r600g: Implement GL_ARB_sample_shading

Also fixes two sided lighting which was broken at least
on pre-evergreen by commit b1eb00.

Signed-off-by: Glenn Kennard <glenn.kennard@gmail.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>

10 files changed, 383 insertions, 119 deletions

diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c
index 27a9ad9e396..78c1b680acc 100644
--- a/src/gallium/drivers/r600/evergreen_state.c
+++ b/src/gallium/drivers/r600/evergreen_state.c
@@ -1400,7 +1400,7 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx,
 
 	/* MSAA. */
 	if (rctx->b.chip_class == EVERGREEN)
-		rctx->framebuffer.atom.num_dw += 14; /* Evergreen */
+		rctx->framebuffer.atom.num_dw += 17; /* Evergreen */
 	else
 		rctx->framebuffer.atom.num_dw += 28; /* Cayman */
 
@@ -1420,8 +1420,22 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx,
 	}
 
 	rctx->framebuffer.atom.dirty = true;
+
+	r600_set_sample_locations_constant_buffer(rctx);
 }
 
+static void evergreen_set_min_samples(struct pipe_context *ctx, unsigned min_samples)
+{
+	struct r600_context *rctx = (struct r600_context *)ctx;
+
+	if (rctx->ps_iter_samples == min_samples)
+		return;
+
+	rctx->ps_iter_samples = min_samples;
+	if (rctx->framebuffer.nr_samples > 1) {
+		rctx->framebuffer.atom.dirty = true;
+	}
+}
 
 /* 8xMSAA */
 static uint32_t sample_locs_8x[] = {
@@ -1475,7 +1489,7 @@ static void evergreen_get_sample_position(struct pipe_context *ctx,
 	}
 }
 
-static void evergreen_emit_msaa_state(struct r600_context *rctx, int nr_samples)
+static void evergreen_emit_msaa_state(struct r600_context *rctx, int nr_samples, int ps_iter_samples)
 {
 
 	struct radeon_winsys_cs *cs = rctx->b.rings.gfx.cs;
@@ -1508,10 +1522,12 @@ static void evergreen_emit_msaa_state(struct r600_context *rctx, int nr_samples)
 				     S_028C00_EXPAND_LINE_WIDTH(1)); /* R_028C00_PA_SC_LINE_CNTL */
 		radeon_emit(cs, S_028C04_MSAA_NUM_SAMPLES(util_logbase2(nr_samples)) |
 				     S_028C04_MAX_SAMPLE_DIST(max_dist)); /* R_028C04_PA_SC_AA_CONFIG */
+		r600_write_context_reg(cs, EG_R_028A4C_PA_SC_MODE_CNTL_1, EG_S_028A4C_PS_ITER_SAMPLE(ps_iter_samples > 1));
 	} else {
 		r600_write_context_reg_seq(cs, R_028C00_PA_SC_LINE_CNTL, 2);
 		radeon_emit(cs, S_028C00_LAST_PIXEL(1)); /* R_028C00_PA_SC_LINE_CNTL */
 		radeon_emit(cs, 0); /* R_028C04_PA_SC_AA_CONFIG */
+		r600_write_context_reg(cs, EG_R_028A4C_PA_SC_MODE_CNTL_1, 0);
 	}
 }
 
@@ -1672,10 +1688,10 @@ static void evergreen_emit_framebuffer_state(struct r600_context *rctx, struct r
 	radeon_emit(cs, br); /* R_028208_PA_SC_WINDOW_SCISSOR_BR */
 
 	if (rctx->b.chip_class == EVERGREEN) {
-		evergreen_emit_msaa_state(rctx, rctx->framebuffer.nr_samples);
+		evergreen_emit_msaa_state(rctx, rctx->framebuffer.nr_samples, rctx->ps_iter_samples);
 	} else {
 		cayman_emit_msaa_sample_locs(cs, rctx->framebuffer.nr_samples);
-		cayman_emit_msaa_config(cs, rctx->framebuffer.nr_samples, 1);
+		cayman_emit_msaa_config(cs, rctx->framebuffer.nr_samples, rctx->ps_iter_samples);
 	}
 }
 
@@ -2432,8 +2448,6 @@ void evergreen_init_common_regs(struct r600_command_buffer *cb,
 		r600_store_value(cb, tmp); /* R_008C0C_SQ_GPR_RESOURCE_MGMT_3 */
 	}
 
-	r600_store_context_reg(cb, R_028A4C_PA_SC_MODE_CNTL_1, 0);
-
 	/* The cs checker requires this register to be set. */
 	r600_store_context_reg(cb, R_028800_DB_DEPTH_CONTROL, 0);
 
@@ -2786,11 +2800,19 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 	struct r600_command_buffer *cb = &shader->command_buffer;
 	struct r600_shader *rshader = &shader->shader;
 	unsigned i, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z, spi_ps_in_control_1, db_shader_control = 0;
-	int pos_index = -1, face_index = -1;
+	int pos_index = -1, face_index = -1, fixed_pt_position_index = -1;
 	int ninterp = 0;
-	boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE;
-	unsigned spi_baryc_cntl, sid, tmp, num = 0;
-	unsigned z_export = 0, stencil_export = 0;
+	boolean have_perspective = FALSE, have_linear = FALSE;
+	static const unsigned spi_baryc_enable_bit[6] = {
+		S_0286E0_PERSP_SAMPLE_ENA(1),
+		S_0286E0_PERSP_CENTER_ENA(1),
+		S_0286E0_PERSP_CENTROID_ENA(1),
+		S_0286E0_LINEAR_SAMPLE_ENA(1),
+		S_0286E0_LINEAR_CENTER_ENA(1),
+		S_0286E0_LINEAR_CENTROID_ENA(1)
+	};
+	unsigned spi_baryc_cntl = 0, sid, tmp, num = 0;
+	unsigned z_export = 0, stencil_export = 0, mask_export = 0;
 	unsigned sprite_coord_enable = rctx->rasterizer ? rctx->rasterizer->sprite_coord_enable : 0;
 	uint32_t spi_ps_input_cntl[32];
 
