diff options
| author | 2015-05-26 21:40:54 +0200 | |
|---|---|---|
| committer | 2015-05-26 21:40:54 +0200 | |
| commit | 4922311dd0e40b1ee632ff9404871505e45a138b (patch) | |
| tree | ce37055c681ef50dc1688fd96d30f9db951e0125 | |
| parent | 583ea9dccb301ecb64c65777ad190c4216b950a5 (diff) | |
- n64: Improve texture filtering performance in the RDP. [MooglyGuy]
| -rw-r--r-- | src/mame/video/n64.c | 9 | ||||
| -rw-r--r-- | src/mame/video/n64.h | 3 | ||||
| -rw-r--r-- | src/mame/video/rdpspn16.c | 18 | ||||
| -rw-r--r-- | src/mame/video/rdptpipe.c | 513 | ||||
| -rw-r--r-- | src/mame/video/rdptpipe.h | 29 |
5 files changed, 174 insertions, 398 deletions
diff --git a/src/mame/video/n64.c b/src/mame/video/n64.c index 1d76af57e30..28c095bc72c 100644 --- a/src/mame/video/n64.c +++ b/src/mame/video/n64.c @@ -3001,6 +3001,15 @@ void n64_rdp::cmd_set_tile(UINT32 w1, UINT32 w2) tex_tile->mask_s = (w2 >> 4) & 0xf; tex_tile->shift_s = (w2 >> 0) & 0xf; + tex_tile->lshift_s = (tex_tile->shift_s >= 11) ? (16 - tex_tile->shift_s) : 0; + tex_tile->rshift_s = (tex_tile->shift_s < 11) ? tex_tile->shift_s : 0; + tex_tile->lshift_t = (tex_tile->shift_t >= 11) ? (16 - tex_tile->shift_t) : 0; + tex_tile->rshift_t = (tex_tile->shift_t < 11) ? tex_tile->shift_t : 0; + tex_tile->wrapped_mask_s = (tex_tile->mask_s > 10 ? 10 : tex_tile->mask_s); + tex_tile->wrapped_mask_t = (tex_tile->mask_t > 10 ? 10 : tex_tile->mask_t); + tex_tile->clamp_s = tex_tile->cs || !tex_tile->mask_s; + tex_tile->clamp_t = tex_tile->ct || !tex_tile->mask_t; + if (tex_tile->format == FORMAT_I && tex_tile->size > PIXEL_SIZE_8BIT) { tex_tile->format = FORMAT_RGBA; // Used by Supercross 2000 (in-game) diff --git a/src/mame/video/n64.h b/src/mame/video/n64.h index b670ac8e7cd..5cdb7ee0ebf 100644 --- a/src/mame/video/n64.h +++ b/src/mame/video/n64.h @@ -185,6 +185,9 @@ struct n64_tile_t INT32 palette; // Palette number for 4b CI texels INT32 ct, mt, cs, ms; // Clamp / mirror enable bits for S / T direction INT32 mask_t, shift_t, mask_s, shift_s; // Mask values / LOD shifts + INT32 lshift_s, rshift_s, lshift_t, rshift_t; + INT32 wrapped_mask_s, wrapped_mask_t; + bool clamp_s, clamp_t; UINT16 sl, tl, sh, th; // 10.2 fixed-point, starting and ending texel row / column INT32 num; }; diff --git a/src/mame/video/rdpspn16.c b/src/mame/video/rdpspn16.c index 1c9ada892ba..bc6fe817530 100644 --- a/src/mame/video/rdpspn16.c +++ b/src/mame/video/rdpspn16.c @@ -340,9 +340,7 @@ void n64_rdp::span_draw_1cycle(INT32 scanline, const extent_t &extent, const rdp #endif rdp_span_aux* userdata = (rdp_span_aux*)extent.userdata; - INT32 clamp_s_diff[8]; - INT32 clamp_t_diff[8]; - m_tex_pipe.calculate_clamp_diffs(tilenum, userdata, object, clamp_s_diff, clamp_t_diff); + m_tex_pipe.calculate_clamp_diffs(tilenum, object); const bool partialreject = (userdata->m_color_inputs.blender2b_a[0] == &userdata->m_inv_pixel_color.i.a && userdata->m_color_inputs.blender1b_a[0] == &userdata->m_pixel_color.i.a); const INT32 sel0 = (userdata->m_color_inputs.blender2b_a[0] == &userdata->m_memory_color.i.a) ? 1 : 0; @@ -437,8 +435,8 @@ void n64_rdp::span_draw_1cycle(INT32 scanline, const extent_t &extent, const rdp rgbaz_correct_triangle(offx, offy, &sr, &sg, &sb, &sa, &sz, userdata, object); rgbaz_clip(sr, sg, sb, sa, &sz, userdata); - ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle0]))(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tilenum, 0, userdata, object, clamp_s_diff, clamp_t_diff); - //m_tex_pipe.Cycle(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tilenum, 0, userdata, object, clamp_s_diff, clamp_t_diff); + ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle0]))(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tilenum, 0, userdata, object); + //m_tex_pipe.Cycle(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tilenum, 0, userdata, object); userdata->m_noise_color.i.r = userdata->m_noise_color.i.g = userdata->m_noise_color.i.b = rand() << 3; // Not accurate @@ -531,9 +529,7 @@ void n64_rdp::span_draw_2cycle(INT32 scanline, const extent_t &extent, const rdp #endif rdp_span_aux* userdata = (rdp_span_aux*)extent.userdata; - INT32 clamp_s_diff[8]; - INT32 clamp_t_diff[8]; - m_tex_pipe.calculate_clamp_diffs(tile1, userdata, object, clamp_s_diff, clamp_t_diff); + m_tex_pipe.calculate_clamp_diffs(tile1, object); bool partialreject = (userdata->m_color_inputs.blender2b_a[1] == &userdata->m_inv_pixel_color.i.a && userdata->m_color_inputs.blender1b_a[1] == &userdata->m_pixel_color.i.a); INT32 sel0 = (userdata->m_color_inputs.blender2b_a[0] == &userdata->m_memory_color.i.a) ? 1 : 0; @@ -644,9 +640,9 @@ void n64_rdp::span_draw_2cycle(INT32 scanline, const extent_t &extent, const rdp rgbaz_correct_triangle(offx, offy, &sr, &sg, &sb, &sa, &sz, userdata, object); rgbaz_clip(sr, sg, sb, sa, &sz, userdata); - ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle0]))(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tile1, 0, userdata, object, clamp_s_diff, clamp_t_diff); - ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle1]))(&userdata->m_texel1_color, &userdata->m_texel0_color, sss, sst, tile2, 1, userdata, object, clamp_s_diff, clamp_t_diff); - ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle1]))(&userdata->m_next_texel_color, &userdata->m_next_texel_color, sss, sst, tile2, 1, userdata, object, clamp_s_diff, clamp_t_diff); + ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle0]))(&userdata->m_texel0_color, &userdata->m_texel0_color, sss, sst, tile1, 0, userdata, object); + ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle1]))(&userdata->m_texel1_color, &userdata->m_texel0_color, sss, sst, tile2, 1, userdata, object); + ((m_tex_pipe).*(m_tex_pipe.m_cycle[cycle1]))(&userdata->m_next_texel_color, &userdata->m_next_texel_color, sss, sst, tile2, 1, userdata, object); userdata->m_noise_color.i.r = userdata->m_noise_color.i.g = userdata->m_noise_color.i.b = rand() << 3; // Not accurate userdata->m_combined_color.i.r = color_combiner_equation(*userdata->m_color_inputs.combiner_rgbsub_a_r[0], diff --git a/src/mame/video/rdptpipe.c b/src/mame/video/rdptpipe.c index 1ddeb3e479d..adfe550d57d 100644 --- a/src/mame/video/rdptpipe.c +++ b/src/mame/video/rdptpipe.c @@ -34,65 +34,69 @@ void n64_texture_pipe_t::set_machine(running_machine &machine) c.i.a = (i & 1) ? 0xff : 0x00; m_expand_16to32_table[i] = c.c; } + + for(UINT32 i = 0; i < 0x80000; i++) + { + if (i & 0x40000) + { + m_lod_lookup[i] = 0x7fff; + } + else if (i & 0x20000) + { + m_lod_lookup[i] = 0x8000; + } + else + { + INT32 tempanded = i & 0x18000; + if ((i & 0x18000) == 0x8000) + { + m_lod_lookup[i] = 0x7fff; + } + else if ((i & 0x18000) == 0x10000) + { + m_lod_lookup[i] = 0x8000; + } + else + { + m_lod_lookup[i] = i & 0xffff; + } + } + } } -void n64_texture_pipe_t::mask(INT32* S, INT32* T, INT32 num, const rdp_poly_state& object) +void n64_texture_pipe_t::mask(INT32* S, INT32* T, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - - if (tiles[num].mask_s) + if (tile.mask_s) { - INT32 wrap = *S >> (tiles[num].mask_s > 10 ? 10 : tiles[num].mask_s); + INT32 wrap = *S >> tile.wrapped_mask_s; wrap &= 1; - if (tiles[num].ms && wrap) + if (tile.ms && wrap) { *S = (~(*S)); } - *S &= m_maskbits_table[tiles[num].mask_s]; + *S &= m_maskbits_table[tile.mask_s]; } - if (tiles[num].mask_t) + if (tile.mask_t) { - INT32 wrap = *T >> (tiles[num].mask_t > 10 ? 