examples: add texture example

Makefile

@@ -2,7 +2,10 @@ OPT = -Os
 
 all: cartridge.bin
 
-phony:
+phony: res
+
+res:
+	make -C res/
 
 arm9/%.bin: phony
 	make -C arm9/ $(notdir $@)
@@ -14,6 +17,7 @@ DEFAULT = header.o arm7/arm7.bin.o
 
 triangle.elf: $(DEFAULT) arm9/triangle.bin.o
 triangle_rotating.elf: $(DEFAULT) arm9/triangle_rotating.bin.o
+texture.elf: $(DEFAULT) arm9/texture.bin.o
 
 TARGET = arm-none-eabi-
 AARCH = -march=armv4t -mlittle-endian
@@ -21,4 +25,4 @@ OBJARCH = -O elf32-littlearm -B armv4t
 LDSCRIPT = cartridge.lds
 include common.mk
 
-.PHONY: phony
+.PHONY: phony res

arm9/Makefile

@@ -20,6 +20,8 @@ triangle.elf: start.o examples/triangle.o
 
 triangle_rotating.elf: start.o examples/triangle_rotating.o ../math/cos_table_fp12.o ../math/cos.o
 
+texture.elf: start.o examples/texture.o ../res/hotaru_futaba.data.o ../res/hotaru_futaba.data.pal.o
+
 CFLAGS += -I../include -I../math
 
