radeonsi: use a single descriptor for the GSVS ring

We can hardcode all of the fields for swizzling in the geometry shader.

The advantage is that we use fewer descriptor slots and we no longer have to
update any of the (ring) descriptors when the geometry shader changes.

Reviewed-by: Marek Olšák <[email protected]>

1 files changed, 65 insertions, 8 deletions

diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 48ccd83b396..9b495925a6e 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -5817,6 +5817,7 @@ static void preload_ring_buffers(struct si_shader_context *ctx)
 {
 	struct gallivm_state *gallivm =
 		ctx->soa.bld_base.base.gallivm;
+	LLVMBuilderRef builder = gallivm->builder;
 
 	LLVMValueRef buf_ptr = LLVMGetParam(ctx->main_fn,
 					    SI_PARAM_RW_BUFFERS);
@@ -5836,18 +5837,74 @@ static void preload_ring_buffers(struct si_shader_context *ctx)
 	}
 
 	if (ctx->shader->is_gs_copy_shader) {
-		LLVMValueRef offset = lp_build_const_int32(gallivm, SI_VS_RING_GSVS);
+		LLVMValueRef offset = lp_build_const_int32(gallivm, SI_RING_GSVS);
 
 		ctx->gsvs_ring[0] =
 			build_indexed_load_const(ctx, buf_ptr, offset);
-	}
-	if (ctx->type == PIPE_SHADER_GEOMETRY) {
-		int i;
-		for (i = 0; i < 4; i++) {
-			LLVMValueRef offset = lp_build_const_int32(gallivm, SI_GS_RING_GSVS0 + i);
+	} else if (ctx->type == PIPE_SHADER_GEOMETRY) {
+		struct lp_build_context *uint = &ctx->soa.bld_base.uint_bld;
+		LLVMValueRef offset = lp_build_const_int32(gallivm, SI_RING_GSVS);
+		LLVMValueRef base_ring;
+
+		base_ring = build_indexed_load_const(ctx, buf_ptr, offset);
+
+		/* The conceptual layout of the GSVS ring is
+		 *   v0c0 .. vLv0 v0c1 .. vLc1 ..
+		 * but the real memory layout is swizzled across
+		 * threads:
+		 *   t0v0c0 .. t15v0c0 t0v1c0 .. t15v1c0 ... t15vLcL
+		 *   t16v0c0 ..
+		 * Override the buffer descriptor accordingly.
+		 */
+		LLVMTypeRef v2i64 = LLVMVectorType(ctx->i64, 2);
+		unsigned max_gsvs_emit_size = ctx->shader->selector->max_gsvs_emit_size;
+		unsigned num_records;
+
+		num_records = 64;
+		if (ctx->screen->b.chip_class >= VI)
+			num_records *= max_gsvs_emit_size;
+
+		for (unsigned stream = 0; stream < 4; ++stream) {
+			LLVMValueRef ring, tmp;
+
+			if (!ctx->shader->selector->info.num_stream_output_components[stream])
+				continue;
 
-			ctx->gsvs_ring[i] =
-				build_indexed_load_const(ctx, buf_ptr, offset);
+			/* Limit on the stride field for <= CIK. */
+			assert(max_gsvs_emit_size < (1 << 14));
+
+			ring = LLVMBuildBitCast(builder, base_ring, v2i64, "");
+			tmp = LLVMBuildExtractElement(builder, ring, uint->zero, "");
+			tmp = LLVMBuildAdd(builder, tmp,
+					   LLVMConstInt(ctx->i64,
+							max_gsvs_emit_size * 64 * stream, 0), "");
+			ring = LLVMBuildInsertElement(builder, ring, tmp, uint->zero, "");
+			ring = LLVMBuildBitCast(builder, ring, ctx->v4i32, "");
+			tmp = LLVMBuildExtractElement(builder, ring, uint->one, "");
+			tmp = LLVMBuildOr(builder, tmp,
+				LLVMConstInt(ctx->i32,
+					     S_008F04_STRIDE(max_gsvs_emit_size) |
+					     S_008F04_SWIZZLE_ENABLE(1), 0), "");
+			ring = LLVMBuildInsertElement(builder, ring, tmp, uint->one, "");
+			ring = LLVMBuildInsertElement(builder, ring,
+					LLVMConstInt(ctx->i32, num_records, 0),
+					LLVMConstInt(ctx->i32, 2, 0), "");
+			ring = LLVMBuildInsertElement(builder, ring,
+				LLVMConstInt(ctx->i32,
+					     S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X) |
+					     S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
+					     S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
+					     S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
+					     S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+					     S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
+					     S_008F0C_ELEMENT_SIZE(1) | /* element_size = 4 (bytes) */
+					     S_008F0C_INDEX_STRIDE(1) | /* index_stride = 16 (elements) */
+					     S_008F0C_ADD_TID_ENABLE(1),
+					     0),
+				LLVMConstInt(ctx->i32, 3, 0), "");
+			ring = LLVMBuildBitCast(builder, ring, ctx->v16i8, "");
+
+			ctx->gsvs_ring[stream] = ring;
 		}
 	}
 }