@@ -2813,14 +2835,19 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 			if (face_index == -1)
 				face_index = i; /* lives in same register, same enable bit */
 		}
+		else if (rshader->input[i].name == TGSI_SEMANTIC_SAMPLEID) {
+			fixed_pt_position_index = i;
+		}
 		else {
 			ninterp++;
-			if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR)
-				have_linear = TRUE;
-			if (rshader->input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
-				have_perspective = TRUE;
-			if (rshader->input[i].centroid)
-				have_centroid = TRUE;
+			int k = eg_get_interpolator_index(
+				rshader->input[i].interpolate,
+				rshader->input[i].interpolate_location);
+			if (k >= 0) {
+				spi_baryc_cntl |= spi_baryc_enable_bit[k];
+				have_perspective |= k < 3;
+				have_linear |= !(k < 3);
+			}
 		}
 
 		sid = rshader->input[i].spi_sid;
@@ -2852,17 +2879,22 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 			z_export = 1;
 		if (rshader->output[i].name == TGSI_SEMANTIC_STENCIL)
 			stencil_export = 1;
+		if (rshader->output[i].name == TGSI_SEMANTIC_SAMPLEMASK &&
+			rctx->framebuffer.nr_samples > 1 && rctx->ps_iter_samples > 0)
+			mask_export = 1;
 	}
 	if (rshader->uses_kill)
 		db_shader_control |= S_02880C_KILL_ENABLE(1);
 
 	db_shader_control |= S_02880C_Z_EXPORT_ENABLE(z_export);
 	db_shader_control |= S_02880C_STENCIL_EXPORT_ENABLE(stencil_export);
+	db_shader_control |= S_02880C_MASK_EXPORT_ENABLE(mask_export);
 
 	exports_ps = 0;
 	for (i = 0; i < rshader->noutput; i++) {
 		if (rshader->output[i].name == TGSI_SEMANTIC_POSITION ||
-		    rshader->output[i].name == TGSI_SEMANTIC_STENCIL)
+		    rshader->output[i].name == TGSI_SEMANTIC_STENCIL ||
+		    rshader->output[i].name == TGSI_SEMANTIC_SAMPLEMASK)
 			exports_ps |= 1;
 	}
 
@@ -2878,6 +2910,8 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 		ninterp = 1;
 		have_perspective = TRUE;
 	}
+	if (!spi_baryc_cntl)
+		spi_baryc_cntl |= spi_baryc_enable_bit[0];
 
 	if (!have_perspective && !have_linear)
 		have_perspective = TRUE;
@@ -2888,7 +2922,7 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 	spi_input_z = 0;
 	if (pos_index != -1) {
 		spi_ps_in_control_0 |=  S_0286CC_POSITION_ENA(1) |
-			S_0286CC_POSITION_CENTROID(rshader->input[pos_index].centroid) |
+			S_0286CC_POSITION_CENTROID(rshader->input[pos_index].interpolate_location == TGSI_INTERPOLATE_LOC_CENTROID) |
 			S_0286CC_POSITION_ADDR(rshader->input[pos_index].gpr);
 		spi_input_z |= S_0286D8_PROVIDE_Z_TO_SPI(1);
 	}
@@ -2898,14 +2932,10 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 		spi_ps_in_control_1 |= S_0286D0_FRONT_FACE_ENA(1) |
 			S_0286D0_FRONT_FACE_ADDR(rshader->input[face_index].gpr);
 	}
-
-	spi_baryc_cntl = 0;
-	if (have_perspective)
-		spi_baryc_cntl |= S_0286E0_PERSP_CENTER_ENA(1) |
-				  S_0286E0_PERSP_CENTROID_ENA(have_centroid);
-	if (have_linear)
-		spi_baryc_cntl |= S_0286E0_LINEAR_CENTER_ENA(1) |
-				  S_0286E0_LINEAR_CENTROID_ENA(have_centroid);
+	if (fixed_pt_position_index != -1) {
+		spi_ps_in_control_1 |= S_0286D0_FIXED_PT_POSITION_ENA(1) |
+			S_0286D0_FIXED_PT_POSITION_ADDR(rshader->input[fixed_pt_position_index].gpr);
+	}
 
 	r600_store_context_reg_seq(cb, R_0286CC_SPI_PS_IN_CONTROL_0, 2);
 	r600_store_value(cb, spi_ps_in_control_0); /* R_0286CC_SPI_PS_IN_CONTROL_0 */
@@ -2924,7 +2954,7 @@ void evergreen_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader
 	/* After that, the NOP relocation packet must be emitted (shader->bo, RADEON_USAGE_READ). */
 
 	shader->db_shader_control = db_shader_control;
-	shader->ps_depth_export = z_export | stencil_export;
+	shader->ps_depth_export = z_export | stencil_export | mask_export;
 
 	shader->sprite_coord_enable = sprite_coord_enable;
 	if (rctx->rasterizer)
@@ -3446,6 +3476,7 @@ void evergreen_init_state_functions(struct r600_context *rctx)
 	rctx->b.b.create_sampler_view = evergreen_create_sampler_view;
 	rctx->b.b.set_framebuffer_state = evergreen_set_framebuffer_state;
 	rctx->b.b.set_polygon_stipple = evergreen_set_polygon_stipple;
+	rctx->b.b.set_min_samples = evergreen_set_min_samples;
 	rctx->b.b.set_scissor_states = evergreen_set_scissor_states;
 
 	if (rctx->b.chip_class == EVERGREEN)
diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h
index 784d495a40f..49899960367 100644
--- a/src/gallium/drivers/r600/evergreend.h
+++ b/src/gallium/drivers/r600/evergreend.h
@@ -803,6 +803,9 @@
 #define   S_02880C_KILL_ENABLE(x)                      (((x) & 0x1) << 6)
 #define   G_02880C_KILL_ENABLE(x)                      (((x) >> 6) & 0x1)
 #define   C_02880C_KILL_ENABLE                         0xFFFFFFBF
+#define   S_02880C_MASK_EXPORT_ENABLE(x)               (((x) & 0x1) << 8)
+#define   G_02880C_MASK_EXPORT_ENABLE(x)               (((x) >> 8) & 0x1)
+#define   C_02880C_MASK_EXPORT_ENABLE                  0XFFFFFEFF
 #define   S_02880C_DUAL_EXPORT_ENABLE(x)               (((x) & 0x1) << 9)
 #define   G_02880C_DUAL_EXPORT_ENABLE(x)               (((x) >> 9) & 0x1)
 #define   C_02880C_DUAL_EXPORT_ENABLE                  0xFFFFFDFF
diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c
index c6459d81209..3962fee4a43 100644
--- a/src/gallium/drivers/r600/r600_pipe.c
+++ b/src/gallium/drivers/r600/r600_pipe.c
@@ -265,6 +265,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 	case PIPE_CAP_BUFFER_MAP_PERSISTENT_COHERENT:
 	case PIPE_CAP_TGSI_VS_WINDOW_SPACE_POSITION:
 	case PIPE_CAP_TGSI_VS_LAYER_VIEWPORT:
+	case PIPE_CAP_SAMPLE_SHADING:
 		return 1;
 
