X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fswr%2Fswr_state.cpp;h=b0cbc21964ed33f9307c18c9ddfa421633df6471;hb=2b7bdff83fd6765ddfc3ee0f32f75e51ae2c5206;hp=ded51a9b196ed72ca1b08364a3dfbd835ee926d1;hpb=12f88ba32a14ea79134f4e995a55149f078a2f27;p=mesa.git diff --git a/src/gallium/drivers/swr/swr_state.cpp b/src/gallium/drivers/swr/swr_state.cpp index ded51a9b196..b0cbc21964e 100644 --- a/src/gallium/drivers/swr/swr_state.cpp +++ b/src/gallium/drivers/swr/swr_state.cpp @@ -21,9 +21,14 @@ * IN THE SOFTWARE. ***************************************************************************/ +// llvm redefines DEBUG +#pragma push_macro("DEBUG") +#undef DEBUG +#include "JitManager.h" +#pragma pop_macro("DEBUG") + #include "common/os.h" #include "jit_api.h" -#include "JitManager.h" #include "state_llvm.h" #include "gallivm/lp_bld_tgsi.h" @@ -33,6 +38,7 @@ #include "util/u_inlines.h" #include "util/u_helpers.h" #include "util/u_framebuffer.h" +#include "util/u_viewport.h" #include "swr_state.h" #include "swr_context.h" @@ -230,7 +236,7 @@ swr_create_sampler_state(struct pipe_context *pipe, static void swr_bind_sampler_states(struct pipe_context *pipe, - unsigned shader, + enum pipe_shader_type shader, unsigned start, unsigned num, void **samplers) @@ -239,7 +245,7 @@ swr_bind_sampler_states(struct pipe_context *pipe, unsigned i; assert(shader < PIPE_SHADER_TYPES); - assert(start + num <= Elements(ctx->samplers[shader])); + assert(start + num <= ARRAY_SIZE(ctx->samplers[shader])); /* set the new samplers */ ctx->num_samplers[shader] = num; @@ -277,7 +283,7 @@ swr_create_sampler_view(struct pipe_context *pipe, static void swr_set_sampler_views(struct pipe_context *pipe, - unsigned shader, + enum pipe_shader_type shader, unsigned start, unsigned num, struct pipe_sampler_view **views) @@ -288,7 +294,7 @@ swr_set_sampler_views(struct pipe_context *pipe, assert(num <= PIPE_MAX_SHADER_SAMPLER_VIEWS); assert(shader < PIPE_SHADER_TYPES); - assert(start + num <= Elements(ctx->sampler_views[shader])); + assert(start + num <= ARRAY_SIZE(ctx->sampler_views[shader])); /* set the new sampler views */ ctx->num_sampler_views[shader] = num; @@ -365,6 +371,10 @@ swr_delete_vs_state(struct pipe_context *pipe, void *vs) { struct swr_vertex_shader *swr_vs = (swr_vertex_shader *)vs; FREE((void *)swr_vs->pipe.tokens); + struct swr_screen *screen = swr_screen(pipe->screen); + if (!swr_is_fence_pending(screen->flush_fence)) + swr_fence_submit(swr_context(pipe), screen->flush_fence); + swr_fence_finish(pipe->screen, NULL, screen->flush_fence, 0); delete swr_vs; } @@ -401,6 +411,10 @@ swr_delete_fs_state(struct pipe_context *pipe, void *fs) { struct swr_fragment_shader *swr_fs = (swr_fragment_shader *)fs; FREE((void *)swr_fs->pipe.tokens); + struct swr_screen *screen = swr_screen(pipe->screen); + if (!swr_is_fence_pending(screen->flush_fence)) + swr_fence_submit(swr_context(pipe), screen->flush_fence); + swr_fence_finish(pipe->screen, NULL, screen->flush_fence, 0); delete swr_fs; } @@ -409,13 +423,13 @@ static void swr_set_constant_buffer(struct pipe_context *pipe, uint shader, uint index, - struct pipe_constant_buffer *cb) + const struct pipe_constant_buffer *cb) { struct swr_context *ctx = swr_context(pipe); struct pipe_resource *constants = cb ? cb->buffer : NULL; assert(shader < PIPE_SHADER_TYPES); - assert(index < Elements(ctx->constants[shader])); + assert(index < ARRAY_SIZE(ctx->constants[shader])); /* note: reference counting */ util_copy_constant_buffer(&ctx->constants[shader][index], cb); @@ -441,6 +455,7 @@ swr_create_vertex_elements_state(struct pipe_context *pipe, assert(num_elements <= PIPE_MAX_ATTRIBS); velems = CALLOC_STRUCT(swr_vertex_element_state); if (velems) { + velems->fsState.bVertexIDOffsetEnable = true; velems->fsState.numAttribs = num_elements; for (unsigned i = 0; i < num_elements; i++) { // XXX: we should do this keyed on the VS usage info @@ -570,6 +585,10 @@ swr_set_scissor_states(struct pipe_context *pipe, struct swr_context *ctx = swr_context(pipe); ctx->scissor = *scissor; + ctx->swr_scissor.xmin = scissor->minx; + ctx->swr_scissor.xmax = scissor->maxx; + ctx->swr_scissor.ymin = scissor->miny; + ctx->swr_scissor.ymax = scissor->maxy; ctx->dirty |= SWR_NEW_SCISSOR; } @@ -663,12 +682,29 @@ swr_update_resource_status(struct pipe_context *pipe, swr_resource_read(ib->buffer); } + /* transform feedback buffers */ + for (uint32_t i = 0; i < ctx->num_so_targets; i++) { + struct pipe_stream_output_target *target = ctx->so_targets[i]; + if (target && target->buffer) + swr_resource_write(target->buffer); + } + /* texture sampler views */ - for (uint32_t i = 0; i < PIPE_MAX_SHADER_SAMPLER_VIEWS; i++) { - struct pipe_sampler_view *view = - ctx->sampler_views[PIPE_SHADER_FRAGMENT][i]; - if (view) - swr_resource_read(view->texture); + for (uint32_t j : {PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT}) { + for (uint32_t i = 0; i < ctx->num_sampler_views[j]; i++) { + struct pipe_sampler_view *view = ctx->sampler_views[j][i]; + if (view) + swr_resource_read(view->texture); + } + } + + /* constant buffers */ + for (uint32_t j : {PIPE_SHADER_VERTEX, PIPE_SHADER_FRAGMENT}) { + for (uint32_t i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) { + struct pipe_constant_buffer *cb = &ctx->constants[j][i]; + if (cb->buffer) + swr_resource_read(cb->buffer); + } } } @@ -681,25 +717,46 @@ swr_update_texture_state(struct swr_context *ctx, for (unsigned i = 0; i < num_sampler_views; i++) { struct pipe_sampler_view *view = ctx->sampler_views[shader_type][i]; + struct swr_jit_texture *jit_tex = &textures[i]; + memset(jit_tex, 0, sizeof(*jit_tex)); if (view) { struct pipe_resource *res = view->texture; struct swr_resource *swr_res = swr_resource(res); - struct swr_jit_texture *jit_tex = &textures[i]; - memset(jit_tex, 0, sizeof(*jit_tex)); + SWR_SURFACE_STATE *swr = &swr_res->swr; + size_t *mip_offsets = swr_res->mip_offsets; + if (swr_res->has_depth && swr_res->has_stencil && + !util_format_has_depth(util_format_description(view->format))) { + swr = &swr_res->secondary; + mip_offsets = swr_res->secondary_mip_offsets; + } + jit_tex->width = res->width0; jit_tex->height = res->height0; - jit_tex->depth = res->depth0; - jit_tex->first_level = view->u.tex.first_level; - jit_tex->last_level = view->u.tex.last_level; - jit_tex->base_ptr = swr_res->swr.pBaseAddress; + jit_tex->base_ptr = swr->pBaseAddress; + if (view->target != PIPE_BUFFER) { + jit_tex->first_level = view->u.tex.first_level; + jit_tex->last_level = view->u.tex.last_level; + if (view->target == PIPE_TEXTURE_3D) + jit_tex->depth = res->depth0; + else + jit_tex->depth = + view->u.tex.last_layer - view->u.tex.first_layer + 