radeonsi: set FLUSH_ON_BINNING_TRANSITION when needed

Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Acked-by: Dave Airlie <airlied@redhat.com>

diff --git a/src/gallium/drivers/radeonsi/si_gfx_cs.c b/src/gallium/drivers/radeonsi/si_gfx_cs.c
index 277a25a0b3ee47304d528333544c67556f0887b6..843f8206149566014db6acc8a288112bdc37bb55 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_gfx_cs.c
+++ b/src/gallium/drivers/radeonsi/si_gfx_cs.c
@@ -449,6 +449,7 @@ void si_begin_new_gfx_cs(struct si_context *ctx)
        ctx->last_tes_sh_base = -1;
        ctx->last_num_tcs_input_cp = -1;
        ctx->last_ls_hs_config = -1; /* impossible value */
+       ctx->last_binning_enabled = -1;
 
        ctx->prim_discard_compute_ib_initialized = false;
 

diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index d04e7e155fce5552187f954ac611b3c7cc65b40d..cd8fb5d5df41a36142a32dcc96b2714f42a9581d 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -1061,6 +1061,7 @@ struct si_context {
        int                     last_multi_vgt_param;
        int                     last_rast_prim;
        int                     last_flatshade_first;
+       int                     last_binning_enabled;
        unsigned                last_sc_line_stipple;
        unsigned                current_vs_state;
        unsigned                last_vs_state;

diff --git a/src/gallium/drivers/radeonsi/si_state_binning.c b/src/gallium/drivers/radeonsi/si_state_binning.c
index 4d884316a612cfab0a2dd1453adf75655cf08c59..0720e269c8503dd2553e80ff2db77f257a5d699f 100644 (file)
--- a/src/gallium/drivers/radeonsi/si_state_binning.c
+++ b/src/gallium/drivers/radeonsi/si_state_binning.c
@@ -331,12 +331,17 @@ static void si_emit_dpbb_disable(struct si_context *sctx)
                        S_028C44_BIN_SIZE_Y(bin_size.y == 16) |
                        S_028C44_BIN_SIZE_X_EXTEND(bin_size_extend.x) |
                        S_028C44_BIN_SIZE_Y_EXTEND(bin_size_extend.y) |
-                       S_028C44_DISABLE_START_OF_PRIM(1));
+                       S_028C44_DISABLE_START_OF_PRIM(1) |
+                       S_028C44_FLUSH_ON_BINNING_TRANSITION(sctx->last_binning_enabled != 0));
        } else {
                radeon_opt_set_context_reg(sctx, R_028C44_PA_SC_BINNER_CNTL_0,
                        SI_TRACKED_PA_SC_BINNER_CNTL_0,
                        S_028C44_BINNING_MODE(V_028C44_DISABLE_BINNING_USE_LEGACY_SC) |
-                       S_028C44_DISABLE_START_OF_PRIM(1));
+                       S_028C44_DISABLE_START_OF_PRIM(1) |
+                       S_028C44_FLUSH_ON_BINNING_TRANSITION((sctx->family == CHIP_VEGA12 ||
+                                                             sctx->family == CHIP_VEGA20 ||
+                                                             sctx->family >= CHIP_RAVEN2) &&
+                                                            sctx->last_binning_enabled != 0));
        }
 
        unsigned db_dfsm_control = sctx->chip_class >= GFX10 ? R_028038_DB_DFSM_CONTROL
@@ -347,6 +352,8 @@ static void si_emit_dpbb_disable(struct si_context *sctx)
                                   S_028060_POPS_DRAIN_PS_ON_OVERLAP(1));
        if (initial_cdw != sctx->gfx_cs->current.cdw)
                sctx->context_roll = true;
+
+       sctx->last_binning_enabled = false;
 }
 
 void si_emit_dpbb_state(struct si_context *sctx)
@@ -452,7 +459,8 @@ void si_emit_dpbb_state(struct si_context *sctx)
                S_028C44_PERSISTENT_STATES_PER_BIN(persistent_states_per_bin) |
                S_028C44_DISABLE_START_OF_PRIM(disable_start_of_prim) |
                S_028C44_FPOVS_PER_BATCH(fpovs_per_batch) |
-               S_028C44_OPTIMAL_BIN_SELECTION(1));
+               S_028C44_OPTIMAL_BIN_SELECTION(1) |
+               G_028C44_FLUSH_ON_BINNING_TRANSITION(sctx->last_binning_enabled != 1));
 
        unsigned db_dfsm_control = sctx->chip_class >= GFX10 ? R_028038_DB_DFSM_CONTROL
                                                             : R_028060_DB_DFSM_CONTROL;
@@ -462,4 +470,6 @@ void si_emit_dpbb_state(struct si_context *sctx)
                                   S_028060_POPS_DRAIN_PS_ON_OVERLAP(1));
        if (initial_cdw != sctx->gfx_cs->current.cdw)
                sctx->context_roll = true;
+
+       sctx->last_binning_enabled = true;
 }