diff --git a/game/CMakeLists.txt b/game/CMakeLists.txt index 8637875eba..1039870e32 100644 --- a/game/CMakeLists.txt +++ b/game/CMakeLists.txt @@ -81,6 +81,8 @@ set(RUNTIME_SOURCE graphics/opengl_renderer/DirectRenderer.cpp graphics/opengl_renderer/OpenGLRenderer.cpp graphics/opengl_renderer/Shader.cpp + graphics/texture/TextureConverter.cpp + graphics/texture/TexturePool.cpp system/vm/dmac.cpp system/vm/vm.cpp) diff --git a/game/graphics/gfx.cpp b/game/graphics/gfx.cpp index ca04dc124e..fcbc2b3f99 100644 --- a/game/graphics/gfx.cpp +++ b/game/graphics/gfx.cpp @@ -16,6 +16,7 @@ #include "common/goal_constants.h" #include "common/util/Timer.h" #include "game/graphics/opengl_renderer/OpenGLRenderer.h" +#include "game/graphics/texture/TexturePool.h" namespace Gfx { @@ -31,10 +32,16 @@ struct GraphicsData { bool has_data_to_render = false; FixedChunkDmaCopier dma_copier; + // texture pool + std::shared_ptr texture_pool; + // temporary opengl renderer OpenGLRenderer ogl_renderer; - GraphicsData() : dma_copier(EE_MAIN_MEM_SIZE) {} + GraphicsData() + : dma_copier(EE_MAIN_MEM_SIZE), + texture_pool(std::make_shared()), + ogl_renderer(texture_pool) {} }; std::unique_ptr g_gfx_data; @@ -186,4 +193,13 @@ void send_chain(const void* data, u32 offset) { } } +void texture_upload_now(const u8* tpage, int mode, u32 s7_ptr) { + if (g_gfx_data) { + // just pass it to the texture pool. + // the texture pool will take care of locking. + // we don't want to lock here for the entire duration of the conversion. + g_gfx_data->texture_pool->handle_upload_now(tpage, mode, g_ee_main_mem, s7_ptr); + } +} + } // namespace Gfx diff --git a/game/graphics/gfx.h b/game/graphics/gfx.h index 2a0ec445b4..7130724b59 100644 --- a/game/graphics/gfx.h +++ b/game/graphics/gfx.h @@ -21,6 +21,7 @@ u32 Exit(); u32 vsync(); void send_chain(const void* data, u32 offset); +void texture_upload_now(const u8* tpage, int mode, u32 s7_ptr); } // namespace Gfx diff --git a/game/graphics/opengl_renderer/BucketRenderer.h b/game/graphics/opengl_renderer/BucketRenderer.h index b2f6023f83..d89c33d58a 100644 --- a/game/graphics/opengl_renderer/BucketRenderer.h +++ b/game/graphics/opengl_renderer/BucketRenderer.h @@ -1,8 +1,10 @@ #pragma once #include +#include #include "game/graphics/dma/dma_chain_read.h" #include "game/graphics/opengl_renderer/Shader.h" +#include "game/graphics/texture/TexturePool.h" /*! * Matches the bucket-id enum in GOAL @@ -21,7 +23,10 @@ enum class BucketId { * This allows bucket renders to share textures and shaders. */ struct SharedRenderState { + explicit SharedRenderState(std::shared_ptr _texture_pool) + : texture_pool(_texture_pool) {} ShaderLibrary shaders; + std::shared_ptr texture_pool; u32 buckets_base = 0; // address of buckets array. u32 next_bucket = 0; // address of next bucket that we haven't started rendering in buckets u32 default_regs_buffer = 0; // address of the default regs chain. diff --git a/game/graphics/opengl_renderer/OpenGLRenderer.cpp b/game/graphics/opengl_renderer/OpenGLRenderer.cpp index 78ff34e34b..cfe30ac81b 100644 --- a/game/graphics/opengl_renderer/OpenGLRenderer.cpp +++ b/game/graphics/opengl_renderer/OpenGLRenderer.cpp @@ -20,7 +20,8 @@ void GLAPIENTRY opengl_error_callback(GLenum /*source*/, lg::error("OpenGL error: {}", message); } -OpenGLRenderer::OpenGLRenderer() { +OpenGLRenderer::OpenGLRenderer(std::shared_ptr texture_pool) + : m_render_state(texture_pool) { // setup OpenGL errors glEnable(GL_DEBUG_OUTPUT); glDebugMessageCallback(opengl_error_callback, nullptr); diff --git a/game/graphics/opengl_renderer/OpenGLRenderer.h b/game/graphics/opengl_renderer/OpenGLRenderer.h index 313f369d4d..e1128c8574 100644 --- a/game/graphics/opengl_renderer/OpenGLRenderer.h +++ b/game/graphics/opengl_renderer/OpenGLRenderer.h @@ -9,7 +9,7 @@ class OpenGLRenderer { public: - OpenGLRenderer(); + OpenGLRenderer(std::shared_ptr texture_pool); void render(DmaFollower dma, int window_width_px, int window_height_px); private: diff --git a/game/graphics/texture/TextureConverter.cpp b/game/graphics/texture/TextureConverter.cpp new file mode 100644 index 0000000000..3da14a6a36 --- /dev/null +++ b/game/graphics/texture/TextureConverter.cpp @@ -0,0 +1,414 @@ +#include "TextureConverter.h" +#include "common/util/assert.h" +#include "third-party/fmt/core.h" +#include "common/util/FileUtil.h" + +/*! + * Convert from a pixel location in a texture (x, y, texture buffer width) to VRAM address (byte). + * Uses the PSMCT32 format. + * This format is used either to store 8-bit RGBA (texture palettes) or to copy memory. + * See Ch. 8, Details of GS Local Memory for these tables. + */ +u32 psmct32_addr(u32 x, u32 y, u32 width) { + // XXX_col refers to which XXX you're in (screen) + // XXX refers to which XXX you're in (memory) + // XXX_x refers to the pixel within the XXX + + // first, determine the page + u32 pages_per_row = width / 64; + u32 page_col = x / 64; + u32 page_row = y / 32; + u32 page_x = x % 64; + u32 page_y = y % 32; + u32 page = page_col + page_row * pages_per_row; + + // next the block + u32 block_col = page_x / 8; + u32 block_row = page_y / 8; + u32 block_x = page_x % 8; + u32 block_y = page_y % 8; + const u32 psm32_table[4][8] = {{0, 1, 4, 5, 16, 17, 20, 21}, + {2, 3, 6, 7, 18, 19, 22, 23}, + {8, 9, 12, 13, 24, 25, 28, 29}, + {10, 11, 14, 15, 26, 27, 30, 31}}; + + u32 block = psm32_table[block_row][block_col]; + + // next the column (there's only one "column" per column) + u32 col_row = block_y / 2; + u32 col_y = block_y % 2; + u32 col_x = block_x; + + // next the pixel + const u32 psm32_pix_table[2][8] = {{0, 1, 4, 5, 8, 9, 12, 13}, {2, 3, 6, 7, 10, 11, 14, 15}}; + u32 pixel = psm32_pix_table[col_y][col_x]; + + // now the sum + return ((page * 64 * 32) + (block * 8 * 8) + (col_row * 8 * 2) + pixel) * 4; +} + +/*! + * Convert from a pixel location in a texture (x, y, texture buffer width) to VRAM address (byte). + * Uses the PSMT8 format. + * This format is used either to store 8-bit palette indices, used in most textures. + * See Ch. 8, Details of GS Local Memory for these tables. + */ +u32 psmt8_addr(u32 x, u32 y, u32 width) { + // page is 128, 64 + // block is 16, 16 + // column is 16, 4 + + // first determine the page + u32 pages_per_row = width / 128; + u32 page_col = x / 128; + u32 page_row = y / 64; + u32 page_x = x % 128; + u32 page_y = y % 64; + u32 page = page_col + page_row * pages_per_row; + + // next block + u32 block_col = page_x / 16; + u32 block_row = page_y / 16; + u32 block_x = page_x % 16; + u32 block_y = page_y % 16; + const u32 psm32_table[4][8] = {{0, 1, 4, 5, 16, 17, 20, 21}, + {2, 3, 6, 7, 18, 19, 22, 23}, + {8, 9, 12, 13, 24, 25, 28, 29}, + {10, 11, 14, 15, 26, 27, 30, 31}}; + u32 block = psm32_table[block_row][block_col]; // it's the same table!!! + + // both columns and pixels within columns. + const uint8_t pix_table[16][16] = { + {0, 4, 16, 20, 32, 36, 48, 52, 2, 6, 