X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=source%2Fgles2glide64%2Fsrc%2FGlide64%2Fucode02.h;fp=source%2Fgles2glide64%2Fsrc%2FGlide64%2Fucode02.h;h=bb8eeda68e915e79284a77672a6a2b85bcee7b18;hb=98e75f2d18c02c233da543560f76282f04fc796c;hp=0000000000000000000000000000000000000000;hpb=0ced54f867d36e8b324155bef49e8abfebfc3237;p=mupen64plus-pandora.git diff --git a/source/gles2glide64/src/Glide64/ucode02.h b/source/gles2glide64/src/Glide64/ucode02.h new file mode 100755 index 0000000..bb8eeda --- /dev/null +++ b/source/gles2glide64/src/Glide64/ucode02.h @@ -0,0 +1,849 @@ +/* +* Glide64 - Glide video plugin for Nintendo 64 emulators. +* Copyright (c) 2002 Dave2001 +* Copyright (c) 2003-2009 Sergey 'Gonetz' Lipski +* +* This program is free software; you can redistribute it and/or modify +* it under the terms of the GNU General Public License as published by +* the Free Software Foundation; either version 2 of the License, or +* any later version. +* +* This program is distributed in the hope that it will be useful, +* but WITHOUT ANY WARRANTY; without even the implied warranty of +* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +* GNU General Public License for more details. +* +* You should have received a copy of the GNU General Public License +* along with this program; if not, write to the Free Software +* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA +*/ + +//**************************************************************** +// +// Glide64 - Glide Plugin for Nintendo 64 emulators +// Project started on December 29th, 2001 +// +// Authors: +// Dave2001, original author, founded the project in 2001, left it in 2002 +// Gugaman, joined the project in 2002, left it in 2002 +// Sergey 'Gonetz' Lipski, joined the project in 2002, main author since fall of 2002 +// Hiroshi 'KoolSmoky' Morii, joined the project in 2007 +// +//**************************************************************** +// +// To modify Glide64: +// * Write your name and (optional)email, commented by your work, so I know who did it, and so that you can find which parts you modified when it comes time to send it to me. +// * Do NOT send me the whole project or file that you modified. Take out your modified code sections, and tell me where to put them. If people sent the whole thing, I would have many different versions, but no idea how to combine them all. +// +//**************************************************************** + +static void calc_point_light (VERTEX *v, float * vpos) +{ +Check_FrameSkip; + + float light_intensity = 0.0f; + register float color[3] = {rdp.light[rdp.num_lights].r, rdp.light[rdp.num_lights].g, rdp.light[rdp.num_lights].b}; + for (wxUint32 l=0; l 0.0f) + light_intensity = 1/at;//DotProduct (lvec, nvec) / (light_len * normal_len * at); + else + light_intensity = 0.0f; + } + else + { + light_intensity = 0.0f; + } + if (light_intensity > 0.0f) + { + color[0] += rdp.light[l].r * light_intensity; + color[1] += rdp.light[l].g * light_intensity; + color[2] += rdp.light[l].b * light_intensity; + } + } + if (color[0] > 1.0f) color[0] = 1.0f; + if (color[1] > 1.0f) color[1] = 1.0f; + if (color[2] > 1.0f) color[2] = 1.0f; + + v->r = (wxUint8)(color[0]*255.0f); + v->g = (wxUint8)(color[1]*255.0f); + v->b = (wxUint8)(color[2]*255.0f); +} + +static void uc6_obj_rectangle(); + +static void uc2_vertex () +{ +Check_FrameSkip; + + if (!(rdp.cmd0 & 0x00FFFFFF)) + { + uc6_obj_rectangle(); + return; + } + + // This is special, not handled in update(), but here + // * Matrix Pre-multiplication idea by Gonetz (Gonetz@ngs.ru) + if (rdp.update & UPDATE_MULT_MAT) + { + rdp.update ^= UPDATE_MULT_MAT; + MulMatrices(rdp.model, rdp.proj, rdp.combined); + } + if (rdp.update & UPDATE_LIGHTS) + { + rdp.update ^= UPDATE_LIGHTS; + + // Calculate light vectors + for (wxUint32 l=0; l> 12) & 0xFF; + rdp.v0 = v0 = ((rdp.cmd0 >> 1) & 0x7F) - n; + + FRDP ("uc2:vertex n: %d, v0: %d, from: %08lx\n", n, v0, addr); + + if (v0 < 0) + { + RDP_E ("** ERROR: uc2:vertex v0 < 0\n"); + LRDP("** ERROR: uc2:vertex v0 < 0\n"); + return; + } + + wxUint32 geom_mode = rdp.geom_mode; + if ((settings.hacks&hack_Fzero) && (rdp.geom_mode & 0x40000)) + { + if (((short*)gfx.RDRAM)[(((addr) >> 1) + 4)^1] || ((short*)gfx.RDRAM)[(((addr) >> 1) + 5)^1]) + rdp.geom_mode ^= 0x40000; + } + #ifdef __ARM_NEON__ + float32x4_t comb0, comb1, comb2, comb3; + float32x4_t v_xyzw; + comb0 = vld1q_f32(rdp.combined[0]); + comb1 = vld1q_f32(rdp.combined[1]); + comb2 = vld1q_f32(rdp.combined[2]); + comb3 = vld1q_f32(rdp.combined[3]); + #endif + for (i=0; i < (n<<4); i+=16) + { + VERTEX *v = &rdp.vtx[v0 + (i>>4)]; + x = (float)((short*)gfx.RDRAM)[(((addr+i) >> 1) + 0)^1]; + y = (float)((short*)gfx.RDRAM)[(((addr+i) >> 1) + 1)^1]; + z = (float)((short*)gfx.RDRAM)[(((addr+i) >> 1) + 2)^1]; + v->flags = ((wxUint16*)gfx.RDRAM)[(((addr+i) >> 1) + 3)^1]; + v->ou = (float)((short*)gfx.RDRAM)[(((addr+i) >> 1) + 4)^1]; + v->ov = (float)((short*)gfx.RDRAM)[(((addr+i) >> 1) + 5)^1]; + v->uv_scaled = 0; + v->a = ((wxUint8*)gfx.RDRAM)[(addr+i + 15)^3]; + + #ifdef __ARM_NEON__ + v_xyzw = x*comb0+y*comb1+z*comb2+comb3; + //vst1q_f32((float*)v, v_xyzw); + v->x=v_xyzw[0]; + v->y=v_xyzw[1]; + v->z=v_xyzw[2]; + v->w=v_xyzw[3]; + #else + v->x = x*rdp.combined[0][0] + y*rdp.combined[1][0] + z*rdp.combined[2][0] + rdp.combined[3][0]; + v->y = x*rdp.combined[0][1] + y*rdp.combined[1][1] + z*rdp.combined[2][1] + rdp.combined[3][1]; + v->z = x*rdp.combined[0][2] + y*rdp.combined[1][2] + z*rdp.combined[2][2] + rdp.combined[3][2]; + v->w = x*rdp.combined[0][3] + y*rdp.combined[1][3] + z*rdp.combined[2][3] + rdp.combined[3][3]; + #endif + + v->uv_calculated = 0xFFFFFFFF; + v->screen_translated = 