screen_horizontal_coordinate_increment_dx = (1 << 16);
+  table->screen_horizontal_coordinate_increment_dy = 0;
+  table->rotation_matrix_parameter_a = (-1 << 16);
+  table->rotation_matrix_parameter_b = 0;
+  table->rotation_matrix_parameter_c = 0;
+  table->rotation_matrix_parameter_d = 0;
+  table->rotation_matrix_parameter_e = (1 << 16);
+  table->rotation_matrix_parameter_f = 0;
+  table->viewpoint_coordinate_px = 0;
+  table->viewpoint_coordinate_py = 0;
+  table->viewpoint_coordinate_pz = 0;
+  table->center_point_coordinate_px = 0;
+  table->center_point_coordinate_py = 0;
+  table->center_point_coordinate_pz = 0;
+  table->horizontal_shift_mx = 0;
+  table->horizontal_shift_my = 0;
+  table->scaling_coefficient_kx = (1 << 16);
+  table->scaling_coefficient_ky = (1 << 16);
+
+  vdp2.reg.RPTA = (((uint32_t)table) >> 1) & 0x7ffff;
+
+  /* */
+
+  for (int i = 0; i < 64 * 64; i++) {
+    vdp2.vram.u32[(plane_a_offset / 4) + i] = ' ' - 0x20;
+  }
+}
+
+void main()
+{
+  vdp2_init();
+
+  scu.reg.PPAF = 0; // stop execution
+  scu.reg.PPAF = (1 << 15) | 0; // reset PC
+
+  uint32_t * program = (uint32_t *)&_binary_scu_dsp_div10_vdp2_dsp_start;
+  int program_length = ((int)&_binary_scu_dsp_div10_vdp2_dsp_size) / 4;
+
+  for (int i = 0; i < program_length; i++) {
+    uint32_t p = program[i];
+    scu.reg.PPD = p;
+    //serial_integer(p, 8, '\n');
+  }
+
+  scu.reg.PPAF = (1 << 16); // execute
+
+  int end_flag = 0;
+  while (end_flag == 0) {
+    end_flag = (scu.reg.PPAF >> 18) & 1;
+  }
+
+  scu.reg.PPAF = 0;
+  scu.reg.PDA = 64 * 3; // m3[0]
+
+  serial_string("answer:\n");
+  for (int i = 0; i < 8; i++) {
+    serial_integer(scu.reg.PDD, 8, '\n');
+  }
+  serial_string("answer2:\n");
+  volatile uint32_t * buf = &vdp2.vram.u32[(8 * 0x4000) / 4 + 64];
+  for (int i = 0; i < 8; i++) {
+    uint32_t addr = (uint32_t)&buf[-(i + 1)];
+    serial_integer(addr, 8, ' ');
+    serial_integer(buf[-(i + 1)], 8, '\n');
+  }
+}

scu-dsp/div10_vdp2.dsp.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_scu_dsp_div10_vdp2_dsp_start __asm("_binary_scu_dsp_div10_vdp2_dsp_start");
+extern uint32_t _binary_scu_dsp_div10_vdp2_dsp_end __asm("_binary_scu_dsp_div10_vdp2_dsp_end");
+extern uint32_t _binary_scu_dsp_div10_vdp2_dsp_size __asm("_binary_scu_dsp_div10_vdp2_dsp_size");

scu-dsp/parse10.asm

@@ -0,0 +1,93 @@
+        ;; input m0[0] start address
+        ;;   [X       ]   [Y       ]   [D1       ]
+        ;;                             mov 0,ct0
+        ;;                             mov mc0,ra0
+
+        ;; variables:
+        ;; m0[0]
+        ;; m0[1] dma read value (local); post-shift 4-byte remainder (global)
+        ;; m0[2] saved accumulator (global)
+
+setup:
+        ;;   [X       ]   [Y       ]   [D1       ]
+                                       mov 1,ct0
+                                       mov 0,ct3
+                                       mov 0,mc0
+                                       mov 0,mc0
+        mvi parse_base10,pc
+                                       mov 1,lop
+        ;; parse_base10 return
+                                       mov all,mc3
+
+        mvi parse_base10,pc
+                                       mov 1,lop
+        ;; parse_base10 return
+                                       mov all,mc3
+
+        nop
+        nop
+        nop
+
+parse_base10:
+        ;;   [X       ]   [Y       ]   [D1       ]
+                                       mov 1,ct0
+             mov m0,p     mov m0,a
+        add
+        jmp nz,parse_loop
+                          clr a        ; delay slot
+dma_next_chunk:
+        ;; dma d0,mc0,1
+        ;; begin fake number 0x39322033 "92 3"
+        mvi 0x393220,pl
+        add               mov alu,a    mov 0x33,pl
+        rl8               mov alu,a    mov 1,ct0
+        add                            mov all,mc0
+        ;; end fake number
+end_dma_next_chunk:
+parse_loop:
+        ;; extract first (leftmost) digit
+                                       mov 1,ct0
+                          mov mc0,a    mov 0x7f,pl
+        or
+        jmp s,end_of_input
+        rl8               mov alu,a    mov 10,rx     ; m0[2] 0x00000000
+        jmp z,dma_next_chunk
+        and               mov m0,y     mov all,pl
+
+        ;; MUL: post-multiplication accumulator
+        ;; P: post-mask digit
+        ;; A: pre-xor string
+                                       mov 1,ct0
+        xor               clr a        mov all,mc0 ; m0[1] 0x20323300
+
+        add               mov alu,a    mov 0x30,pl ; 0x30 ascii '0'
+
+        ;; A: post-mask digit
+        ;;
+        sub               mov alu,a    mov 10,pl
+
+        ;; if ((0x39 - 0x30) < 0) goto non_digit;
+        jmp s,non_digit
+        ;; possibly set S flag for `jmp ns,non_digit`
+        sub                            mov 2,ct0   ; sub in delay slot
+        ;; if (((0x39 - 0x30) - 10) >= 0) goto non_digit;
+        jmp ns,non_digit
+                          mov mul,p    mov 2,ct0   ; mov in delay slot
+        ;; P: accumulator * 10 result
+digit:
+        ;; we have a valid digit; add it to the accumulator; save to m0[2]
+        ;;   [X       ]   [Y       ]   [D1       ]
+        jmp parse_loop
+        add                            mov all,mc0 ; add in delay slot ; m0[2] 0x00000009
+
+non_digit:
+        ;; this is not a valid digit; return and clear the accumulator as A
+        ;;   [X       ]   [Y       ]   [D1       ]
+        btm
+                          mov m0,a     mov 0,mc0   ; clr in delay slot ; m0[2]
+
+        ;; m0:
+        ;; 00000000 32330000 00000009 0000000
+end_of_input:
+        btm
+                          clr a