18, 22, 34, 38, 50, 54}, + {8, 12, 24, 28, 40, 44, 56, 60, 10, 14, 26, 30, 42, 46, 58, 62}, + {33, 37, 49, 53, 1, 5, 17, 21, 35, 39, 51, 55, 3, 7, 19, 23}, + {41, 45, 57, 61, 9, 13, 25, 29, 43, 47, 59, 63, 11, 15, 27, 31}, + {96, 100, 112, 116, 64, 68, 80, 84, 98, 102, 114, 118, 66, 70, 82, 86}, + {104, 108, 120, 124, 72, 76, 88, 92, 106, 110, 122, 126, 74, 78, 90, 94}, + {65, 69, 81, 85, 97, 101, 113, 117, 67, 71, 83, 87, 99, 103, 115, 119}, + {73, 77, 89, 93, 105, 109, 121, 125, 75, 79, 91, 95, 107, 111, 123, 127}, + {128, 132, 144, 148, 160, 164, 176, 180, 130, 134, 146, 150, 162, 166, 178, 182}, + {136, 140, 152, 156, 168, 172, 184, 188, 138, 142, 154, 158, 170, 174, 186, 190}, + {161, 165, 177, 181, 129, 133, 145, 149, 163, 167, 179, 183, 131, 135, 147, 151}, + {169, 173, 185, 189, 137, 141, 153, 157, 171, 175, 187, 191, 139, 143, 155, 159}, + {224, 228, 240, 244, 192, 196, 208, 212, 226, 230, 242, 246, 194, 198, 210, 214}, + {232, 236, 248, 252, 200, 204, 216, 220, 234, 238, 250, 254, 202, 206, 218, 222}, + {193, 197, 209, 213, 225, 229, 241, 245, 195, 199, 211, 215, 227, 231, 243, 247}, + {201, 205, 217, 221, 233, 237, 249, 253, 203, 207, 219, 223, 235, 239, 251, 255}, + }; + + u32 pixel = pix_table[block_y][block_x]; + return (page * 128 * 64) + (block * 16 * 16) + pixel; +} + +u32 psmct16_addr(u32 x, u32 y, u32 width) { + // page is 64x64 + // block is 16x8 + // column is 16x2 + + // page + u32 pages_per_row = width / 64; + u32 page_col = x / 64; + u32 page_row = y / 64; + u32 page_x = x % 64; + u32 page_y = y % 64; + u32 page = page_col + page_row * pages_per_row; + + // block + u32 block_col = page_x / 16; + u32 block_row = page_y / 8; + u32 block_x = page_x % 16; + u32 block_y = page_y % 8; + const u32 psm16_table[8][4] = {{0, 2, 8, 10}, {1, 3, 9, 11}, {4, 6, 12, 14}, + {5, 7, 13, 15}, {16, 18, 24, 26}, {17, 19, 25, 27}, + {20, 22, 28, 30}, {21, 23, 29, 31}}; + u32 block = psm16_table[block_row][block_col]; + + const uint8_t pix_tabel[8][16] = { + {0, 2, 8, 10, 16, 18, 24, 26, 1, 3, 9, 11, 17, 19, 25, 27}, + {4, 6, 12, 14, 20, 22, 28, 30, 5, 7, 13, 15, 21, 23, 29, 31}, + {32, 34, 40, 42, 48, 50, 56, 58, 33, 35, 41, 43, 49, 51, 57, 59}, + {36, 38, 44, 46, 52, 54, 60, 62, 37, 39, 45, 47, 53, 55, 61, 63}, + {64, 66, 72, 74, 80, 82, 88, 90, 65, 67, 73, 75, 81, 83, 89, 91}, + {68, 70, 76, 78, 84, 86, 92, 94, 69, 71, 77, 79, 85, 87, 93, 95}, + {96, 98, 104, 106, 112, 114, 120, 122, 97, 99, 105, 107, 113, 115, 121, 123}, + {100, 102, 108, 110, 116, 118, 124, 126, 101, 103, 109, 111, 117, 119, 125, 127}, + }; + u32 pixel = pix_tabel[block_y][block_x]; + return 2 * ((page * 64 * 64) + (block * 16 * 8) + pixel); +} + +u32 psmt4_addr_half_byte(u32 x, u32 y, u32 width) { + // page is 128, 128 + // block is 32, 16 + // column is 32, 4 + + // first determine the page + u32 pages_per_row = width / 128; + u32 page_col = x / 128; + u32 page_row = y / 128; + u32 page_x = x % 128; + u32 page_y = y % 128; + u32 page = page_col + page_row * pages_per_row; + + // next block + u32 block_col = page_x / 32; + u32 block_row = page_y / 16; + u32 block_x = page_x % 32; + u32 block_y = page_y % 16; + const u32 psm4_table[8][4] = {{0, 2, 8, 10}, {1, 3, 9, 11}, {4, 6, 12, 14}, + {5, 7, 13, 15}, {16, 18, 24, 26}, {17, 19, 25, 27}, + {20, 22, 28, 30}, {21, 23, 29, 31}}; + assert(block_row < 8); + assert(block_col < 4); + u32 block = psm4_table[block_row][block_col]; // it's the same table!!! + + // both columns