The "fail" flag is set if loop_unroll_count encounters a nested loop; calling the flag "nested_loop" is a bit clearer.
The original reasoning was that count is inaccurate (too small) if there are nested loops, as we don't do any sort of analysis on the inner loop. Signed-off-by: Kenneth Graunke <kenn...@whitecape.org> --- src/glsl/loop_unroll.cpp | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) diff --git a/src/glsl/loop_unroll.cpp b/src/glsl/loop_unroll.cpp index 11680b3..1ce4d58 100644 --- a/src/glsl/loop_unroll.cpp +++ b/src/glsl/loop_unroll.cpp @@ -63,12 +63,13 @@ is_break(ir_instruction *ir) class loop_unroll_count : public ir_hierarchical_visitor { public: int nodes; - bool fail; + /* If there are nested loops, the node count will be inaccurate. */ + bool nested_loop; loop_unroll_count(exec_list *list) { nodes = 0; - fail = false; + nested_loop = false; run(list); } @@ -87,7 +88,7 @@ public: virtual ir_visitor_status visit_enter(ir_loop *) { - fail = true; + nested_loop = true; return visit_continue; } }; @@ -258,7 +259,7 @@ loop_unroll_visitor::visit_leave(ir_loop *ir) */ loop_unroll_count count(&ir->body_instructions); - if (count.fail || count.nodes * iterations > max_iterations * 5) + if (count.nested_loop || count.nodes * iterations > max_iterations * 5) return visit_continue; /* Note: the limiting terminator contributes 1 to ls->num_loop_jumps. -- 1.9.1 _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev