assembly/amd: rename to better names (#14384)

* assembly/amd: rename to better names

* might help fuzzing segfault

* emu2 -> emu

.github/workflows/test.yml

@@ -679,7 +679,7 @@ jobs:
           python-version: '3.13'
       - name: Verify AMD autogen is up to date
         run: |
-          python -m extra.assembly.amd.amdxml
+          python -m extra.assembly.amd.generate
           git diff --exit-code extra/assembly/amd/autogen/
       - name: Install LLVM 21
         run: |

extra/assembly/amd/emu.py

@@ -48,13 +48,13 @@ from tinygrad.runtime.autogen import hsa
 from tinygrad.helpers import Context, DEBUG, colored
 from tinygrad.engine.realize import get_runner
 
-from extra.assembly.amd.decode import decode_inst
+from extra.assembly.amd import decode_inst
 from extra.assembly.amd.autogen.rdna3.str_pcode import PCODE
 from extra.assembly.amd.autogen.rdna3.ins import (SOP1, SOP2, SOPC, SOPK, SOPP, SMEM, VOP1, VOP1_SDST, VOP2, VOP3, VOP3_SDST, VOP3SD, VOP3P, VOPC,
   DS, FLAT, GLOBAL, SCRATCH, VOPD, SOPPOp, SMEMOp, VOP1Op, VOP2Op, VOP3Op, VOPDOp)
 from extra.assembly.amd.dsl import VCC_LO, EXEC_LO, SCC
 from extra.assembly.amd.autogen.common import OpType
-from extra.assembly.amd.expr_parser import parse_block, _FUNCS
+from extra.assembly.amd.pcode import parse_block, _FUNCS
 
 MASK32 = 0xFFFFFFFF
 
@@ -1035,7 +1035,7 @@ def _get_runner(inst_bytes: bytes):
       if cls in _INST_HANDLERS:
         handler = _INST_HANDLERS[cls]
         break
-  if handler is None: raise RuntimeError(f"[emu2] unimplemented instruction type: {type(inst).__name__} {_op_name(inst)}")
+  if handler is None: raise RuntimeError(f"[emu] unimplemented instruction type: {type(inst).__name__} {_op_name(inst)}")
 
   ctx = _Ctx(inst_size)
   sink = handler(inst, ctx)
@@ -1061,14 +1061,13 @@ def decode_program(data: bytes) -> dict[int, tuple[str, Callable, list[int], Any
       if DEBUG >= 3:
         try: inst_str = repr(inst)
         except Exception: inst_str = f"<{type(inst).__name__} at PC={i}>"
-        msg = f"[emu2] PC={i}: {inst_str}"
+        msg = f"[emu] PC={i}: {inst_str}"
         print(colored(msg, 'green') if is_new else msg)
-        if DEBUG >= 4: print(f"{colored(runner.p.src, 'BLACK')}")
       result[i] = (runner.p.function_name, runner._prg.fxn, runner.p.globals, runner)
     except Exception as e:
       try: inst_str = repr(inst)
       except Exception: inst_str = f"<{type(inst).__name__}>"
-      raise RuntimeError(f"[emu2] Failed to compile PC={i} {inst_str}: {type(e).__name__}: {e}") from e
+      raise RuntimeError(f"[emu] Failed to compile PC={i} {inst_str}: {type(e).__name__}: {e}") from e
     i += inst.size()
   return result
 
@@ -1160,11 +1159,11 @@ def run_asm(lib: int, lib_sz: int, gx: int, gy: int, gz: int, lx: int, ly: int,
             for inst_count in range(1_000_000):
               if (pc := st.pc) == 0xFFFFFFFFFFFFFFFF or pc not in program: break
               name, fxn, globals_list, _ = program[pc]
-              assert fxn is not None, f"[emu2] No fxn for {name} at PC={pc}"
+              assert fxn is not None, f"[emu] No fxn for {name} at PC={pc}"
               assert 4 not in globals_list or scratch_buf, f"SCRATCH instruction {name} but scratch_size=0"
-              if DEBUG >= 5:
+              if DEBUG >= 6:
                 inst = decode_inst(bytes((ctypes.c_char * 12).from_address(pc).raw))
-                print(f"[emu2] exec PC={pc:X}: {inst!r}")
+                print(f"[emu] exec PC={pc:X}: {inst!r}")
               fxn(*[c_bufs[g] for g in globals_list])
             else: raise RuntimeError("exceeded 1M instructions, likely infinite loop")
   return 0

