radeonsi: set number of userdata SGPRs of GS copy shader to 4

It only needs the constant buffer with clip planes and read-write resources
for the GS->VS ring and streamout. That's 2 pointers.

Reviewed-by: Michel Dänzer <[email protected]>

3 files changed, 23 insertions, 10 deletions

diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 4e8f80f20dc..8680824076b 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -2402,8 +2402,8 @@ static void create_function(struct si_shader_context *si_shader_ctx)
 	v8i32 = LLVMVectorType(i32, 8);
 	v16i8 = LLVMVectorType(i8, 16);
 
-	params[SI_PARAM_CONST] = const_array(v16i8, SI_NUM_CONST_BUFFERS);
 	params[SI_PARAM_RW_BUFFERS] = const_array(v16i8, SI_NUM_RW_BUFFERS);
+	params[SI_PARAM_CONST] = const_array(v16i8, SI_NUM_CONST_BUFFERS);
 	params[SI_PARAM_SAMPLER] = const_array(v4i32, SI_NUM_SAMPLER_STATES);
 	params[SI_PARAM_RESOURCE] = const_array(v8i32, SI_NUM_SAMPLER_VIEWS);
 	last_array_pointer = SI_PARAM_RESOURCE;
@@ -2415,10 +2415,16 @@ static void create_function(struct si_shader_context *si_shader_ctx)
 		params[SI_PARAM_BASE_VERTEX] = i32;
 		params[SI_PARAM_START_INSTANCE] = i32;
 		num_params = SI_PARAM_START_INSTANCE+1;
+
 		if (shader->key.vs.as_es) {
 			params[SI_PARAM_ES2GS_OFFSET] = i32;
 			num_params++;
 		} else {
+			if (shader->is_gs_copy_shader) {
+				last_array_pointer = SI_PARAM_CONST;
+				num_params = SI_PARAM_CONST+1;
+			}
+
 			/* The locations of the other parameters are assigned dynamically. */
 
 			/* Streamout SGPRs. */
@@ -2716,6 +2722,7 @@ static int si_generate_gs_copy_shader(struct si_screen *sscreen,
 	outputs = MALLOC(gs->noutput * sizeof(outputs[0]));
 
 	si_shader_ctx->type = TGSI_PROCESSOR_VERTEX;
+	shader->is_gs_copy_shader = true;
 
 	radeon_llvm_context_init(&si_shader_ctx->radeon_bld);
 
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index c0e5cf40534..11e5ae0334a 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -33,10 +33,10 @@
 #include "tgsi/tgsi_scan.h"
 #include "si_state.h"
 
-#define SI_SGPR_CONST		0
-#define SI_SGPR_SAMPLER		2
-#define SI_SGPR_RESOURCE	4
-#define SI_SGPR_RW_BUFFERS	6  /* rings (& stream-out, VS only) */
+#define SI_SGPR_RW_BUFFERS	0  /* rings (& stream-out, VS only) */
+#define SI_SGPR_CONST		2
+#define SI_SGPR_SAMPLER		4
+#define SI_SGPR_RESOURCE	6
 #define SI_SGPR_VERTEX_BUFFER	8  /* VS only */
 #define SI_SGPR_BASE_VERTEX	10 /* VS only */
 #define SI_SGPR_START_INSTANCE	11 /* VS only */
@@ -44,13 +44,14 @@
 
 #define SI_VS_NUM_USER_SGPR	12
 #define SI_GS_NUM_USER_SGPR	8
+#define SI_GSCOPY_NUM_USER_SGPR	4
 #define SI_PS_NUM_USER_SGPR	9
 
 /* LLVM function parameter indices */
-#define SI_PARAM_CONST		0
-#define SI_PARAM_SAMPLER	1
-#define SI_PARAM_RESOURCE	2
-#define SI_PARAM_RW_BUFFERS	3
+#define SI_PARAM_RW_BUFFERS	0
+#define SI_PARAM_CONST		1
+#define SI_PARAM_SAMPLER	2
+#define SI_PARAM_RESOURCE	3
 
 /* VS only parameters */
 #define SI_PARAM_VERTEX_BUFFER	4
@@ -183,6 +184,7 @@ struct si_shader {
 	bool			vs_out_layer;
 	unsigned		nr_pos_exports;
 	unsigned		clip_dist_write;
+	bool			is_gs_copy_shader;
 };
 
 static inline struct si_shader* si_get_vs_state(struct si_context *sctx)
diff --git a/src/gallium/drivers/radeonsi/si_state_draw.c b/src/gallium/drivers/radeonsi/si_state_draw.c
index 6ad2df0dbe0..e8d84a90d57 100644
--- a/src/gallium/drivers/radeonsi/si_state_draw.c
+++ b/src/gallium/drivers/radeonsi/si_state_draw.c
@@ -166,7 +166,11 @@ static void si_shader_vs(struct pipe_context *ctx, struct si_shader *shader)
 
 	vgpr_comp_cnt = shader->uses_instanceid ? 3 : 0;
 
-	num_user_sgprs = SI_VS_NUM_USER_SGPR;
+	if (shader->is_gs_copy_shader)
+		num_user_sgprs = SI_GSCOPY_NUM_USER_SGPR;
+	else
+		num_user_sgprs = SI_VS_NUM_USER_SGPR;
+
 	num_sgprs = shader->num_sgprs;
 	if (num_user_sgprs > num_sgprs) {
 		/* Last 2 reserved SGPRs are used for VCC */