1; + jit_tex->base_ptr += view->u.tex.first_layer * + swr->qpitch * swr->pitch; + } else { + unsigned view_blocksize = util_format_get_blocksize(view->format); + jit_tex->base_ptr += view->u.buf.offset; + jit_tex->width = view->u.buf.size / view_blocksize; + jit_tex->depth = 1; + } for (unsigned level = jit_tex->first_level; level <= jit_tex->last_level; level++) { - jit_tex->row_stride[level] = swr_res->row_stride[level]; - jit_tex->img_stride[level] = swr_res->img_stride[level]; - jit_tex->mip_offsets[level] = swr_res->mip_offsets[level]; + jit_tex->row_stride[level] = swr->pitch; + jit_tex->img_stride[level] = swr->qpitch * swr->pitch; + jit_tex->mip_offsets[level] = mip_offsets[level]; } } } @@ -724,6 +781,106 @@ swr_update_sampler_state(struct swr_context *ctx, } } +static void +swr_update_constants(struct swr_context *ctx, enum pipe_shader_type shaderType) +{ + swr_draw_context *pDC = &ctx->swrDC; + + const float **constant; + uint32_t *num_constants; + struct swr_scratch_space *scratch; + + switch (shaderType) { + case PIPE_SHADER_VERTEX: + constant = pDC->constantVS; + num_constants = pDC->num_constantsVS; + scratch = &ctx->scratch->vs_constants; + break; + case PIPE_SHADER_FRAGMENT: + constant = pDC->constantFS; + num_constants = pDC->num_constantsFS; + scratch = &ctx->scratch->fs_constants; + break; + default: + debug_printf("Unsupported shader type constants\n"); + return; + } + + for (UINT i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) { + const pipe_constant_buffer *cb = &ctx->constants[shaderType][i]; + num_constants[i] = cb->buffer_size; + if (cb->buffer) { + constant[i] = + (const float *)(swr_resource_data(cb->buffer) + + cb->buffer_offset); + } else { + /* Need to copy these constants to scratch space */ + if (cb->user_buffer && cb->buffer_size) { + const void *ptr = + ((const uint8_t *)cb->user_buffer + cb->buffer_offset); + uint32_t size = AlignUp(cb->buffer_size, 4); + ptr = swr_copy_to_scratch_space(ctx, scratch, ptr, size); + constant[i] = (const float *)ptr; + } + } + } +} + +static bool +swr_change_rt(struct swr_context *ctx, + unsigned attachment, + const struct pipe_surface *sf) +{ + swr_draw_context *pDC = &ctx->swrDC; + struct SWR_SURFACE_STATE *rt = &pDC->renderTargets[attachment]; + + /* Do nothing if the render target hasn't changed */ + if ((!sf || !sf->texture) && rt->pBaseAddress == nullptr) + return false; + + /* Deal with disabling RT up front */ + if (!sf || !sf->texture) { + /* If detaching attachment, mark tiles as RESOLVED so core + * won't try to load from non-existent target. */ + swr_store_render_target(&ctx->pipe, attachment, SWR_TILE_RESOLVED); + *rt = {0}; + return true; + } + + const struct swr_resource *swr = swr_resource(sf->texture); + const SWR_SURFACE_STATE *swr_surface = &swr->swr; + SWR_FORMAT fmt = mesa_to_swr_format(sf->format); + + if (attachment == SWR_ATTACHMENT_STENCIL && swr->secondary.pBaseAddress) { + swr_surface = &swr->secondary; + fmt = swr_surface->format; + } + + if (rt->pBaseAddress == swr_surface->pBaseAddress && + rt->format == fmt && + rt->lod == sf->u.tex.level && + rt->arrayIndex == sf->u.tex.first_layer) + return false; + + bool need_fence = false; + + /* StoreTile for changed target */ + if (rt->pBaseAddress) { + /* If changing attachment to a new target, mark tiles as + * INVALID so they are