extra/assembly/amd/sqttmap.py

@@ -6,7 +6,7 @@ from tinygrad.runtime.support.elf import elf_loader
 
 from extra.assembly.amd.sqtt import decode, print_packets, INST, VALUINST, IMMEDIATE, WAVESTART, WAVEEND, InstOp, PacketType, IMMEDIATE_MASK
 from extra.assembly.amd.dsl import Inst
-from extra.assembly.amd.decode import decode_inst
+from extra.assembly.amd import decode_inst
 from extra.assembly.amd.autogen.rdna3.ins import SOPP, s_endpgm
 from extra.assembly.amd.autogen.rdna3.enum import SOPPOp
 

extra/assembly/amd/test/bench_emu.py

@@ -6,8 +6,8 @@ from pathlib import Path
 # Set AMD=1 before importing tinygrad
 os.environ["AMD"] = "1"
 
-from extra.assembly.amd.emu2 import run_asm as python_run_asm, decode_program
-from extra.assembly.amd.decode import decode_inst
+from extra.assembly.amd.emu import run_asm as python_run_asm, decode_program
+from extra.assembly.amd import decode_inst
 from extra.assembly.amd.autogen.rdna3.ins import SOPP, SOPPOp
 
 REMU_PATH = Path(__file__).parents[3] / "remu/target/release/libremu.so"
@@ -68,7 +68,7 @@ def benchmark_emulator(name: str, run_fn, kernel: bytes, global_size, local_size
 
 def profile_instructions(kernel: bytes):
   """Profile individual instruction compile times."""
-  from extra.assembly.amd.emu2 import _get_runner, _canonical_runner_cache
+  from extra.assembly.amd.emu import _get_runner, _canonical_runner_cache
   from tinygrad.helpers import Context
   _get_runner.cache_clear()
   _canonical_runner_cache.clear()
@@ -98,7 +98,7 @@ def profile_instructions(kernel: bytes):
 
 def benchmark_python_split(kernel: bytes, global_size, local_size, args_ptr, rsrc2: int, iterations: int = 5):
   """Benchmark Python emulator with compile and execution times."""
-  from extra.assembly.amd.emu2 import _get_runner, _canonical_runner_cache
+  from extra.assembly.amd.emu import _get_runner, _canonical_runner_cache
   from tinygrad.helpers import Context
   _get_runner.cache_clear()
   _canonical_runner_cache.clear()

extra/assembly/amd/test/hw/helpers.py

@@ -6,7 +6,7 @@ Set USE_HW=1 to run on both emulator and hardware, comparing results.
 import ctypes, math, os, struct
 from extra.assembly.amd.autogen.rdna3.ins import *
 
-from extra.assembly.amd.emu2 import run_asm
+from extra.assembly.amd.emu import run_asm
 from extra.assembly.amd.dsl import NULL, SCC, VCC_LO, VCC_HI, EXEC_LO, EXEC_HI, M0
 
 def _i32(f: float) -> int: return struct.unpack('<I', struct.pack('<f', f))[0]

extra/assembly/amd/test/test_compare_emulators.py

@@ -2,8 +2,8 @@
 import unittest, ctypes
 from dataclasses import dataclass
 
-from extra.assembly.amd.emu2 import WaveState, decode_program, WAVE_SIZE, VCC_LO, EXEC_LO, SCC
-from extra.assembly.amd.decode import decode_inst
+from extra.assembly.amd.emu import WaveState, decode_program, WAVE_SIZE, VCC_LO, EXEC_LO, SCC
+from extra.assembly.amd import decode_inst
 from extra.assembly.amd.test.helpers import KernelInfo
 from extra.assembly.amd.test.bench_emu import REMU_PATH
 

extra/assembly/amd/test/test_emu2_pcode.py

@@ -4,8 +4,8 @@ from collections import defaultdict
 from tinygrad.helpers import DEBUG
 from tinygrad.dtype import dtypes
 from tinygrad.uop.ops import UOp, Ops
-from extra.assembly.amd.emu2 import parse_pcode
-from extra.assembly.amd.expr_parser import parse_expr
+from extra.assembly.amd.emu import parse_pcode
+from extra.assembly.amd.pcode import parse_expr
 from extra.assembly.amd.autogen.rdna3.str_pcode import PCODE
 from extra.assembly.amd.autogen.rdna3.enum import VOP1Op, VOP2Op, VOP3Op, SOP1Op, SOP2Op, DSOp
 

extra/assembly/amd/test/test_formats.py

@@ -7,7 +7,7 @@ import unittest
 from extra.assembly.amd.autogen.rdna3.ins import *
 from extra.assembly.amd.dsl import VCC_HI, EXEC_LO, NULL
 OFF = NULL  # OFF is alias for NULL
-from extra.assembly.amd.decode import detect_format
+from extra.assembly.amd import detect_format
 
 
 class TestDS(unittest.TestCase):

extra/assembly/amd/test/test_llvm.py

@@ -11,7 +11,7 @@ Only compute-relevant instruction formats are tested. Graphics-only formats not
 import unittest, re, subprocess, functools
 from tinygrad.helpers import fetch
 from extra.assembly.amd.disasm import disasm
-from extra.assembly.amd.decode import decode_inst, detect_format
+from extra.assembly.amd import decode_inst, detect_format
 from extra.assembly.amd.test.helpers import get_llvm_mc, get_target, get_mattr
 