10 : tiles[num].mask_t); + INT32 wrap = *T >> tile.wrapped_mask_t; wrap &= 1; - if (tiles[num].mt && wrap) + if (tile.mt && wrap) { *T = (~(*T)); } - *T &= m_maskbits_table[tiles[num].mask_t]; + *T &= m_maskbits_table[tile.mask_t]; } } -#define MASK_COUPLED(param, param1, num, coord) \ - if (tiles[num].mask_##coord) \ - { \ - INT32 maskbits_##coord = m_maskbits_table[tiles[num].mask_##coord]; \ - if (tiles[num].m##coord) \ - { \ - INT32 wrapthreshold = tiles[num].mask_##coord > 10 ? 10 : tiles[num].mask_##coord; \ - if ((param >> wrapthreshold) & 1) \ - { \ - param = ~param; \ - } \ - if ((param1 >> wrapthreshold) & 1) \ - { \ - param1 = ~param1; \ - } \ - } \ - param &= maskbits_##coord; \ - param1 &= maskbits_##coord; \ - } - -void n64_texture_pipe_t::mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, INT32 num, const rdp_poly_state& object) +void n64_texture_pipe_t::mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - - if (tiles[num].mask_s) + if (tile.mask_s) { - const INT32 maskbits_s = m_maskbits_table[tiles[num].mask_s]; - if (tiles[num].ms) + const INT32 maskbits_s = m_maskbits_table[tile.mask_s]; + if (tile.ms) { - const INT32 swrapthreshold = tiles[num].mask_s > 10 ? 10 : tiles[num].mask_s; + const INT32 swrapthreshold = tile.mask_s > 10 ? 10 : tile.mask_s; const INT32 wrap = (*S >> swrapthreshold) & 1; const INT32 wrap1 = (*S1 >> swrapthreshold) & 1; if (wrap) @@ -108,12 +112,12 @@ void n64_texture_pipe_t::mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, *S1 &= maskbits_s; } - if (tiles[num].mask_t) + if (tile.mask_t) { - const INT32 maskbits_t = m_maskbits_table[tiles[num].mask_t]; - if (tiles[num].mt) + const INT32 maskbits_t = m_maskbits_table[tile.mask_t]; + if (tile.mt) { - const INT32 twrapthreshold = tiles[num].mask_t > 10 ? 10 : tiles[num].mask_t; + const INT32 twrapthreshold = tile.mask_t > 10 ? 10 : tile.mask_t; const INT32 wrap = (*T >> twrapthreshold) & 1; const INT32 wrap1 = (*T1 >> twrapthreshold) & 1; if (wrap) @@ -130,100 +134,39 @@ void n64_texture_pipe_t::mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, } } -#define SHIFT_CYCLE(param, max, num, coord) \ - param = SIGN16(param); \ - if (tiles[num].shift_##coord < 11) \ - { \ - param >>= tiles[num].shift_##coord; \ - } \ - else \ - { \ - param <<= (16 - tiles[num].shift_##coord); \ - } \ - param = SIGN16(param); \ - max = ((param >> 3) >= tiles[num].coord##h); - -void n64_texture_pipe_t::shift_cycle(INT32* S, INT32* T, INT32* maxs, INT32* maxt, UINT32 num, const rdp_poly_state& object) +void n64_texture_pipe_t::shift_cycle(INT32* S, INT32* T, bool* maxs, bool* maxt, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - *S = SIGN16(*S); - if (tiles[num].shift_s < 11) - { - *S >>= tiles[num].shift_s; - } - else - { - *S <<= (16 - tiles[num].shift_s); - } - *S = SIGN16(*S); - *maxs = ((*S >> 3) >= tiles[num].sh); - - *T = SIGN16(*T); - if (tiles[num].shift_t < 11) - { - *T >>= tiles[num].shift_t; - } - else - { - *T <<= (16 - tiles[num].shift_t); - } - *T = SIGN16(*T); - *maxt = ((*T >> 3) >= tiles[num].th); + INT32 sss = (INT32)(INT16)*S; + sss >>= tile.rshift_s; + sss <<= tile.lshift_s; + sss = (INT32)(INT16)sss; + *maxs = ((sss >> 3) >= tile.sh); + *S = (((sss >> 3) - tile.sl) << 3) | (sss & 7); + + INT32 sst = (INT32)(INT16)*T; + sst >>= tile.rshift_t; + sst <<= tile.lshift_t; + *T = (INT32)(INT16)sst; + *maxt = ((*T >> 3) >= tile.th); + *T = (((sst >> 3) - tile.tl) << 3) | (sst & 7); } -void n64_texture_pipe_t::shift_copy(INT32* S, INT32* T, UINT32 num, const rdp_poly_state& object) +void n64_texture_pipe_t::shift_copy(INT32* S, INT32* T, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - *S = SIGN16(*S); - *T = SIGN16(*T); - if (tiles[num].shift_s < 11)//?