0; + v->shade_mod = 0; + + if (fabs(v->w) < 0.001) v->w = 0.001f; + v->oow = 1.0f / v->w; + #ifdef __ARM_NEON__ + v_xyzw *= v->oow; + v->x_w=v_xyzw[0]; + v->y_w=v_xyzw[1]; + v->z_w=v_xyzw[2]; + #else + v->x_w = v->x * v->oow; + v->y_w = v->y * v->oow; + v->z_w = v->z * v->oow; + #endif + CalculateFog (v); + + + v->scr_off = 0; + if (v->x < -v->w) v->scr_off |= 1; + if (v->x > v->w) v->scr_off |= 2; + if (v->y < -v->w) v->scr_off |= 4; + if (v->y > v->w) v->scr_off |= 8; + if (v->w < 0.1f) v->scr_off |= 16; +// if (v->z_w > 1.0f) v->scr_off |= 32; + + if (rdp.geom_mode & 0x00020000) + { + v->vec[0] = ((char*)gfx.RDRAM)[(addr+i + 12)^3]; + v->vec[1] = ((char*)gfx.RDRAM)[(addr+i + 13)^3]; + v->vec[2] = ((char*)gfx.RDRAM)[(addr+i + 14)^3]; + // FRDP("Calc light. x: %f, y: %f z: %f\n", v->vec[0], v->vec[1], v->vec[2]); + // if (!(rdp.geom_mode & 0x800000)) + { + if (rdp.geom_mode & 0x40000) + { + if (rdp.geom_mode & 0x80000) + { + calc_linear (v); +#ifdef EXTREME_LOGGING + FRDP ("calc linear: v%d - u: %f, v: %f\n", i>>4, v->ou, v->ov); +#endif + } + else + { + calc_sphere (v); +#ifdef EXTREME_LOGGING + FRDP ("calc sphere: v%d - u: %f, v: %f\n", i>>4, v->ou, v->ov); +#endif + } + } + } + if (rdp.geom_mode & 0x00400000) + { + float tmpvec[3] = {x, y, z}; + calc_point_light (v, tmpvec); + } + else + { + NormalizeVector (v->vec); + calc_light (v); + } + } + else + { + v->r = ((wxUint8*)gfx.RDRAM)[(addr+i + 12)^3]; + v->g = ((wxUint8*)gfx.RDRAM)[(addr+i + 13)^3]; + v->b = ((wxUint8*)gfx.RDRAM)[(addr+i + 14)^3]; + } +#ifdef EXTREME_LOGGING + FRDP ("v%d - x: %f, y: %f, z: %f, w: %f, u: %f, v: %f, f: %f, z_w: %f, r=%d, g=%d, b=%d, a=%d\n", i>>4, v->x, v->y, v->z, v->w, v->ou*rdp.tiles[rdp.cur_tile].s_scale, v->ov*rdp.tiles[rdp.cur_tile].t_scale, v->f, v->z_w, v->r, v->g, v->b, v->a); +#endif + } + rdp.geom_mode = geom_mode; +} + +static void uc2_modifyvtx () +{ +Check_FrameSkip; + + wxUint8 where = (wxUint8)((rdp.cmd0 >> 16) & 0xFF); + wxUint16 vtx = (wxUint16)((rdp.cmd0 >> 1) & 0xFFFF); + + FRDP ("uc2:modifyvtx: vtx: %d, where: 0x%02lx, val: %08lx - ", vtx, where, rdp.cmd1); + uc0_modifyvtx(where, vtx, rdp.cmd1); +} + +static void uc2_culldl () +{ + wxUint16 vStart = (wxUint16)(rdp.cmd0 & 0xFFFF) >> 1; + wxUint16 vEnd = (wxUint16)(rdp.cmd1 & 0xFFFF) >> 1; + wxUint32 cond = 0; + FRDP ("uc2:culldl start: %d, end: %d\n", vStart, vEnd); + + if (vEnd < vStart) return; + for (wxUint16 i=vStart; i<=vEnd; i++) + { + /* + VERTEX v = &rdp.vtx[i]; + // Check if completely off the screen (quick frustrum clipping for 90 FOV) + if (v->x >= -v->w) + cond |= 0x01; + if (v->x <= v->w) + cond |= 0x02; + if (v->y >= -v->w) + cond |= 0x04; + if (v->y <= v->w) + cond |= 0x08; + if (v->w >= 0.1f) + cond |= 0x10; + + if (cond == 0x1F) + return; + //*/ + +#ifdef EXTREME_LOGGING + FRDP (" v[%d] = (%02f, %02f, %02f, 0x%02lx)\n", i, rdp.vtx[i].x, rdp.vtx[i].y, rdp.vtx[i].w, rdp.vtx[i].scr_off); +#endif + + cond |= (~rdp.vtx[i].scr_off) & 0x1F; + if (cond == 0x1F) + return; + } + + LRDP(" - "); // specify that the enddl is not a real command + uc0_enddl (); +} + +static void uc6_obj_loadtxtr (); + +static void uc2_tri1() +{ +Check_FrameSkip; + + if ((rdp.cmd0 & 0x00FFFFFF) == 0x17) + { + uc6_obj_loadtxtr (); + return; + } + if (rdp.skip_drawing) + { + LRDP("uc2:tri1. skipped\n"); + return; + } + + FRDP("uc2:tri1 #%d - %d, %d, %d\n", rdp.tri_n, + ((rdp.cmd0 >> 17) & 0x7F), + ((rdp.cmd0 >> 9) & 0x7F), + ((rdp.cmd0 >> 1) & 0x7F)); + + VERTEX *v[3] = { + &rdp.vtx[(rdp.cmd0 >> 17) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 9) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 1) & 0x7F] + }; + + rsp_tri1(v); +} + +static void uc6_obj_ldtx_sprite (); +static void uc6_obj_ldtx_rect (); + +static void uc2_quad () +{ +Check_FrameSkip; + + if ((rdp.cmd0 & 0x00FFFFFF) == 0x2F) + { + wxUint32 command = rdp.cmd0>>24; + if (command == 0x6) + { + uc6_obj_ldtx_sprite (); + return; + } + if (command == 0x7) + { + uc6_obj_ldtx_rect (); + return; + } + } + + if (rdp.skip_drawing) + { + LRDP("uc2_quad. skipped\n"); + return; + } + + LRDP("uc2:quad"); + + FRDP(" #%d, #%d - %d, %d, %d - %d, %d, %d\n", rdp.tri_n, rdp.tri_n+1, + ((rdp.cmd0 >> 17) & 0x7F), + ((rdp.cmd0 >> 9) & 0x7F), + ((rdp.cmd0 >> 1) & 0x7F), + ((rdp.cmd1 >> 17) & 0x7F), + ((rdp.cmd1 >> 9) & 0x7F), + ((rdp.cmd1 >> 1) & 0x7F)); + + VERTEX *v[6] = { + &rdp.vtx[(rdp.cmd0 >> 17) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 9) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 1) & 0x7F], + &rdp.vtx[(rdp.cmd1 >> 17) & 0x7F], + &rdp.vtx[(rdp.cmd1 >> 9) & 0x7F], + &rdp.vtx[(rdp.cmd1 >> 1) & 0x7F] + }; + + rsp_tri2(v); +} + +static void uc6_ldtx_rect_r (); + +static void uc2_line3d () +{ +Check_FrameSkip; + + if ( (rdp.cmd0&0xFF) == 0x2F ) + uc6_ldtx_rect_r (); + else + { + FRDP("uc2:line3d #%d, #%d - %d, %d\n", rdp.tri_n, rdp.tri_n+1, + (rdp.cmd0 >> 17) & 0x7F, + (rdp.cmd0 >> 9) & 0x7F); + + VERTEX *v[3] = { + &rdp.vtx[(rdp.cmd0 >> 17) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 9) & 0x7F], + &rdp.vtx[(rdp.cmd0 >> 9) & 0x7F] + }; + wxUint16 width = (wxUint16)(rdp.cmd0 + 3)&0xFF; + wxUint32 cull_mode = (rdp.flags & CULLMASK) >> CULLSHIFT; + rdp.flags |= CULLMASK; + rdp.update |= UPDATE_CULL_MODE; + rsp_tri1(v, width); + rdp.flags ^= CULLMASK; + rdp.flags |= cull_mode << CULLSHIFT; + rdp.update |= UPDATE_CULL_MODE; + } +} + +static void uc2_special3 () +{ + LRDP("uc2:special3\n"); +} + +static void uc2_special2 () +{ + LRDP("uc2:special2\n"); +} + +static void uc2_dma_io () +{ + LRDP("uc2:dma_io\n"); +} + +static void uc2_pop_matrix () +{ + FRDP ("uc2:pop_matrix %08lx, %08lx\n", rdp.cmd0, rdp.cmd1); + + // Just pop the modelview matrix + modelview_pop (rdp.cmd1 >> 6); +} + +static void uc2_geom_mode () +{ + // Switch around some things + wxUint32 clr_mode = (rdp.cmd0 & 0x00DFC9FF) | + ((rdp.cmd0 & 0x00000600) << 3) | + ((rdp.cmd0 & 0x00200000) >> 12) | 0xFF000000; + wxUint32 set_mode = (rdp.cmd1 & 0xFFDFC9FF) | + ((rdp.cmd1 & 0x00000600) << 3) | + ((rdp.cmd1 & 0x00200000) >> 12); + + FRDP("uc2:geom_mode c:%08lx, s:%08lx ", clr_mode, set_mode); + + rdp.geom_mode &= clr_mode; + rdp.geom_mode |= set_mode; + + FRDP ("result:%08lx\n", rdp.geom_mode); + + if (rdp.geom_mode & 0x00000001) // Z-Buffer enable + { + if (!(rdp.flags & ZBUF_ENABLED)) + { + rdp.flags |= ZBUF_ENABLED; + rdp.update |= UPDATE_ZBUF_ENABLED; + } + } + else + { + if ((rdp.flags & ZBUF_ENABLED)) + { + if (!settings.flame_corona || (rdp.rm != 0x00504341)) //hack for flame's corona + rdp.flags ^= ZBUF_ENABLED; + rdp.update |= UPDATE_ZBUF_ENABLED; + } + } + if (rdp.geom_mode & 0x00001000) // Front culling + { + if (!(rdp.flags & CULL_FRONT)) + { + rdp.flags |= CULL_FRONT; + rdp.update |= UPDATE_CULL_MODE; + } + } + else + { + if (rdp.flags & CULL_FRONT) + { + rdp.flags ^= CULL_FRONT; + rdp.update |= UPDATE_CULL_MODE; + } + } + if (rdp.geom_mode & 0x00002000) // Back culling + { + if (!(rdp.flags & CULL_BACK)) + { + rdp.flags |= CULL_BACK; + rdp.update |= UPDATE_CULL_MODE; + } + } + else + { + if (rdp.flags & CULL_BACK) + { + rdp.flags ^= CULL_BACK; + rdp.update |= UPDATE_CULL_MODE; + } + } + + //Added by Gonetz + if (rdp.geom_mode & 0x00010000) // Fog enable + { + if (!(rdp.flags & FOG_ENABLED)) + { + rdp.flags |= FOG_ENABLED; + rdp.update |= UPDATE_FOG_ENABLED; + } + } + else + { + if (rdp.flags & FOG_ENABLED) + { + rdp.flags ^= FOG_ENABLED; + rdp.update |= UPDATE_FOG_ENABLED; + } + } +} + +static void uc6_obj_rectangle_r (); + +static void uc2_matrix () +{ + if (!(rdp.cmd0 & 0x00FFFFFF)) + { + uc6_obj_rectangle_r(); + return; + } + LRDP("uc2:matrix\n"); + + DECLAREALIGN16VAR(m[4][4]); + load_matrix(m, segoffset(rdp.cmd1)); + + wxUint8 command = (wxUint8)((rdp.cmd0 ^ 1) & 0xFF); + switch (command) + { + case 0: // modelview mul nopush + LRDP("modelview mul\n"); + modelview_mul (m); + break; + + case 1: // modelview mul push + LRDP("modelview mul push\n"); + modelview_mul_push (m); + break; + + case 2: // modelview load nopush + LRDP("modelview load\n"); + modelview_load (m); + break; + + case 3: // modelview load push + LRDP("modelview load push\n"); + modelview_load_push (m); + break; + + case 4: // projection mul nopush + case 5: // projection mul push, can't push projection + LRDP("projection mul\n"); + projection_mul (m); + break; + + case 6: // projection load nopush + case 7: // projection load push, can't push projection + LRDP("projection load\n"); + projection_load (m); + break; + + default: + FRDP_E ("Unknown matrix command, %02lx", command); + FRDP ("Unknown matrix command, %02lx", command); + } + +#ifdef