From 8ee7370c9bed0be40c5d134f0b92eb3782c6b7e9 Mon Sep 17 00:00:00 2001 From: Andreas Hartmetz Date: Sat, 7 Dec 2013 04:42:24 +0100 Subject: [PATCH] radeonsi: Write htile state to hardware. --- src/gallium/drivers/radeonsi/si_state.c | 67 ++++++++++++++++++-- src/gallium/drivers/radeonsi/si_state.h | 4 +- src/gallium/drivers/radeonsi/si_state_draw.c | 7 +- 3 files changed, 65 insertions(+), 13 deletions(-) diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c index 7bae72a4c0c..52c9d53ebb0 100644 --- a/src/gallium/drivers/radeonsi/si_state.c +++ b/src/gallium/drivers/radeonsi/si_state.c @@ -708,7 +708,7 @@ static void *si_create_dsa_state(struct pipe_context *ctx, struct si_state_dsa *dsa = CALLOC_STRUCT(si_state_dsa); struct si_pm4_state *pm4 = &dsa->pm4; unsigned db_depth_control; - unsigned db_render_override, db_render_control; + unsigned db_render_control; uint32_t db_stencil_control = 0; if (dsa == NULL) { @@ -754,10 +754,6 @@ static void *si_create_dsa_state(struct pipe_context *ctx, /* misc */ db_render_control = 0; - db_render_override = S_02800C_FORCE_HIZ_ENABLE(V_02800C_FORCE_DISABLE) | - S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) | - S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE); - /* TODO db_render_override depends on query */ si_pm4_set_reg(pm4, R_028020_DB_DEPTH_BOUNDS_MIN, 0x00000000); si_pm4_set_reg(pm4, R_028024_DB_DEPTH_BOUNDS_MAX, 0x00000000); si_pm4_set_reg(pm4, R_028028_DB_STENCIL_CLEAR, 0x00000000); @@ -765,12 +761,10 @@ static void *si_create_dsa_state(struct pipe_context *ctx, //si_pm4_set_reg(pm4, R_028410_SX_ALPHA_TEST_CONTROL, alpha_test_control); si_pm4_set_reg(pm4, R_028800_DB_DEPTH_CONTROL, db_depth_control); si_pm4_set_reg(pm4, R_028000_DB_RENDER_CONTROL, db_render_control); - si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, db_render_override); si_pm4_set_reg(pm4, R_02842C_DB_STENCIL_CONTROL, db_stencil_control); si_pm4_set_reg(pm4, R_028AC0_DB_SRESULTS_COMPARE_STATE0, 0x0); si_pm4_set_reg(pm4, R_028AC4_DB_SRESULTS_COMPARE_STATE1, 0x0); si_pm4_set_reg(pm4, R_028AC8_DB_PRELOAD_CONTROL, 0x0); - dsa->db_render_override = db_render_override; return dsa; } @@ -1742,6 +1736,47 @@ static void si_cb(struct r600_context *rctx, struct si_pm4_state *pm4, } } +/* Update register(s) containing depth buffer and draw state. */ +void si_update_db_draw_state(struct r600_context *rctx, struct r600_surface *zsbuf) +{ + struct si_pm4_state *pm4; + uint32_t db_render_override; + boolean hiz_enable = false; + + pm4 = si_pm4_alloc_state(rctx); + if (pm4 == NULL) { + return; + } + + /* db */ + + /* TODO HiS aka stencil buffer htile goes here */ + db_render_override = S_02800C_FORCE_HIS_ENABLE0(V_02800C_FORCE_DISABLE) | + S_02800C_FORCE_HIS_ENABLE1(V_02800C_FORCE_DISABLE); + + /* HiZ aka depth buffer htile */ + if (zsbuf && zsbuf->base.texture) { + struct r600_texture *rtex = (struct r600_texture*)zsbuf->base.texture; + uint level = zsbuf->base.u.tex.level; + /* use htile only for first level */ + hiz_enable = rtex->htile_buffer && !level; + } + if (hiz_enable) { + db_render_override |= S_02800C_FORCE_HIZ_ENABLE(V_02800C_FORCE_OFF); + } else { + db_render_override |= S_02800C_FORCE_HIZ_ENABLE(V_02800C_FORCE_DISABLE); + } + + /* draw */ + + if (rctx->num_cs_dw_nontimer_queries_suspend) { + db_render_override |= S_02800C_NOOP_CULL_DISABLE(1); + } + + si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, db_render_override); + si_pm4_set_state(rctx, db_draw, pm4); +} + static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4, const struct pipe_framebuffer_state *state) { @@ -1752,6 +1787,7 @@ static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4, unsigned macro_aspect, tile_split, stile_split, bankh, bankw, nbanks, pipe_config; uint32_t z_info, s_info, db_depth_info; uint64_t z_offs, s_offs; + uint32_t db_htile_data_base, db_htile_surface; if (state->zsbuf == NULL) { si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, S_028040_FORMAT(V_028040_Z_INVALID)); @@ -1836,9 +1872,23 @@ static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4, s_info |= S_028044_TILE_MODE_INDEX(tile_mode_index); } + /* HiZ aka depth buffer htile */ + /* use htile only for first level */ + if (rtex->htile_buffer && !level) { + z_info |= S_028040_TILE_SURFACE_ENABLE(1); + /* Force off means no force, DB_SHADER_CONTROL decides */ + uint64_t va = r600_resource_va(&rctx->screen->b.b, &rtex->htile_buffer->b.b); + db_htile_data_base = va >> 8; + db_htile_surface = S_028ABC_FULL_CACHE(1); + } else { + db_htile_data_base = 0; + db_htile_surface = 0; + } + si_pm4_set_reg(pm4, R_028008_DB_DEPTH_VIEW, S_028008_SLICE_START(state->zsbuf->u.tex.first_layer) | S_028008_SLICE_MAX(state->zsbuf->u.tex.last_layer)); + si_pm4_set_reg(pm4, R_028014_DB_HTILE_DATA_BASE, db_htile_data_base); si_pm4_set_reg(pm4, R_02803C_DB_DEPTH_INFO, db_depth_info); si_pm4_set_reg(pm4, R_028040_DB_Z_INFO, z_info); @@ -1852,6 +1902,8 @@ static void si_db(struct r600_context *rctx, struct si_pm4_state *pm4, si_pm4_set_reg(pm4, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch)); si_pm4_set_reg(pm4, R_02805C_DB_DEPTH_SLICE, S_02805C_SLICE_TILE_MAX(slice)); + + si_pm4_set_reg(pm4, R_028ABC_DB_HTILE_SURFACE, db_htile_surface); } #define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \ @@ -2122,6 +2174,7 @@ static void si_set_framebuffer_state(struct pipe_context *ctx, si_pm4_set_state(rctx, framebuffer, pm4); si_update_fb_rs_state(rctx); si_update_fb_blend_state(rctx); + si_update_db_draw_state(rctx, (struct r600_surface *)state->zsbuf); } /* diff --git a/src/gallium/drivers/radeonsi/si_state.h b/src/gallium/drivers/radeonsi/si_state.h index 6774e57e21f..d44c4f3bf9c 100644 --- a/src/gallium/drivers/radeonsi/si_state.h +++ b/src/gallium/drivers/radeonsi/si_state.h @@ -61,7 +61,6 @@ struct si_state_dsa { struct si_pm4_state pm4; float alpha_ref; unsigned alpha_func; - unsigned db_render_override; unsigned db_render_control; uint8_t valuemask[2]; uint8_t writemask[2]; @@ -88,6 +87,7 @@ union si_state { struct si_state_dsa *dsa; struct si_pm4_state *fb_rs; struct si_pm4_state *fb_blend; + struct si_pm4_state *db_draw; struct si_pm4_state *dsa_stencil_ref; struct si_pm4_state *vs; struct si_pm4_state *vs_sampler; @@ -203,12 +203,14 @@ void si_copy_buffer(struct r600_context *rctx, /* si_state.c */ struct si_pipe_shader_selector; +struct r600_surface; boolean si_is_format_supported(struct pipe_screen *screen, enum pipe_format format, enum pipe_texture_target target, unsigned sample_count, unsigned usage); +void si_update_db_draw_state(struct r600_context *rctx, struct r600_surface *zsbuf); int si_shader_select(struct pipe_context *ctx, struct si_pipe_shader_selector *sel, unsigned *dirty); diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c index be5a5c8e030..0a08be15b00 100644 --- a/src/gallium/drivers/radeonsi/si_state_draw.c +++ b/src/gallium/drivers/radeonsi/si_state_draw.c @@ -574,8 +574,6 @@ static void si_state_draw(struct r600_context *rctx, /* queries need some special values * (this is non-zero if any query is active) */ if (rctx->num_cs_dw_nontimer_queries_suspend) { - struct si_state_dsa *dsa = rctx->queued.named.dsa; - if (rctx->b.chip_class >= CIK) { si_pm4_set_reg(pm4, R_028004_DB_COUNT_CONTROL, S_028004_PERFECT_ZPASS_COUNTS(1) | @@ -588,9 +586,6 @@ static void si_state_draw(struct r600_context *rctx, S_028004_PERFECT_ZPASS_COUNTS(1) | S_028004_SAMPLE_RATE(rctx->fb_log_samples)); } - si_pm4_set_reg(pm4, R_02800C_DB_RENDER_OVERRIDE, - dsa->db_render_override | - S_02800C_NOOP_CULL_DISABLE(1)); } if (info->count_from_stream_output) { @@ -647,7 +642,9 @@ static void si_state_draw(struct r600_context *rctx, initiator |= S_0287F0_USE_OPAQUE(!!info->count_from_stream_output); si_cmd_draw_index_auto(pm4, info->count, initiator, rctx->predicate_drawing); } + si_pm4_set_state(rctx, draw, pm4); + si_update_db_draw_state(rctx, (struct r600_surface *)rctx->framebuffer.zsbuf); } void si_emit_cache_flush(struct r600_common_context *rctx, struct r600_atom *atom) -- 2.30.2