 include arm9.mk

arm9/examples/texture.c

@@ -0,0 +1,272 @@
+#include "io_registers.h"
+#include "bits.h"
+
+#include "../res/hotaru_futaba.data.h"
+#include "../res/hotaru_futaba.data.pal.h"
+
+static inline uint16_t rgb565(const uint8_t * buf)
+{
+  uint8_t r = buf[0] >> 3;
+  uint8_t g = buf[1] >> 3;
+  uint8_t g_l = (buf[1] >> 2) & 1;
+  uint8_t b = buf[2] >> 3;
+
+  return (g_l << 15) | (b << 10) | (g << 5) | (r << 0);
+}
+
+void main()
+{
+  // power control
+  io_registers.a.POWCNT = 0
+    | POWCNT__lcd_output_destination__a_to_upper__b_to_lower
+    | POWCNT__geometry_engine__enable
+    | POWCNT__rendering_engine__enable
+    | POWCNT__lcd__enable;
+
+  // enable bg0 and 3d graphics
+  io_registers.a.DISPCNT = 0
+    | DISPCNT__display_mode__graphics_display
+    | DISPCNT__bg0__enable
+    | DISPCNT__display_selection_for_bg0__3d_graphics
+    ;
+
+  // disable all 3d effects, enable texture mapping
+  io_registers.a.DISP3DCNT = 0
+    | DISP3DCNT__clear_image__disable
+    | DISP3DCNT__fog_master__disable
+    | DISP3DCNT__edge_marking__disable
+    | DISP3DCNT__anti_aliasing__disable
+    | DISP3DCNT__alpha_blending__disable
+    | DISP3DCNT__alpha_test__disable
+    | DISP3DCNT__texture_mapping__enable;
+
+  // memory bank allocation
+  // use VRAM-A for texture pixel data
+  // use VRAM-E for texture palette data
+
+  // temporarily map VRAM-A (128KB) to the arm9 address space:
+  // 0x06800000 - 0x0681FFFF
+  io_registers.a.VRAMCNT = 0
+    | VRAMCNT__vram_a__enable
+    | VRAMCNT__vram_a__mst(0b00); // arm9
+
+  // temporarily map VRAM-F (16KB) to the arm9 address space:
+  // 0x06890000 - 0x06893FFF
+  io_registers.a.WVRAMCNT = 0
+    | WVRAMCNT__vram_f__enable
+    | WVRAMCNT__vram_f__mst(0b000); // arm9
+
+  // at this point, VRAM-A/VRAM-E are not accessible by the 3d engine.
+
+  // copy palette data
+  const uint8_t * pal = (const uint8_t *)&_binary_hotaru_futaba_data_pal_start;
+  volatile uint16_t * vram_f = (volatile uint16_t *)(0x06890000);
+  for (int i = 0; i < 16; i++) {
+    // as exported by GIMP, the palette data is three bytes per pixel, packed
+    // RGB24
+    vram_f[i] = rgb565(&pal[i * 3]);
+  }
+
+  // copy pixel data
+  const uint8_t * pixels = (const uint8_t *)&_binary_hotaru_futaba_data_start;
+  volatile uint16_t * vram_a = (volatile uint16_t *)(0x06800000);
+
+  // notice the += 2 increment
+  for (int i = 0; i < 96 * 64; i += 2) {
+    // as exported by GIMP, the pixel data is two bytes per pixel.
+    // [0]: color index (0-15)
+    // [1]: alpha (0 or 255)
+
+    // generate a3i5 pixel data:
+    // | 15 14 13 | 12 10 09 08 | 07 06 05 | 04 03 02 01 00 |
+    // | t1 alpha | t1 index    | t0 alpha | t0 index       |
+
+    // process two pixels per loop iteration
+    uint8_t t0_index = pixels[(i + 0) * 2 + 0];
+    uint8_t t0_alpha = pixels[(i + 0) * 2 + 1];
+
+    uint8_t t1_index = pixels[(i + 1) * 2 + 0];
+    uint8_t t1_alpha = pixels[(i + 1) * 2 + 1];
+
+    // scale alpha from 8-bit to 3-bit
+    t0_alpha >>= 5;
+    t1_alpha >>= 5;
+
+    // mask index to 4-bit (though the data should already be 16-color, just in case)
+    t0_index &= 0xf;
+    t1_index &= 0xf;
+
+    // the data is now one byte per a3i5 texel, two a3i5 texels per uint16_t
+    uint16_t a3i5_texel = (t1_alpha << 13) | (t1_index << 8) | (t0_alpha << 5) | (t0_index << 0);
+    vram_a[i / 2] = a3i5_texel;
+  }
+
+  // arm9 no longer needs read/write access to texture data, remap vram to the 3d engine
+
+  // map VRAM-A (128KB) to the 3d-engine "texture image slot 0":
+  // 0x00000 - 0x1ffff (3d engine texture image address space)
+  io_registers.a.VRAMCNT = 0
+    | VRAMCNT__vram_a__enable
+    | VRAMCNT__vram_a__ofs(0)     // slot 0
+    | VRAMCNT__vram_a__mst(0b11); // texture image
+
+  // map VRAM-F (16KB) to the 3d-engine "texture palette slot 0":
+  // 0x0000 - 0x3fff (3d engine texture palette address space)
+  io_registers.a.WVRAMCNT = 0
+    | WVRAMCNT__vram_f__enable
+    | WVRAMCNT__vram_f__ofs(0)      // slot 0
+    | WVRAMCNT__vram_f__mst(0b011); // texture palette
+
+  // set the texture palette base address
+  io_registers.a.TEXPLTT_BASE = TEXPLTT_BASE__base_address(0);
+
+  io_registers.a.TEXIMAGE_PARAM = 0
+    | TEXIMAGE_PARAM__texture_coordinate_transformation_mode__texcoord_source
+    | TEXIMAGE_PARAM__texture_format__a3i5_translucent
+    | TEXIMAGE_PARAM__t_size__128_texels // actually 96 pixels
+    | TEXIMAGE_PARAM__s_size__64_texels