tools/ttf_bitmap2.cpp

@@ -0,0 +1,164 @@
+#include <assert.h>
+#include <stdint.h>
+#include <stdio.h>
+
+#include <ft2build.h>
+#include FT_FREETYPE_H
+
+int
+load_bitmap_char(FT_Face face,
+		 FT_ULong char_code,
+                 bool hflip,
+		 uint8_t * buf)
+{
+  FT_Error error;
+  FT_UInt glyph_index = FT_Get_Char_Index(face, char_code);
+
+  error = FT_Load_Glyph(face, glyph_index, FT_LOAD_DEFAULT);
+  if (error) {
+    fprintf(stderr, "FT_Load_Glyph %s\n", FT_Error_String(error));
+    return -1;
+  }
+
+  printf("horiBearingX: %ld\n", face->glyph->metrics.horiBearingX >> 6);
+  printf("horiBearingY: %ld\n", face->glyph->metrics.horiBearingY >> 6);
+  printf("horiAdvance: %ld\n", face->glyph->metrics.horiAdvance >> 6);
+  printf("width: %ld\n", face->glyph->metrics.width >> 6);
+  printf("height: %ld\n", face->glyph->metrics.height >> 6);
+
+  assert(face->glyph->format == FT_GLYPH_FORMAT_BITMAP);
+  assert(face->glyph->bitmap.num_grays == 2);
+
+  for (int y = 0; y < (int)face->glyph->bitmap.rows; y++) {
+    uint8_t * row = &face->glyph->bitmap.buffer[y * face->glyph->bitmap.pitch];
+    for (int x = 0; x < (int)face->glyph->bitmap.width; x += 1) {
+      const int bit = (row[x / 8] >> (7 - (x % 8))) & 1;
+      //std::cerr << (bit ? "█" : " ");
+      if (hflip)
+        buf[y * face->glyph->bitmap.width + (7 - x)] = bit;
+      else
+        buf[y * face->glyph->bitmap.width + x] = bit;
+    }
+    //std::cerr << "|\n";
+  }
+
+  return face->glyph->bitmap.rows * face->glyph->bitmap.width;
+}
+
+int load_font(FT_Library * library, FT_Face * face, const char * font_file_path)
+{
+  FT_Error error;
+
+  error = FT_Init_FreeType(library);
+  if (error) {
+    fprintf(stderr, "FT_Init_FreeType\n");
+    return -1;
+  }
+
+  error = FT_New_Face(*library, font_file_path, 0, face);
+  if (error) {
+    fprintf(stderr, "FT_New_Face\n");
+    return -1;
+  }
+
+  error = FT_Select_Size(*face, 0);
+  if (error) {
+    fprintf(stderr, "FT_Select_Size: %d: %s\n", error, FT_Error_String(error));
+    return -1;
+  }
+
+  return 0;
+}
+
+void usage(const char * argv_0)
+{
+  printf("%s [start-hex] [end-hex] [font-width] [font-height] [hflip] [font-file-path] [output-file-path]\n", argv_0);
+}
+
+void pack_4bit(const uint8_t * src, int width, int height, int size, uint8_t * dst)
+{
+  int offset = 0;
+  while (offset < size) {
+    for (int y = 0; y < height; y++) {
+      for (int x = 0; x < width; x += 2) {
+        int px0 = src[offset + y * width + x + 0];
+        int px1 = src[offset + y * width + x + 1];
+        dst[(offset / 2) + y * (width / 2) + (x / 2)] = (px0 << 4) | (px1 << 0);
+      }
+    }
+    offset += width * height;
+  }
+}
+
+int main(int argc, const char * argv[])
+{
+  if (argc != 8) {
+    usage(argv[0]);
+    return -1;
+  }
+
+  char * endptr;
+
+  int start_hex = strtol(argv[1], &endptr, 16);
+  assert(*endptr == 0);
+  int end_hex = strtol(argv[2], &endptr, 16);
+  assert(*endptr == 0);
+
+  int font_width = strtol(argv[3], &endptr, 10);
+  assert(*endptr == 0);
+  int font_height = strtol(argv[4], &endptr, 10);
+  assert(*endptr == 0);
+
+  int hflip = strtol(argv[5], &endptr, 10);
+  assert(*endptr == 0);
+  assert(hflip == 0 || hflip == 1);
+
+  const char * font_file_path = argv[6];
+  const char * output_file_path = argv[7];
+
+  printf("start_hex %x\n", start_hex);
+  printf("end_hex %x\n", start_hex);
+  printf("font_width %d\n", font_width);
+  printf("font_height %d\n", font_height);
+  printf("font_file_path %s\n", font_file_path);
+  printf("output_file_path %s\n", output_file_path);
+
+  FT_Library library;
+  FT_Face face;
+  int res;
+  res = load_font(&library, &face, font_file_path);
+  if (res < 0)
+    return -1;
+
+  int texture_buf_size = font_width * font_height * ((end_hex - start_hex) + 1);
+  uint8_t * texture = (uint8_t *)malloc(texture_buf_size);
+
+  int offset = 0;
+
+  for (int char_code = start_hex; char_code <= end_hex; char_code++) {
+    assert(offset < texture_buf_size);
+
+    res = load_bitmap_char(face,
+                           char_code,
+                           hflip,
+                           &texture[offset]);
+    if (res < 0)
+      return - 1;
+
+    assert(res == font_width * font_height);
+
+    offset += res;
+  }
+
+  uint8_t * pack = (uint8_t *)malloc(texture_buf_size / 2);
+  pack_4bit(texture, font_width, font_height, offset, pack);
+
+  FILE * out = fopen(output_file_path, "w");
+  if (out == NULL) {
+    perror("fopen(w)");
+    return -1;
+  }
+  //fwrite((void *)texture, texture_buf_size, 1, out);
+  fwrite((void *)pack, texture_buf_size / 2, 1, out);
+  fclose(out);
+}