and pixels within columns. + const uint16_t pix_table[16][32] = { + {0, 8, 32, 40, 64, 72, 96, 104, 2, 10, 34, 42, 66, 74, 98, 106, + 4, 12, 36, 44, 68, 76, 100, 108, 6, 14, 38, 46, 70, 78, 102, 110}, + {16, 24, 48, 56, 80, 88, 112, 120, 18, 26, 50, 58, 82, 90, 114, 122, + 20, 28, 52, 60, 84, 92, 116, 124, 22, 30, 54, 62, 86, 94, 118, 126}, + {65, 73, 97, 105, 1, 9, 33, 41, 67, 75, 99, 107, 3, 11, 35, 43, + 69, 77, 101, 109, 5, 13, 37, 45, 71, 79, 103, 111, 7, 15, 39, 47}, + {81, 89, 113, 121, 17, 25, 49, 57, 83, 91, 115, 123, 19, 27, 51, 59, + 85, 93, 117, 125, 21, 29, 53, 61, 87, 95, 119, 127, 23, 31, 55, 63}, + {192, 200, 224, 232, 128, 136, 160, 168, 194, 202, 226, 234, 130, 138, 162, 170, + 196, 204, 228, 236, 132, 140, 164, 172, 198, 206, 230, 238, 134, 142, 166, 174}, + {208, 216, 240, 248, 144, 152, 176, 184, 210, 218, 242, 250, 146, 154, 178, 186, + 212, 220, 244, 252, 148, 156, 180, 188, 214, 222, 246, 254, 150, 158, 182, 190}, + {129, 137, 161, 169, 193, 201, 225, 233, 131, 139, 163, 171, 195, 203, 227, 235, + 133, 141, 165, 173, 197, 205, 229, 237, 135, 143, 167, 175, 199, 207, 231, 239}, + {145, 153, 177, 185, 209, 217, 241, 249, 147, 155, 179, 187, 211, 219, 243, 251, + 149, 157, 181, 189, 213, 221, 245, 253, 151, 159, 183, 191, 215, 223, 247, 255}, + {256, 264, 288, 296, 320, 328, 352, 360, 258, 266, 290, 298, 322, 330, 354, 362, + 260, 268, 292, 300, 324, 332, 356, 364, 262, 270, 294, 302, 326, 334, 358, 366}, + {272, 280, 304, 312, 336, 344, 368, 376, 274, 282, 306, 314, 338, 346, 370, 378, + 276, 284, 308, 316, 340, 348, 372, 380, 278, 286, 310, 318, 342, 350, 374, 382}, + {321, 329, 353, 361, 257, 265, 289, 297, 323, 331, 355, 363, 259, 267, 291, 299, + 325, 333, 357, 365, 261, 269, 293, 301, 327, 335, 359, 367, 263, 271, 295, 303}, + {337, 345, 369, 377, 273, 281, 305, 313, 339, 347, 371, 379, 275, 283, 307, 315, + 341, 349, 373, 381, 277, 285, 309, 317, 343, 351, 375, 383, 279, 287, 311, 319}, + {448, 456, 480, 488, 384, 392, 416, 424, 450, 458, 482, 490, 386, 394, 418, 426, + 452, 460, 484, 492, 388, 396, 420, 428, 454, 462, 486, 494, 390, 398, 422, 430}, + {464, 472, 496, 504, 400, 408, 432, 440, 466, 474, 498, 506, 402, 410, 434, 442, + 468, 476, 500, 508, 404, 412, 436, 444, 470, 478, 502, 510, 406, 414, 438, 446}, + {385, 393, 417, 425, 449, 457, 481, 489, 387, 395, 419, 427, 451, 459, 483, 491, + 389, 397, 421, 429, 453, 461, 485, 493, 391, 399, 423, 431, 455, 463, 487, 495}, + {401, 409, 433, 441, 465, 473, 497, 505, 403, 411, 435, 443, 467, 475, 499, 507, + 405, 413, 437, 445, 469, 477, 501, 509, 407, 415, 439, 447, 471, 479, 503, 511}, + }; + + assert(block_y < 16); + assert(block_x < 32); + u32 pixel = pix_table[block_y][block_x]; + return (page * 128 * 128) + (block * 32 * 16) + pixel; +} + +u32 rgba16_to_rgba32(u32 in) { + float ratio = 255.0 / 31.0; + u32 r = (in & 0b11111) * ratio; + u32 g = ((in >> 5) & 0b11111) * ratio; + u32 b = ((in >> 10) & 0b11111) * ratio; + u32 a = (in & 0x8000) * 0x1FE00; + + return a | (b << 16) | (g << 8) | r; +} + +TextureConverter::TextureConverter() { + m_vram.resize(4 * 1024 * 1024); +} + +void TextureConverter::upload(const u8* data, u32 dest, u32 size_vram_words) { + // all textures are copied to vram 128 pixels wide, regardless of actual width + int copy_width = 128; + // scale the copy height to be whatever it needs to be to transfer the right amount of data. + int copy_height = size_vram_words / copy_width; + fmt::print("upload