r600g: rework rasterizer discard for evergreen

Implement it right using STRMOUT_CONFIG.RAST_STREAM. This fixes rasterizer
discard with points and lines.

This also adds another derived state. It's a combination of rasterizer discard
and streamout enable.

Reviewed-by: Alex Deucher <[email protected]>
Reviewed-by: Christian König <[email protected]>

5 files changed, 46 insertions, 16 deletions

diff --git a/src/gallium/drivers/r600/evergreen_hw_context.c b/src/gallium/drivers/r600/evergreen_hw_context.c
index 352aeb8b4a1..161709cc233 100644
--- a/src/gallium/drivers/r600/evergreen_hw_context.c
+++ b/src/gallium/drivers/r600/evergreen_hw_context.c
@@ -829,18 +829,16 @@ void evergreen_flush_vgt_streamout(struct r600_context *ctx)
 void evergreen_set_streamout_enable(struct r600_context *ctx, unsigned buffer_enable_bit)
 {
 	struct radeon_winsys_cs *cs = ctx->cs;
+	bool enable = buffer_enable_bit != 0;
 
-	if (buffer_enable_bit) {
-		cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, 1, 0);
-		cs->buf[cs->cdw++] = (R_028B94_VGT_STRMOUT_CONFIG - EVERGREEN_CONTEXT_REG_OFFSET) >> 2;
-		cs->buf[cs->cdw++] = S_028B94_STREAMOUT_0_EN(1);
+	if (enable != ctx->atom_eg_strmout_config.stream0_enable) {
+		ctx->atom_eg_strmout_config.stream0_enable = enable;
+		r600_emit_atom(ctx, &ctx->atom_eg_strmout_config.atom);
+	}
 
+	if (buffer_enable_bit) {
 		cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, 1, 0);
 		cs->buf[cs->cdw++] = (R_028B98_VGT_STRMOUT_BUFFER_CONFIG - EVERGREEN_CONTEXT_REG_OFFSET) >> 2;
 		cs->buf[cs->cdw++] = S_028B98_STREAM_0_BUFFER_EN(buffer_enable_bit);
-	} else {
-		cs->buf[cs->cdw++] = PKT3(PKT3_SET_CONTEXT_REG, 1, 0);
-		cs->buf[cs->cdw++] = (R_028B94_VGT_STRMOUT_CONFIG - EVERGREEN_CONTEXT_REG_OFFSET) >> 2;
-		cs->buf[cs->cdw++] = S_028B94_STREAMOUT_0_EN(0);
 	}
 }
diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 65becc41b9f..adc400f1e79 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -781,6 +781,7 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx,
 		prov_vtx = 0;
 
 	rstate = &rs->rstate;
+	rs->rasterizer_discard = state->rasterizer_discard;
 	rs->flatshade = state->flatshade;
 	rs->sprite_coord_enable = state->sprite_coord_enable;
 	rs->two_side = state->light_twoside;
@@ -849,8 +850,8 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx,
 	r600_pipe_state_add_reg(rstate, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, fui(state->offset_clamp), NULL, 0);
 	r600_pipe_state_add_reg(rstate, R_028814_PA_SU_SC_MODE_CNTL,
 				S_028814_PROVOKING_VTX_LAST(prov_vtx) |
-				S_028814_CULL_FRONT(state->rasterizer_discard || (state->cull_face & PIPE_FACE_FRONT) ? 1 : 0) |
-				S_028814_CULL_BACK(state->rasterizer_discard || (state->cull_face & PIPE_FACE_BACK) ? 1 : 0) |
+				S_028814_CULL_FRONT(state->cull_face & PIPE_FACE_FRONT ? 1 : 0) |
+				S_028814_CULL_BACK(state->cull_face & PIPE_FACE_BACK ? 1 : 0) |
 				S_028814_FACE(!state->front_ccw) |
 				S_028814_POLY_OFFSET_FRONT_ENABLE(state->offset_tri) |
 				S_028814_POLY_OFFSET_BACK_ENABLE(state->offset_tri) |
@@ -862,6 +863,16 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx,
 	return rstate;
 }
 
+void evergreen_set_rasterizer_discard(struct pipe_context *ctx, boolean discard)
+{
+	struct r600_context *rctx = (struct r600_context*)ctx;
+
+	if (discard != rctx->atom_eg_strmout_config.rasterizer_discard) {
+		rctx->atom_eg_strmout_config.rasterizer_discard = discard;
+		r600_atom_dirty(rctx, &rctx->atom_eg_strmout_config.atom);
+	}
+}
+
 static void *evergreen_create_sampler_state(struct pipe_context *ctx,
 					const struct pipe_sampler_state *state)
 {
@@ -1676,9 +1687,22 @@ static void evergreen_emit_db_misc_state(struct r600_context *rctx, struct r600_
 	r600_write_context_reg(cs, R_02800C_DB_RENDER_OVERRIDE, db_render_override);
 }
 