vdp1/bear.cpp

@@ -0,0 +1,243 @@
+#include <stdint.h>
+#include "vdp2.h"
+#include "vdp1.h"
+
+#include <concepts>
+
+#include "common/vdp2_func.hpp"
+
+#include "math/fp.hpp"
+#include "math/vec3.hpp"
+#include "math/mat4x4.hpp"
+
+#include "cos.hpp"
+
+#include "model/model.h"
+#include "model/bear/material.h"
+#include "model/bear/model.h"
+
+// |--
+// |
+// |
+
+using vec3 = vec<3, fp16_16>;
+using mat4x4 = mat<4, 4, fp16_16>;
+
+struct canvas {
+  fp16_16 width;
+  fp16_16 height;
+};
+
+constexpr struct canvas canvas = { 240, 240 };
+
+template <typename T>
+vec<3, T> viewport_to_canvas(T x, T y)
+{
+  return vec<3, T>(x * canvas.width, y * canvas.height, T(1));
+}
+
+template <typename T>
+inline constexpr vec<3, T> project_vertex(vec<3, T> const& v)
+{
+  // / (v.z - T(5))
+  // / (v.z - T(5))
+  return viewport_to_canvas<T>((v.x * T(0.5) + T(2.0/3.0)),
+			       (v.y * T(0.5) + T(0.5)));
+}
+
+constexpr inline uint16_t rgb15(int32_t r, int32_t g, int32_t b)
+{
+  return ((b & 31) << 10) | ((g & 31) << 5) | ((r & 31) << 0);
+}
+
+constexpr uint16_t colors[] = {
+  rgb15(31,  0,  0), // red
+  rgb15( 0, 31,  0), // green
+  rgb15( 0,  0, 31), // blue
+  rgb15(31,  0, 31), // magenta
+  rgb15( 0, 31, 31), // cyan
+  rgb15(31, 31,  0), // yellow
+};
+
+static int32_t tick = 0;
+
+static inline void render_quad(int ix, const vec3 a, const vec3 b, const vec3 c, const vec3 d)
+{
+  vdp1.vram.cmd[ix].CTRL = CTRL__JP__JUMP_NEXT | CTRL__COMM__POLYGON;
+  vdp1.vram.cmd[ix].LINK = 0;
+  vdp1.vram.cmd[ix].PMOD = PMOD__ECD | PMOD__SPD;
+  vdp1.vram.cmd[ix].COLR = COLR__RGB | colors[ix & 3];
+
+  vdp1.vram.cmd[ix].A.X = static_cast<int>(a.x);
+  vdp1.vram.cmd[ix].A.Y = static_cast<int>(a.y);
+  vdp1.vram.cmd[ix].B.X = static_cast<int>(b.x);
+  vdp1.vram.cmd[ix].B.Y = static_cast<int>(b.y);
+  vdp1.vram.cmd[ix].C.X = static_cast<int>(c.x);
+  vdp1.vram.cmd[ix].C.Y = static_cast<int>(c.y);
+  vdp1.vram.cmd[ix].D.X = static_cast<int>(d.x);
+  vdp1.vram.cmd[ix].D.Y = static_cast<int>(d.y);
+}
+
+int render_object(int ix,
+                  const mat4x4& screen,
+                  const struct model * model,
+                  const struct object * object0)
+{
+  mat4x4 trans = screen;
+
+  for (int i = 0; i < object0->quadrilateral_count; i++) {
+
+    const union quadrilateral * quad0 = &object0->quadrilateral[i];
+
+    vec3 a0 = model->position[quad0->v[0].position];
+    vec3 b0 = model->position[quad0->v[1].position];
+    vec3 c0 = model->position[quad0->v[2].position];
+    vec3 d0 = model->position[quad0->v[3].position];
+
+    vec3 a = trans * a0;
+    vec3 b = trans * b0;
+    vec3 c = trans * c0;
+    vec3 d = trans * d0;
+
+    vec3 an0 = trans * model->normal[quad0->v[0].normal];
+    //vec3 l = {0, 0, 1};
+    fp16_16 cull = dot(a, an0);
+
+    if (cull.value > 0) {
+      // `origin` above is p==0 below; the `origin` calculation could
+      // be reused, though it would hurt readability slightly
+      /*
+      for (int p = 0; p < 4; p++) {
+	const vec3& v0 = vertices[face[p]];
+
+	// rotation
+	const vec3 v1 = transform * v0;
+	// translation
+	const vec3 v2{v1.x, v1.y + fp16_16(1), v1.z + fp16_16(2)};
+
+	const vec3 v3 = project_vertex(v2 / v2.z);
+      }
+      */
+
+      render_quad(ix,
+                  project_vertex(a / a.z),