EXTREME_LOGGING + FRDP ("{%f,%f,%f,%f}\n", m[0][0], m[0][1], m[0][2], m[0][3]); + FRDP ("{%f,%f,%f,%f}\n", m[1][0], m[1][1], m[1][2], m[1][3]); + FRDP ("{%f,%f,%f,%f}\n", m[2][0], m[2][1], m[2][2], m[2][3]); + FRDP ("{%f,%f,%f,%f}\n", m[3][0], m[3][1], m[3][2], m[3][3]); + FRDP ("\nmodel\n{%f,%f,%f,%f}\n", rdp.model[0][0], rdp.model[0][1], rdp.model[0][2], rdp.model[0][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.model[1][0], rdp.model[1][1], rdp.model[1][2], rdp.model[1][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.model[2][0], rdp.model[2][1], rdp.model[2][2], rdp.model[2][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.model[3][0], rdp.model[3][1], rdp.model[3][2], rdp.model[3][3]); + FRDP ("\nproj\n{%f,%f,%f,%f}\n", rdp.proj[0][0], rdp.proj[0][1], rdp.proj[0][2], rdp.proj[0][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.proj[1][0], rdp.proj[1][1], rdp.proj[1][2], rdp.proj[1][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.proj[2][0], rdp.proj[2][1], rdp.proj[2][2], rdp.proj[2][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.proj[3][0], rdp.proj[3][1], rdp.proj[3][2], rdp.proj[3][3]); +#endif +} + +static void uc2_moveword () +{ + wxUint8 index = (wxUint8)((rdp.cmd0 >> 16) & 0xFF); + wxUint16 offset = (wxUint16)(rdp.cmd0 & 0xFFFF); + wxUint32 data = rdp.cmd1; + + FRDP ("uc2:moveword "); + + switch (index) + { + // NOTE: right now it's assuming that it sets the integer part first. This could + // be easily fixed, but only if i had something to test with. + + case 0x00: // moveword matrix + { + // do matrix pre-mult so it's re-updated next time + if (rdp.update & UPDATE_MULT_MAT) + { + rdp.update ^= UPDATE_MULT_MAT; + MulMatrices(rdp.model, rdp.proj, rdp.combined); + } + + if (rdp.cmd0 & 0x20) // fractional part + { + int index_x = (rdp.cmd0 & 0x1F) >> 1; + int index_y = index_x >> 2; + index_x &= 3; + + float fpart = (rdp.cmd1>>16)/65536.0f; + rdp.combined[index_y][index_x] = (float)(int)rdp.combined[index_y][index_x]; + rdp.combined[index_y][index_x] += fpart; + + fpart = (rdp.cmd1&0xFFFF)/65536.0f; + rdp.combined[index_y][index_x+1] = (float)(int)rdp.combined[index_y][index_x+1]; + rdp.combined[index_y][index_x+1] += fpart; + } + else + { + int index_x = (rdp.cmd0 & 0x1F) >> 1; + int index_y = index_x >> 2; + index_x &= 3; + + rdp.combined[index_y][index_x] = (short)(rdp.cmd1>>16); + rdp.combined[index_y][index_x+1] = (short)(rdp.cmd1&0xFFFF); + } + + LRDP("matrix\n"); + } + break; + + case 0x02: + rdp.num_lights = data / 24; + rdp.update |= UPDATE_LIGHTS; + FRDP ("numlights: %d\n", rdp.num_lights); + break; + + case 0x04: + if (offset == 0x04) + { + rdp.clip_ratio = sqrt((float)rdp.cmd1); + rdp.update |= UPDATE_VIEWPORT; + } + FRDP ("mw_clip %08lx, %08lx\n", rdp.cmd0, rdp.cmd1); + break; + + case 0x06: // moveword SEGMENT + { + FRDP ("SEGMENT %08lx -> seg%d\n", data, offset >> 2); + if ((data&BMASK)> 2) & 0xF] = data; + } + break; + + + case 0x08: + { + rdp.fog_multiplier = (short)(rdp.cmd1 >> 16); + rdp.fog_offset = (short)(rdp.cmd1 & 0x0000FFFF); + FRDP ("fog: multiplier: %f, offset: %f\n", rdp.fog_multiplier, rdp.fog_offset); + + //offset must be 0 for move_fog, but it can be non zero in Nushi Zuri 64 - Shiokaze ni Notte + //low-level display list has setothermode commands in this place, so this is obviously not move_fog. + if (offset == 0x04) + rdp.tlut_mode = (data == 0xffffffff) ? 0 : 2; + } + break; + + case 0x0a: // moveword LIGHTCOL + { + int n = offset / 24; + FRDP ("lightcol light:%d, %08lx\n", n, data); + + rdp.light[n].r = (float)((data >> 24) & 0xFF) / 255.0f; + rdp.light[n].g = (float)((data >> 16) & 0xFF) / 255.0f; + rdp.light[n].b = (float)((data >> 8) & 0xFF) / 255.0f; + rdp.light[n].a = 255; + } + break; + + case 0x0c: + RDP_E ("uc2:moveword forcemtx - IGNORED\n"); + LRDP("forcemtx - IGNORED\n"); + break; + + case 0x0e: + LRDP("perspnorm - IGNORED\n"); + break; + + default: + FRDP_E("uc2:moveword unknown (index: 0x%08lx, offset 0x%08lx)\n", index, offset); + FRDP ("unknown (index: 0x%08lx, offset 0x%08lx)\n", index, offset); + } +} + +static void uc6_obj_movemem (); + +static void uc2_movemem () +{ + int idx = rdp.cmd0 & 0xFF; + wxUint32 addr = segoffset(rdp.cmd1); + int ofs = (rdp.cmd0 >> 5) & 0x7F8; + + FRDP ("uc2:movemem ofs:%d ", ofs); + + switch (idx) + { + case 0: + case 2: + uc6_obj_movemem (); + break; + + case 8: // VIEWPORT + { + wxUint32 a = addr >> 1; + short scale_x = ((short*)gfx.RDRAM)[(a+0)^1] >> 2; + short scale_y = ((short*)gfx.RDRAM)[(a+1)^1] >> 2; + short scale_z = ((short*)gfx.RDRAM)[(a+2)^1]; + short trans_x = ((short*)gfx.RDRAM)[(a+4)^1] >> 2; + short trans_y = ((short*)gfx.RDRAM)[(a+5)^1] >> 2; + short trans_z = ((short*)gfx.RDRAM)[(a+6)^1]; + rdp.view_scale[0] = scale_x * rdp.scale_x; + rdp.view_scale[1] = -scale_y * rdp.scale_y; + rdp.view_scale[2] = 32.0f * scale_z; + rdp.view_trans[0] = trans_x * rdp.scale_x; + rdp.view_trans[1] = trans_y * rdp.scale_y; + rdp.view_trans[2] = 32.0f * trans_z; + + rdp.update |= UPDATE_VIEWPORT; + + FRDP ("viewport scale(%d, %d, %d), trans(%d, %d, %d), from:%08lx\n", scale_x, scale_y, scale_z, + trans_x, trans_y, trans_z, a); + } + break; + + case 10: // LIGHT + { + int n = ofs / 24; + + if (n < 2) + { + char dir_x = ((char*)gfx.RDRAM)[(addr+8)^3]; + rdp.lookat[n][0] = (float)(dir_x) / 127.0f; + char dir_y = ((char*)gfx.RDRAM)[(addr+9)^3]; + rdp.lookat[n][1] = (float)(dir_y) / 127.0f; + char dir_z = ((char*)gfx.RDRAM)[(addr+10)^3]; + rdp.lookat[n][2] = (float)(dir_z) / 127.0f; + rdp.use_lookat = TRUE; + if (n == 1) + { + if (!dir_x && !dir_y) + rdp.use_lookat = FALSE; + } + FRDP("lookat_%d (%f, %f, %f)\n", n, rdp.lookat[n][0], rdp.lookat[n][1], rdp.lookat[n][2]); + return; + } + n -= 2; + if (n > 7) return; + + // Get the data + wxUint8 col = gfx.RDRAM[(addr+0)^3]; + rdp.light[n].r = (float)col / 255.0f; + rdp.light[n].nonblack = col; + col = gfx.RDRAM[(addr+1)^3]; + rdp.light[n].g = (float)col / 255.0f; + rdp.light[n].nonblack += col; + col = gfx.RDRAM[(addr+2)^3]; + rdp.light[n].b = (float)col / 255.0f; + rdp.light[n].nonblack += col; + rdp.light[n].a = 1.0f; + // ** Thanks to Icepir8 for pointing this out ** + // Lighting must be signed byte instead of byte + rdp.light[n].dir_x = (float)(((char*)gfx.RDRAM)[(addr+8)^3]) / 127.0f; + rdp.light[n].dir_y = (float)(((char*)gfx.RDRAM)[(addr+9)^3]) / 127.0f; + rdp.light[n].dir_z = (float)(((char*)gfx.RDRAM)[(addr+10)^3]) / 127.0f; + wxUint32 a = addr >> 1; + rdp.light[n].x = (float)(((short*)gfx.RDRAM)[(a+4)^1]); + rdp.light[n].y = (float)(((short*)gfx.RDRAM)[(a+5)^1]); + rdp.light[n].z = (float)(((short*)gfx.RDRAM)[(a+6)^1]); + rdp.light[n].ca = (float)(gfx.RDRAM[(addr+3)^3]) / 16.0f; + rdp.light[n].la = (float)(gfx.RDRAM[(addr+7)^3]); + rdp.light[n].qa = (float)(gfx.RDRAM[(addr+14)^3]) / 8.0f; +#ifdef EXTREME_LOGGING + FRDP ("light: n: %d, pos: x: %f, y: %f, z: %f, ca: %f, la:%f, qa: %f\n", + n, rdp.light[n].x, rdp.light[n].y, rdp.light[n].z, rdp.light[n].ca, rdp.light[n].la, rdp.light[n].qa); +#endif + FRDP ("light: n: %d, r: %.3f, g: %.3f, b: %.3f. dir: x: %.3f, y: %.3f, z: %.3f\n", + n, rdp.light[n].r, rdp.light[n].g, rdp.light[n].b, + rdp.light[n].dir_x, rdp.light[n].dir_y, rdp.light[n].dir_z); + } + break; + + case 14: // matrix + { + // do not update the combined matrix! + rdp.update &= ~UPDATE_MULT_MAT; + load_matrix(rdp.combined, segoffset(rdp.cmd1)); + +#ifdef EXTREME_LOGGING + FRDP ("{%f,%f,%f,%f}\n", rdp.combined[0][0], rdp.combined[0][1], rdp.combined[0][2], rdp.combined[0][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.combined[1][0], rdp.combined[1][1], rdp.combined[1][2], rdp.combined[1][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.combined[2][0], rdp.combined[2][1], rdp.combined[2][2], rdp.combined[2][3]); + FRDP ("{%f,%f,%f,%f}\n", rdp.combined[3][0], rdp.combined[3][1], rdp.combined[3][2], rdp.combined[3][3]); +#endif + } + break; + + default: + FRDP ("uc2:matrix unknown (%d)\n", idx); + FRDP ("** UNKNOWN %d\n", idx); + } +} + +static void uc2_load_ucode () +{ + LRDP("uc2:load_ucode\n"); +} + +static void uc2_rdphalf_2 () +{ + LRDP("uc2:rdphalf_2\n"); +} + +static void uc2_dlist_cnt () +{ + wxUint32 addr = segoffset(rdp.cmd1) & BMASK; + int count = rdp.cmd0 & 0x000000FF; + FRDP ("dl_count - addr: %08lx, count: %d\n", addr, count); + if (addr == 0) + return; + + if (rdp.pc_i >= 9) { + RDP_E ("** DL stack overflow **\n"); + LRDP("** DL stack overflow **\n"); + return; + } + rdp.pc_i ++; // go to the next PC in the stack + rdp.pc[rdp.pc_i] = addr; // jump to the address + rdp.dl_count = count + 1; +}