 	case PIPE_CAP_COMPUTE:
@@ -319,7 +320,6 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 	case PIPE_CAP_FRAGMENT_COLOR_CLAMPED:
 	case PIPE_CAP_VERTEX_COLOR_CLAMPED:
 	case PIPE_CAP_USER_VERTEX_BUFFERS:
-	case PIPE_CAP_SAMPLE_SHADING:
 	case PIPE_CAP_TEXTURE_GATHER_OFFSETS:
 	case PIPE_CAP_DRAW_INDIRECT:
 	case PIPE_CAP_CONDITIONAL_RENDER_INVERTED:
diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h
index 2df168f01b5..fa9d34b0d71 100644
--- a/src/gallium/drivers/r600/r600_pipe.h
+++ b/src/gallium/drivers/r600/r600_pipe.h
@@ -52,6 +52,14 @@
 #define R600_TXQ_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
 #define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
 #define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 3)
+/* Currently R600_MAX_CONST_BUFFERS is too large, the hardware only has 16 buffers, but the driver is
+ * trying to use 17. Avoid accidentally aliasing with user UBOs for SAMPLE_POSITIONS by using an id<16.
+ * UCP/SAMPLE_POSITIONS are never accessed by same shader stage so they can use the same id.
+ *
+ * Fixing this properly would require the driver to combine its buffers into a single hardware buffer,
+ * which would also allow supporting the d3d 11 mandated minimum of 15 user const buffers.
+ */
+#define R600_SAMPLE_POSITIONS_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
 
 #define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
 
@@ -452,6 +460,7 @@ struct r600_context {
 	bool				force_blend_disable;
 	boolean				dual_src_blend;
 	unsigned			zwritemask;
+	int					ps_iter_samples;
 
 	/* Index buffer. */
 	struct pipe_index_buffer	index_buffer;
@@ -639,6 +648,7 @@ void r600_sampler_views_dirty(struct r600_context *rctx,
 void r600_sampler_states_dirty(struct r600_context *rctx,
 			       struct r600_sampler_states *state);
 void r600_constant_buffers_dirty(struct r600_context *rctx, struct r600_constbuf_state *state);
+void r600_set_sample_locations_constant_buffer(struct r600_context *rctx);
 uint32_t r600_translate_stencil_op(int s_op);
 uint32_t r600_translate_fill(uint32_t func);
 unsigned r600_tex_wrap(unsigned wrap);
diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c
index 9f10c20c4c3..9e9a557e867 100644
--- a/src/gallium/drivers/r600/r600_shader.c
+++ b/src/gallium/drivers/r600/r600_shader.c
@@ -64,6 +64,7 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 				 struct r600_pipe_shader *pipeshader,
 				 struct r600_shader_key key);
 
+
 static void r600_add_gpr_array(struct r600_shader *ps, int start_gpr,
                            int size, unsigned comp_mask) {
 
@@ -267,6 +268,11 @@ struct r600_shader_src {
 	uint32_t				value[4];
 };
 
+struct eg_interp {
+	boolean					enabled;
+	unsigned				ij_index;
+};
+
 struct r600_shader_ctx {
 	struct tgsi_shader_info			info;
 	struct tgsi_parse_context		parse;
@@ -283,13 +289,11 @@ struct r600_shader_ctx {
 	uint32_t				max_driver_temp_used;
 	boolean use_llvm;
 	/* needed for evergreen interpolation */
-	boolean                                 input_centroid;
-	boolean                                 input_linear;
-	boolean                                 input_perspective;
-	int					num_interp_gpr;
+	struct eg_interp		eg_interpolators[6]; // indexed by Persp/Linear * 3 + sample/center/centroid
 	/* evergreen/cayman also store sample mask in face register */
 	int					face_gpr;
-	boolean					has_samplemask;
+	/* sample id is .w component stored in fixed point position register */
+	int					fixed_pt_position_gpr;
 	int					colors_used;
 	boolean                 clip_vertex_write;
 	unsigned                cv_output;
@@ -320,6 +324,12 @@ static int tgsi_endif(struct r600_shader_ctx *ctx);
 static int tgsi_bgnloop(struct r600_shader_ctx *ctx);
 static int tgsi_endloop(struct r600_shader_ctx *ctx);
 static int tgsi_loop_brk_cont(struct r600_shader_ctx *ctx);
+static int tgsi_fetch_rel_const(struct r600_shader_ctx *ctx,
+                                unsigned int cb_idx, unsigned int offset, unsigned ar_chan,
+                                unsigned int dst_reg);
+static void r600_bytecode_src(struct r600_bytecode_alu_src *bc_src,
+			const struct r600_shader_src *shader_src,
+			unsigned chan);
 
 static int tgsi_is_supported(struct r600_shader_ctx *ctx)
 {
@@ -364,27 +374,41 @@ static int tgsi_is_supported(struct r600_shader_ctx *ctx)
 	return 0;
 }
 
-static void evergreen_interp_assign_ij_index(struct r600_shader_ctx *ctx,
-		int input)
+int eg_get_interpolator_index(unsigned interpolate, unsigned location)
 {
-	int ij_index = 0;
+	if (interpolate == TGSI_INTERPOLATE_COLOR ||
+		interpolate == TGSI_INTERPOLATE_LINEAR ||
+		interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
+	{
+		int is_linear = interpolate == TGSI_INTERPOLATE_LINEAR;
+		int loc;
 
