X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fauxiliary%2Fdraw%2Fdraw_vs_varient.c;h=6eb26927f27366130d8185f05b649758bbb3e103;hb=8d93f360c582297b9ced11c234ab4bd53103a8a6;hp=4a155251b53c02c075a20569f54c75e1c32c3326;hpb=3869c3c87a40d9f41414082ad8dd3d747e6b34d8;p=mesa.git diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c index 4a155251b53..6eb26927f27 100644 --- a/src/gallium/auxiliary/draw/draw_vs_varient.c +++ b/src/gallium/auxiliary/draw/draw_vs_varient.c @@ -30,14 +30,14 @@ * Keith Whitwell */ -#include "pipe/p_util.h" +#include "util/u_memory.h" +#include "util/u_math.h" #include "draw/draw_context.h" #include "draw/draw_private.h" #include "draw/draw_vbuf.h" #include "draw/draw_vertex.h" #include "draw/draw_vs.h" #include "translate/translate.h" -#include "translate/translate_cache.h" /* A first pass at incorporating vertex fetch/emit functionality into */ @@ -55,23 +55,27 @@ struct draw_vs_varient_generic { */ struct translate *fetch; struct translate *emit; + + unsigned temp_vertex_stride; }; -static void vsvg_set_input( struct draw_vs_varient *varient, - unsigned buffer, - const void *ptr, - unsigned stride ) +static void vsvg_set_buffer( struct draw_vs_varient *varient, + unsigned buffer, + const void *ptr, + unsigned stride, + unsigned max_index ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; vsvg->fetch->set_buffer(vsvg->fetch, buffer, ptr, - stride); + stride, + max_index ); } @@ -84,9 +88,11 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, char *ptr = (char *)output_buffer; const float *scale = vsvg->base.vs->draw->viewport.scale; const float *trans = vsvg->base.vs->draw->viewport.translate; - unsigned stride = vsvg->base.key.output_stride; + unsigned stride = vsvg->temp_vertex_stride; unsigned j; + ptr += vsvg->base.vs->position_output * 4 * sizeof(float); + for (j = 0; j < count; j++, ptr += stride) { float *data = (float *)ptr; float w = 1.0f / data[3]; @@ -99,15 +105,17 @@ static void do_rhw_viewport( struct draw_vs_varient_generic *vsvg, } static void do_viewport( struct draw_vs_varient_generic *vsvg, - unsigned count, - void *output_buffer ) + unsigned count, + void *output_buffer ) { char *ptr = (char *)output_buffer; const float *scale = vsvg->base.vs->draw->viewport.scale; const float *trans = vsvg->base.vs->draw->viewport.translate; - unsigned stride = vsvg->base.key.output_stride; + unsigned stride = vsvg->temp_vertex_stride; unsigned j; + ptr += vsvg->base.vs->position_output * 4 * sizeof(float); + for (j = 0; j < count; j++, ptr += stride) { float *data = (float *)ptr; @@ -124,6 +132,10 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, void *output_buffer) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + unsigned temp_vertex_stride = vsvg->temp_vertex_stride; + void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride ); + + if (0) debug_printf("%s %d \n", __FUNCTION__, count); /* Want to do this in small batches for cache locality? */ @@ -131,46 +143,52 @@ static void PIPE_CDECL vsvg_run_elts( struct draw_vs_varient *varient, vsvg->fetch->run_elts( vsvg->fetch, elts, count, - output_buffer ); - - //if (!vsvg->base.vs->is_passthrough) - { - vsvg->base.vs->run_linear( vsvg->base.vs, - output_buffer, - output_buffer, - (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, - count, - vsvg->base.key.output_stride, - vsvg->base.key.output_stride); - - - if (vsvg->base.key.clip) { - /* not really handling clipping, just do the rhw so we can - * see the results... - */ - do_rhw_viewport( vsvg, - count, - output_buffer ); - } - else if (vsvg->base.key.viewport) { - do_viewport( vsvg, - count, - output_buffer ); - } + vsvg->draw->instance_id, + temp_buffer ); + + vsvg->base.vs->run_linear( vsvg->base.vs, + temp_buffer, + temp_buffer, + vsvg->base.vs->draw->pt.user.vs_constants, + count, + temp_vertex_stride, + temp_vertex_stride); + + /* FIXME: geometry shading? */ + + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + temp_buffer ); + } + else if (vsvg->base.key.viewport) { + do_viewport( vsvg, + count, + temp_buffer ); + } - //if (!vsvg->already_in_emit_format) + vsvg->emit->set_buffer( vsvg->emit, + 0, + temp_buffer, + temp_vertex_stride, + ~0 ); - vsvg->emit->set_buffer( vsvg->emit, - 0, - output_buffer, - vsvg->base.key.output_stride ); + vsvg->emit->set_buffer( vsvg->emit, + 1, + &vsvg->draw->rasterizer->point_size, + 0, + ~0 ); + vsvg->emit->run( vsvg->emit, + 0, count, + vsvg->draw->instance_id, + output_buffer ); - vsvg->emit->run( vsvg->emit, - 0, count, - output_buffer ); - } + FREE(temp_buffer); } @@ -180,54 +198,59 @@ static void PIPE_CDECL vsvg_run_linear( struct draw_vs_varient *varient, void *output_buffer ) { struct draw_vs_varient_generic *vsvg = (struct draw_vs_varient_generic *)varient; + unsigned temp_vertex_stride = vsvg->temp_vertex_stride; + void *temp_buffer = MALLOC( align(count,4) * temp_vertex_stride ); - //debug_printf("%s %d %d\n", __FUNCTION__, start, count); - - + if (0) debug_printf("%s %d %d (sz %d, %d)\n", __FUNCTION__, start, count, + vsvg->base.key.output_stride, + temp_vertex_stride); + vsvg->fetch->run( vsvg->fetch, start, count, - output_buffer ); - - //if (!vsvg->base.vs->is_passthrough) - { - vsvg->base.vs->run_linear( vsvg->base.vs, - output_buffer, - output_buffer, - (const float (*)[4])vsvg->base.vs->draw->pt.user.constants, - count, - vsvg->base.key.output_stride, - vsvg->base.key.output_stride); - - if (vsvg->base.key.clip) { - /* not really handling clipping, just do the rhw so we can - * see the results... - */ - do_rhw_viewport( vsvg, - count, - output_buffer ); - } - else if (vsvg->base.key.viewport) { - do_viewport( vsvg, - count, - output_buffer ); - } - - //if (!vsvg->already_in_emit_format) - vsvg->emit->set_buffer( vsvg->emit, - 0, - output_buffer, - vsvg->base.key.output_stride ); + vsvg->draw->instance_id, + temp_buffer ); + + vsvg->base.vs->run_linear( vsvg->base.vs, + temp_buffer, + temp_buffer, + vsvg->base.vs->draw->pt.user.vs_constants, + count, + temp_vertex_stride, + temp_vertex_stride); + + if (vsvg->base.key.clip) { + /* not really handling clipping, just do the rhw so we can + * see the results... + */ + do_rhw_viewport( vsvg, + count, + temp_buffer ); + } + else if (vsvg->base.key.viewport) { + do_viewport( vsvg, + count, + temp_buffer ); + } - vsvg->emit->set_buffer( vsvg->emit, - 1, - &vsvg->draw->rasterizer->point_size, - 0); + vsvg->emit->set_buffer( vsvg->emit, + 0, + temp_buffer, + temp_vertex_stride, + ~0 ); + + vsvg->emit->set_buffer( vsvg->emit, + 1, + &vsvg->draw->rasterizer->point_size, + 0, + ~0 ); + + vsvg->emit->run( vsvg->emit, + 0, count, + vsvg->draw->instance_id, + output_buffer ); - vsvg->emit->run( vsvg->emit, - 0, count, - output_buffer ); - } + FREE(temp_buffer); } @@ -240,8 +263,9 @@ static void vsvg_destroy( struct draw_vs_varient *varient ) } -struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, - const struct draw_vs_varient_key *key ) +struct draw_vs_varient * +draw_vs_create_varient_generic( struct draw_vertex_shader *vs, + const struct draw_vs_varient_key *key ) { unsigned i; struct translate_key fetch, emit; @@ -252,24 +276,29 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, vsvg->base.key = *key; vsvg->base.vs = vs; - vsvg->base.set_input = vsvg_set_input; + vsvg->base.set_buffer = vsvg_set_buffer; vsvg->base.run_elts = vsvg_run_elts; vsvg->base.run_linear = vsvg_run_linear; vsvg->base.destroy = vsvg_destroy; + vsvg->draw = vs->draw; + vsvg->temp_vertex_stride = MAX2(key->nr_inputs, + vsvg->base.vs->info.num_outputs) * 4 * sizeof(float); /* Build free-standing fetch and emit functions: */ fetch.nr_elements = key->nr_inputs; - fetch.output_stride = 0; + fetch.output_stride = vsvg->temp_vertex_stride; for (i = 0; i < key->nr_inputs; i++) { + fetch.element[i].type = TRANSLATE_ELEMENT_NORMAL; fetch.element[i].input_format = key->element[i].in.format; fetch.element[i].input_buffer = key->element[i].in.buffer; fetch.element[i].input_offset = key->element[i].in.offset; + fetch.element[i].instance_divisor = 0; fetch.element[i].output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - fetch.element[i].output_offset = fetch.output_stride; - fetch.output_stride += 4 * sizeof(float); + fetch.element[i].output_offset = i * 4 * sizeof(float); + assert(fetch.element[i].output_offset < fetch.output_stride); } @@ -278,16 +307,21 @@ struct draw_vs_varient *draw_vs_varient_generic( struct draw_vertex_shader *vs, for (i = 0; i < key->nr_outputs; i++) { if (key->element[i].out.format != EMIT_1F_PSIZE) { + emit.element[i].type = TRANSLATE_ELEMENT_NORMAL; emit.element[i].input_format = PIPE_FORMAT_R32G32B32A32_FLOAT; emit.element[i].input_buffer = 0; emit.element[i].input_offset = key->element[i].out.vs_output * 4 * sizeof(float); + emit.element[i].instance_divisor = 0; emit.element[i].output_format = draw_translate_vinfo_format(key->element[i].out.format); emit.element[i].output_offset = key->element[i].out.offset; + assert(emit.element[i].input_offset <= fetch.output_stride); } else { + emit.element[i].type = TRANSLATE_ELEMENT_NORMAL; emit.element[i].input_format = PIPE_FORMAT_R32_FLOAT; emit.element[i].input_buffer = 1; emit.element[i].input_offset = 0; + emit.element[i].instance_divisor = 0; emit.element[i].output_format = PIPE_FORMAT_R32_FLOAT; emit.element[i].output_offset = key->element[i].out.offset; }