-? tcu_tile - { - *S >>= tiles[num].shift_s; - } - else - { - *S <<= (16 - tiles[num].shift_s); - } - *S = SIGN16(*S); - if (tiles[num].shift_t < 11) - { - *T >>= tiles[num].shift_t; - } - else - { - *T <<= (16 - tiles[num].shift_t); - } - *T = SIGN16(*T); + INT32 sss = (INT32)(INT16)*S; + sss >>= tile.rshift_s; + sss <<= tile.lshift_s; + *S = (INT32)(INT16)sss; + + INT32 sst = (INT32)(INT16)*T; + sst >>= tile.rshift_t; + sst <<= tile.lshift_t; + *T = (INT32)(INT16)sst; } -#define CLAMP_CYCLE(param, frac, max, num, coord) \ - if (tiles[num].c##coord || !tiles[num].mask_##coord) \ - { \ - if (param & 0x10000) \ - { \ - param = 0; \ - frac = 0; \ - } \ - else if (max) \ - { \ - param = m_clamp_##coord##_diff[num]; \ - frac = 0; \ - } \ - else \ - { \ - param = (SIGN17(param) >> 5) & 0x1fff; \ - } \ - } \ - else \ - { \ - param = (SIGN17(param) >> 5) & 0x1fff; \ - } - -void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TFRAC, INT32 maxs, INT32 maxt, INT32 num, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TFRAC, const bool maxs, const bool maxt, const INT32 tilenum, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - - if (tiles[num].cs || !tiles[num].mask_s) + if (tile.clamp_s) { if (*S & 0x10000) { @@ -232,7 +175,7 @@ void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TF } else if (maxs) { - *S = m_clamp_s_diff[num]; + *S = m_clamp_s_diff[tilenum]; *SFRAC = 0; } else @@ -245,7 +188,7 @@ void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TF *S = (SIGN17(*S) >> 5) & 0x1fff; } - if (tiles[num].ct || !tiles[num].mask_t) + if (tile.clamp_t) { if (*T & 0x10000) { @@ -254,7 +197,7 @@ void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TF } else if (maxt) { - *T = m_clamp_t_diff[num]; + *T = m_clamp_t_diff[tilenum]; *TFRAC = 0; } else @@ -268,11 +211,9 @@ void n64_texture_pipe_t::clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TF } } -void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool maxt, INT32 num, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, const bool maxs, const bool maxt, const INT32 tilenum, const n64_tile_t& tile) { - const n64_tile_t* tiles = object.m_tiles; - - if (tiles[num].cs || !tiles[num].mask_s) + if (tile.clamp_s) { if (*S & 0x10000) { @@ -280,7 +221,7 @@ void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool m } else if (maxs) { - *S = m_clamp_s_diff[num]; + *S = m_clamp_s_diff[tilenum]; } else { @@ -292,7 +233,7 @@ void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool m *S = (SIGN17(*S) >> 5) & 0x1fff; } - if (tiles[num].ct || !tiles[num].mask_t) + if (tile.clamp_t) { if (*T & 0x10000) { @@ -300,7 +241,7 @@ void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool m } else if (maxt) { - *T = m_clamp_t_diff[num]; + *T = m_clamp_t_diff[tilenum]; } else { @@ -313,30 +254,22 @@ void n64_texture_pipe_t::clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool m } } -void n64_texture_pipe_t::cycle_nearest(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::cycle_nearest(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object) { const n64_tile_t* tiles = object.m_tiles; - const n64_tile_t& tile = tiles[tilenum]; + const n64_tile_t& tile = object.m_tiles[tilenum]; const UINT32 tformat = tile.format; const UINT32 tsize = tile.size; const UINT32 tpal = tile.palette; const UINT32 index = (tformat << 4) | (tsize << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; -#define