-	if (ctx->shader->input[input].interpolate == TGSI_INTERPOLATE_PERSPECTIVE) {
-		if (ctx->shader->input[input].centroid)
-			ij_index++;
-	} else if (ctx->shader->input[input].interpolate == TGSI_INTERPOLATE_LINEAR) {
-		/* if we have perspective add one */
-		if (ctx->input_perspective)  {
-			ij_index++;
-			/* if we have perspective centroid */
-			if (ctx->input_centroid)
-				ij_index++;
+		switch(location) {
+		case TGSI_INTERPOLATE_LOC_CENTER:
+			loc = 1;
+			break;
+		case TGSI_INTERPOLATE_LOC_CENTROID:
+			loc = 2;
+			break;
+		case TGSI_INTERPOLATE_LOC_SAMPLE:
+		default:
+			loc = 0; break;
 		}
-		if (ctx->shader->input[input].centroid)
-			ij_index++;
+
+		return is_linear * 3 + loc;
 	}
 
-	ctx->shader->input[input].ij_index = ij_index;
+	return -1;
+}
+
+static void evergreen_interp_assign_ij_index(struct r600_shader_ctx *ctx,
+		int input)
+{
+	int i = eg_get_interpolator_index(
+		ctx->shader->input[input].interpolate,
+		ctx->shader->input[input].interpolate_location);
+	assert(i >= 0);
+	ctx->shader->input[input].ij_index = ctx->eg_interpolators[i].ij_index;
 }
 
 static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int input)
@@ -582,13 +606,15 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx)
 		ctx->shader->input[i].name = d->Semantic.Name;
 		ctx->shader->input[i].sid = d->Semantic.Index;
 		ctx->shader->input[i].interpolate = d->Interp.Interpolate;
-		ctx->shader->input[i].centroid = d->Interp.Location == TGSI_INTERPOLATE_LOC_CENTROID;
+		ctx->shader->input[i].interpolate_location = d->Interp.Location;
 		ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + d->Range.First;
 		if (ctx->type == TGSI_PROCESSOR_FRAGMENT) {
 			ctx->shader->input[i].spi_sid = r600_spi_sid(&ctx->shader->input[i]);
 			switch (ctx->shader->input[i].name) {
 			case TGSI_SEMANTIC_FACE:
-				if (ctx->face_gpr == -1)
+				if (ctx->face_gpr != -1)
+					ctx->shader->input[i].gpr = ctx->face_gpr; /* already allocated by allocate_system_value_inputs */
+				else
 					ctx->face_gpr = ctx->shader->input[i].gpr;
 				break;
 			case TGSI_SEMANTIC_COLOR:
@@ -679,14 +705,11 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx)
 		break;
 
 	case TGSI_FILE_SYSTEM_VALUE:
-		if (d->Semantic.Name == TGSI_SEMANTIC_SAMPLEMASK) {
-			ctx->has_samplemask = true;
-			/* lives in Front Face GPR */
-			if (ctx->face_gpr == -1)
-				ctx->face_gpr = ctx->file_offset[TGSI_FILE_SYSTEM_VALUE] + d->Range.First;
-			break;
-		}
-		else if (d->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) {
+		if (d->Semantic.Name == TGSI_SEMANTIC_SAMPLEMASK ||
+			d->Semantic.Name == TGSI_SEMANTIC_SAMPLEID ||
+			d->Semantic.Name == TGSI_SEMANTIC_SAMPLEPOS) {
+			break; /* Already handled from allocate_system_value_inputs */
+		} else if (d->Semantic.Name == TGSI_SEMANTIC_INSTANCEID) {
 			if (!ctx->native_integers) {
 				struct r600_bytecode_alu alu;
 				memset(&alu, 0, sizeof(struct r600_bytecode_alu));
@@ -720,12 +743,69 @@ static int r600_get_temp(struct r600_shader_ctx *ctx)
 	return ctx->temp_reg + ctx->max_driver_temp_used++;
 }
 
+static int allocate_system_value_inputs(struct r600_shader_ctx *ctx, int gpr_offset)
+{
+	struct tgsi_parse_context parse;
+	struct {
+		boolean enabled;
+		int *reg;
+		unsigned name, alternate_name;
+	} inputs[2] = {
+		{ false, &ctx->face_gpr, TGSI_SEMANTIC_SAMPLEMASK, ~0u }, /* lives in Front Face GPR.z */
+
+		{ false, &ctx->fixed_pt_position_gpr, TGSI_SEMANTIC_SAMPLEID, TGSI_SEMANTIC_SAMPLEPOS } /* SAMPLEID is in Fixed Point Position GPR.w */
+	};
+	int i, k, num_regs = 0;
+
+	if (tgsi_parse_init(&parse, ctx->tokens) != TGSI_PARSE_OK) {
+		return 0;
+	}
+
+	while (!tgsi_parse_end_of_tokens(&parse)) {
+		tgsi_parse_token(&parse);
+
+		if (parse.FullToken.Token.Type == TGSI_TOKEN_TYPE_DECLARATION) {
+			struct tgsi_full_declaration *d = &parse.FullToken.FullDeclaration;
+			if (d->Declaration.File == TGSI_FILE_SYSTEM_VALUE) {
+				for (k = 0; k < Elements(inputs); k++) {
+					if (d->Semantic.Name == inputs[k].name ||
+						d->Semantic.Name == inputs[k].alternate_name) {
+						inputs[k].enabled = true;
+					}
+				}
+			}
+		}
+	}
+
+	tgsi_parse_free(&parse);
+
+	for (i = 0; i < Elements(inputs); i++) {
+		boolean enabled = inputs[i].enabled;
+		int *reg = inputs[i].reg;
+		unsigned name = inputs[i].name;
+
+		if (enabled) {
+			int gpr = gpr_offset + num_regs++;
+
+			// add to inputs, allocate a gpr
+			k = ctx->shader->ninput ++;
+			ctx->shader->input[k].name = name;
+			ctx->shader->input[k].sid = 0;
+			ctx->shader->input[k].interpolate = TGSI_INTERPOLATE_CONSTANT;
+			ctx->shader->input[k].interpolate_location = TGSI_INTERPOLATE_LOC_CENTER;
+			*reg = ctx->shader->input[k].gpr = gpr;
+		}
+	}
+
+	return gpr_offset + num_regs;
+}
+
 /*
  * for evergreen we need to scan the shader to find the number of GPRs we need to
- * reserve for interpolation.
+ * reserve for interpolation and system values
  *
  * we need to know if we are going to emit
- * any centroid inputs
+ * any sample or centroid inputs
  * if perspective and linear are required
 */
 static int evergreen_gpr_count(struct r600_shader_ctx *ctx)
@@ -733,39 +813,92 @@ static int evergreen_gpr_count(struct r600_shader_ctx *ctx)
 	int i;
 	int num_baryc;
 