+                  project_vertex(b / b.z),
+                  project_vertex(c / c.z),
+                  project_vertex(d / d.z));
+
+      ix++;
+    }
+  }
+
+  return ix;
+}
+
+void render()
+{
+  tick++;
+  int ix = 2;
+
+  const mat4x4 scale {
+    1, 0, 0, 0,
+    0, -1, 0, 0,
+    0, 0, 1, 0,
+    0, 0, 0, 1,
+  };
+
+  const int rx = tick >> 2;
+  const mat4x4 rotationX {
+    1,        0,       0, 0,
+    0,  cos(rx), sin(rx), 0,
+    0, -sin(rx), cos(rx), 0,
+    0,        0,       0, 1,
+  };
+
+  const int ry = tick >> 1;
+  const mat4x4 rotationY {
+    cos(ry), 0, -sin(ry), 0,
+          0, 1,        0, 0,
+    sin(ry), 0,  cos(ry), 0,
+          0, 0,        0, 1,
+  };
+
+  const mat4x4 translation = {
+    1, 0, 0, 0,
+    0, 1, 0, 4.5,
+    0, 0, 1, 6.5,
+    0, 0, 0, 1,
+  };
+
+  const mat4x4 screen = translation * rotationY * scale;
+
+  //const vec3 camera = {0, 0, 0};
+
+  const int frame_ix0 = 0;
+  const struct model * model = &bear_model;
+  const struct object * object0 = bear_object[frame_ix0 * 2];
+  //const struct object * object1 = bear_object[frame_ix1 * 2];
+
+  ix = render_object(ix, screen, model, object0);
+
+  vdp1.vram.cmd[ix].CTRL = CTRL__END;
+}
+
+void main()
+{
+  v_blank_in();
+
+  // DISP: Please make sure to change this bit from 0 to 1 during V blank.
+  vdp2.reg.TVMD = ( TVMD__DISP | TVMD__LSMD__NON_INTERLACE
+                  | TVMD__VRESO__240 | TVMD__HRESO__NORMAL_320);
+
+  // disable all VDP2 backgrounds (e.g: the Sega bios logo)
+  vdp2.reg.BGON = 0;
+
+  // VDP2 User's Manual:
+  // "When sprite data is in an RGB format, sprite register 0 is selected"
+  // "When the value of a priority number is 0h, it is read as transparent"
+  //
+  // The power-on value of PRISA is zero. Set the priority for sprite register 0
+  // to some number greater than zero, so that the color data is not interpreted
+  // as "transparent".
+  vdp2.reg.PRISA = PRISA__S0PRIN(1); // Sprite register 0 Priority Number
+
+  /* TVM settings must be performed from the second H-blank IN interrupt after the
+  V-blank IN interrupt to the H-blank IN interrupt immediately after the V-blank
+  OUT interrupt. */
+  // "normal" display resolution, 16 bits per pixel, 512x256 framebuffer
+  vdp1.reg.TVMR = TVMR__TVM__NORMAL;
+
+  // swap framebuffers every 1 cycle; non-interlace
+  vdp1.reg.FBCR = 0;
+
+  // during a framebuffer erase cycle, write the color "black" to each pixel
+  constexpr uint16_t black = 0x0000;
+  vdp1.reg.EWDR = black;
+
+  // the EWLR/EWRR macros use somewhat nontrivial math for the X coordinates
+  // erase upper-left coordinate
+  vdp1.reg.EWLR = EWLR__16BPP_X1(0) | EWLR__Y1(0);
+
+  // erase lower-right coordinate
+  vdp1.reg.EWRR = EWRR__16BPP_X3(319) | EWRR__Y3(239);
+
+  vdp1.vram.cmd[0].CTRL = CTRL__JP__JUMP_NEXT | CTRL__COMM__SYSTEM_CLIP_COORDINATES;
+  vdp1.vram.cmd[0].LINK = 0;
+  vdp1.vram.cmd[0].XC = 319;
+  vdp1.vram.cmd[0].YC = 239;
+
+  vdp1.vram.cmd[1].CTRL = CTRL__JP__JUMP_NEXT | CTRL__COMM__LOCAL_COORDINATE;
+  vdp1.vram.cmd[1].LINK = 0;
+  vdp1.vram.cmd[1].XA = 0;
+  vdp1.vram.cmd[1].YA = 0;
+
+  vdp1.vram.cmd[2].CTRL = CTRL__END;
+
+  // start drawing (execute the command list) on every frame
+  vdp1.reg.PTMR = PTMR__PTM__FRAME_CHANGE;
+
+  while (true) {
+    v_blank_in();
+    render();
+  }
+}

