[Mesa-dev] [PATCH 4/4] r600g: copy IA_MULTI_VGT_PARAM programming from radeonsi for Cayman

[Marek Olšák]

From: Marek Olšák <marek.olsak at amd.com>

---
 src/gallium/drivers/r600/evergreen_state.c   |  2 --
 src/gallium/drivers/r600/r600_pipe.h         |  2 +-
 src/gallium/drivers/r600/r600_state_common.c | 24 ++++++++++++++++++++++++
 src/gallium/drivers/r600/r600d.h             | 11 +++++++++++
 4 files changed, 36 insertions(+), 3 deletions(-)

diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index e6e9f49..841ad0c 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -2243,8 +2243,6 @@ static void cayman_init_atom_start_cs(struct r600_context *rctx)
 
 	r600_store_config_reg(cb, R_008A14_PA_CL_ENHANCE, (3 << 1) | 1);
 
-	r600_store_context_reg(cb, CM_R_028AA8_IA_MULTI_VGT_PARAM, S_028AA8_SWITCH_ON_EOP(1) | S_028AA8_PARTIAL_VS_WAVE_ON(1) | S_028AA8_PRIMGROUP_SIZE(63));
-
 	r600_store_context_reg_seq(cb, CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0, 2);
 	r600_store_value(cb, 0x76543210); /* CM_R_028BD4_PA_SC_CENTROID_PRIORITY_0 */
 	r600_store_value(cb, 0xfedcba98); /* CM_R_028BD8_PA_SC_CENTROID_PRIORITY_1 */
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index ee836b7..e277269 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -40,7 +40,7 @@
 
 /* the number of CS dwords for flushing and drawing */
 #define R600_MAX_FLUSH_CS_DWORDS	16
-#define R600_MAX_DRAW_CS_DWORDS		37
+#define R600_MAX_DRAW_CS_DWORDS		40
 #define R600_TRACE_CS_DWORDS		7
 
 #define R600_MAX_USER_CONST_BUFFERS 13
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index d2f0d17..7594d0e 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1418,6 +1418,30 @@ static void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info
 		r600_emit_atom(rctx, rctx->atoms[i]);
 	}
 
+	if (rctx->b.chip_class == CAYMAN) {
+		/* Copied from radeonsi. */
+		unsigned primgroup_size = 128; /* recommended without a GS */
+		bool ia_switch_on_eop = false;
+		bool partial_vs_wave = false;
+
+		if (rctx->gs_shader)
+			primgroup_size = 64; /* recommended with a GS */
+
+		if ((rctx->rasterizer && rctx->rasterizer->pa_sc_line_stipple) ||
+		    (rctx->b.screen->debug_flags & DBG_SWITCH_ON_EOP)) {
+			ia_switch_on_eop = true;
+		}
+
+		if (rctx->b.streamout.streamout_enabled ||
+		    rctx->b.streamout.prims_gen_query_enabled)
+			partial_vs_wave = true;
+
+		r600_write_context_reg(cs, CM_R_028AA8_IA_MULTI_VGT_PARAM,
+				       S_028AA8_SWITCH_ON_EOP(ia_switch_on_eop) |
+				       S_028AA8_PARTIAL_VS_WAVE_ON(partial_vs_wave) |
+				       S_028AA8_PRIMGROUP_SIZE(primgroup_size - 1));
+	}
+
 	/* On R6xx, CULL_FRONT=1 culls all points, lines, and rectangles,
 	 * even though it should have no effect on those. */
 	if (rctx->b.chip_class == R600 && rctx->rasterizer) {
diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h
index 8405fbb..17568ab 100644
--- a/src/gallium/drivers/r600/r600d.h
+++ b/src/gallium/drivers/r600/r600d.h
@@ -3747,6 +3747,17 @@
 #define SQ_TEX_INST_SAMPLE_C_G_LB	0x1E
 #define SQ_TEX_INST_SAMPLE_C_G_LZ	0x1F
 
+#define CM_R_028AA8_IA_MULTI_VGT_PARAM                0x028AA8
+#define   S_028AA8_PRIMGROUP_SIZE(x)                   (((x) & 0xFFFF) << 0)
+#define   G_028AA8_PRIMGROUP_SIZE(x)                   (((x) >> 0) & 0xFFFF)
+#define   C_028AA8_PRIMGROUP_SIZE                      0xFFFF0000
+#define   S_028AA8_PARTIAL_VS_WAVE_ON(x)               (((x) & 0x1) << 16)
+#define   G_028AA8_PARTIAL_VS_WAVE_ON(x)               (((x) >> 16) & 0x1)
+#define   C_028AA8_PARTIAL_VS_WAVE_ON                  0xFFFEFFFF
+#define   S_028AA8_SWITCH_ON_EOP(x)                    (((x) & 0x1) << 17)
+#define   G_028AA8_SWITCH_ON_EOP(x)                    (((x) >> 17) & 0x1)
+#define   C_028AA8_SWITCH_ON_EOP                       0xFFFDFFFF
+
 /* async DMA packets */
 #define DMA_PACKET(cmd, t, s, n)	((((cmd) & 0xF) << 28) |	\
 					(((t) & 0x1) << 23) |		\
-- 
1.9.1