/** * \brief Call this after drawing to mark which buffers need resolving * * If the depth buffer was written to and if it has an accompanying HiZ * buffer, then mark that it needs a depth resolve. * * (In the future, this will also mark needed MSAA resolves). */ static void brw_postdraw_set_buffers_need_resolve(struct brw_context *brw) { struct gl_context *ctx = &brw->intel.ctx; struct gl_framebuffer *fb = ctx->DrawBuffer; struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH); if (depth_irb && ctx->Depth.Mask) { intel_renderbuffer_set_needs_depth_resolve(depth_irb); } }
/** * \brief Call this after drawing to mark which buffers need resolving * * If the depth buffer was written to and if it has an accompanying HiZ * buffer, then mark that it needs a depth resolve. * * If the color buffer is a multisample window system buffer, then * mark that it needs a downsample. */ static void brw_postdraw_set_buffers_need_resolve(struct brw_context *brw) { struct gl_context *ctx = &brw->ctx; struct gl_framebuffer *fb = ctx->DrawBuffer; struct intel_renderbuffer *front_irb = NULL; struct intel_renderbuffer *back_irb = intel_get_renderbuffer(fb, BUFFER_BACK_LEFT); struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH); if (brw->is_front_buffer_rendering) front_irb = intel_get_renderbuffer(fb, BUFFER_FRONT_LEFT); if (front_irb) intel_renderbuffer_set_needs_downsample(front_irb); if (back_irb) intel_renderbuffer_set_needs_downsample(back_irb); if (depth_irb && ctx->Depth.Mask) intel_renderbuffer_set_needs_depth_resolve(depth_irb); }
/** * Implements fast depth clears on gen6+. * * Fast clears basically work by setting a flag in each of the subspans * represented in the HiZ buffer that says "When you need the depth values for * this subspan, it's the hardware's current clear value." Then later rendering * can just use the static clear value instead of referencing memory. * * The tricky part of the implementation is that you have to have the clear * value that was used on the depth buffer in place for all further rendering, * at least until a resolve to the real depth buffer happens. */ static bool brw_fast_clear_depth(struct gl_context *ctx) { struct intel_context *intel = intel_context(ctx); struct gl_framebuffer *fb = ctx->DrawBuffer; struct intel_renderbuffer *depth_irb = intel_get_renderbuffer(fb, BUFFER_DEPTH); struct intel_mipmap_tree *mt = depth_irb->mt; if (intel->gen < 6) return false; if (!mt->hiz_mt) return false; /* We only handle full buffer clears -- otherwise you'd have to track whether * a previous clear had happened at a different clear value and resolve it * first. */ if (ctx->Scissor.Enabled && !noop_scissor(ctx, fb)) { perf_debug("Failed to fast clear depth due to scissor being enabled. " "Possible 5%% performance win if avoided.\n"); return false; } /* The rendered area has to be 8x4 samples, not resolved pixels, so we look * at the miptree slice dimensions instead of renderbuffer size. */ if (mt->level[depth_irb->mt_level].width % 8 != 0 || mt->level[depth_irb->mt_level].height % 4 != 0) { perf_debug("Failed to fast clear depth due to width/height %d,%d not " "being aligned to 8,4. Possible 5%% performance win if " "avoided\n", mt->level[depth_irb->mt_level].width, mt->level[depth_irb->mt_level].height); return false; } uint32_t depth_clear_value; switch (mt->format) { case MESA_FORMAT_Z32_FLOAT_X24S8: case MESA_FORMAT_S8_Z24: /* From the Sandy Bridge PRM, volume 2 part 1, page 314: * * "[DevSNB+]: Several cases exist where Depth Buffer Clear cannot be * enabled (the legacy method of clearing must be performed): * * - If the depth buffer format is D32_FLOAT_S8X24_UINT or * D24_UNORM_S8_UINT. */ return false; case MESA_FORMAT_Z32_FLOAT: depth_clear_value = float_as_int(ctx->Depth.Clear); break; case MESA_FORMAT_Z16: /* From the Sandy Bridge PRM, volume 2 part 1, page 314: * * "[DevSNB+]: Several cases exist where Depth Buffer Clear cannot be * enabled (the legacy method of clearing must be performed): * * - DevSNB{W/A}]: When depth buffer format is D16_UNORM and the * width of the map (LOD0) is not multiple of 16, fast clear * optimization must be disabled. */ if (intel->gen == 6 && (mt->level[depth_irb->mt_level].width % 16) != 0) return false; /* FALLTHROUGH */ default: depth_clear_value = fb->_DepthMax * ctx->Depth.Clear; break; } /* If we're clearing to a new clear value, then we need to resolve any clear * flags out of the HiZ buffer into the real depth buffer. */ if (mt->depth_clear_value != depth_clear_value) { intel_miptree_all_slices_resolve_depth(intel, mt); mt->depth_clear_value = depth_clear_value; } /* From the Sandy Bridge PRM, volume 2 part 1, page 313: * * "If other rendering operations have preceded this clear, a * PIPE_CONTROL with write cache flush enabled and Z-inhibit disabled * must be issued before the rectangle primitive used for the depth * buffer clear operation. */ intel_batchbuffer_emit_mi_flush(intel); intel_hiz_exec(intel, mt, depth_irb->mt_level, depth_irb->mt_layer, GEN6_HIZ_OP_DEPTH_CLEAR); if (intel->gen == 6) { /* From the Sandy Bridge PRM, volume 2 part 1, page 314: * * "DevSNB, DevSNB-B{W/A}]: Depth buffer clear pass must be followed * by a PIPE_CONTROL command with DEPTH_STALL bit set and Then * followed by Depth FLUSH' */ intel_batchbuffer_emit_mi_flush(intel); } /* Now, the HiZ buffer contains data that needs to be resolved to the depth * buffer. */ intel_renderbuffer_set_needs_depth_resolve(depth_irb); return true; }