+    | TEXIMAGE_PARAM__texture_starting_address(0);
+
+  // the following polygons are fully opaque and are not
+  // backface-culled
+  // use
+  io_registers.a.POLYGON_ATTR = 0
+    | POLYGON_ATTR__alpha_value(31)
+    | POLYGON_ATTR__render_front_surface__enable
+    | POLYGON_ATTR__render_back_surface__enable;
+
+  // clear matrix stack status
+  io_registers.a.GXSTAT |= GXSTAT__matrix_stack_status__overflow_or_underflow;
+
+  // load identity matrices
+  io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__projection;
+  io_registers.a.MTX_IDENTITY = 0;
+  // scale the x-axis by the ratio of the display height by the display width.
+  io_registers.a.MTX_SCALE = (192 << 12) / 256;
+  io_registers.a.MTX_SCALE = 1 << 12;
+  io_registers.a.MTX_SCALE = 1 << 12;
+
+  io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__position_and_vector;
+  io_registers.a.MTX_IDENTITY = 0;
+
+  io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__position;
+  io_registers.a.MTX_IDENTITY = 0;
+  // scale the x-axis by the ratio of the hotaru sprite dimensions
+  io_registers.a.MTX_SCALE = (64 << 12) / 96;
+  io_registers.a.MTX_SCALE = 1 << 12;
+  io_registers.a.MTX_SCALE = 1 << 12;
+
+  io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__texture;
+  io_registers.a.MTX_IDENTITY = 0;
+
+  // set the 3d clear color to a dark red
+  io_registers.a.CLEAR_COLOR = 0
+    | CLEAR_COLOR__clear_polygon_id(31)
+    | CLEAR_COLOR__alpha_value(31)
+    | CLEAR_COLOR__blue(1)
+    | CLEAR_COLOR__green(1)
+    | CLEAR_COLOR__red(10);
+
+  // set the depth buffer clear value to the maximum value
+  io_registers.a.CLEAR_DEPTH = CLEAR_DEPTH__value(0x7fff);
+
+  // the 3d viewport is the entire display area
+  io_registers.a.VIEWPORT = 0
+    | VIEWPORT__y2(191)
+    | VIEWPORT__x2(255)
+    | VIEWPORT__y1(0)
+    | VIEWPORT__x1(0);
+
+  // VTX_10 uses signed 4.6 floating point (10 bit)
+  // | 9 | 8 7 6 | 5 4 3 2 1 0 |
+  // | s |  int  | decimal     |
+  int vtx_10_fixed_point = 64; // == 2⁶
+
+  // TEXCOORD uses signed 12.4 floating point (16 bit)
+  int texcoord_fixed_point = 16; // == 2⁴
+
+  // quadrilateral; top-left corner is top-left of screen
+  /*
+      A--D
+      |  |
+      |  |
+      B--C
+
+      abcd
+  */
+  int ax = -1.0 * vtx_10_fixed_point;
+  int ay =  1.0 * vtx_10_fixed_point;
+  int as =  0.0 * texcoord_fixed_point;
+  int at =  0.0 * texcoord_fixed_point;
+
+  int bx = -1.0 * vtx_10_fixed_point;
+  int by = -1.0 * vtx_10_fixed_point;
+  int bs =  0.0 * texcoord_fixed_point;
+  int bt = 95.0 * texcoord_fixed_point;
+
+  int cx =  1.0 * vtx_10_fixed_point;
+  int cy = -1.0 * vtx_10_fixed_point;
+  int cs = 63.0 * texcoord_fixed_point;
+  int ct = 95.0 * texcoord_fixed_point;
+
+  int dx =  1.0 * vtx_10_fixed_point;
+  int dy =  1.0 * vtx_10_fixed_point;
+  int ds = 63.0 * texcoord_fixed_point;
+  int dt =  0.0 * texcoord_fixed_point;
+
+  int z  =  1.0 * vtx_10_fixed_point;
+
+  while (1) {
+    io_registers.a.COLOR = 0
+      | COLOR__blue(31)
+      | COLOR__green(31)
+      | COLOR__red(31);
+
+    // first quadrilateral: ABCD
+    io_registers.a.BEGIN_VTXS = BEGIN_VTXS__type__quadrilateral;
+
+    io_registers.a.TEXCOORD = 0
+      | TEXCOORD__t_coordinate(at)
+      | TEXCOORD__s_coordinate(as);
+    io_registers.a.VTX_10 = 0
+      | VTX_10__z_coordinate(z)
+      | VTX_10__y_coordinate(ay)
+      | VTX_10__x_coordinate(ax);
+
+    io_registers.a.TEXCOORD = 0
+      | TEXCOORD__t_coordinate(bt)
+      | TEXCOORD__s_coordinate(bs);
+    io_registers.a.VTX_10 = 0
+      | VTX_10__z_coordinate(z)
+      | VTX_10__y_coordinate(by)
+      | VTX_10__x_coordinate(bx);
+
+    io_registers.a.TEXCOORD = 0
+      | TEXCOORD__t_coordinate(ct)
+      | TEXCOORD__s_coordinate(cs);
+    io_registers.a.VTX_10 = 0
+      | VTX_10__z_coordinate(z)
+      | VTX_10__y_coordinate(cy)
+      | VTX_10__x_coordinate(cx);
+
+    io_registers.a.TEXCOORD = 0
+      | TEXCOORD__t_coordinate(dt)
+      | TEXCOORD__s_coordinate(ds);
+    io_registers.a.VTX_10 = 0
+      | VTX_10__z_coordinate(z)
+      | VTX_10__y_coordinate(dy)
+      | VTX_10__x_coordinate(dx);
+
+    // end of the ABCD quadrilateral
+    io_registers.a.END_VTXS = 0;
+
+    // wait for the geometry engine
+    while (io_registers.a.GXSTAT & GXSTAT__geometry_engine_busy);
+
+    while (io_registers.a.VCOUNT != 262);
+    while (io_registers.a.VCOUNT != 0);
+
+    // swap buffers
+    io_registers.a.SWAP_BUFFERS = 0;
+  }
+}

