roc: install sqtt decoder (#13091)

* roc: install?

* msg

* 0.1.4

autogen_stubs.sh

@@ -432,11 +432,13 @@ generate_sqtt() {
     $ROCPROF_SRC/include/rocprof_trace_decoder.h \
     $ROCPROF_SRC/include/trace_decoder_instrument.h \
     $ROCPROF_SRC/include/trace_decoder_types.h \
-    -o extra/sqtt/rocprof/rocprof.py
-  fixup extra/sqtt/rocprof/rocprof.py
-  sed -i '1s/^/# pylint: skip-file\n/' extra/sqtt/rocprof/rocprof.py
-  sed -i "s/import ctypes/import ctypes, ctypes.util/g" extra/sqtt/rocprof/rocprof.py
-  sed -i "s|FunctionFactoryStub()|ctypes.CDLL(ctypes.util.find_library('rocprof-trace-decoder'))|g" extra/sqtt/rocprof/rocprof.py
+    -o $BASE/rocprof.py
+  fixup $BASE/rocprof.py
+  sed -i '1s/^/# pylint: skip-file\n/' $BASE/rocprof.py
+  sed -i "s/import ctypes/import ctypes, ctypes.util/g" $BASE/rocprof.py
+  patch_dlopen $BASE/rocprof.py rocprof-trace-decoder "'/usr/local/lib/rocprof-trace-decoder.so'" "'/usr/local/lib/rocprof-trace-decoder.dylib'"
+  sed -i "s/def _try_dlopen_rocprof-trace-decoder():/def _try_dlopen_rocprof_trace_decoder():/g" $BASE/rocprof.py
+  sed -i "s|FunctionFactoryStub()|_try_dlopen_rocprof_trace_decoder()|g" $BASE/rocprof.py
 }
 
 generate_webgpu() {

extra/sqtt/install_sqtt_decoder.py

@@ -13,6 +13,6 @@ if __name__ == "__main__":
     os.chmod(fp, 0o755)
     os.system(f"sudo {fp} --prefix={fp.parent} --include-subdir")
   else:
-    lib = fetch("https://github.com/ROCm/rocprof-trace-decoder/raw/5420409ad0963b2d76450add067b9058493ccbd0/releases/linux_glibc_2_28_x86_64/librocprof-trace-decoder.so", name="librocprof-trace-decoder.so")
+    lib = fetch("https://github.com/ROCm/rocprof-trace-decoder/raw/43bf0fef74a83c3c25badfc5a09c0bd39ed8c6f9/releases/linux_glibc_2_28_x86_64/librocprof-trace-decoder.so", name="librocprof-trace-decoder.so")
   shutil.copy2(lib, DEST)
   print(f"Installed {lib.name} to", DEST)

extra/sqtt/roc.py

@@ -1,9 +1,8 @@
 import ctypes, pathlib, argparse, pickle, re, functools, dataclasses, itertools
-from extra.sqtt.rocprof import rocprof
 from tinygrad.helpers import temp, unwrap, DEBUG
 from tinygrad.device import ProfileEvent, ProfileDeviceEvent, ProfileProgramEvent
 from tinygrad.runtime.ops_amd import ProfileSQTTEvent, ProfilePMCEvent
-from tinygrad.runtime.autogen import llvm
+from tinygrad.runtime.autogen import llvm, rocprof
 from tinygrad.runtime.support.elf import elf_loader
 
 # to pass NULL to callbacks
@@ -122,7 +121,7 @@ def decode(profile:list[ProfileEvent]) -> _ROCParseCtx:
 
   try:
     rocprof.rocprof_trace_decoder_parse_data(copy_cb, trace_cb, isa_cb, None)
-  except Exception as e: print("Error in sqtt decoder:", e)
+  except AttributeError as e: raise RuntimeError("Failed to find rocprof-trace-decoder. Run ./extra/sqtt/install_sqtt_decoder.py to install") from e
   return ROCParseCtx
 
 if __name__ == "__main__":

tinygrad/runtime/autogen/rocprof.py

@@ -8,11 +8,20 @@
 # LONGDOUBLE_SIZE is: 16
 #
 import ctypes, ctypes.util
+PATHS_TO_TRY = [
+  '/usr/local/lib/rocprof-trace-decoder.so',
+  '/usr/local/lib/rocprof-trace-decoder.dylib',
+]
+def _try_dlopen_rocprof_trace_decoder():
+  library = ctypes.util.find_library("rocprof-trace-decoder")
+  if library: return ctypes.CDLL(library)
+  for candidate in PATHS_TO_TRY:
+    try: return ctypes.CDLL(candidate)
+    except OSError: pass
+  return None
 
 
 class AsDictMixin:
-    import sys
-    if sys.version_info >= (3, 14): _layout_ = 'ms'
     @classmethod
     def as_dict(cls, self):
         result = {}
@@ -157,7 +166,7 @@ class FunctionFactoryStub:
 # You can either re-run clan2py with -l /path/to/library.so
 # Or manually fix this by comment the ctypes.CDLL loading
 _libraries = {}
-_libraries['FIXME_STUB'] = ctypes.CDLL(ctypes.util.find_library('rocprof-trace-decoder')) #  ctypes.CDLL('FIXME_STUB')
+_libraries['FIXME_STUB'] = _try_dlopen_rocprof_trace_decoder() #  ctypes.CDLL('FIXME_STUB')