 LLVM_BASE = "https://raw.githubusercontent.com/llvm/llvm-project/llvmorg-21.1.0/llvm/test/MC/AMDGPU"

extra/assembly/amd/test/test_pdf.py

@@ -1,7 +1,7 @@
 #!/usr/bin/env python3
-"""Test PDF pseudocode extraction from amdxml.py."""
+"""Test PDF pseudocode extraction from generate.py."""
 import unittest
-from extra.assembly.amd.amdxml import extract_pdf_text, extract_pcode, parse_xml, ARCHS, FIXES
+from extra.assembly.amd.generate import extract_pdf_text, extract_pcode, parse_xml, ARCHS, FIXES
 
 EXPECTED_PAGES = {"rdna3": 655, "rdna4": 711, "cdna": 610}
 

extra/assembly/amd/test/test_roundtrip.py

@@ -2,7 +2,7 @@
 """Roundtrip tests: generate tinygrad kernels, decode instructions, re-encode, verify match."""
 import unittest, io, sys, re, subprocess, os
 from extra.assembly.amd.dsl import Inst
-from extra.assembly.amd.decode import decode_inst, detect_format
+from extra.assembly.amd import decode_inst, detect_format
 from extra.assembly.amd.test.helpers import get_llvm_mc, get_llvm_objdump, get_target, get_mattr
 
 def disassemble_lib(lib: bytes, compiler) -> list[tuple[str, bytes]]:

extra/assembly/amd/test/test_sqtt_examples.py

@@ -5,7 +5,7 @@ from pathlib import Path
 from tinygrad.helpers import DEBUG
 from tinygrad.runtime.autogen import rocprof
 from tinygrad.runtime.support.elf import elf_loader
-from extra.assembly.amd.decode import decode_inst
+from extra.assembly.amd import decode_inst
 from extra.assembly.amd.autogen.rdna3.ins import SOPP
 from extra.assembly.amd.autogen.rdna3.enum import SOPPOp
 from extra.assembly.amd.sqtt import (decode, LAYOUT_HEADER, WAVESTART, WAVESTART_L4, WAVEEND, INST, INST_L4, VALUINST, IMMEDIATE, IMMEDIATE_MASK,

test/external/fuzz_symbolic.py

@@ -55,7 +55,7 @@ if __name__ == "__main__":
     with Context(CORRECT_DIVMOD_FOLDING=1):
       simplified_expr = expr.simplify()
 
-    solver = z3.Solver()
+    solver = z3.Solver(ctx=z3.Context())
     solver.set(timeout=5000)  # some expressions take very long verify, but its very unlikely they actually return sat
     z3_expr, z3_simplified_expr, v1, v2, v3 = uops_to_z3(solver, expr, simplified_expr, u1, u2, u3)
     check = solver.check(z3_simplified_expr != z3_expr)

test/mockgpu/helpers.py

@@ -22,7 +22,7 @@ class PythonRemu:
   scratch_size: int = 0  # private_segment_fixed_size from kernel descriptor
 
   def run_asm(self, lib: int, lib_sz: int, gx: int, gy: int, gz: int, lx: int, ly: int, lz: int, args_ptr: int) -> int:
-    from extra.assembly.amd.emu2 import run_asm
+    from extra.assembly.amd.emu import run_asm
     return run_asm(lib, lib_sz, gx, gy, gz, lx, ly, lz, args_ptr, self.rsrc2, self.scratch_size)
 
 def _try_dlopen_remu():

tinygrad/viz/serve.py

@@ -434,7 +434,7 @@ def amd_readelf(lib:bytes) -> list[dict]:
 
 def amd_decode(target:int, lib:bytes) -> dict[int, Any]: # Any is the Inst class from extra.assembly.amd.dsl
   from tinygrad.runtime.support.elf import elf_loader
-  from extra.assembly.amd.decode import detect_format
+  from extra.assembly.amd import detect_format
   from extra.assembly.amd.dsl import Inst
   image, sections, _ = elf_loader(lib)
   text = next((sh for sh in sections if sh.name == ".text"), None)