reloaded from surface. */ + swr_store_render_target(&ctx->pipe, attachment, SWR_TILE_INVALID); + need_fence = true; + } + + /* Make new attachment */ + *rt = *swr_surface; + rt->format = fmt; + rt->lod = sf->u.tex.level; + rt->arrayIndex = sf->u.tex.first_layer; + + return need_fence; +} + void swr_update_derived(struct pipe_context *pipe, const struct pipe_draw_info *p_draw_info) @@ -731,6 +888,10 @@ swr_update_derived(struct pipe_context *pipe, struct swr_context *ctx = swr_context(pipe); struct swr_screen *screen = swr_screen(ctx->pipe.screen); + /* Update screen->pipe to current pipe context. */ + if (screen->pipe != pipe) + screen->pipe = pipe; + /* Any state that requires dirty flags to be re-triggered sets this mask */ /* For example, user_buffer vertex and index buffers. */ unsigned post_update_dirty_flags = 0; @@ -738,64 +899,30 @@ swr_update_derived(struct pipe_context *pipe, /* Render Targets */ if (ctx->dirty & SWR_NEW_FRAMEBUFFER) { struct pipe_framebuffer_state *fb = &ctx->framebuffer; - SWR_SURFACE_STATE *new_attachment[SWR_NUM_ATTACHMENTS] = {0}; - UINT i; + const struct util_format_description *desc = NULL; + bool need_fence = false; /* colorbuffer targets */ - if (fb->nr_cbufs) - for (i = 0; i < fb->nr_cbufs; ++i) - if (fb->cbufs[i]) { - struct swr_resource *colorBuffer = - swr_resource(fb->cbufs[i]->texture); - new_attachment[SWR_ATTACHMENT_COLOR0 + i] = &colorBuffer->swr; - } - - /* depth/stencil target */ - if (fb->zsbuf) { - struct swr_resource *depthStencilBuffer = - swr_resource(fb->zsbuf->texture); - if (depthStencilBuffer->has_depth) { - new_attachment[SWR_ATTACHMENT_DEPTH] = &depthStencilBuffer->swr; - - if (depthStencilBuffer->has_stencil) - new_attachment[SWR_ATTACHMENT_STENCIL] = - &depthStencilBuffer->secondary; - - } else if (depthStencilBuffer->has_stencil) - new_attachment[SWR_ATTACHMENT_STENCIL] = &depthStencilBuffer->swr; + if (fb->nr_cbufs) { + for (unsigned i = 0; i < fb->nr_cbufs; ++i) + need_fence |= swr_change_rt( + ctx, SWR_ATTACHMENT_COLOR0 + i, fb->cbufs[i]); } + for (unsigned i = fb->nr_cbufs; i < SWR_NUM_RENDERTARGETS; ++i) + need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_COLOR0 + i, NULL); - /* Make the attachment updates */ - swr_draw_context *pDC = &ctx->swrDC; - SWR_SURFACE_STATE *renderTargets = pDC->renderTargets; - unsigned need_fence = FALSE; - for (i = 0; i < SWR_NUM_ATTACHMENTS; i++) { - void *new_base = nullptr; - if (new_attachment[i]) - new_base = new_attachment[i]->pBaseAddress; - - /* StoreTile for changed target */ - if (renderTargets[i].pBaseAddress != new_base) { - if (renderTargets[i].pBaseAddress) { - /* If changing attachment to a new target, mark tiles as - * INVALID so they are reloaded from surface. - * If detaching attachment, mark tiles as RESOLVED so core - * won't try to load from non-existent target. */ - enum SWR_TILE_STATE post_state = (new_attachment[i] - ? SWR_TILE_INVALID : SWR_TILE_RESOLVED); - swr_store_render_target(pipe, i, post_state); - - need_fence |= TRUE; - } + /* depth/stencil target */ + if (fb->zsbuf) + desc = util_format_description(fb->zsbuf->format); + if (fb->zsbuf && util_format_has_depth(desc)) + need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_DEPTH, fb->zsbuf); + else + need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_DEPTH, NULL); - /* Make new attachment */ - if (new_attachment[i]) - renderTargets[i] = *new_attachment[i]; - else - if (renderTargets[i].pBaseAddress) - renderTargets[i] = {0}; - } - } + if (fb->zsbuf && util_format_has_stencil(desc)) + need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_STENCIL, fb->zsbuf); + else + need_fence |= swr_change_rt(ctx, SWR_ATTACHMENT_STENCIL, NULL); /* This fence ensures any attachment changes are resolved before the * next draw */ @@ -804,7 +931,9 @@ swr_update_derived(struct pipe_context *pipe, } /* Raster state */ - if (ctx->dirty & (SWR_NEW_RASTERIZER | SWR_NEW_FRAMEBUFFER)) { + if (ctx->dirty & (SWR_NEW_RASTERIZER | + SWR_NEW_VS | // clipping + SWR_NEW_FRAMEBUFFER)) { pipe_rasterizer_state *rasterizer = ctx->rasterizer; pipe_framebuffer_state *fb = &ctx->framebuffer; @@ -831,7 +960,7 @@ swr_update_derived(struct pipe_context *pipe, rastState->msaaRastEnable = false; rastState->rastMode = SWR_MSAA_RASTMODE_OFF_PIXEL; rastState->sampleCount = SWR_MULTISAMPLE_1X; - rastState->bForcedSampleCount = false; + rastState->forcedSampleCount = false; bool do_offset = false; switch (rasterizer->fill_front) { @@ -860,16 +989,22 @@ swr_update_derived(struct pipe_context *pipe, rastState->depthFormat = swr_resource(zb->texture)->swr.format; rastState->depthClipEnable = rasterizer->depth_clip; + rastState->clipHalfZ = rasterizer->clip_halfz; + + rastState->clipDistanceMask = + ctx->vs->info.base.num_written_clipdistance ? + ctx->vs->info.base.clipdist_writemask & rasterizer->clip_plane_enable : + rasterizer->clip_plane_enable; + + rastState->cullDistanceMask = + ctx->vs->info.base.culldist_writemask << ctx->vs->info.base.num_written_clipdistance; SwrSetRastState(ctx->swrContext, rastState); } /* Scissor */ if (ctx->dirty & SWR_NEW_SCISSOR) { - pipe_scissor_state *scissor = &ctx->scissor; - BBOX bbox(scissor->miny, scissor->maxy, - scissor->minx, scissor->maxx); - SwrSetScissorRects(ctx->swrContext, 1, &bbox); + SwrSetScissorRects(ctx->swrContext, 1, &ctx->swr_scissor); } /* Viewport */ @@ -880,33 +1015,34 @@ swr_update_derived(struct pipe_context *pipe, pipe_rasterizer_state *rasterizer = ctx->rasterizer; SWR_VIEWPORT *vp = &ctx->derived.vp; - SWR_VIEWPORT_MATRIX *vpm = &ctx->derived.vpm; + SWR_VIEWPORT_MATRICES *vpm = &ctx->derived.vpm; vp->x = state->translate[0] - state->scale[0]; - vp->width = state->translate[0] + state->scale[0]; + vp->width = 2 * state->scale[0]; vp->y = state->translate[1] - fabs(state->scale[1]); - vp->height = state->translate[1] + fabs(state->scale[1]); - if (rasterizer->clip_halfz == 0) { - vp->minZ = state->translate[2] - state->scale[2]; - vp->maxZ = state->translate[2] + state->scale[2]; - } else { - vp->minZ = state->translate[2]; - vp->maxZ = state->translate[2] + state->scale[2]; - } + vp->height = 2 * fabs(state->scale[1]); + util_viewport_zmin_zmax(state, rasterizer->clip_halfz, + &vp->minZ, &vp->maxZ); - vpm->m00 = state->scale[0]; - vpm->m11 = state->scale[1]; - vpm->m22 = state->scale[2]; - vpm->m30 = state->translate[0]; - vpm->m31 = state->translate[1]; - vpm->m32 = state->translate[2]; + vpm->m00[0] = state->scale[0]; + vpm->m11[0] = state->scale[1]; + vpm->m22[0] = state->scale[2]; + vpm->m30[0] = state->translate[0]; + vpm->m31[0] = state->translate[1]; + vpm->m32[0] = state->translate[2]; /* Now that the matrix is calculated, clip the view coords