-	ctx->input_linear = FALSE;
-	ctx->input_perspective = FALSE;
-	ctx->input_centroid = FALSE;
-	ctx->num_interp_gpr = 1;
+	memset(&ctx->eg_interpolators, 0, sizeof(ctx->eg_interpolators));
 
-	/* any centroid inputs */
 	for (i = 0; i < ctx->info.num_inputs; i++) {
-		/* skip position/face */
+		int k;
+		/* skip position/face/mask/sampleid */
 		if (ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_POSITION ||
 		    ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_FACE ||
-		    ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_SAMPLEMASK)
+		    ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_SAMPLEMASK ||
+		    ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_SAMPLEID)
 			continue;
-		if (ctx->info.input_interpolate[i] == TGSI_INTERPOLATE_LINEAR)
-			ctx->input_linear = TRUE;
-		if (ctx->info.input_interpolate[i] == TGSI_INTERPOLATE_PERSPECTIVE)
-			ctx->input_perspective = TRUE;
-		if (ctx->info.input_interpolate_loc[i] == TGSI_INTERPOLATE_LOC_CENTROID)
-			ctx->input_centroid = TRUE;
+
+		k = eg_get_interpolator_index(
+			ctx->info.input_interpolate[i],
+			ctx->info.input_interpolate_loc[i]);
+		if (k >= 0)
+			ctx->eg_interpolators[k].enabled = TRUE;
 	}
 
+	/* assign gpr to each interpolator according to priority */
 	num_baryc = 0;
-	/* ignoring sample for now */
-	if (ctx->input_perspective)
-		num_baryc++;
-	if (ctx->input_linear)
-		num_baryc++;
-	if (ctx->input_centroid)
-		num_baryc *= 2;
-
-	ctx->num_interp_gpr += (num_baryc + 1) >> 1;
-
-	/* XXX PULL MODEL and LINE STIPPLE, FIXED PT POS */
-	return ctx->num_interp_gpr;
+	for (i = 0; i < Elements(ctx->eg_interpolators); i++) {
+		if (ctx->eg_interpolators[i].enabled) {
+			ctx->eg_interpolators[i].ij_index = num_baryc;
+			num_baryc ++;
+		}
+	}
+
+	/* XXX PULL MODEL and LINE STIPPLE */
+
+	num_baryc = (num_baryc + 1) >> 1;
+	return allocate_system_value_inputs(ctx, num_baryc);
+}
+
+/* sample_id_sel == NULL means fetch for current sample */
+static int load_sample_position(struct r600_shader_ctx *ctx, struct r600_shader_src *sample_id, int chan_sel)
+{
+	struct r600_bytecode_vtx vtx;
+	int r, t1;
+
+	assert(ctx->fixed_pt_position_gpr != -1);
+
+	t1 = r600_get_temp(ctx);
+
+	memset(&vtx, 0, sizeof(struct r600_bytecode_vtx));
+	vtx.op = FETCH_OP_VFETCH;
+	vtx.buffer_id = R600_SAMPLE_POSITIONS_CONST_BUFFER;
+	vtx.fetch_type = 2;	/* VTX_FETCH_NO_INDEX_OFFSET */
+	if (sample_id == NULL) {
+		vtx.src_gpr = ctx->fixed_pt_position_gpr; // SAMPLEID is in .w;
+		vtx.src_sel_x = 3;
+	}
+	else {
+		struct r600_bytecode_alu alu;
+
+		memset(&alu, 0, sizeof(struct r600_bytecode_alu));
+		alu.op = ALU_OP1_MOV;
+		r600_bytecode_src(&alu.src[0], sample_id, chan_sel);
+		alu.dst.sel = t1;
+		alu.dst.write = 1;
+		alu.last = 1;
+		r = r600_bytecode_add_alu(ctx->bc, &alu);
+		if (r)
+			return r;
+
+		vtx.src_gpr = t1;
+		vtx.src_sel_x = 0;
+	}
+	vtx.mega_fetch_count = 16;
+	vtx.dst_gpr = t1;
+	vtx.dst_sel_x = 0;
+	vtx.dst_sel_y = 1;
+	vtx.dst_sel_z = 7;
+	vtx.dst_sel_w = 7;
+	vtx.data_format = FMT_32_32_32_32_FLOAT;
+	vtx.num_format_all = 2;
+	vtx.format_comp_all = 1;
+	vtx.use_const_fields = 0;
+	vtx.offset = 1; // first element is size of buffer
+	vtx.endian = r600_endian_swap(32);
+	vtx.srf_mode_all = 1; /* SRF_MODE_NO_ZERO */
+
+	r = r600_bytecode_add_vtx(ctx->bc, &vtx);
+	if (r)
+		return r;
+
+	return t1;
 }
 
 static void tgsi_src(struct r600_shader_ctx *ctx,
@@ -797,10 +930,22 @@ static void tgsi_src(struct r600_shader_ctx *ctx,
 	} else if (tgsi_src->Register.File == TGSI_FILE_SYSTEM_VALUE) {
 		if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_SAMPLEMASK) {
 			r600_src->swizzle[0] = 2; // Z value
-			r600_src->swizzle[0] = 2;
-			r600_src->swizzle[0] = 2;
-			r600_src->swizzle[0] = 2;
+			r600_src->swizzle[1] = 2;
+			r600_src->swizzle[2] = 2;
+			r600_src->swizzle[3] = 2;
 			r600_src->sel = ctx->face_gpr;
+		} else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_SAMPLEID) {
+			r600_src->swizzle[0] = 3; // W value
+			r600_src->swizzle[1] = 3;
+			r600_src->swizzle[2] = 3;
+			r600_src->swizzle[3] = 3;
+			r600_src->sel = ctx->fixed_pt_position_gpr;
+		} else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_SAMPLEPOS) {
+			r600_src->swizzle[0] = 0;
+			r600_src->swizzle[1] = 1;
+			r600_src->swizzle[2] = 4;
+			r600_src->swizzle[3] = 4;
+			r600_src->sel = load_sample_position(ctx, NULL, -1);
 		} else if (ctx->info.system_value_semantic_name[tgsi_src->Register.Index] == TGSI_SEMANTIC_INSTANCEID) {
 			r600_src->swizzle[0] = 3;
 			r600_src->swizzle[1] = 3;
@@ -1612,7 +1757,7 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 	ctx.gs_next_vertex = 0;
 