arm9/examples/triangle.c

@@ -17,6 +17,16 @@ void main()
     | DISPCNT__display_selection_for_bg0__3d_graphics
     ;
 
+  // disable all 3d effects
+  io_registers.a.DISP3DCNT = 0
+    | DISP3DCNT__clear_image__disable
+    | DISP3DCNT__fog_master__disable
+    | DISP3DCNT__edge_marking__disable
+    | DISP3DCNT__anti_aliasing__disable
+    | DISP3DCNT__alpha_blending__disable
+    | DISP3DCNT__alpha_test__disable
+    | DISP3DCNT__texture_mapping__disable;
+
   // clear matrix stack status
   io_registers.a.GXSTAT |= GXSTAT__matrix_stack_status__overflow_or_underflow;
 
@@ -30,16 +40,6 @@ void main()
   io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__position_and_vector;
   io_registers.a.MTX_IDENTITY = 0;
 
-  // disable all 3d effects
-  io_registers.a.DISP3DCNT = 0
-    | DISP3DCNT__clear_image__disable
-    | DISP3DCNT__fog_master__disable
-    | DISP3DCNT__edge_marking__disable
-    | DISP3DCNT__anti_aliasing__disable
-    | DISP3DCNT__alpha_blending__disable
-    | DISP3DCNT__alpha_test__disable
-    | DISP3DCNT__texture_mapping__disable;
-
   // set the 3d clear color to a dark red
   io_registers.a.CLEAR_COLOR = 0
     | CLEAR_COLOR__clear_polygon_id(31)

arm9/examples/triangle_rotating.c

@@ -19,6 +19,16 @@ void main()
     | DISPCNT__display_selection_for_bg0__3d_graphics
     ;
 
+  // disable all 3d effects
+  io_registers.a.DISP3DCNT = 0
+    | DISP3DCNT__clear_image__disable
+    | DISP3DCNT__fog_master__disable
+    | DISP3DCNT__edge_marking__disable
+    | DISP3DCNT__anti_aliasing__disable
+    | DISP3DCNT__alpha_blending__disable
+    | DISP3DCNT__alpha_test__disable
+    | DISP3DCNT__texture_mapping__disable;
+
   // clear matrix stack status
   io_registers.a.GXSTAT |= GXSTAT__matrix_stack_status__overflow_or_underflow;
 
@@ -36,16 +46,6 @@ void main()
   io_registers.a.MTX_MODE = MTX_MODE__matrix_mode__position_and_vector;
   io_registers.a.MTX_IDENTITY = 0;
 
-  // disable all 3d effects
-  io_registers.a.DISP3DCNT = 0
-    | DISP3DCNT__clear_image__disable
-    | DISP3DCNT__fog_master__disable
-    | DISP3DCNT__edge_marking__disable
-    | DISP3DCNT__anti_aliasing__disable
-    | DISP3DCNT__alpha_blending__disable
-    | DISP3DCNT__alpha_test__disable
-    | DISP3DCNT__texture_mapping__disable;
-
   // set the 3d clear color to a dark red
   io_registers.a.CLEAR_COLOR = 0
     | CLEAR_COLOR__clear_polygon_id(31)