vdp2/line_color_screen.cpp

@@ -0,0 +1,189 @@
+/*
+  color data
+
+  255 colors, 3 bytes per channel, packed
+*/
+/*
+  image data
+
+  64*64 px, 1 byte per palette index
+*/
+
+#include <stdint.h>
+
+#include "vdp2.h"
+#include "../common/vdp2_func.hpp"
+
+inline constexpr uint16_t rgb555(int r, int g, int b)
+{
+  if (r > 255) r = 255;
+  if (g > 255) g = 255;
+  if (b > 255) b = 255;
+  if (r < 0) r = 0;
+  if (g < 0) g = 0;
+  if (b < 0) b = 0;
+
+  return ((b >> 3) << 10) // blue
+       | ((g >> 3) << 5)  // green
+       | ((r >> 3) << 0); // red
+}
+
+template <typename T>
+void fill(T * buf, T v, int32_t n) noexcept
+{
+  while (n > 0) {
+    *buf++ = v;
+    n -= (sizeof (T));
+  }
+}
+
+struct color {
+  uint8_t r;
+  uint8_t g;
+  uint8_t b;
+};
+
+struct interp {
+  int line;
+  struct color color;
+};
+
+struct interp gradient[] = {
+  {  0, {0x59, 0x6b, 0x88}},
+  { 40, {0x59, 0x6b, 0x88}},
+  { 80, {0xa8, 0x34, 0x14}},
+  {110, {0xac, 0xa5, 0x30}},
+  {120, {0xac, 0xa5, 0x30}},
+  {140, {0xa8, 0x34, 0x14}},
+  {999, {0xa8, 0x34, 0x14}},
+};
+
+const int gradient_len = (sizeof (gradient)) / (sizeof (gradient[0]));
+
+struct color lerp(struct color a, struct color b, int start, int end, int y)
+{
+  if (start == end) {
+    return (struct color){a.r, a.g, a.b};
+  }
+  if (end == 999) {
+    return (struct color){a.r, a.g, a.b};
+  }
+
+  int dr = ((int)b.r - (int)a.r);
+  int dg = ((int)b.g - (int)a.g);
+  int db = ((int)b.b - (int)a.b);
+
+  int t1 = end - start;
+  int t0 = y - start;
+
+  return (struct color){
+    (uint8_t)((int)a.r + dr * t0 / t1),
+    (uint8_t)((int)a.g + dg * t0 / t1),
+    (uint8_t)((int)a.b + db * t0 / t1),
+  };
+}
+
+void gen_gradient()
+{
+  struct interp * prev = &gradient[0];
+  struct interp * cur = prev;
+
+  int i = 0;
+
+  for (int y = 0; y < 240; y++) {
+    if (y >= cur->line) {
+      prev = cur;
+      i += 1;
+      cur = &gradient[i];
+    }
+
+    struct color c = lerp(prev->color, cur->color, prev->line, cur->line, y);
+    vdp2.vram.u16[y] = rgb555(c.r, c.g, c.b);
+    //vdp2.vram.u16[y] = rgb555(cur->color.r, cur->color.g, cur->color.b);
+  }
+}
+
+void main()
+{
+  v_blank_in();
+
+  // DISP: Please make sure to change this bit from 0 to 1 during V blank.
+  vdp2.reg.TVMD = ( TVMD__DISP | TVMD__LSMD__NON_INTERLACE
+                  | TVMD__VRESO__240 | TVMD__HRESO__NORMAL_320);
+
+
+  /* set the color mode to 5bits per channel, 1024 colors */
+  vdp2.reg.RAMCTL = RAMCTL__CRMD__RGB_5BIT_1024;
+
+  /* enable display of NBG0 */
+  //vdp2.reg.BGON = BGON__N0ON;
+
+  /* set character format for NBG0 to palettized 2048 color
+     set enable "cell format" for NBG0
+     set character size for NBG0 to 1x1 cell */
+  vdp2.reg.CHCTLA = CHCTLA__N0CHCN__2048_COLOR
+                  | CHCTLA__N0BMEN__CELL_FORMAT
+                  | CHCTLA__N0CHSZ__1x1_CELL;
+  /* "Note: In color RAM modes 0 and 2, 2048-color becomes 1024-color" */
+
+  /* use 1-word (16-bit) pattern names */
+  vdp2.reg.PNCN0 = PNCN0__N0PNB__1WORD;
+
+  /* plane size */
+  vdp2.reg.PLSZ = PLSZ__N0PLSZ__1x1;
+
+  /* map plane offset
+     1-word: value of bit 6-0 * 0x2000
+     2-word: value of bit 5-0 * 0x4000
+  */
+  constexpr int plane_a = 2;
+  constexpr int plane_a_offset = plane_a * 0x2000;
+
+  constexpr int page_size = 64 * 64 * 2; // N0PNB__1WORD (16-bit)
+  constexpr int plane_size = page_size * 1;
+
+  vdp2.reg.MPOFN = MPOFN__N0MP(0); // bits 8~6
+  vdp2.reg.MPABN0 = MPABN0__N0MPB(plane_a) | MPABN0__N0MPA(plane_a); // bits 5~0
+  vdp2.reg.MPCDN0 = MPCDN0__N0MPD(plane_a) | MPCDN0__N0MPC(plane_a); // bits 5~0
+
+  constexpr int cell_size = (8 * 8) * 2; // N0CHCN__2048_COLOR (16-bit)
+  constexpr int character_size = cell_size * (1 * 1); // N0CHSZ__1x1_CELL
+  constexpr int character_offset = character_size / 0x20;
+
+  // zeroize character/cell data from 0 up to plane_a_offset
+  fill<uint32_t>(&vdp2.vram.u32[(0 / 2)], 0, plane_a_offset);
+
+  // "zeroize" plane_a to the 0x40th (64th) character index (an unused/transparent character)
+  // this creates the '40' indexes in the below picture
+  fill<uint16_t>(&vdp2.vram.u16[(plane_a_offset / 2)], character_offset * 0x40, plane_size);
+
+  constexpr int pixel_width = 64;
+  constexpr int pixel_height = 64;
+  constexpr int cell_horizontal = pixel_width / 8;
+  constexpr int cell_vertical = pixel_height / 8;
+
+  constexpr int page_width = 64;
+
+  /*
+  set plane_a character index data to this (hex):
+
+                            |
+     00 01 02 03 04 05 06 07 40 40 ..
+     08 09 0a 0b 0c 0d 0e 0f 40 40 ..
+     10 11 12 13 14 15 16 17 40 40 ..
+     18 19 1a 1b 1c 1d 1e 1f 40 40 ..
+     20 21 22 23 24 25 26 27 40 40 ..
+     28 29 2a 2b 2c 2d 2e 2f 40 40 ..
+     30 31 32 33 34 35 36 37 40 40 ..
+  -- 38 39 3a 3b 3c 3d 3e 3f 40 40 ..
+     40 40 40 40 40 40 40 40 40 40 ..
+     40 40 40 40 40 40 40 40 40 40 ..
+     .. .. .. .. .. .. .. .. .. .. ..
+
+  (the above numbers are not pre-multiplied by character_offset)
+  */
+
+  vdp2.reg.BKTA = BKTA__BKCLMD_PER_LINE;
+
+  gen_gradient();
+}

