From 5cf665afa1250d50856130f889570f7227f20eb0 Mon Sep 17 00:00:00 2001 From: Jason Ekstrand Date: Mon, 6 Jun 2016 20:25:21 -0700 Subject: [PATCH] i965/state: Add a helper for emitting a surface state using isl Signed-off-by: Jason Ekstrand Reviewed-by: Topi Pohjolainen Reviewed-by: Chad Versace --- src/mesa/drivers/dri/i965/brw_state.h | 7 ++ src/mesa/drivers/dri/i965/brw_wm_surface_state.c | 81 ++++++++++++++++++++++++ 2 files changed, 88 insertions(+) diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index a16e876..599aa25 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -274,6 +274,13 @@ GLuint translate_tex_format(struct brw_context *brw, int brw_get_texture_swizzle(const struct gl_context *ctx, const struct gl_texture_object *t); +void brw_emit_surface_state(struct brw_context *brw, + struct intel_mipmap_tree *mt, + const struct isl_view *view, + uint32_t mocs, bool for_gather, + uint32_t *surf_offset, int surf_index, + unsigned read_domains, unsigned write_domains); + void brw_update_renderbuffer_surfaces(struct brw_context *brw, const struct gl_framebuffer *fb, uint32_t render_target_start, diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index c101e05..fcaaddd 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -35,6 +35,7 @@ #include "main/mtypes.h" #include "main/samplerobj.h" #include "main/shaderimage.h" +#include "main/teximage.h" #include "program/prog_parameter.h" #include "program/prog_instruction.h" #include "main/framebuffer.h" @@ -52,6 +53,86 @@ #include "brw_defines.h" #include "brw_wm.h" +struct surface_state_info { + unsigned num_dwords; + unsigned ss_align; /* Required alignment of RENDER_SURFACE_STATE in bytes */ + unsigned reloc_dw; + unsigned aux_reloc_dw; + unsigned tex_mocs; + unsigned rb_mocs; +}; + +static const struct surface_state_info surface_state_infos[] = { + [4] = {6, 32, 1, 0}, + [5] = {6, 32, 1, 0}, + [6] = {6, 32, 1, 0}, + [7] = {8, 32, 1, 6, GEN7_MOCS_L3, GEN7_MOCS_L3}, + [8] = {13, 64, 8, 10, BDW_MOCS_WB, BDW_MOCS_PTE}, + [9] = {16, 64, 8, 10, SKL_MOCS_WB, SKL_MOCS_PTE}, +}; + +void +brw_emit_surface_state(struct brw_context *brw, + struct intel_mipmap_tree *mt, + const struct isl_view *view, + uint32_t mocs, bool for_gather, + uint32_t *surf_offset, int surf_index, + unsigned read_domains, unsigned write_domains) +{ + const struct surface_state_info ss_info = surface_state_infos[brw->gen]; + + struct isl_surf surf; + intel_miptree_get_isl_surf(brw, mt, &surf); + + union isl_color_value clear_color = { .u32 = { 0, 0, 0, 0 } }; + + struct isl_surf *aux_surf = NULL, aux_surf_s; + uint64_t aux_offset = 0; + enum isl_aux_usage aux_usage = ISL_AUX_USAGE_NONE; + if (mt->mcs_mt && + ((view->usage & ISL_SURF_USAGE_RENDER_TARGET_BIT) || + mt->fast_clear_state != INTEL_FAST_CLEAR_STATE_RESOLVED)) { + intel_miptree_get_aux_isl_surf(brw, mt, &aux_surf_s, &aux_usage); + aux_surf = &aux_surf_s; + assert(mt->mcs_mt->offset == 0); + aux_offset = mt->mcs_mt->bo->offset64; + + /* We only really need a clear color if we also have an auxiliary + * surfacae. Without one, it does nothing. + */ + clear_color = intel_miptree_get_isl_clear_color(brw, mt); + } + + uint32_t *dw = __brw_state_batch(brw, AUB_TRACE_SURFACE_STATE, + ss_info.num_dwords * 4, ss_info.ss_align, + surf_index, surf_offset); + + isl_surf_fill_state(&brw->isl_dev, dw, .surf = &surf, .view = view, + .address = mt->bo->offset64 + mt->offset, + .aux_surf = aux_surf, .aux_usage = aux_usage, + .aux_address = aux_offset, + .mocs = mocs, .clear_color = clear_color); + + drm_intel_bo_emit_reloc(brw->batch.bo, + *surf_offset + 4 * ss_info.reloc_dw, + mt->bo, mt->offset, + read_domains, write_domains); + + if (aux_surf) { + /* On gen7 and prior, the upper 20 bits of surface state DWORD 6 are the + * upper 20 bits of the GPU address of the MCS buffer; the lower 12 bits + * contain other control information. Since buffer addresses are always + * on 4k boundaries (and thus have their lower 12 bits zero), we can use + * an ordinary reloc to do the necessary address translation. + */ + assert((aux_offset & 0xfff) == 0); + drm_intel_bo_emit_reloc(brw->batch.bo, + *surf_offset + 4 * ss_info.aux_reloc_dw, + mt->mcs_mt->bo, dw[ss_info.aux_reloc_dw] & 0xfff, + read_domains, write_domains); + } +} + GLuint translate_tex_target(GLenum target) { -- 2.7.4