dest is {}, size is {}\n", dest, copy_height); + + for (int y = 0; y < copy_height; y++) { + for (int x = 0; x < copy_width; x++) { + // VRAM address (bytes) + auto addr32 = psmct32_addr(x, y, copy_width) + dest * 4; + + *(u32*)(m_vram.data() + addr32) = *((const u32*)(data) + (x + y * copy_width)); + // fmt::print("{} = {}\n", addr32, *(const u32*)(data + (x + y * copy_width))); + } + } +} + +// texture format enums +enum class PSM { PSMCT16 = 0x02, PSMT8 = 0x13, PSMT4 = 0x14 }; +// clut format enums +enum class CPSM { PSMCT32 = 0x0, PSMCT16 = 0x02 }; + +void TextureConverter::download_rgba8888(u8* result, + u32 vram_addr, + u32 goal_tex_width, + u32 w, + u32 h, + u32 psm, + u32 clut_psm, + u32 clut_vram_addr) { + fmt::print("psm: {} cpsm: {}\n", psm, clut_psm); + if (psm == int(PSM::PSMT8) && clut_psm == int(CPSM::PSMCT32)) { + // this is the only supported texture format for now. + + // will store output pixels, rgba (8888) + // std::vector out; + u32 out_offset = 0; + + // width is like the TEX0 register, in 64 texel units. + // not sure what the other widths are yet. + int read_width = 64 * goal_tex_width; + // loop over pixels in output texture image + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + // read as the PSMT8 type. The dest field tells us a block offset. + auto addr8 = psmt8_addr(x, y, read_width) + vram_addr * 256; + u8 value = *(u8*)(m_vram.data() + addr8); + // fmt::print("{} {} = {}\n", x, y, value); + + // there's yet another scramble from the CLUT. The palette index turns into an X, Y value + // See GS manual 2.7.3 CLUT Storage Mode, IDTEX8 in CSM1 mode. + u32 clut_chunk = value / 16; + u32 off_in_chunk = value % 16; + u8 clx = 0, cly = 0; + if (clut_chunk & 1) { + clx = 8; + } + cly = (clut_chunk >> 1) * 2; + if (off_in_chunk >= 8) { + off_in_chunk -= 8; + cly++; + } + clx += off_in_chunk; + + // the x, y CLUT value is looked up in PSMCT32 mode + u32 clut_addr = psmct32_addr(clx, cly, 64) + clut_vram_addr * 256; + u32 clut_value = *(u32*)(m_vram.data() + clut_addr); + // fmt::print("{} {}\n", clut_addr, clut_value); + // out.push_back(clut_value); + memcpy(result + out_offset, &clut_value, 4); + out_offset += 4; + } + } + + } else if (psm == int(PSM::PSMT8) && clut_psm == int(CPSM::PSMCT16)) { + // will store output pixels, rgba (8888) + u32 out_offset = 0; + + // width is like the TEX0 register, in 64 texel units. + // not sure what the other widths are yet. + int read_width = 64 * goal_tex_width; + + // loop over pixels in output texture image + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + // read as the PSMT8 type. The dest field tells us a block offset. + auto addr8 = psmt8_addr(x, y, read_width) + vram_addr * 256; + u8 value = *(u8*)(m_vram.data() + addr8); + + // there's yet another scramble from the CLUT. The palette index turns into an X, Y value + // See GS manual 2.7.3 CLUT Storage Mode, IDTEX8 in CSM1 mode. + u32 clut_chunk = value / 16; + u32 off_in_chunk = value % 16; + u8 clx = 0, cly = 0; + if (clut_chunk & 1) { + clx = 8; + } + cly = (clut_chunk >> 1) * 2; + if (off_in_chunk >= 8) { + off_in_chunk -= 8; + cly++; + } + clx += off_in_chunk; + + // the x, y CLUT value is looked up in PSMCT32 mode + u32 clut_addr = psmct16_addr(clx, cly, 64) + clut_vram_addr * 256; + u32 clut_value = *(u16*)(m_vram.data() + clut_addr); + u32 rgba32 = rgba16_to_rgba32(clut_value); + memcpy(result + out_offset, &rgba32, 4); + out_offset += 4; + } + } + + } else