vdp2/nbg0_font.cpp

@@ -0,0 +1,86 @@
+#include <stdint.h>
+
+#include "vdp2.h"
+#include "../common/vdp2_func.hpp"
+
+#include "font/hp_100lx_4bit.data.h"
+
+void cell_data()
+{
+  const uint32_t * start = reinterpret_cast<uint32_t *>(&_binary_font_hp_100lx_4bit_data_start);
+  const int size = reinterpret_cast<uint32_t>(&_binary_font_hp_100lx_4bit_data_size);
+
+  for (int i = 0; i < (size / 4); i++) {
+    vdp2.vram.u32[i] = start[i];
+  }
+}
+
+void palette_data()
+{
+  vdp2.cram.u16[0] = 0x0000;
+  vdp2.cram.u16[1] = 0xffff;
+}
+
+void main()
+{
+  v_blank_in();
+
+  // DISP: Please make sure to change this bit from 0 to 1 during V blank.
+  vdp2.reg.TVMD = ( TVMD__DISP | TVMD__LSMD__NON_INTERLACE
+                  | TVMD__VRESO__240 | TVMD__HRESO__NORMAL_320);
+
+  /* set the color mode to 5bits per channel, 1024 colors */
+  vdp2.reg.RAMCTL = RAMCTL__CRMD__RGB_5BIT_1024
+                  | RAMCTL__VRAMD | RAMCTL__VRBMD;
+
+  vdp2.reg.VRSIZE = 0;
+
+  /* enable display of NBG0 */
+  vdp2.reg.BGON = BGON__N0ON | BGON__N0TPON;
+
+  /* set character format for NBG0 to palettized 16 color
+     set enable "cell format" for NBG0
+     set character size for NBG0 to 1x1 cell */
+  vdp2.reg.CHCTLA = CHCTLA__N0CHCN__16_COLOR
+                  | CHCTLA__N0BMEN__CELL_FORMAT
+                  | CHCTLA__N0CHSZ__1x1_CELL;
+
+  /* plane size */
+  vdp2.reg.PLSZ = PLSZ__N0PLSZ__1x1;
+
+  /* map plane offset
+     1-word: value of bit 6-0 * 0x2000
+     2-word: value of bit 5-0 * 0x4000
+  */
+  constexpr int plane_a = 1;
+  constexpr int plane_a_offset = plane_a * 0x4000;
+
+  constexpr int page_size = 64 * 64 * 2; // N0PNB__1WORD (16-bit)
+  constexpr int plane_size = page_size * 1;
+
+  vdp2.reg.CYCA0 = 0x0F44F99F;
+  vdp2.reg.CYCA1 = 0x0F44F99F;
+  vdp2.reg.CYCB0 = 0x0F44F99F;
+  vdp2.reg.CYCB1 = 0x0F44F99F;
+
+  vdp2.reg.MPOFN = MPOFN__N0MP(0); // bits 8~6
+  vdp2.reg.MPABN0 = MPABN0__N0MPB(plane_a) | MPABN0__N0MPA(plane_a); // bits 5~0
+  vdp2.reg.MPCDN0 = MPCDN0__N0MPD(plane_a) | MPCDN0__N0MPC(plane_a); // bits 5~0
+
+  palette_data();
+  cell_data();
+
+  vdp2.reg.PNCN0 = PNCN0__N0PNB__2WORD;
+  for (int i = 0; i < 64 * 64; i++) {
+    vdp2.vram.u32[(plane_a_offset / 4) + i] = ' ' - 0x20;
+  }
+
+  const char * test = "conversion from 8";
+  int ix = 0;
+  while (*test) {
+    uint8_t c = *test++;
+    vdp2.vram.u32[(plane_a_offset / 4) + ix++] = c - 0x20;
+  }
+
+  while (1);
+}