TRELATIVE(x, y) ((((x) >> 3) - (y)) << 3) | (x & 7); color_t t0; - INT32 sss1 = SSS; - INT32 maxs; - SHIFT_CYCLE(sss1, maxs, tilenum, s); - sss1 = TRELATIVE(sss1, tile.sl); - - INT32 sst1 = SST; - INT32 maxt; - SHIFT_CYCLE(sst1, maxt, tilenum, t); - sst1 = TRELATIVE(sst1, tile.tl); - - clamp_cycle_light(&sss1, &sst1, maxs, maxt, tilenum, userdata, object, m_clamp_s_diff, m_clamp_t_diff); - mask(&sss1, &sst1, tilenum, object); + INT32 sss1 = SSS, sst1 = SST; + bool maxs, maxt; + shift_cycle(&sss1, &sst1, &maxs, &maxt, tile); + clamp_cycle_light(&sss1, &sst1, maxs, maxt, tilenum, tile); + mask(&sss1, &sst1, tile); UINT32 tbase = tile.tmem + ((tile.line * sst1) & 0x1ff); @@ -367,7 +300,7 @@ void n64_texture_pipe_t::cycle_nearest(color_t* TEX, color_t* prev, INT32 SSS, I TEX->i.a &= 0x1ff; } -void n64_texture_pipe_t::cycle_nearest_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::cycle_nearest_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object) { const n64_tile_t* tiles = object.m_tiles; const n64_tile_t& tile = tiles[tilenum]; @@ -376,60 +309,47 @@ void n64_texture_pipe_t::cycle_nearest_lerp(color_t* TEX, color_t* prev, INT32 S const UINT32 tpal = tile.palette; const UINT32 index = (tformat << 4) | (tsize << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; -#define TRELATIVE(x, y) ((((x) >> 3) - (y)) << 3) | (x & 7); color_t t0; - INT32 sss1 = SSS; - INT32 maxs; - SHIFT_CYCLE(sss1, maxs, tilenum, s); - sss1 = TRELATIVE(sss1, tile.sl); - - INT32 sst1 = SST; - INT32 maxt; - SHIFT_CYCLE(sst1, maxt, tilenum, t); - sst1 = TRELATIVE(sst1, tile.tl); - - clamp_cycle_light(&sss1, &sst1, maxs, maxt, tilenum, userdata, object, m_clamp_s_diff, m_clamp_t_diff); - mask(&sss1, &sst1, tilenum, object); + INT32 sss1 = SSS, sst1 = SST; + bool maxs, maxt; + shift_cycle(&sss1, &sst1, &maxs, &maxt, tile); + clamp_cycle_light(&sss1, &sst1, maxs, maxt, tilenum, tile); + mask(&sss1, &sst1, tile); UINT32 tbase = tile.tmem + ((tile.line * sst1) & 0x1ff); (*TEX).c = ((this)->*(m_texel_fetch[index]))(sss1, sst1, tbase, tpal, userdata); } -void n64_texture_pipe_t::cycle_linear(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::cycle_linear(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object) { const n64_tile_t* tiles = object.m_tiles; const n64_tile_t& tile = tiles[tilenum]; -#define TRELATIVE(x, y) ((((x) >> 3) - (y)) << 3) | (x & 7); const UINT32 tpal = tile.palette; const UINT32 index = (tile.format << 4) | (tile.size << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; - INT32 invsf = 0; - INT32 sss1 = SSS; - INT32 maxs; - SHIFT_CYCLE(sss1, maxs, tilenum, s); - sss1 = TRELATIVE(sss1, tile.sl); - INT32 sfrac = sss1 & 0x1f; - CLAMP_CYCLE(sss1, sfrac, maxs, tilenum, s); - INT32 sss2 = sss1 + 1; - MASK_COUPLED(sss1, sss2, tilenum, s); + INT32 sss1 = SSS, sst1 = SST; + bool maxs, maxt; + shift_cycle(&sss1, &sst1, &maxs, &maxt, tile); - INT32 invtf = 0; - INT32 sst1 = SST; - INT32 maxt; - SHIFT_CYCLE(sst1, maxt, tilenum, t); - sst1 = TRELATIVE(sst1, tile.tl); + INT32 sfrac = sss1 & 0x1f; INT32 tfrac = sst1 & 0x1f; - CLAMP_CYCLE(sst1, tfrac, maxt, tilenum, t); + + clamp_cycle(&sss1, &sst1, &sfrac, &tfrac, maxs, maxt, tilenum, tile); + + INT32 sss2 = sss1 + 1; INT32 sst2 = sst1 + 1; - MASK_COUPLED(sst1, sst2, tilenum, t); - UINT32 tbase = tile.tmem + ((tile.line * sst1) & 0x1ff); + mask_coupled(&sss1, &sss2, &sst1, &sst2, tile); + + const UINT32 tbase = tile.tmem + ((tile.line * sst1) & 0x1ff); bool upper = ((sfrac + tfrac) >= 0x20); + INT32 invsf = 0; + INT32 invtf = 0; if (upper) { invsf = 0x20 - sfrac; @@ -467,40 +387,35 @@ void n64_texture_pipe_t::cycle_linear(color_t* TEX, color_t* prev, INT32 SSS, IN TEX->i.a &= 0x1ff; } -void n64_texture_pipe_t::cycle_linear_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::cycle_linear_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object) { const n64_tile_t* tiles = object.m_tiles; const n64_tile_t& tile = tiles[tilenum]; -#define TRELATIVE(x, y) ((((x) >> 3) - (y)) << 3) | (x & 7); UINT32 tpal = tile.palette; UINT32 index = (tile.format << 4) | (tile.size << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; - INT32 invsf = 0; - INT32 sss1 = SSS; - INT32 maxs; - SHIFT_CYCLE(sss1, maxs, tilenum, s); - sss1 = TRELATIVE(sss1, tile.sl); - INT32 sfrac = sss1 & 0x1f; - CLAMP_CYCLE(sss1, sfrac, maxs, tilenum, s); - INT32 sss2 = sss1 + 1; - MASK_COUPLED(sss1, sss2, tilenum, s); + INT32 sss1 = SSS, sst1 = SST; + bool maxs, maxt; + shift_cycle(&sss1, &sst1, &maxs, &maxt, tile); - INT32 invtf = 0; - INT32 sst1 = SST; - INT32 maxt; - SHIFT_CYCLE(sst1, maxt, tilenum, t); - sst1 = TRELATIVE(sst1, tile.tl); + INT32 sfrac = sss1 & 0x1f; INT32 tfrac = sst1 & 0x1f; - CLAMP_CYCLE(sst1, tfrac, maxt, tilenum, t); + + clamp_cycle(&sss1, &sst1, &sfrac, &tfrac, maxs, maxt, tilenum, tile); + + INT32 sss2 = sss1 + 1; INT32 sst2 = sst1 + 1; - MASK_COUPLED(sst1, sst2, tilenum, t); - UINT32 tbase1 = tile.tmem + ((tile.line * sst1) & 0x1ff); - UINT32 tbase2 = tile.tmem + ((tile.line * sst2) & 0x1ff); + mask_coupled(&sss1, &sss2, &sst1, &sst2, tile); + + const UINT32 tbase1 = tile.tmem + ((tile.line * sst1) & 0x1ff); + const UINT32 tbase2 = tile.tmem + ((tile.line * sst2) & 0x1ff); bool upper = ((sfrac + tfrac) >= 0x20); + INT32 invsf = 0; + INT32 invtf = 0; if (upper) { invsf = 0x20 - sfrac; @@ -561,15 +476,20 @@ void n64_texture_pipe_t::copy(color_t* TEX, INT32 SSS, INT32 SST, UINT32 tilenum { const n64_tile_t* tiles = object.m_tiles; const n64_tile_t& tile = tiles[tilenum]; + INT32 sss1 = SSS; INT32 sst1 = SST; - shift_copy(&sss1, &sst1, tilenum, object); - sss1 = TRELATIVE(sss1, tile.sl); - sst1 = TRELATIVE(sst1, tile.tl); + shift_copy(&sss1, &sst1, tile); + sss1 = (((sss1 >> 3) - tile.sl) << 3) | (sss1 & 7); sss1 = (SIGN17(sss1) >> 5) & 0x1fff; + + sst1 = (((sst1 >> 3) - tile.tl) << 3) | (sst1 & 7); sst1 = (SIGN17(sst1) >> 5) & 0x1fff; - mask(&sss1, &sst1, tilenum, object); - TEX->c = fetch(sss1, sst1, tilenum, object, userdata); + mask(&sss1, &sst1, tile); + + const UINT32 index = (tile.format << 4) | (tile.size << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; + const UINT32 tbase = tile.tmem + ((tile.line * sst1) & 0x1ff); + TEX->c = ((this)->*(m_texel_fetch[index]))(sss1, sst1, tbase, tile.palette, userdata); } void n64_texture_pipe_t::lod_1cycle(INT32* sss, INT32* sst, const INT32 s, const INT32 t, const INT32 w, const INT32 dsinc, const INT32 dtinc, const INT32 dwinc, rdp_span_aux* userdata, const rdp_poly_state& object) @@ -606,55 +526,8 @@ void n64_texture_pipe_t::lod_1cycle(INT32* sss, INT32* sst, const INT32 s, const INT32 lod = (horstep >= vertstep) ? horstep : vertstep; - if (*sss & 0x40000) - { - *sss = 0x7fff; - } - else if (*sss & 0x20000) - { - *sss = 0x8000; - } - else - { - INT32 tempanded = *sss & 0x18000; - if (tempanded == 0x8000) - { - *sss = 0x7fff; - } - else if (tempanded == 0x10000) - { - *sss = 0x8000; - } - else - { - *sss &= 0xffff; - } - } - - if (*sst & 0x40000) - { - *sst = 0x7fff; - } - else if (*sst & 0x20000) - { - *sst = 0x8000; - } - else - { - INT32 tempanded = *sst & 0x18000; - if (tempanded == 0x8000) - { - *sst = 0x7fff; - } - else if (tempanded == 0x10000) - { - *sst = 0x8000; - } - else - { - *sst &= 