if (psm == int(PSM::PSMT4) && clut_psm == int(CPSM::PSMCT16)) { + // will store output pixels, rgba (8888) + u32 out_offset = 0; + + // width is like the TEX0 register, in 64 texel units. + // not sure what the other widths are yet. + int read_width = 64 * goal_tex_width; + + // loop over pixels in output texture image + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + // read as the PSMT4 type, use half byte addressing + auto addr4 = psmt4_addr_half_byte(x, y, read_width) + vram_addr * 512; + + // read (half bytes) + u8 value = *(u8*)(m_vram.data() + addr4 / 2); + if (addr4 & 1) { + value >>= 4; + } else { + value = value & 0x0f; + } + + // there's yet another scramble from the CLUT. The palette index turns into an X, Y value + // See GS manual 2.7.3 CLUT Storage Mode, IDTEX4 in CSM1 mode. + + u8 clx = value & 0x7; + u8 cly = value >> 3; + + // the x, y CLUT value is looked up in PSMCT16 mode + u32 clut_addr = psmct16_addr(clx, cly, 64) + clut_vram_addr * 256; + u32 clut_value = *(u16*)(m_vram.data() + clut_addr); + u32 rgba32 = rgba16_to_rgba32(clut_value); + memcpy(result + out_offset, &rgba32, 4); + out_offset += 4; + } + } + } else if (psm == int(PSM::PSMT4) && clut_psm == int(CPSM::PSMCT32)) { + // will store output pixels, rgba (8888) + u32 out_offset = 0; + + // width is like the TEX0 register, in 64 texel units. + // not sure what the other widths are yet. + int read_width = 64 * goal_tex_width; + + // loop over pixels in output texture image + for (int y = 0; y < h; y++) { + for (int x = 0; x < w; x++) { + // read as the PSMT4 type, use half byte addressing + auto addr4 = psmt4_addr_half_byte(x, y, read_width) + vram_addr * 512; +// fmt::print("addr: {}\n", addr4/8); + + // read (half bytes) + u8 value = *(u8*)(m_vram.data() + addr4 / 2); + if (addr4 & 1) { + value >>= 4; + } else { + value = value & 0x0f; + } + + // there's yet another scramble from the CLUT. The palette index turns into an X, Y value + // See GS manual 2.7.3 CLUT Storage Mode, IDTEX4 in CSM1 mode. + + u8 clx = value & 0x7; + u8 cly = value >> 3; + + // the x, y CLUT value is looked up in PSMCT16 mode + u32 clut_addr = psmct32_addr(clx, cly, 64) + clut_vram_addr * 256; + u32 clut_value = *(u32*)(m_vram.data() + clut_addr); +// fmt::print("{} {}\n", value, clut_value); + memcpy(result + out_offset, &clut_value, 4); + out_offset += 4; + } + } + } + + else { + assert(false); + } +} \ No newline at end of file diff --git a/game/graphics/texture/TextureConverter.h b/game/graphics/texture/TextureConverter.h new file mode 100644 index 0000000000..44ecb3f0f7 --- /dev/null +++ b/game/graphics/texture/TextureConverter.h @@ -0,0 +1,22 @@ +#pragma once + +#include + +#include "common/common_types.h" + +class TextureConverter { + public: + TextureConverter(); + void upload(const u8* data, u32 dest, u32 size_vram_words); + void download_rgba8888(u8* result, + u32 vram_addr, + u32 goal_tex_width, + u32 w, + u32 h, + u32 psm, + u32 clut_psm, + u32 clut_vram_addr); + + private: + std::vector m_vram; +}; diff --git a/game/graphics/texture/TexturePool.cpp b/game/graphics/texture/TexturePool.cpp new file mode 100644 index 0000000000..c4114b902c --- /dev/null +++ b/game/graphics/texture/TexturePool.cpp @@ -0,0 +1,153 @@ +#include "TexturePool.h" + +#include "third-party/fmt/core.h" +#include "common/util/assert.h" +#include "common/util/FileUtil.h" + +//////////////////////////////// +// Extraction of textures +// Note: this is intended to be temporary until we have a better system. +// this simply converts the PS2 format textures loaded by the game, then puts them into the PC +// port texture pool. + +const char empty_string[] = ""; +const char* goal_string(u32 ptr, const u8* memory_base) { + if (ptr == 0) { + return empty_string; + } + return (const char*)(memory_base + ptr + 4); +} + +struct GoalTexture { + s16 w; + s16 h; + u8 num_mips; + u8 tex1_control; + u8 psm; + u8 mip_shift; + u16 clutpsm; + u16 dest[7]; + u16 clut_dest; + u8 width[7]; + u32 name_ptr; + u32 size; + float uv_dist; + u32 masks[3]; + + s32 segment_of_mip(s32 mip) const { + if (2 >= num_mips) { + return num_mips - mip - 1; + } else { + return std::max(0, 2 - mip); + } + } +}; + +static_assert(sizeof(GoalTexture) == 60, "GoalTexture size"); +static_assert(offsetof(GoalTexture, clutpsm) == 8); +static_assert(offsetof(GoalTexture, clut_dest) == 24); + +struct GoalTexturePage { + struct Seg { + u32 block_data_ptr; + u32 size; + u32 dest; + }; + u32 file_info_ptr; + u32 name_ptr; + u32 id; + s32 length; // texture count + u32 mip0_size; + u32 size; + Seg segment[3]; + u32 pad[16]; + // start of array. + + std::string print() const { + return fmt::format("Tpage id {} textures {} seg0 {} {} seg1 {} {} seg2 {} {}\n", id, length, + segment[0].size, segment[0].dest, segment[1].size, segment[1].dest, + segment[2].size, segment[2].dest); + } + + bool try_copy_texture_description(GoalTexture* dest, + int idx, + const u8* memory_base, + const u8* tpage, + u32 s7_ptr) { + u32 ptr; + memcpy(&ptr, tpage + sizeof(GoalTexturePage) + 4 * idx, 4); + if (ptr == s7_ptr) { + return false; + } + memcpy(dest, memory_base + ptr, sizeof(GoalTexture)); + return true; + } +}; + + +/*! + * Handle a GOAL texture-page object being uploaded to VRAM. + * The strategy: + * - upload the texture-age to a fake 4MB VRAM, like the GOAL code would have done. + * - "download" each texture in a reasonable format for the PC Port (currently RGBA8888) + * - add this to the PC pool. + * + * The textures are scrambled around in a confusing way. + */ +void TexturePool::handle_upload_now(const u8* tpage, int mode, const u8* memory_base, u32 s7_ptr) { + fmt::print("[TexutrePool C++] Got upload now! {} {}\n", (const void*)tpage, mode); + + // extract the texture-page object. This is just a description of the page data. + GoalTexturePage texture_page; + memcpy(&texture_page, tpage, sizeof(GoalTexturePage)); + + std::vector output_buffer; + output_buffer.resize(1024 * 1024); + + u32 sizes[3] = {texture_page.segment[0].size, texture_page.segment[1].size, + texture_page.segment[2].size}; + if (mode == -1) { + // I don't really understand what's going on here with the size. + // the sizes given aren't the actual sizes in memory, so if you just use that, you get the + // wrong answer. I solved this in the decompiler by using + u32 size = ((sizes[0] + sizes[1] + sizes[2] + 255) / 256) * 256; + m_tex_converter.upload(memory_base + texture_page.segment[0].block_data_ptr, + texture_page.segment[0].dest, size); + } else { + assert(false); + } + + + // loop over all texture in the tpage and download them. + for (int tex_idx = 0; tex_idx < texture_page.length; tex_idx++) { + GoalTexture tex; + if (texture_page.try_copy_texture_description(&tex, tex_idx, memory_base, tpage, s7_ptr)) { + + // each texture may have multiple mip levels. + for (int mip_idx = 0; mip_idx < tex.num_mips; mip_idx++) { + s32 