vdp2/rbg0_font.cpp

@@ -0,0 +1,139 @@
+#include <stdint.h>
+
+#include "vdp2.h"
+#include "../common/vdp2_func.hpp"
+
+#include "font/hp_100lx_4bit.data.h"
+
+void cell_data()
+{
+  const uint32_t * start = reinterpret_cast<uint32_t *>(&_binary_font_hp_100lx_4bit_data_start);
+  const int size = reinterpret_cast<uint32_t>(&_binary_font_hp_100lx_4bit_data_size);
+
+  // the start of VRAM-A0
+  for (int i = 0; i < (size / 4); i++) {
+    vdp2.vram.u32[i] = start[i];
+  }
+}
+
+void palette_data()
+{
+  vdp2.cram.u16[0] = 0x0000;
+  vdp2.cram.u16[1] = 0xffff;
+}
+
+void main()
+{
+  v_blank_in();
+
+  // DISP: Please make sure to change this bit from 0 to 1 during V blank.
+  vdp2.reg.TVMD = ( TVMD__DISP | TVMD__LSMD__NON_INTERLACE
+                  | TVMD__VRESO__240 | TVMD__HRESO__NORMAL_320);
+
+  /* set the color mode to 5bits per channel, 1024 colors */
+  vdp2.reg.RAMCTL = RAMCTL__CRMD__RGB_5BIT_1024
+                  | RAMCTL__VRAMD
+                  | RAMCTL__VRBMD
+                  | RAMCTL__RDBSA0__CHARACTER_PATTERN_TABLE // VRAM-A0 0x000000
+                  | RAMCTL__RDBSA1__PATTERN_NAME_TABLE;     // VRAM-A1 0x020000
+
+  vdp2.reg.VRSIZE = 0;
+
+  /* enable display of NBG0 */
+  vdp2.reg.BGON = BGON__R0ON | BGON__R0TPON;
+
+  /* set character format for NBG0 to palettized 16 color
+     set enable "cell format" for NBG0
+     set character size for NBG0 to 1x1 cell */
+  vdp2.reg.CHCTLB = CHCTLB__R0CHCN__16_COLOR
+                  | CHCTLB__R0BMEN__CELL_FORMAT
+                  | CHCTLB__R0CHSZ__1x1_CELL;
+
+  /* plane size */
+  vdp2.reg.PLSZ = PLSZ__RAPLSZ__1x1
+                | PLSZ__RBPLSZ__1x1;
+
+  /* map plane offset
+     1-word: value of bit 6-0 * 0x2000
+     2-word: value of bit 5-0 * 0x4000
+  */
+  // plane_a_offset is at the start of VRAM-A1
+  constexpr int plane_a = 8;
+  constexpr int plane_a_offset = plane_a * 0x4000;
+
+  constexpr int page_size = 64 * 64 * 2; // N0PNB__1WORD (16-bit)
+  constexpr int plane_size = page_size * 1;
+
+  /* cycle pattern table not used for RBG0 ? */
+  vdp2.reg.CYCA0 = 0x0F44F99F;
+  vdp2.reg.CYCA1 = 0x0F44F99F;
+  vdp2.reg.CYCB0 = 0x0F44F99F;
+  vdp2.reg.CYCB1 = 0x0F44F99F;
+
+  vdp2.reg.MPOFR = MPOFR__RAMP(0); // bits 8~6
+  vdp2.reg.MPABRA = MPABRA__RAMPB(plane_a) | MPABRA__RAMPA(plane_a); // bits 5~0
+  vdp2.reg.MPCDRA = MPCDRA__RAMPD(plane_a) | MPCDRA__RAMPC(plane_a); // bits 5~0
+  vdp2.reg.MPEFRA = MPEFRA__RAMPF(plane_a) | MPEFRA__RAMPE(plane_a); // bits 5~0
+  vdp2.reg.MPGHRA = MPGHRA__RAMPH(plane_a) | MPGHRA__RAMPG(plane_a); // bits 5~0
+  vdp2.reg.MPIJRA = MPIJRA__RAMPJ(plane_a) | MPIJRA__RAMPI(plane_a); // bits 5~0
+  vdp2.reg.MPKLRA = MPKLRA__RAMPL(plane_a) | MPKLRA__RAMPK(plane_a); // bits 5~0
+  vdp2.reg.MPMNRA = MPMNRA__RAMPN(plane_a) | MPMNRA__RAMPM(plane_a); // bits 5~0
+  vdp2.reg.MPOPRA = MPOPRA__RAMPP(plane_a) | MPOPRA__RAMPO(plane_a); // bits 5~0
+
+  vdp2.reg.PNCR = PNCR__R0PNB__2WORD;
+
+  vdp2.reg.RPMD = RPMD__ROTATION_PARAMETER_A;
+
+  //vdp2.reg.RPRCTL = 0;
+
+  vdp2.reg.KTCTL = 0;
+
+  vdp2.reg.KTAOF = 0;
+
+  vdp2.reg.PRIR = 3;
+
+  palette_data();
+  cell_data();
+
+  volatile struct vdp2_rotation_parameter_table * table = (struct vdp2_rotation_parameter_table *)&vdp2.vram.u32[0x4000 / 4];
+  table->screen_start_coordinate_xst = 0;
+  table->screen_start_coordinate_yst = 0;
+  table->screen_start_coordinate_zst = 0;
+  table->screen_vertical_coordinate_increment_dxst = 0;
+  table->screen_vertical_coordinate_increment_dyst = (1 << 16);
+  table->screen_horizontal_coordinate_increment_dx = (1 << 16);
+  table->screen_horizontal_coordinate_increment_dy = 0;
+  table->rotation_matrix_parameter_a = (-1 << 16);
+  table->rotation_matrix_parameter_b = 0;
+  table->rotation_matrix_parameter_c = 0;
+  table->rotation_matrix_parameter_d = 0;
+  table->rotation_matrix_parameter_e = (1 << 16);
+  table->rotation_matrix_parameter_f = 0;
+  table->viewpoint_coordinate_px = 0;
+  table->viewpoint_coordinate_py = 0;
+  table->viewpoint_coordinate_pz = 0;
+  table->center_point_coordinate_px = 0;
+  table->center_point_coordinate_py = 0;
+  table->center_point_coordinate_pz = 0;
+  table->horizontal_shift_mx = 0;
+  table->horizontal_shift_my = 0;
+  table->scaling_coefficient_kx = (1 << 16);
+  table->scaling_coefficient_ky = (1 << 16);
+
+  vdp2.reg.RPTA = (((uint32_t)table) >> 1) & 0x7ffff;
+
+  /* */
+
+  for (int i = 0; i < 64 * 64; i++) {
+    vdp2.vram.u32[(plane_a_offset / 4) + i] = ' ' - 0x20;
+  }
+
+  const char * test = "If sound level changes progressively, volume on 3 bit makes audi";
+  int ix = 0;
+  while (*test) {
+    uint8_t c = *test++;
+    vdp2.vram.u32[(plane_a_offset / 4) + ix++] = c - 0x20;
+  }
+
+  while (1);
+}