 	ctx.face_gpr = -1;
-	ctx.has_samplemask = false;
+	ctx.fixed_pt_position_gpr = -1;
 	ctx.fragcoord_input = -1;
 	ctx.colors_used = 0;
 	ctx.clip_vertex_write = 0;
@@ -1661,8 +1806,11 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 			r600_bytecode_add_cfinst(ctx.bc, CF_OP_CALL_FS);
 		}
 	}
-	if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chip_class >= EVERGREEN) {
-		ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx);
+	if (ctx.type == TGSI_PROCESSOR_FRAGMENT) {
+		if (ctx.bc->chip_class >= EVERGREEN)
+			ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx);
+		else
+			ctx.file_offset[TGSI_FILE_INPUT] = allocate_system_value_inputs(&ctx, ctx.file_offset[TGSI_FILE_INPUT]);
 	}
 	if (ctx.type == TGSI_PROCESSOR_GEOMETRY) {
 		/* FIXME 1 would be enough in some cases (3 or less input vertices) */
@@ -1775,14 +1923,6 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 	
 	shader->ring_item_size = ctx.next_ring_offset;
 
-	/* Need to tell setup to program FACE register */
-	if (ctx.has_samplemask && ctx.face_gpr != -1) {
-		i = ctx.shader->ninput++;
-		ctx.shader->input[i].name = TGSI_SEMANTIC_SAMPLEMASK;
-		ctx.shader->input[i].spi_sid = 0;
-		ctx.shader->input[i].gpr = ctx.face_gpr;
-	}
-
 	/* Process two side if needed */
 	if (shader->two_side && ctx.colors_used) {
 		int i, count = ctx.shader->ninput;
@@ -1795,6 +1935,7 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 		int gpr = ctx.file_offset[TGSI_FILE_INPUT] +
 				ctx.info.file_max[TGSI_FILE_INPUT] + 1;
 
+		/* if two sided and neither face or sample mask is used by shader, ensure face_gpr is emitted */
 		if (ctx.face_gpr == -1) {
 			i = ctx.shader->ninput++;
 			ctx.shader->input[i].name = TGSI_SEMANTIC_FACE;
@@ -2162,6 +2303,13 @@ static int r600_shader_from_tgsi(struct r600_context *rctx,
 					output[j].swizzle_y = 1;
 					output[j].swizzle_z = output[j].swizzle_w = 7;
 					output[j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
+				} else if (shader->output[i].name == TGSI_SEMANTIC_SAMPLEMASK) {
+					output[j].array_base = 61;
+					output[j].swizzle_x = 7;
+					output[j].swizzle_y = 7;
+					output[j].swizzle_z = 0;
+					output[j].swizzle_w = 7;
+					output[j].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL;
 				} else {
 					R600_ERR("unsupported fragment output name %d\n", shader->output[i].name);
 					r = -EINVAL;
diff --git a/src/gallium/drivers/r600/r600_shader.h b/src/gallium/drivers/r600/r600_shader.h
index 4b27ede96dd..20829fd9fdb 100644
--- a/src/gallium/drivers/r600/r600_shader.h
+++ b/src/gallium/drivers/r600/r600_shader.h
@@ -33,7 +33,7 @@ struct r600_shader_io {
 	int			spi_sid;
 	unsigned		interpolate;
 	unsigned		ij_index;
-	boolean                 centroid;
+	unsigned        interpolate_location; //  TGSI_INTERPOLATE_LOC_CENTER, CENTROID, SAMPLE
 	unsigned		lds_pos; /* for evergreen */
 	unsigned		back_color_input;
 	unsigned		write_mask;
@@ -115,4 +115,8 @@ struct r600_pipe_shader {
 	unsigned		ps_depth_export;
 };
 
+/* return the table index 0-5 for TGSI_INTERPOLATE_LINEAR/PERSPECTIVE and
+ TGSI_INTERPOLATE_LOC_CENTER/SAMPLE/COUNT. Other input values return -1. */
+int eg_get_interpolator_index(unsigned interpolate, unsigned location);
+
 #endif
diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c
index 9ca61718aba..1f933efca9e 100644
--- a/src/gallium/drivers/r600/r600_state.c
+++ b/src/gallium/drivers/r600/r600_state.c
@@ -486,7 +486,12 @@ static void *r600_create_rs_state(struct pipe_context *ctx,
 
 	sc_mode_cntl = S_028A4C_MSAA_ENABLE(state->multisample) |
 		       S_028A4C_LINE_STIPPLE_ENABLE(state->line_stipple_enable) |
-		       S_028A4C_FORCE_EOV_CNTDWN_ENABLE(1);
+		       S_028A4C_FORCE_EOV_CNTDWN_ENABLE(1) |
+		       S_028A4C_PS_ITER_SAMPLE(state->multisample && rctx->ps_iter_samples > 1);
+	if (rctx->b.family == CHIP_RV770) {
+		/* workaround possible rendering corruption on RV770 with hyperz together with sample shading */
+		sc_mode_cntl |= S_028A4C_TILE_COVER_DISABLE(state->multisample && rctx->ps_iter_samples > 1);
+	}
 	if (rctx->b.chip_class >= R700) {
 		sc_mode_cntl |= S_028A4C_FORCE_EOV_REZ_ENABLE(1) |
 				S_028A4C_R700_ZMM_LINE_OFFSET(1) |
@@ -1245,6 +1250,8 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx,
 	}
 
 	rctx->framebuffer.atom.dirty = true;
+
+	r600_set_sample_locations_constant_buffer(rctx);
 }
 
 static uint32_t sample_locs_2x[] = {
@@ -1524,6 +1531,21 @@ static void r600_emit_framebuffer_state(struct r600_context *rctx, struct r600_a
 	r600_emit_msaa_state(rctx, rctx->framebuffer.nr_samples);
 }
 
