diff options
Diffstat (limited to 'src/gallium/drivers/cell/spu/spu_tri.c')
| -rw-r--r-- | src/gallium/drivers/cell/spu/spu_tri.c | 149 |
1 files changed, 90 insertions, 59 deletions
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c index 8b93878192..87991c3136 100644 --- a/src/gallium/drivers/cell/spu/spu_tri.c +++ b/src/gallium/drivers/cell/spu/spu_tri.c @@ -118,6 +118,8 @@ struct setup_stage { float oneoverarea; + uint facing; + uint tx, ty; int cliprect_minx, cliprect_maxx, cliprect_miny, cliprect_maxy; @@ -241,6 +243,19 @@ eval_coeff(uint slot, float x, float y, vector float result[4]) } +/** + * As above, but return 4 vectors in SOA format. + * XXX this will all be re-written someday. + */ +static INLINE void +eval_coeff_soa(uint slot, float x, float y, vector float result[4]) +{ + eval_coeff(slot, x, y, result); + _transpose_matrix4x4(result, result); +} + + + static INLINE vector float eval_z(float x, float y) { @@ -261,20 +276,23 @@ eval_z(float x, float y) * overall. */ static INLINE void -emit_quad( int x, int y, mask_t mask ) +emit_quad( int x, int y, mask_t mask) { /* If any bits in mask are set... */ if (spu_extract(spu_orx(mask), 0)) { const int ix = x - setup.cliprect_minx; const int iy = y - setup.cliprect_miny; - vector float colors[4]; spu.cur_ctile_status = TILE_STATUS_DIRTY; spu.cur_ztile_status = TILE_STATUS_DIRTY; - if (spu.texture[0].start) { - /* texture mapping */ + if (0/*spu.texture[0].start*/) { + /* + * Temporary texture mapping path + * This will go away when fragment programs support TEX inst. + */ const uint unit = 0; + vector float colors[4]; vector float texcoords[4]; eval_coeff(2, (float) x, (float) y, texcoords); @@ -311,70 +329,62 @@ emit_quad( int x, int y, mask_t mask ) colors[3] = spu_mul(colors[3], colors1[3]); } + { + /* Convert fragment data from AoS to SoA format. + * I.e. (RGBA,RGBA,RGBA,RGBA) -> (RRRR,GGGG,BBBB,AAAA) + * This is temporary! + */ + vector float soa_frag[4]; + _transpose_matrix4x4(soa_frag, colors); + + vector float fragZ = eval_z((float) x, (float) y); + + /* Do all per-fragment/quad operations here, including: + * alpha test, z test, stencil test, blend and framebuffer writing. + */ + spu.fragment_ops(ix, iy, &spu.ctile, &spu.ztile, + fragZ, + soa_frag[0], soa_frag[1], + soa_frag[2], soa_frag[3], + mask, + setup.facing); + } + } else { - /* simple shading */ -#if 0 - eval_coeff(1, (float) x, (float) y, colors); + /* + * Run fragment shader, execute per-fragment ops, update fb/tile. + */ + vector float inputs[4*4], outputs[2*4]; + vector float fragZ = eval_z((float) x, (float) y); + /* setup inputs */ +#if 0 + eval_coeff_soa(1, (float) x, (float) y, inputs); #else - /* XXX new fragment program code */ - - if (spu.fragment_program) { - vector float inputs[4*4], outputs[2*4]; - - /* setup inputs */ - eval_coeff(1, (float) x, (float) y, inputs); - - /* Execute the current fragment program */ - spu.fragment_program(inputs, outputs, spu.constants); - - /* Copy outputs */ - colors[0] = outputs[0*4+0]; - colors[1] = outputs[0*4+1]; - colors[2] = outputs[0*4+2]; - colors[3] = outputs[0*4+3]; - - if (0 && spu.init.id==0 && y == 48) { - printf("colors[0] = %f %f %f %f\n", - spu_extract(colors[0], 0), - spu_extract(colors[0], 1), - spu_extract(colors[0], 2), - spu_extract(colors[0], 3)); - printf("colors[1] = %f %f %f %f\n", - spu_extract(colors[1], 0), - spu_extract(colors[1], 1), - spu_extract(colors[1], 2), - spu_extract(colors[1], 3)); - } - + uint i; + for (i = 0; i < spu.vertex_info.num_attribs; i++) { + eval_coeff_soa(i+1, (float) x, (float) y, inputs + i * 4); } #endif - } - - - { - /* Convert fragment data from AoS to SoA format. - * I.e. (RGBA,RGBA,RGBA,RGBA) -> (RRRR,GGGG,BBBB,AAAA) - * This is temporary! - */ - vector float soa_frag[4]; - _transpose_matrix4x4(soa_frag, colors); + ASSERT(spu.fragment_program); + ASSERT(spu.fragment_ops); - float4 fragZ; + /* Execute the current fragment program */ + spu.fragment_program(inputs, outputs, spu.constants); - fragZ.v = eval_z((float) x, (float) y); - - /* Do all per-fragment/quad operations here, including: - * alpha test, z test, stencil test, blend and framebuffer writing. + /* Execute per-fragment/quad operations, including: + * alpha test, z test, stencil test, blend and framebuffer writing. */ spu.fragment_ops(ix, iy, &spu.ctile, &spu.ztile, - fragZ.v, - soa_frag[0], soa_frag[1], - soa_frag[2], soa_frag[3], - mask); + fragZ, + outputs[0*4+0], + outputs[0*4+1], + outputs[0*4+2], + outputs[0*4+3], + mask, + setup.facing); } - } } @@ -477,7 +487,7 @@ static void flush_spans( void ) */ for (x = block(minleft); x <= block(maxright); x += 2) { #if 1 - emit_quad( x, setup.span.y, calculate_mask( x ) ); + emit_quad( x, setup.span.y, calculate_mask( x )); #endif } @@ -896,13 +906,28 @@ static void subtriangle( struct edge *eleft, eright->sy += lines; } +static float +determinant( const float *v0, + const float *v1, + const float *v2 ) +{ + /* edge vectors e = v0 - v2, f = v1 - v2 */ + const float ex = v0[0] - v2[0]; + const float ey = v0[1] - v2[1]; + const float fx = v1[0] - v2[0]; + const float fy = v1[1] - v2[1]; + + /* det = cross(e,f).z */ + return ex * fy - ey * fx; +} + /** * Draw triangle into tile at (tx, ty) (tile coords) * The tile data should have already been fetched. */ boolean -tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) +tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty, uint front_winding) { setup.tx = tx; setup.ty = ty; @@ -913,6 +938,12 @@ tri_draw(const float *v0, const float *v1, const float *v2, uint tx, uint ty) setup.cliprect_maxx = (tx + 1) * TILE_SIZE; setup.cliprect_maxy = (ty + 1) * TILE_SIZE; + /* Before we sort vertices, determine the facing of the triangle, + * which will be needed for front/back-face stencil application + */ + float det = determinant(v0, v1, v2); + setup.facing = (det > 0.0) ^ (front_winding == PIPE_WINDING_CW); + if (!setup_sort_vertices((struct vertex_header *) v0, (struct vertex_header *) v1, (struct vertex_header *) v2)) { |