segment = tex.segment_of_mip(mip_idx); + + u32 ww = tex.w >> mip_idx; + u32 hh = tex.h >> mip_idx; + m_tex_converter.download_rgba8888(output_buffer.data(), tex.dest[mip_idx], + tex.width[mip_idx], ww, hh, tex.psm, tex.clutpsm, + tex.clut_dest); + + // Debug output. + const char* tpage_name = goal_string(texture_page.name_ptr, memory_base); + const char* tex_name = goal_string(tex.name_ptr, memory_base); + file_util::create_dir_if_needed( + file_util::get_file_path({"debug_out", "textures", tpage_name})); + file_util::write_rgba_png( + fmt::format( + file_util::get_file_path({"debug_out", "textures", tpage_name, "{}-{}-{}.png"}), + tex_idx, tex_name, mip_idx), + output_buffer.data(), ww, hh); + } + + // TODO the clut + } else { + fmt::print("[{}] #f ------------\n", tex_idx); + } + } +} \ No newline at end of file diff --git a/game/graphics/texture/TexturePool.h b/game/graphics/texture/TexturePool.h new file mode 100644 index 0000000000..31b659b26f --- /dev/null +++ b/game/graphics/texture/TexturePool.h @@ -0,0 +1,12 @@ +#pragma once + +#include "common/common_types.h" +#include "game/graphics/texture/TextureConverter.h" + +class TexturePool { + public: + void handle_upload_now(const u8* tpage, int mode, const u8* memory_base, u32 s7_ptr); + + private: + TextureConverter m_tex_converter; +}; diff --git a/game/kernel/kmachine.cpp b/game/kernel/kmachine.cpp index aa7e92beb8..75d987b6f1 100644 --- a/game/kernel/kmachine.cpp +++ b/game/kernel/kmachine.cpp @@ -442,6 +442,10 @@ void send_gfx_dma_chain(u32 /*bank*/, u32 chain) { Gfx::send_chain(g_ee_main_mem, chain); } +void pc_texture_upload_now(u32 page, u32 mode) { + Gfx::texture_upload_now(Ptr(page).c(), mode, s7.offset); +} + /*! * Open a file-stream. Name is a GOAL string. Mode is a GOAL symbol. Use 'read for readonly * and anything else for write only. @@ -601,6 +605,7 @@ void InitMachine_PCPort() { make_function_symbol_from_c("__read-ee-timer", (void*)read_ee_timer); make_function_symbol_from_c("__mem-move", (void*)c_memmove); make_function_symbol_from_c("__send-gfx-dma-chain", (void*)send_gfx_dma_chain); + make_function_symbol_from_c("__pc-texture-upload-now", (void*)pc_texture_upload_now); } /*! diff --git a/goal_src/engine/gfx/texture.gc b/goal_src/engine/gfx/texture.gc index 6913e640af..f18885b5f4 100644 --- a/goal_src/engine/gfx/texture.gc +++ b/goal_src/engine/gfx/texture.gc @@ -1687,6 +1687,12 @@ (defmethod upload-now! texture-page ((obj texture-page) (arg0 int)) "Immediately upload the texture-page to the given buffer, using arg0 mode." + + (#when PC_PORT + ;; load it to the PC Port's texture pool. + (__pc-texture-upload-now obj arg0) + ) + (let ((gp-0 *txt-dma-list*)) ;; Set up the DMA buffer (let ((v1-0 gp-0)) @@ -2094,6 +2100,7 @@ (-> obj info maya-file-name) (-> obj info tool-debug) (-> obj info mdb-file-name)) + (tex-dbg "block data #x~X~%" (-> obj segment 0 block-data)) (local-vars (s4-0 texture-page-dir-entry)) (cond ((or (not obj) diff --git a/goal_src/kernel-defs.gc b/goal_src/kernel-defs.gc index 73d6e95340..a76bb6ce97 100644 --- a/goal_src/kernel-defs.gc +++ b/goal_src/kernel-defs.gc @@ -182,6 +182,7 @@ (define-extern __read-ee-timer (function uint)) (define-extern __mem-move (function pointer pointer uint none)) (define-extern __send-gfx-dma-chain (function object object none)) +(define-extern __pc-texture-upload-now (function object object none)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;; ksound - InitSoundScheme