+static void r600_set_min_samples(struct pipe_context *ctx, unsigned min_samples)
+{
+	struct r600_context *rctx = (struct r600_context *)ctx;
+
+	if (rctx->ps_iter_samples == min_samples)
+		return;
+
+	rctx->ps_iter_samples = min_samples;
+	if (rctx->framebuffer.nr_samples > 1) {
+		rctx->rasterizer_state.atom.dirty = true;
+		if (rctx->b.chip_class == R600)
+			rctx->db_misc_state.atom.dirty = true;
+	}
+}
+
 static void r600_emit_cb_misc_state(struct r600_context *rctx, struct r600_atom *atom)
 {
 	struct radeon_winsys_cs *cs = rctx->b.rings.gfx.cs;
@@ -1603,6 +1625,10 @@ static void r600_emit_db_misc_state(struct r600_context *rctx, struct r600_atom
 	} else {
 		db_render_override |= S_028D10_FORCE_HIZ_ENABLE(V_028D10_FORCE_DISABLE);
 	}
+	if (rctx->b.chip_class == R600 && rctx->framebuffer.nr_samples > 1 && rctx->ps_iter_samples > 0) {
+		/* sample shading and hyperz causes lockups on R6xx chips */
+		db_render_override |= S_028D10_FORCE_HIZ_ENABLE(V_028D10_FORCE_DISABLE);
+	}
 	if (a->flush_depthstencil_through_cb) {
 		assert(a->copy_depth || a->copy_stencil);
 
@@ -2418,10 +2444,10 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 	struct r600_command_buffer *cb = &shader->command_buffer;
 	struct r600_shader *rshader = &shader->shader;
 	unsigned i, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z, spi_ps_in_control_1, db_shader_control;
-	int pos_index = -1, face_index = -1;
+	int pos_index = -1, face_index = -1, fixed_pt_position_index = -1;
 	unsigned tmp, sid, ufi = 0;
 	int need_linear = 0;
-	unsigned z_export = 0, stencil_export = 0;
+	unsigned z_export = 0, stencil_export = 0, mask_export = 0;
 	unsigned sprite_coord_enable = rctx->rasterizer ? rctx->rasterizer->sprite_coord_enable : 0;
 
 	if (!cb->buf) {
@@ -2434,8 +2460,10 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 	for (i = 0; i < rshader->ninput; i++) {
 		if (rshader->input[i].name == TGSI_SEMANTIC_POSITION)
 			pos_index = i;
-		if (rshader->input[i].name == TGSI_SEMANTIC_FACE)
+		if (rshader->input[i].name == TGSI_SEMANTIC_FACE && face_index == -1)
 			face_index = i;
+		if (rshader->input[i].name == TGSI_SEMANTIC_SAMPLEID)
+			fixed_pt_position_index = i;
 
 		sid = rshader->input[i].spi_sid;
 
@@ -2452,9 +2480,12 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 			tmp |= S_028644_PT_SPRITE_TEX(1);
 		}
 
-		if (rshader->input[i].centroid)
+		if (rshader->input[i].interpolate_location == TGSI_INTERPOLATE_LOC_CENTROID)
 			tmp |= S_028644_SEL_CENTROID(1);
 
+		if (rshader->input[i].interpolate_location == TGSI_INTERPOLATE_LOC_SAMPLE)
+			tmp |= S_028644_SEL_SAMPLE(1);
+
 		if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR) {
 			need_linear = 1;
 			tmp |= S_028644_SEL_LINEAR(1);
@@ -2469,16 +2500,21 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 			z_export = 1;
 		if (rshader->output[i].name == TGSI_SEMANTIC_STENCIL)
 			stencil_export = 1;
+		if (rshader->output[i].name == TGSI_SEMANTIC_SAMPLEMASK &&
+			rctx->framebuffer.nr_samples > 1 && rctx->ps_iter_samples > 0)
+			mask_export = 1;
 	}
 	db_shader_control |= S_02880C_Z_EXPORT_ENABLE(z_export);
 	db_shader_control |= S_02880C_STENCIL_REF_EXPORT_ENABLE(stencil_export);
+	db_shader_control |= S_02880C_MASK_EXPORT_ENABLE(mask_export);
 	if (rshader->uses_kill)
 		db_shader_control |= S_02880C_KILL_ENABLE(1);
 
 	exports_ps = 0;
 	for (i = 0; i < rshader->noutput; i++) {
 		if (rshader->output[i].name == TGSI_SEMANTIC_POSITION ||
-		    rshader->output[i].name == TGSI_SEMANTIC_STENCIL) {
+		    rshader->output[i].name == TGSI_SEMANTIC_STENCIL ||
+		    rshader->output[i].name == TGSI_SEMANTIC_SAMPLEMASK) {
 			exports_ps |= 1;
 		}
 	}
@@ -2497,9 +2533,10 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 	spi_input_z = 0;
 	if (pos_index != -1) {
 		spi_ps_in_control_0 |= (S_0286CC_POSITION_ENA(1) |
-					S_0286CC_POSITION_CENTROID(rshader->input[pos_index].centroid) |
+					S_0286CC_POSITION_CENTROID(rshader->input[pos_index].interpolate_location == TGSI_INTERPOLATE_LOC_CENTROID) |
 					S_0286CC_POSITION_ADDR(rshader->input[pos_index].gpr) |
-					S_0286CC_BARYC_SAMPLE_CNTL(1));
+					S_0286CC_BARYC_SAMPLE_CNTL(1)) |
+					S_0286CC_POSITION_SAMPLE(rshader->input[pos_index].interpolate_location == TGSI_INTERPOLATE_LOC_SAMPLE);
 		spi_input_z |= S_0286D8_PROVIDE_Z_TO_SPI(1);
 	}
 
@@ -2508,6 +2545,10 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 		spi_ps_in_control_1 |= S_0286D0_FRONT_FACE_ENA(1) |
 			S_0286D0_FRONT_FACE_ADDR(rshader->input[face_index].gpr);
 	}
+	if (fixed_pt_position_index != -1) {
+		spi_ps_in_control_1 |= S_0286D0_FIXED_PT_POSITION_ENA(1) |
+			S_0286D0_FIXED_PT_POSITION_ADDR(rshader->input[fixed_pt_position_index].gpr);
+	}
 