+static void evergreen_emit_streamout_config(struct r600_context *rctx, struct r600_atom *atom)
+{
+	struct radeon_winsys_cs *cs = rctx->cs;
+	struct r600_atom_eg_strmout_config *a = (struct r600_atom_eg_strmout_config*)atom;
+
+	r600_write_context_reg(cs, R_028B94_VGT_STRMOUT_CONFIG,
+			       S_028B94_STREAMOUT_0_EN(a->stream0_enable) |
+			       S_028B94_RAST_STREAM(a->rasterizer_discard ? 4 : 0));
+}
+
 void evergreen_init_state_functions(struct r600_context *rctx)
 {
 	r600_init_atom(&rctx->atom_db_misc_state.atom, evergreen_emit_db_misc_state, 6, 0);
+	r600_atom_dirty(rctx, &rctx->atom_db_misc_state.atom);
+	r600_init_atom(&rctx->atom_eg_strmout_config.atom, evergreen_emit_streamout_config, 6, 0);
+	r600_atom_dirty(rctx, &rctx->atom_eg_strmout_config.atom);
 
 	rctx->context.create_blend_state = evergreen_create_blend_state;
 	rctx->context.create_depth_stencil_alpha_state = evergreen_create_dsa_state;
@@ -1763,9 +1787,7 @@ static void cayman_init_atom_start_cs(struct r600_context *rctx)
 	r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */
 	r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE */
 
-	r600_store_context_reg_seq(cb, R_028B94_VGT_STRMOUT_CONFIG, 2);
-	r600_store_value(cb, 0); /* R_028B94_VGT_STRMOUT_CONFIG */
-	r600_store_value(cb, 0); /* R_028B98_VGT_STRMOUT_BUFFER_CONFIG */
+	r600_store_context_reg(cb, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0);
 
 	r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2);
 	r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */
@@ -2258,9 +2280,7 @@ void evergreen_init_atom_start_cs(struct r600_context *rctx)
 	r600_store_value(cb, 0); /* R_028A3C_VGT_GROUP_VECT_1_FMT_CNTL */
 	r600_store_value(cb, 0); /* R_028A40_VGT_GS_MODE */
 
-	r600_store_context_reg_seq(cb, R_028B94_VGT_STRMOUT_CONFIG, 2);
-	r600_store_value(cb, 0); /* R_028B94_VGT_STRMOUT_CONFIG */
-	r600_store_value(cb, 0); /* R_028B98_VGT_STRMOUT_BUFFER_CONFIG */
+	r600_store_context_reg(cb, R_028B98_VGT_STRMOUT_BUFFER_CONFIG, 0);
 
 	r600_store_context_reg_seq(cb, R_028AB4_VGT_REUSE_OFF, 2);
 	r600_store_value(cb, 0); /* R_028AB4_VGT_REUSE_OFF */
diff --git a/src/gallium/drivers/r600/r600_hw_context.c b/src/gallium/drivers/r600/r600_hw_context.c
index 6d30aa3ca05..1dc23fdcca5 100644
--- a/src/gallium/drivers/r600/r600_hw_context.c
+++ b/src/gallium/drivers/r600/r600_hw_context.c
@@ -1270,6 +1270,8 @@ void r600_context_flush(struct r600_context *ctx, unsigned flags)
 
 	r600_emit_atom(ctx, &ctx->atom_start_cs.atom);
 	r600_atom_dirty(ctx, &ctx->atom_db_misc_state.atom);
+	if (ctx->chip_class >= EVERGREEN)
+		r600_atom_dirty(ctx, &ctx->atom_eg_strmout_config.atom);
 
 	if (streamout_suspended) {
 		ctx->streamout_start = TRUE;
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index 9550e9e0bae..de73bd8ee68 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -86,6 +86,12 @@ struct r600_atom_db_misc_state {
 	bool flush_depthstencil_enabled;
 };
 
+struct r600_atom_eg_strmout_config {
+	struct r600_atom atom;
+	bool rasterizer_discard;
+	bool stream0_enable;
+};
+
 enum r600_pipe_state_id {
 	R600_PIPE_STATE_BLEND = 0,
 	R600_PIPE_STATE_BLEND_COLOR,
@@ -153,6 +159,7 @@ struct r600_pipe_rasterizer {
 	float				offset_units;
 	float				offset_scale;
 	bool				scissor_enable;
+	bool				rasterizer_discard;
 };
 
 struct r600_pipe_blend {
@@ -294,6 +301,7 @@ struct r600_context {
 	struct r600_atom_surface_sync	atom_surface_sync;
 	struct r600_atom		atom_r6xx_flush_and_inv;
 	struct r600_atom_db_misc_state	atom_db_misc_state;
+	struct r600_atom_eg_strmout_config atom_eg_strmout_config;
 
 	/* Below are variables from the old r600_context.
 	 */
@@ -384,6 +392,7 @@ boolean evergreen_is_format_supported(struct pipe_screen *screen,
 				      enum pipe_texture_target target,
 				      unsigned sample_count,
 				      unsigned usage);
+void evergreen_set_rasterizer_discard(struct pipe_context *ctx, boolean discard);
 
 /* r600_blit.c */
 void r600_init_blit_functions(struct r600_context *rctx);
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index 565af35c7e5..2f643cf0acd 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -301,6 +301,7 @@ void r600_bind_rs_state(struct pipe_context *ctx, void *state)
 
 	if (rctx->chip_class >= EVERGREEN) {
 		evergreen_polygon_offset_update(rctx);
+		evergreen_set_rasterizer_discard(ctx, rs->rasterizer_discard);
 	} else {
 		r600_polygon_offset_update(rctx);
 	}