to screen * size. OpenGL allows for -ve x,y in the viewport. */ - vp->x = std::max(vp->x, 0.0f); - vp->y = std::max(vp->y, 0.0f); - vp->width = std::min(vp->width, (float)fb->width); - vp->height = std::min(vp->height, (float)fb->height); + if (vp->x < 0.0f) { + vp->width += vp->x; + vp->x = 0.0f; + } + if (vp->y < 0.0f) { + vp->height += vp->y; + vp->y = 0.0f; + } + vp->width = std::min(vp->width, (float)fb->width - vp->x); + vp->height = std::min(vp->height, (float)fb->height - vp->y); SwrSetViewports(ctx->swrContext, 1, vp, vpm); } @@ -914,14 +1050,39 @@ swr_update_derived(struct pipe_context *pipe, /* Set vertex & index buffers */ /* (using draw info if called by swr_draw_vbo) */ if (ctx->dirty & SWR_NEW_VERTEX) { - uint32_t size, pitch, max_vertex, partial_inbounds; + uint32_t size, pitch, max_vertex, partial_inbounds, scratch_total; const uint8_t *p_data; + uint8_t *scratch = NULL; /* If being called by swr_draw_vbo, copy draw details */ struct pipe_draw_info info = {0}; if (p_draw_info) info = *p_draw_info; + /* We must get all the scratch space in one go */ + scratch_total = 0; + for (UINT i = 0; i < ctx->num_vertex_buffers; i++) { + struct pipe_vertex_buffer *vb = &ctx->vertex_buffer[i]; + + if (!vb->user_buffer) + continue; + + if (vb->stride) { + size = (info.max_index - info.min_index + 1) * vb->stride; + } else { + /* pitch = 0, means constant value + * set size to 1 vertex */ + size = ctx->velems->stream_pitch[i]; + } + + scratch_total += AlignUp(size, 4); + } + + if (scratch_total) { + scratch = (uint8_t *)swr_copy_to_scratch_space( + ctx, &ctx->scratch->vertex_buffer, NULL, scratch_total); + } + /* vertex buffers */ SWR_VERTEX_BUFFER_STATE swrVertexBuffers[PIPE_MAX_ATTRIBS]; for (UINT i = 0; i < ctx->num_vertex_buffers; i++) { @@ -936,8 +1097,7 @@ swr_update_derived(struct pipe_context *pipe, max_vertex = size / pitch; partial_inbounds = size % pitch; - p_data = (const uint8_t *)swr_resource_data(vb->buffer) - + vb->buffer_offset; + p_data = swr_resource_data(vb->buffer) + vb->buffer_offset; } else { /* Client buffer * client memory is one-time use, re-trigger SWR_NEW_VERTEX to @@ -959,8 +1119,9 @@ swr_update_derived(struct pipe_context *pipe, size = AlignUp(size, 4); const void *ptr = (const uint8_t *) vb->user_buffer + info.min_index * pitch; - ptr = swr_copy_to_scratch_space( - ctx, &ctx->scratch->vertex_buffer, ptr, size); + memcpy(scratch, ptr, size); + ptr = scratch; + scratch += size; p_data = (const uint8_t *)ptr - info.min_index * pitch; } @@ -989,8 +1150,7 @@ swr_update_derived(struct pipe_context *pipe, * size is based on buffer->width0 rather than info.count * to prevent having to validate VBO on each draw */ size = ib->buffer->width0; - p_data = - (const uint8_t *)swr_resource_data(ib->buffer) + ib->offset; + p_data = swr_resource_data(ib->buffer) + ib->offset; } else { /* Client buffer * client memory is one-time use, re-trigger SWR_NEW_VERTEX to @@ -1024,6 +1184,7 @@ swr_update_derived(struct pipe_context *pipe, /* VertexShader */ if (ctx->dirty & (SWR_NEW_VS | + SWR_NEW_RASTERIZER | // for clip planes SWR_NEW_SAMPLER | SWR_NEW_SAMPLER_VIEW | SWR_NEW_FRAMEBUFFER)) { @@ -1032,10 +1193,9 @@ swr_update_derived(struct pipe_context *pipe, auto search = ctx->vs->map.find(key); PFN_VERTEX_FUNC func; if (search != ctx->vs->map.end()) { - func = search->second; + func = search->second->shader; } else { func = swr_compile_vs(ctx, key); - ctx->vs->map.insert(std::make_pair(key, func)); } SwrSetVertexFunc(ctx->swrContext, func); @@ -1064,10 +1224,9 @@ swr_update_derived(struct pipe_context *pipe, auto search = ctx->fs->map.find(key); PFN_PIXEL_KERNEL func; if (search != ctx->fs->map.end()) { - func = search->second; + func = search->second->shader; } else { func = swr_compile_fs(ctx, key); - ctx->fs->map.insert(std::make_pair(key, func)); } SWR_PS_STATE psState = {0}; psState.pfnPixelShader = func; @@ -1130,52 +1289,12 @@ swr_update_derived(struct pipe_context *pipe, /* VertexShader Constants */ if (ctx->dirty & SWR_NEW_VSCONSTANTS) { - swr_draw_context *pDC = &ctx->swrDC; - - for (UINT i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) { - const pipe_constant_buffer *cb = - &ctx->constants[PIPE_SHADER_VERTEX][i]; - pDC->num_constantsVS[i] = cb->buffer_size; - if (cb->buffer) - pDC->constantVS[i] = - (const float *)((const uint8_t *)cb->buffer + cb->buffer_offset); - else { - /* Need to copy these constants to scratch space */ - if (cb->user_buffer && cb->buffer_size) { - const void *ptr = - ((const uint8_t *)cb->user_buffer + cb->buffer_offset); - uint32_t size = AlignUp(cb->buffer_size, 4); - ptr = swr_copy_to_scratch_space( - ctx, &ctx->scratch->vs_constants, ptr, size); - pDC->constantVS[i] = (const float *)ptr; - } - } - } + swr_update_constants(ctx, PIPE_SHADER_VERTEX); } /* FragmentShader Constants */ if (ctx->dirty & SWR_NEW_FSCONSTANTS) { - swr_draw_context *pDC = &ctx->swrDC; - - for (UINT i = 0; i < PIPE_MAX_CONSTANT_BUFFERS; i++) { - const pipe_constant_buffer *cb = - &ctx->constants[PIPE_SHADER_FRAGMENT][i]; - pDC->num_constantsFS[i] = cb->buffer_size; - if (cb->buffer) - pDC->constantFS[i] = - (const float *)((const uint8_t *)cb->buffer + cb->buffer_offset); - else { - /* Need to copy these constants to scratch space */ - if (cb->user_buffer && cb->buffer_size) { - const void *ptr = - ((const uint8_t *)cb->user_buffer + cb->buffer_offset); - uint32_t size = AlignUp(cb->buffer_size, 4); - ptr = swr_copy_to_scratch_space( - ctx, &ctx->scratch->fs_constants, ptr, size); - pDC->constantFS[i] = (const float *)ptr; - } - } - } + swr_update_constants(ctx, PIPE_SHADER_FRAGMENT); } /* Depth/stencil state */ @@ -1183,6 +1302,7 @@ swr_update_derived(struct pipe_context *pipe, struct pipe_depth_state *depth = &(ctx->depth_stencil->depth); struct pipe_stencil_state *stencil = ctx->depth_stencil->stencil; SWR_DEPTH_STENCIL_STATE depthStencilState = {{0}}; + SWR_DEPTH_BOUNDS_STATE depthBoundsState = {0}; /* XXX, incomplete. Need to flesh out stencil & alpha test state struct pipe_stencil_state *front_stencil = @@ -1229,6 +1349,11 @@ swr_update_derived(struct pipe_context *pipe, depthStencilState.depthTestFunc = swr_convert_depth_func(depth->func); depthStencilState.depthWriteEnable = depth->writemask; SwrSetDepthStencilState(ctx->swrContext, &depthStencilState); + + depthBoundsState.depthBoundsTestEnable = depth->bounds_test; + depthBoundsState.depthBoundsTestMinValue = depth->bounds_min; + depthBoundsState.depthBoundsTestMaxValue = depth->bounds_max; + SwrSetDepthBoundsState(ctx->swrContext, &depthBoundsState); } /* Blend State */ @@ -1277,8 +1402,18 @@ swr_update_derived(struct pipe_context *pipe, &ctx->blend->compileState[target], sizeof(compileState.blendState)); + const SWR_FORMAT_INFO& info = GetFormatInfo(compileState.format); + if (compileState.blendState.logicOpEnable && + ((info.type[0] == SWR_TYPE_FLOAT) || info.isSRGB)) { + compileState.blendState.logicOpEnable = false; + } + + if (info.type[0] == SWR_TYPE_SINT || info.type[0] == SWR_TYPE_UINT) + compileState.blendState.blendEnable = false; + if (compileState.blendState.blendEnable == false && - compileState.blendState.logicOpEnable == false) { + compileState.blendState.logicOpEnable == false && + ctx->depth_stencil->alpha.enabled == 0) { SwrSetBlendFunc(ctx->swrContext, target, NULL); continue; } @@ -1286,7 +1421,12 @@ swr_update_derived(struct pipe_context *pipe, compileState.desc.alphaTestEnable = ctx->depth_stencil->alpha.enabled; compileState.desc.independentAlphaBlendEnable = - ctx->blend->pipe.independent_blend_enable; + (compileState.blendState.sourceBlendFactor != + compileState.blendState.sourceAlphaBlendFactor) || + (compileState.blendState.destBlendFactor != + compileState.blendState.destAlphaBlendFactor) || + (compileState.blendState.colorBlendFunc != + compileState.blendState.alphaBlendFunc); compileState.desc.alphaToCoverageEnable = ctx->blend->pipe.alpha_to_coverage; compileState.desc.sampleMaskEnable = 0; // XXX @@ -1296,6 +1436,8 @@ swr_update_derived(struct pipe_context *pipe, swr_convert_depth_func(ctx->depth_stencil->alpha.func); compileState.alphaTestFormat = ALPHA_TEST_FLOAT32; // xxx + compileState.Canonicalize(); + PFN_BLEND_JIT_FUNC func = NULL; auto search = ctx->blendJIT->find(compileState); if (search != ctx->blendJIT->end()) { @@ -1340,28 +1482,36 @@ swr_update_derived(struct pipe_context *pipe, } } - uint32_t linkage = ctx->vs->linkageMask; - if (ctx->rasterizer->sprite_coord_enable) - linkage |= (1 << ctx->vs->info.base.num_outputs); - - SwrSetLinkage(ctx->swrContext, linkage, NULL); - - // set up frontend state - SWR_FRONTEND_STATE feState = {0}; - SwrSetFrontendState(ctx->swrContext, &feState); + if (ctx->dirty & SWR_NEW_CLIP) { + // shader exporting clip distances overrides all user clip planes + if (ctx->rasterizer->clip_plane_enable && + !ctx->vs->info.base.num_written_clipdistance) + { + swr_draw_context *pDC = &ctx->swrDC; + memcpy(pDC->userClipPlanes, + ctx->clip.ucp, + sizeof(pDC->userClipPlanes)); + } + } // set up backend state SWR_BACKEND_STATE backendState = {0}; - backendState.numAttributes = 1; - backendState.numComponents[0] = 4; - backendState.constantInterpolationMask = ctx->fs->constantMask; + backendState.numAttributes = + ctx->vs->info.base.num_outputs - 1 + + (ctx->rasterizer->sprite_coord_enable ? 1 : 0); + for (unsigned i = 0; i < backendState.numAttributes; i++) + backendState.numComponents[i] = 4; + backendState.constantInterpolationMask = + ctx->rasterizer->flatshade ? + ctx->fs->flatConstantMask : + ctx->fs->constantMask; backendState.pointSpriteTexCoordMask = ctx->fs->pointSpriteMask; SwrSetBackendState(ctx->swrContext, &backendState); /* Ensure that any in-progress attachment change StoreTiles finish */ if (swr_is_fence_pending(screen->flush_fence)) - swr_fence_finish(pipe->screen, screen->flush_fence, 0); + swr_fence_finish(pipe->screen, NULL, screen->flush_fence, 0); /* Finally, update the in-use status of all resources involved in draw */ swr_update_resource_status(pipe, p_draw_info);