vec4.hpp

@@ -0,0 +1,170 @@
+#pragma once
+
+#include "math.hpp"
+#include "vec.hpp"
+
+//
+// vec4
+//
+
+template <typename T>
+struct vec<4, T>
+{
+  union {
+    struct { T x, y, z, w; };
+    struct { T a, r, g, b; };
+  };
+
+  inline constexpr vec();
+  inline constexpr vec(T scalar);
+  inline constexpr vec(T _x, T _y, T _z, T _w);
+  inline constexpr vec(const vec<3, T>& v);
+
+  constexpr inline vec<4, T> operator-() const;
+  inline constexpr T const& operator[](int i) const;
+  inline constexpr vec<4, T>& operator=(vec<4, T> const& v);
+  inline constexpr vec<4, T>& operator+=(vec<4, T> const& v);
+  inline constexpr vec<4, T>& operator-=(vec<4, T> const& v);
+};
+
+template <typename T>
+inline constexpr vec<4, T>::vec()
+  : x(0), y(0), z(0), w(0)
+{}
+
+template <typename T>
+inline constexpr vec<4, T>::vec(T scalar)
+  : x(scalar), y(scalar), z(scalar), w(scalar)
+{}
+
+template <typename T>
+inline constexpr vec<4, T>::vec(T _x, T _y, T _z, T _w)
+  : x(_x), y(_y), z(_z), w(_w)
+{}
+
+template <typename T>
+inline constexpr vec<4, T>::vec(const vec<3, T>& v)
+  : x(v.x), y(v.y), z(v.z), w(1.f)
+{}
+
+template <typename T>
+constexpr inline vec<4, T> vec<4, T>::operator-() const
+{
+  return vec<4, T>(-x, -y, -z, -w);
+}
+
+template <typename T>
+inline constexpr T const& vec<4, T>::operator[](int i) const
+{
+  switch(i)
+  {
+  default: [[fallthrough]];
+  case 0: return x;
+  case 1: return y;
+  case 2: return z;
+  case 3: return w;
+  }
+}
+
+template <typename T>
+inline constexpr vec<4, T>& vec<4, T>::operator=(vec<4, T> const& v)
+{
+  this->x = static_cast<T>(v.x);
+  this->y = static_cast<T>(v.y);
+  this->z = static_cast<T>(v.z);
+  this->w = static_cast<T>(v.w);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<4, T>& vec<4, T>::operator+=(vec<4, T> const& v)
+{
+  *this = *this + vec<4, T>(v);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<4, T>& vec<4, T>::operator-=(vec<4, T> const& v)
+{
+  *this = *this - vec<4, T>(v);
+  return *this;
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator+(vec<4, T> const& v1, vec<4, T> const& v2)
+{
+  return vec<4, T>(v1.x + v2.x,
+                   v1.y + v2.y,
+                   v1.z + v2.z,
+		   v1.w + v2.w);
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator-(vec<4, T> const& v1, vec<4, T> const& v2)
+{
+  return vec<4, T>(v1.x - v2.x,
+                   v1.y - v2.y,
+                   v1.z - v2.z,
+		   v1.w - v2.w);
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator*(vec<4, T> const& v1, vec<4, T> const& v2)
+{
+  return vec<4, T>(v1.x * v2.x,
+                   v1.y * v2.y,
+                   v1.z * v2.z,
+		   v1.w * v2.w);
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator*(vec<4, T> const& v1, T const& scalar)
+{
+  return v1 * vec<4, T>(scalar);
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator*(T const& scalar, vec<4, T> const& v1)
+{
+  return vec<4, T>(scalar) * v1;
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator/(vec<4, T> const& v1, vec<4, T> const& v2)
+{
+  return vec<4, T>(v1.x / v2.x,
+                   v1.y / v2.y,
+                   v1.z / v2.z,
+		   v1.w / v2.w);
+}
+
+template <typename T>
+inline constexpr vec<4, T> operator/(vec<4, T> const& v1, T const& scalar)
+{
+  return v1 / vec<4, T>(scalar);
+}
+
+template <typename T>
+inline constexpr T dot(vec<4, T> const& v1, vec<4, T> const& v2)
+{
+  vec<4, T> tmp(v1 * v2);
+  return tmp.x + tmp.y + tmp.z + tmp.w;
+}
+
+template <typename T>
+inline constexpr vec<4, T> functor1(T (&func) (T const& x), vec<4, T> const& v)
+{
+  return vec<4, T>(func(v.x), func(v.y), func(v.z), func(v.w));
+}
+
+template <typename T, typename U>
+inline constexpr vec<4, U> functor1(U (&func) (T const& x), vec<4, T> const& v)
+{
+  return vec<4, U>(func(v.x), func(v.y), func(v.z), func(v.w));
+}
+
+template <typename T>
+inline constexpr T magnitude(vec<4, T> const& v)
+{
+  return sqrt(dot(v, v));
+}