 	/* HW bug in original R600 */
 	if (rctx->b.family == CHIP_R600)
@@ -2531,7 +2572,7 @@ void r600_update_ps_state(struct pipe_context *ctx, struct r600_pipe_shader *sha
 
 	/* only set some bits here, the other bits are set in the dsa state */
 	shader->db_shader_control = db_shader_control;
-	shader->ps_depth_export = z_export | stencil_export;
+	shader->ps_depth_export = z_export | stencil_export | mask_export;
 
 	shader->sprite_coord_enable = sprite_coord_enable;
 	if (rctx->rasterizer)
@@ -3046,6 +3087,7 @@ void r600_init_state_functions(struct r600_context *rctx)
 	rctx->b.b.create_sampler_view = r600_create_sampler_view;
 	rctx->b.b.set_framebuffer_state = r600_set_framebuffer_state;
 	rctx->b.b.set_polygon_stipple = r600_set_polygon_stipple;
+	rctx->b.b.set_min_samples = r600_set_min_samples;
 	rctx->b.b.set_scissor_states = r600_set_scissor_states;
 	rctx->b.b.get_sample_position = r600_get_sample_position;
 	rctx->b.dma_copy = r600_dma_copy;
diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c
index d9174a59230..68365f9d9af 100644
--- a/src/gallium/drivers/r600/r600_state_common.c
+++ b/src/gallium/drivers/r600/r600_state_common.c
@@ -1085,6 +1085,26 @@ static void r600_setup_txq_cube_array_constants(struct r600_context *rctx, int s
 	pipe_resource_reference(&cb.buffer, NULL);
 }
 
+/* set sample xy locations as array of fragment shader constants */
+void r600_set_sample_locations_constant_buffer(struct r600_context *rctx)
+{
+	struct pipe_constant_buffer constbuf = {0};
+	float values[4*16] = {0.0f};
+	int i;
+	struct pipe_context *ctx = &rctx->b.b;
+
+	assert(rctx->framebuffer.nr_samples <= Elements(values)/4);
+	for (i = 0; i < rctx->framebuffer.nr_samples; i++) {
+		ctx->get_sample_position(ctx, rctx->framebuffer.nr_samples, i, &values[4*i]);
+	}
+
+	constbuf.user_buffer = values;
+	constbuf.buffer_size = rctx->framebuffer.nr_samples * 4 * 4;
+	ctx->set_constant_buffer(ctx, PIPE_SHADER_FRAGMENT,
+		R600_SAMPLE_POSITIONS_CONST_BUFFER, &constbuf);
+	pipe_resource_reference(&constbuf.buffer, NULL);
+}
+
 static void update_shader_atom(struct pipe_context *ctx,
 			       struct r600_shader_state *state,
 			       struct r600_pipe_shader *shader)
diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h
index 3cf7b8800f4..6a5b9640ee0 100644
--- a/src/gallium/drivers/r600/r600d.h
+++ b/src/gallium/drivers/r600/r600d.h
@@ -841,6 +841,9 @@
 #define   S_02880C_KILL_ENABLE(x)                      (((x) & 0x1) << 6)
 #define   G_02880C_KILL_ENABLE(x)                      (((x) >> 6) & 0x1)
 #define   C_02880C_KILL_ENABLE                         0xFFFFFFBF
+#define   S_02880C_MASK_EXPORT_ENABLE(x)               (((x) & 0x1) << 8)
+#define   G_02880C_MASK_EXPORT_ENABLE(x)               (((x) >> 8) & 0x1)
+#define   C_02880C_MASK_EXPORT_ENABLE                  0xFFFFFEFF
 #define   S_02880C_DUAL_EXPORT_ENABLE(x)               (((x) & 0x1) << 9)
 #define   G_02880C_DUAL_EXPORT_ENABLE(x)               (((x) >> 9) & 0x1)
 #define   C_02880C_DUAL_EXPORT_ENABLE                  0xFFFFFDFF
diff --git a/src/gallium/drivers/r600/sb/sb_bc_parser.cpp b/src/gallium/drivers/r600/sb/sb_bc_parser.cpp
index 346ccc9dbc5..d787e5b1238 100644
--- a/src/gallium/drivers/r600/sb/sb_bc_parser.cpp
+++ b/src/gallium/drivers/r600/sb/sb_bc_parser.cpp
@@ -147,25 +147,28 @@ int bc_parser::parse_decls() {
 	bool ps_interp = ctx.hw_class >= HW_CLASS_EVERGREEN
 			&& sh->target == TARGET_PS;
 
-	unsigned linear = 0, persp = 0, centroid = 1;
+	bool ij_interpolators[6];
+	memset(ij_interpolators, 0, sizeof(ij_interpolators));
 
 	for (unsigned i = 0; i < pshader->ninput; ++i) {
 		r600_shader_io & in = pshader->input[i];
 		bool preloaded = sh->target == TARGET_PS && !(ps_interp && in.spi_sid);
 		sh->add_input(in.gpr, preloaded, /*in.write_mask*/ 0x0F);
 		if (ps_interp && in.spi_sid) {
-			if (in.interpolate == TGSI_INTERPOLATE_LINEAR ||
-					in.interpolate == TGSI_INTERPOLATE_COLOR)
-				linear = 1;
-			else if (in.interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
-				persp = 1;
-			if (in.centroid)
-				centroid = 2;
+			int k = eg_get_interpolator_index(in.interpolate, in.interpolate_location);
+			if (k >= 0)
+				ij_interpolators[k] |= true;
 		}
 	}
 
 	if (ps_interp) {
-		unsigned mask = (1 << (2 * (linear + persp) * centroid)) - 1;
+		/* add the egcm ij interpolators to live inputs */
+		unsigned num_ij = 0;
+		for (unsigned i = 0; i < Elements(ij_interpolators); i++) {
+			num_ij += ij_interpolators[i];
+		}
+
+		unsigned mask = (1 << (2 * num_ij)) - 1;
 		unsigned gpr = 0;
 
 		while (mask) {