diff --git a/tinygrad/runtime/ops_triton.py b/tinygrad/runtime/ops_triton.py
index 435e83e400..01ad141ca9 100644
--- a/tinygrad/runtime/ops_triton.py
+++ b/tinygrad/runtime/ops_triton.py
@@ -7,6 +7,8 @@ from tinygrad.ops import  Compiled
 from tinygrad.runtime.ops_cuda import RawCUDABuffer
 from tinygrad.codegen.linearizer import LinearizerOptions
 from tinygrad.renderer.triton import uops_to_triton
+from tinygrad.helpers import DEBUG
+from tinygrad.runtime.ops_cuda import pretty_ptx 
 
 
 class TritonProgram:
@@ -22,6 +24,7 @@ class TritonProgram:
     codeObject = compile(prg, fn, "exec")
     exec(codeObject, globals())
     self.program = triton_compile(globals()[name], signature=signature, device_type="cuda", debug=True).asm["ptx"]
+    if DEBUG >= 5: print(pretty_ptx(self.program))
     self.program = cuda.module_from_buffer(self.program.encode('utf-8')).get_function(self.program.split(".visible .entry ")[1].split("(")[0])
 
   def __call__(self, global_size, local_size, *args, wait=False) -> Any: