src/gallium/drivers/nv10/nv10_state_emit.c

   1 #include "nv10_context.h"
   2 #include "nv10_state.h"
   3
   4 static void nv10_state_emit_blend(struct nv10_context* nv10)
   5 {
   6         struct nv10_blend_state *b = nv10->blend;
   7
   8         BEGIN_RING(celsius, NV10TCL_DITHER_ENABLE, 1);
   9         OUT_RING  (b->d_enable);
  10
  11         BEGIN_RING(celsius, NV10TCL_BLEND_FUNC_ENABLE, 3);
  12         OUT_RING  (b->b_enable);
  13         OUT_RING  (b->b_srcfunc);
  14         OUT_RING  (b->b_dstfunc);
  15
  16         BEGIN_RING(celsius, NV10TCL_COLOR_MASK, 1);
  17         OUT_RING  (b->c_mask);
  18 }
  19
  20 static void nv10_state_emit_blend_color(struct nv10_context* nv10)
  21 {
  22         struct pipe_blend_color *c = nv10->blend_color;
  23
  24         BEGIN_RING(celsius, NV10TCL_BLEND_COLOR, 1);
  25         OUT_RING  ((float_to_ubyte(c->color[3]) << 24)|
  26                    (float_to_ubyte(c->color[0]) << 16)|
  27                    (float_to_ubyte(c->color[1]) << 8) |
  28                    (float_to_ubyte(c->color[2]) << 0));
  29 }
  30
  31 static void nv10_state_emit_rast(struct nv10_context* nv10)
  32 {
  33         struct nv10_rasterizer_state *r = nv10->rast;
  34
  35         BEGIN_RING(celsius, NV10TCL_SHADE_MODEL, 2);
  36         OUT_RING  (r->shade_model);
  37         OUT_RING  (r->line_width);
  38
  39
  40         BEGIN_RING(celsius, NV10TCL_POINT_SIZE, 1);
  41         OUT_RING  (r->point_size);
  42
  43         BEGIN_RING(celsius, NV10TCL_POLYGON_MODE_FRONT, 2);
  44         OUT_RING  (r->poly_mode_front);
  45         OUT_RING  (r->poly_mode_back);
  46
  47
  48         BEGIN_RING(celsius, NV10TCL_CULL_FACE, 2);
  49         OUT_RING  (r->cull_face);
  50         OUT_RING  (r->front_face);
  51
  52         BEGIN_RING(celsius, NV10TCL_LINE_SMOOTH_ENABLE, 2);
  53         OUT_RING  (r->line_smooth_en);
  54         OUT_RING  (r->poly_smooth_en);
  55
  56         BEGIN_RING(celsius, NV10TCL_CULL_FACE_ENABLE, 1);
  57         OUT_RING  (r->cull_face_en);
  58 }
  59
  60 static void nv10_state_emit_dsa(struct nv10_context* nv10)
  61 {
  62         struct nv10_depth_stencil_alpha_state *d = nv10->dsa;
  63
  64         BEGIN_RING(celsius, NV10TCL_DEPTH_FUNC, 1);
  65         OUT_RING (d->depth.func);
  66
  67         BEGIN_RING(celsius, NV10TCL_DEPTH_WRITE_ENABLE, 1);
  68         OUT_RING (d->depth.write_enable);
  69
  70         BEGIN_RING(celsius, NV10TCL_DEPTH_TEST_ENABLE, 1);
  71         OUT_RING (d->depth.test_enable);
  72
  73 #if 0
  74         BEGIN_RING(celsius, NV10TCL_STENCIL_ENABLE, 1);
  75         OUT_RING (d->stencil.enable);
  76         BEGIN_RING(celsius, NV10TCL_STENCIL_MASK, 7);
  77         OUT_RINGp ((uint32_t *)&(d->stencil.wmask), 7);
  78 #endif
  79
  80         BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_ENABLE, 1);
  81         OUT_RING (d->alpha.enabled);
  82
  83         BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_FUNC, 1);
  84         OUT_RING (d->alpha.func);
  85
  86         BEGIN_RING(celsius, NV10TCL_ALPHA_FUNC_REF, 1);
  87         OUT_RING (d->alpha.ref);
  88 }
  89
  90 static void nv10_state_emit_viewport(struct nv10_context* nv10)
  91 {
  92 }
  93
  94 static void nv10_state_emit_scissor(struct nv10_context* nv10)
  95 {
  96         // XXX this is so not working
  97 /*      struct pipe_scissor_state *s = nv10->scissor;
  98         BEGIN_RING(celsius, NV10TCL_SCISSOR_HORIZ, 2);
  99         OUT_RING  (((s->maxx - s->minx) << 16) | s->minx);
 100         OUT_RING  (((s->maxy - s->miny) << 16) | s->miny);*/
 101 }
 102
 103 static void nv10_state_emit_framebuffer(struct nv10_context* nv10)
 104 {
 105         struct pipe_framebuffer_state* fb = nv10->framebuffer;
 106         struct pipe_surface *rt, *zeta = NULL;
 107         uint32_t rt_format, w, h;
 108         int colour_format = 0, zeta_format = 0;
 109
 110         w = fb->cbufs[0]->width;
 111         h = fb->cbufs[0]->height;