include/bits.h

@@ -113,6 +113,40 @@
 #define BG3CNT__mosaic__enable (0x1 << 6)
 #define BG3CNT__character_base_block(v) (((v) & 0xf) << 2)
 #define BG3CNT__priority(v) (((v) & 0x3) << 0)
+#define VRAMCNT__vram_d__disable (0x0 << 31)
+#define VRAMCNT__vram_d__enable (0x1 << 31)
+#define VRAMCNT__vram_d__ofs(v) (((v) & 0x3) << 27)
+#define VRAMCNT__vram_d__mst(v) (((v) & 0x7) << 24)
+#define VRAMCNT__vram_c__disable (0x0 << 23)
+#define VRAMCNT__vram_c__enable (0x1 << 23)
+#define VRAMCNT__vram_c__ofs(v) (((v) & 0x3) << 19)
+#define VRAMCNT__vram_c__mst(v) (((v) & 0x7) << 16)
+#define VRAMCNT__vram_b__disable (0x0 << 15)
+#define VRAMCNT__vram_b__enable (0x1 << 15)
+#define VRAMCNT__vram_b__ofs(v) (((v) & 0x3) << 11)
+#define VRAMCNT__vram_b__mst(v) (((v) & 0x3) << 8)
+#define VRAMCNT__vram_a__disable (0x0 << 7)
+#define VRAMCNT__vram_a__enable (0x1 << 7)
+#define VRAMCNT__vram_a__ofs(v) (((v) & 0x3) << 3)
+#define VRAMCNT__vram_a__mst(v) (((v) & 0x3) << 0)
+#define WVRAMCNT__wram__bank(v) (((v) & 0x3) << 24)
+#define WVRAMCNT__vram_g__disable (0x0 << 23)
+#define WVRAMCNT__vram_g__enable (0x1 << 23)
+#define WVRAMCNT__vram_g__ofs(v) (((v) & 0x3) << 19)
+#define WVRAMCNT__vram_g__mst(v) (((v) & 0x7) << 16)
+#define WVRAMCNT__vram_f__disable (0x0 << 15)
+#define WVRAMCNT__vram_f__enable (0x1 << 15)
+#define WVRAMCNT__vram_f__ofs(v) (((v) & 0x3) << 11)
+#define WVRAMCNT__vram_f__mst(v) (((v) & 0x7) << 8)
+#define WVRAMCNT__vram_e__disable (0x0 << 7)
+#define WVRAMCNT__vram_e__enable (0x1 << 7)
+#define WVRAMCNT__vram_e__mst(v) (((v) & 0x7) << 0)
+#define VRAM_HI_CNT__vram_i__disable (0x0 << 15)
+#define VRAM_HI_CNT__vram_i__enable (0x1 << 15)
+#define VRAM_HI_CNT__vram_i__mst(v) (((v) & 0x3) << 8)
+#define VRAM_HI_CNT__vram_h__disable (0x0 << 7)
+#define VRAM_HI_CNT__vram_h__enable (0x1 << 7)
+#define VRAM_HI_CNT__vram_h__mst(v) (((v) & 0x3) << 0)
 #define POWCNT__lcd_output_destination__a_to_lower__b_to_upper (0x0 << 15)
 #define POWCNT__lcd_output_destination__a_to_upper__b_to_lower (0x1 << 15)
 #define POWCNT__2d_graphics_engine_b__disable (0x0 << 9)
@@ -171,7 +205,7 @@
 #define NORMAL__y_component(v) (((v) & 0x3ff) << 10)
 #define NORMAL__x_component(v) (((v) & 0x3ff) << 0)
 #define TEXCOORD__t_coordinate(v) (((v) & 0xffff) << 16)
-#define TEXCOORD__x_coordinate(v) (((v) & 0xffff) << 0)
+#define TEXCOORD__s_coordinate(v) (((v) & 0xffff) << 0)
 #define VTX_16__0__y_coordinate(v) (((v) & 0xffff) << 16)
 #define VTX_16__0__x_coordinate(v) (((v) & 0xffff) << 0)
 #define VTX_16__1__z_coordinate(v) (((v) & 0xffff) << 0)