0xffff; - } - } + *sss = m_lod_lookup[*sss & 0x7ffff]; + *sst = m_lod_lookup[*sst & 0x7ffff]; if ((lod & 0x4000) || lodclamp) { @@ -723,55 +596,8 @@ void n64_texture_pipe_t::lod_2cycle(INT32* sss, INT32* sst, const INT32 s, const INT32 lod = (horstep >= vertstep) ? horstep : vertstep; - if (*sss & 0x40000) - { - *sss = 0x7fff; - } - else if (*sss & 0x20000) - { - *sss = 0x8000; - } - else - { - INT32 temp_anded = *sss & 0x18000; - if (temp_anded == 0x8000) - { - *sss = 0x7fff; - } - else if (temp_anded == 0x10000) - { - *sss = 0x8000; - } - else - { - *sss &= 0xffff; - } - } - - if (*sst & 0x40000) - { - *sst = 0x7fff; - } - else if (*sst & 0x20000) - { - *sst = 0x8000; - } - else - { - INT32 temp_anded = *sst & 0x18000; - if (temp_anded == 0x8000) - { - *sst = 0x7fff; - } - else if (temp_anded == 0x10000) - { - *sst = 0x8000; - } - else - { - *sst &= 0xffff; - } - } + *sss = m_lod_lookup[*sss & 0x7ffff]; + *sst = m_lod_lookup[*sst & 0x7ffff]; if ((lod & 0x4000) || lodclamp) { @@ -876,55 +702,8 @@ void n64_texture_pipe_t::lod_2cycle_limited(INT32* sss, INT32* sst, const INT32 INT32 lod = (horstep >= vertstep) ? horstep : vertstep; - if (*sss & 0x40000) - { - *sss = 0x7fff; - } - else if (*sss & 0x20000) - { - *sss = 0x8000; - } - else - { - INT32 tempanded = *sss & 0x18000; - if (tempanded == 0x8000) - { - *sss = 0x7fff; - } - else if (tempanded == 0x10000) - { - *sss = 0x8000; - } - else - { - *sss &= 0xffff; - } - } - - if (*sst & 0x40000) - { - *sst = 0x7fff; - } - else if (*sst & 0x20000) - { - *sst = 0x8000; - } - else - { - INT32 tempanded = *sst & 0x18000; - if (tempanded == 0x8000) - { - *sst = 0x7fff; - } - else if (tempanded == 0x10000) - { - *sst = 0x8000; - } - else - { - *sst &= 0xffff; - } - } + *sss = m_lod_lookup[*sss & 0x7ffff]; + *sst = m_lod_lookup[*sst & 0x7ffff]; if ((lod & 0x4000) || lodclamp) { @@ -964,7 +743,7 @@ void n64_texture_pipe_t::lod_2cycle_limited(INT32* sss, INT32* sst, const INT32 } } -void n64_texture_pipe_t::calculate_clamp_diffs(UINT32 prim_tile, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff) +void n64_texture_pipe_t::calculate_clamp_diffs(UINT32 prim_tile, const rdp_poly_state& object) { const n64_tile_t* tiles = object.m_tiles; if (object.m_other_modes.cycle_type == CYCLE_TYPE_2) @@ -1484,17 +1263,3 @@ UINT32 n64_texture_pipe_t::fetch_i8_raw(INT32 s, INT32 t, INT32 tbase, INT32 tpa return color.c; } - -UINT32 n64_texture_pipe_t::fetch(INT32 s, INT32 t, INT32 tilenum, const rdp_poly_state& object, rdp_span_aux* userdata) -{ - const n64_tile_t* tiles = object.m_tiles; - const n64_tile_t& tile = tiles[tilenum]; - const UINT32 tformat = tile.format; - const UINT32 tsize = tile.size; - const UINT32 tpal = tile.palette; - const UINT32 index = (tformat << 4) | (tsize << 2) | ((UINT32) object.m_other_modes.en_tlut << 1) | (UINT32) object.m_other_modes.tlut_type; - - const UINT32 tbase = tile.tmem + ((tile.line * t) & 0x1ff); - - return ((this)->*(m_texel_fetch[index]))(s, t, tbase, tpal, userdata); -} diff --git a/src/mame/video/rdptpipe.h b/src/mame/video/rdptpipe.h index a01f44e8901..defa2b0feb7 100644 --- a/src/mame/video/rdptpipe.h +++ b/src/mame/video/rdptpipe.h @@ -23,12 +23,13 @@ struct misc_state_t; class color_t; struct rdp_span_aux; struct rdp_poly_state; +struct n64_tile_t; class n64_texture_pipe_t { public: typedef UINT32 (n64_texture_pipe_t::*texel_fetcher_t) (INT32 s, INT32 t, INT32 tbase, INT32 tpal, rdp_span_aux* userdata); - typedef void (n64_texture_pipe_t::*texel_cycler_t) (color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); + typedef void (n64_texture_pipe_t::*texel_cycler_t) (color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object); n64_texture_pipe_t() { @@ -95,16 +96,15 @@ class n64_texture_pipe_t m_cycle[3] = &n64_texture_pipe_t::cycle_linear_lerp; } - void cycle_nearest(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); - void cycle_nearest_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); - void cycle_linear(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); - void cycle_linear_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); + void cycle_nearest(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object); + void cycle_nearest_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object); + void cycle_linear(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object); + void cycle_linear_lerp(color_t* TEX, color_t* prev, INT32 SSS, INT32 SST, UINT32 tilenum, UINT32 cycle, rdp_span_aux* userdata, const rdp_poly_state& object); texel_cycler_t m_cycle[4]; void copy(color_t* TEX, INT32 SSS, INT32 SST, UINT32 tilenum, const rdp_poly_state& object, rdp_span_aux* userdata); - UINT32 fetch(INT32 SSS, INT32 SST, INT32 tile, const rdp_poly_state& object, rdp_span_aux* userdata); - void calculate_clamp_diffs(UINT32 prim_tile, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); + void calculate_clamp_diffs(UINT32 prim_tile, const rdp_poly_state& object); void lod_1cycle(INT32* sss, INT32* sst, const INT32 s, const INT32 t, const INT32 w, const INT32 dsinc, const INT32 dtinc, const INT32 dwinc, rdp_span_aux* userdata, const rdp_poly_state& object); void lod_2cycle(INT32* sss, INT32* sst, const INT32 s, const INT32 t, const INT32 w, const INT32 dsinc, const INT32 dtinc, const INT32 dwinc, const INT32 prim_tile, INT32* t1, INT32* t2, rdp_span_aux* userdata, const rdp_poly_state& object); void lod_2cycle_limited(INT32* sss, INT32* sst, const INT32 s, const INT32 t, INT32 w, const INT32 dsinc, const INT32 dtinc, const INT32 dwinc, const INT32 prim_tile, INT32* t1, const rdp_poly_state& object); @@ -114,14 +114,14 @@ class n64_texture_pipe_t bool m_start_span; private: - void mask(INT32* S, INT32* T, INT32 num, const rdp_poly_state& object); - void mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, INT32 num, const rdp_poly_state& object); + void mask(INT32* S, INT32* T, const n64_tile_t& tile); + void mask_coupled(INT32* S, INT32* S1, INT32* T, INT32* T1, const n64_tile_t& tile); - void shift_cycle(INT32* S, INT32* T, INT32* maxs, INT32* maxt, UINT32 num, const rdp_poly_state& object); - void shift_copy(INT32* S, INT32* T, UINT32 num, const rdp_poly_state& object); + void shift_cycle(INT32* S, INT32* T, bool* maxs, bool* maxt, const n64_tile_t& tile); + void shift_copy(INT32* S, INT32* T, const n64_tile_t& tile); - void clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TFRAC, INT32 maxs, INT32 maxt, INT32 num, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); - void clamp_cycle_light(INT32* S, INT32* T, bool maxs, bool maxt, INT32 num, rdp_span_aux* userdata, const rdp_poly_state& object, INT32* m_clamp_s_diff, INT32* m_clamp_t_diff); + void clamp_cycle(INT32* S, INT32* T, INT32* SFRAC, INT32* TFRAC, const bool maxs, const bool maxt, const INT32 tilenum, const n64_tile_t& tile); + void clamp_cycle_light(INT32* S, INT32* T, const bool maxs, const bool maxt, const INT32 tilenum, const n64_tile_t& tile); UINT32 fetch_nop(INT32 s, INT32 t, INT32 tbase, INT32 tpal, rdp_span_aux* userdata); @@ -164,6 +164,9 @@ class n64_texture_pipe_t INT32 m_maskbits_table[16]; UINT32 m_expand_16to32_table[0x10000]; + UINT16 m_lod_lookup[0x80000]; + INT32 m_clamp_s_diff[8]; + INT32 m_clamp_t_diff[8]; }; #endif // _VIDEO_RDPTEXPIPE_H_ |