 112         colour_format = fb->cbufs[0]->format;
 113         rt = fb->cbufs[0];
 114
 115         if (fb->zsbuf) {
 116                 if (colour_format) {
 117                         assert(w == fb->zsbuf->width);
 118                         assert(h == fb->zsbuf->height);
 119                 } else {
 120                         w = fb->zsbuf->width;
 121                         h = fb->zsbuf->height;
 122                 }
 123
 124                 zeta_format = fb->zsbuf->format;
 125                 zeta = fb->zsbuf;
 126         }
 127
 128         rt_format = NV10TCL_RT_FORMAT_TYPE_LINEAR;
 129
 130         switch (colour_format) {
 131         case PIPE_FORMAT_A8R8G8B8_UNORM:
 132         case 0:
 133                 rt_format |= NV10TCL_RT_FORMAT_COLOR_A8R8G8B8;
 134                 break;
 135         case PIPE_FORMAT_R5G6B5_UNORM:
 136                 rt_format |= NV10TCL_RT_FORMAT_COLOR_R5G6B5;
 137                 break;
 138         default:
 139                 assert(0);
 140         }
 141
 142         if (zeta) {
 143                 BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1);
 144                 OUT_RING  (rt->stride | (zeta->stride << 16));
 145         } else {
 146                 BEGIN_RING(celsius, NV10TCL_RT_PITCH, 1);
 147                 OUT_RING  (rt->stride | (rt->stride << 16));
 148         }
 149
 150         nv10->rt[0] = rt->buffer;
 151
 152         if (zeta_format)
 153         {
 154                 nv10->zeta = zeta->buffer;
 155         }
 156
 157         BEGIN_RING(celsius, NV10TCL_RT_HORIZ, 3);
 158         OUT_RING  ((w << 16) | 0);
 159         OUT_RING  ((h << 16) | 0);
 160         OUT_RING  (rt_format);
 161         BEGIN_RING(celsius, NV10TCL_VIEWPORT_CLIP_HORIZ(0), 2);
 162         OUT_RING  (((w - 1) << 16) | 0 | 0x08000800);
 163         OUT_RING  (((h - 1) << 16) | 0 | 0x08000800);
 164 }
 165
 166 static void nv10_vertex_layout(struct nv10_context *nv10)
 167 {
 168         struct nv10_fragment_program *fp = nv10->fragprog.current;
 169         uint32_t src = 0;
 170         int i;
 171         struct vertex_info vinfo;
 172
 173         memset(&vinfo, 0, sizeof(vinfo));
 174
 175         for (i = 0; i < fp->info.num_inputs; i++) {
 176                 switch (fp->info.input_semantic_name[i]) {
 177                         case TGSI_SEMANTIC_POSITION:
 178                                 draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
 179                                 break;
 180                         case TGSI_SEMANTIC_COLOR:
 181                                 draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_LINEAR, src++);
 182                                 break;
 183                         default:
 184                         case TGSI_SEMANTIC_GENERIC:
 185                                 draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
 186                                 break;
 187                         case TGSI_SEMANTIC_FOG:
 188                                 draw_emit_vertex_attr(&vinfo, EMIT_4F, INTERP_PERSPECTIVE, src++);
 189                                 break;
 190                 }
 191         }
 192         draw_compute_vertex_size(&vinfo);
 193 }
 194
 195 void
 196 nv10_emit_hw_state(struct nv10_context *nv10)
 197 {
 198         int i;
 199
 200         if (nv10->dirty & NV10_NEW_VERTPROG) {
 201                 //nv10_vertprog_bind(nv10, nv10->vertprog.current);
 202                 nv10->dirty &= ~NV10_NEW_VERTPROG;
 203         }
 204
 205         if (nv10->dirty & NV10_NEW_FRAGPROG) {
 206                 nv10_fragprog_bind(nv10, nv10->fragprog.current);
 207                 /*XXX: clear NV10_NEW_FRAGPROG if no new program uploaded */
 208                 nv10->dirty_samplers |= (1<<10);
 209                 nv10->dirty_samplers = 0;
 210         }
 211
 212         if (nv10->dirty_samplers || (nv10->dirty & NV10_NEW_FRAGPROG)) {
 213                 nv10_fragtex_bind(nv10);