registers/graphics_engine_bits.csv

@@ -121,6 +121,44 @@
 "BG3CNT",,"5-2","character_base_block",,"0b1111",
 "BG3CNT",,"1-0","priority",,"0b11",
 ,,,,,,
+"VRAMCNT","vram_d",31,"disable",0,,
+"VRAMCNT","vram_d",31,"enable",1,,
+"VRAMCNT","vram_d","28-27","ofs",,"0b11",
+"VRAMCNT","vram_d","26-24","mst",,"0b111",
+"VRAMCNT","vram_c",23,"disable",0,,
+"VRAMCNT","vram_c",23,"enable",1,,
+"VRAMCNT","vram_c","20-19","ofs",,"0b11",
+"VRAMCNT","vram_c","18-16","mst",,"0b111",
+"VRAMCNT","vram_b",15,"disable",0,,
+"VRAMCNT","vram_b",15,"enable",1,,
+"VRAMCNT","vram_b","12-11","ofs",,"0b11",
+"VRAMCNT","vram_b","9-8","mst",,"0b11",
+"VRAMCNT","vram_a",7,"disable",0,,
+"VRAMCNT","vram_a",7,"enable",1,,
+"VRAMCNT","vram_a","4-3","ofs",,"0b11",
+"VRAMCNT","vram_a","1-0","mst",,"0b11",
+,,,,,,
+"WVRAMCNT","wram","25-24","bank",,"0b11",
+"WVRAMCNT","vram_g",23,"disable",0,,
+"WVRAMCNT","vram_g",23,"enable",1,,
+"WVRAMCNT","vram_g","20-19","ofs",,"0b11",
+"WVRAMCNT","vram_g","18-16","mst",,"0b111",
+"WVRAMCNT","vram_f",15,"disable",0,,
+"WVRAMCNT","vram_f",15,"enable",1,,
+"WVRAMCNT","vram_f","12-11","ofs",,"0b11",
+"WVRAMCNT","vram_f","10-8","mst",,"0b111",
+"WVRAMCNT","vram_e",7,"disable",0,,
+"WVRAMCNT","vram_e",7,"enable",1,,
+"WVRAMCNT","vram_e","2-0","mst",,"0b111",
+,,,,,,
+"VRAM_HI_CNT","vram_i",15,"disable",0,,
+"VRAM_HI_CNT","vram_i",15,"enable",1,,
+"VRAM_HI_CNT","vram_i","9-8","mst",,"0b11",
+"VRAM_HI_CNT","vram_h",7,"disable",0,,
+"VRAM_HI_CNT","vram_h",7,"enable",1,,
+"VRAM_HI_CNT","vram_h","1-0","mst",,"0b11",
+,,,,,,
+,,,,,,
 "POWCNT","lcd_output_destination",15,"a_to_lower__b_to_upper",0,,
 "POWCNT","lcd_output_destination",15,"a_to_upper__b_to_lower",1,,
 "POWCNT","2d_graphics_engine_b",9,"disable",0,,
@@ -192,7 +230,7 @@
 "NORMAL",,"9-0","x_component",,"0x7ff",
 ,,,,,,
 "TEXCOORD",,"31-16","t_coordinate",,"0xffff",
-"TEXCOORD",,"15-0","x_coordinate",,"0xffff",
+"TEXCOORD",,"15-0","s_coordinate",,"0xffff",
 ,,,,,,
 "VTX_16",0,"31-16","y_coordinate",,"0xffff",
 "VTX_16",0,"15-0","x_coordinate",,"0xffff",

res/hotaru_futaba.data.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_hotaru_futaba_data_start __asm("_binary_hotaru_futaba_data_start");
+extern uint32_t _binary_hotaru_futaba_data_end __asm("_binary_hotaru_futaba_data_end");
+extern uint32_t _binary_hotaru_futaba_data_size __asm("_binary_hotaru_futaba_data_size");

res/hotaru_futaba.data.pal.h

@@ -0,0 +1,5 @@
+#pragma once
+#include <stdint.h>
+extern uint32_t _binary_hotaru_futaba_data_pal_start __asm("_binary_hotaru_futaba_data_pal_start");
+extern uint32_t _binary_hotaru_futaba_data_pal_end __asm("_binary_hotaru_futaba_data_pal_end");
+extern uint32_t _binary_hotaru_futaba_data_pal_size __asm("_binary_hotaru_futaba_data_pal_size");