 214                 nv10->dirty &= ~NV10_NEW_FRAGPROG;
 215         }
 216
 217         if (nv10->dirty & NV10_NEW_VTXARRAYS) {
 218                 nv10->dirty &= ~NV10_NEW_VTXARRAYS;
 219                 nv10_vertex_layout(nv10);
 220                 nv10_vtxbuf_bind(nv10);
 221         }
 222
 223         if (nv10->dirty & NV10_NEW_BLEND) {
 224                 nv10->dirty &= ~NV10_NEW_BLEND;
 225                 nv10_state_emit_blend(nv10);
 226         }
 227
 228         if (nv10->dirty & NV10_NEW_BLENDCOL) {
 229                 nv10->dirty &= ~NV10_NEW_BLENDCOL;
 230                 nv10_state_emit_blend_color(nv10);
 231         }
 232
 233         if (nv10->dirty & NV10_NEW_RAST) {
 234                 nv10->dirty &= ~NV10_NEW_RAST;
 235                 nv10_state_emit_rast(nv10);
 236         }
 237
 238         if (nv10->dirty & NV10_NEW_DSA) {
 239                 nv10->dirty &= ~NV10_NEW_DSA;
 240                 nv10_state_emit_dsa(nv10);
 241         }
 242
 243         if (nv10->dirty & NV10_NEW_VIEWPORT) {
 244                 nv10->dirty &= ~NV10_NEW_VIEWPORT;
 245                 nv10_state_emit_viewport(nv10);
 246         }
 247
 248         if (nv10->dirty & NV10_NEW_SCISSOR) {
 249                 nv10->dirty &= ~NV10_NEW_SCISSOR;
 250                 nv10_state_emit_scissor(nv10);
 251         }
 252
 253         if (nv10->dirty & NV10_NEW_FRAMEBUFFER) {
 254                 nv10->dirty &= ~NV10_NEW_FRAMEBUFFER;
 255                 nv10_state_emit_framebuffer(nv10);
 256         }
 257
 258         /* Emit relocs for every referenced buffer.
 259          * This is to ensure the bufmgr has an accurate idea of how
 260          * the buffer is used.  This isn't very efficient, but we don't
 261          * seem to take a significant performance hit.  Will be improved
 262          * at some point.  Vertex arrays are emitted by nv10_vbo.c
 263          */
 264
 265         /* Render target */
 266 // XXX figre out who's who for NV10TCL_DMA_* and fill accordingly
 267 //      BEGIN_RING(celsius, NV10TCL_DMA_COLOR0, 1);
 268 //      OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 269         BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1);
 270         OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 271
 272         if (nv10->zeta) {
 273 // XXX
 274 //              BEGIN_RING(celsius, NV10TCL_DMA_ZETA, 1);
 275 //              OUT_RELOCo(nv10->zeta, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 276                 BEGIN_RING(celsius, NV10TCL_ZETA_OFFSET, 1);
 277                 OUT_RELOCl(nv10->zeta, 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 278                 /* XXX for when we allocate LMA on nv17 */
 279 /*              BEGIN_RING(celsius, NV10TCL_LMA_DEPTH_BUFFER_OFFSET, 1);
 280                 OUT_RELOCl(nv10->zeta + lma_offset);*/
 281         }
 282
 283         /* Vertex buffer */
 284         BEGIN_RING(celsius, NV10TCL_DMA_VTXBUF0, 1);
 285         OUT_RELOCo(nv10->rt[0], NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 286         BEGIN_RING(celsius, NV10TCL_COLOR_OFFSET, 1);
 287         OUT_RELOCl(nv10->rt[0], 0, NOUVEAU_BO_VRAM | NOUVEAU_BO_WR);
 288
 289         /* Texture images */
 290         for (i = 0; i < 2; i++) {
 291                 if (!(nv10->fp_samplers & (1 << i)))
 292                         continue;
 293                 BEGIN_RING(celsius, NV10TCL_TX_OFFSET(i), 1);
 294                 OUT_RELOCl(nv10->tex[i].buffer, 0, NOUVEAU_BO_VRAM |
 295                            NOUVEAU_BO_GART | NOUVEAU_BO_RD);
 296                 BEGIN_RING(celsius, NV10TCL_TX_FORMAT(i), 1);
 297                 OUT_RELOCd(nv10->tex[i].buffer, nv10->tex[i].format,
 298                            NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_RD |
 299                            NOUVEAU_BO_OR, NV10TCL_TX_FORMAT_DMA0,
 300                            NV10TCL_TX_FORMAT_DMA1);
 301         }
 302 }
 303