fix misspellings (#13976)

docs/runtime.md

@@ -70,7 +70,7 @@ AMD backend supports several interfaces for communicating with devices:
 
 * `KFD`: uses the amdgpu driver
 * `PCI`: uses the [AM driver](developer/am.md)
-* `USB`: USB3 interafce for asm24xx chips.
+* `USB`: USB3 interface for asm24xx chips.
 
 You can force an interface by setting `AMD_IFACE` to one of these values. In the case of `AMD_IFACE=PCI`, this may unbind your GPU from the amdgpu driver.
 

extra/sqtt/README.md

@@ -8,7 +8,7 @@ SQTT is implemented on top of normal tinygrad profiling, `VIZ=1 SQTT=1` to get p
 
 `SQTT_ITRACE_SE_MASK=X` to select for which shader engines instruction tracing will be enabled, -1 is all, 0 is none (instruction tracing disabled), >0 is
 bitfield/mask for SEs to enable instruction tracing on. Masking shader engines will give smaller file sizes at a cost of less hits and kernels that
-don't have any wavefront on first simd of shdaer engine with instruction tracing enabled will not have instruction timings.
+don't have any wavefront on first simd of shader engine with instruction tracing enabled will not have instruction timings.
 The default is 2 (second shader engine only), only one for file size reasons, second instead of first because dispatch starts from it so there is
 greater chance that kernels with small global size will have instruction tracing data.
  

test/mockgpu/mockgpu.py

@@ -12,10 +12,10 @@ libc.mmap.restype = ctypes.c_void_p
 drivers = [AMDDriver(), NVDriver()]
 tracked_fds = {}
 
-orignal_memoryview = builtins.memoryview
+original_memoryview = builtins.memoryview
 class TrackedMemoryView:
   def __init__(self, data, rcb, wcb):
-    self.mv = orignal_memoryview(data)
+    self.mv = original_memoryview(data)
     self.rcb, self.wcb = rcb, wcb
 
   def __getitem__(self, index):
@@ -41,7 +41,7 @@ def _memoryview(cls, mem):
     for d in drivers:
       for st,en,rcb,wcb in d.tracked_addresses:
         if st <= addr <= en: return TrackedMemoryView(mem, rcb, wcb)
-  return orignal_memoryview(mem)
+  return original_memoryview(mem)
 builtins.memoryview = type("memoryview", (), {'__new__': _memoryview}) # type: ignore
 
 def _open(path, flags):

test/test_image_dtype.py

@@ -194,7 +194,7 @@ class TestImageDType(unittest.TestCase):
           lst = s.bufs[0].as_buffer().cast("f").tolist()
           print(lst)
           assert not np.any(np.isnan(lst))
-      # NOTE: the w1 grad must realize to a seperate kernel
+      # NOTE: the w1 grad must realize to a separate kernel
       assert w1.grad.uop.is_realized, f"never realized {w1.grad}"
       self.assertEqual(w1.grad.uop.base.buffer.dtype, dtypes.float32)
       self.assertEqual(len(sched), 9)

test/test_ops.py

@@ -1087,7 +1087,7 @@ class TestOps(unittest.TestCase):
     helper_test_op([(2,3,0)], lambda x: torch.cummax(x, dim=2).values, lambda x: Tensor.cummax(x, axis=2))
 
   def test_argmax(self):
-    # check if it returns the first index for multiple occurences
+    # check if it returns the first index for multiple occurrences
     helper_test_op(None, lambda x: x.argmax().type(torch.int32), lambda x: x.argmax(), forward_only=True, vals=[[2, 2]])
     helper_test_op(None, lambda x: x.argmax().type(torch.int32), lambda x: x.argmax(), forward_only=True, vals=[[1, 2, 2]])
     if not COMPILE_ONLY:
@@ -1107,7 +1107,7 @@ class TestOps(unittest.TestCase):
     helper_test_op(None, lambda x: x.type(torch.int32).argmax().type(torch.int32), lambda x: x.argmax(), forward_only=True, vals=[[True, False]])
 
   def test_argmin(self):
-    # check if it returns the first index for multiple occurences
+    # check if it returns the first index for multiple occurrences
     helper_test_op(None, lambda x: x.argmin().type(torch.int32), lambda x: x.argmin(), forward_only=True, vals=[[2, 2]])
     helper_test_op(None, lambda x: x.argmin().type(torch.int32), lambda x: x.argmin(), forward_only=True, vals=[[3, 2, 2]])
     if not COMPILE_ONLY:

test/unit/test_device.py

@@ -28,7 +28,7 @@ class TestDevice(unittest.TestCase):
     self.assertEqual(Device.canonicalize(None), device)
     Device.DEFAULT = device
 
-  @unittest.skipIf(WIN and CI, "skipping windows test") # TODO: subproccess causes memory violation?
+  @unittest.skipIf(WIN and CI, "skipping windows test") # TODO: subprocess causes memory violation?
   def test_env_overwrite_default_compiler(self):
     if Device.DEFAULT == "CPU":
       from tinygrad.runtime.support.compiler_cpu import CPULLVMCompiler, ClangJITCompiler

tinygrad/tensor.py

@@ -1141,7 +1141,7 @@ class Tensor(OpMixin):
           boundary, stride = [start, stop], step
           if all(isinstance(s, int) for s in (start,stop,step)):
             # handle int slicing
-            # if we're slicing a symbolic dimension into a int dimension, we can slice untill the bind size
+            # if we're slicing a symbolic dimension into a int dimension, we can slice until the bind size
             # TODO: right now this is using vmax instead of the bind size because jit doesnt update the bound value of the returned tensor
             if isinstance(size, UOp): size = int(size.vmax)
             *boundary, stride = index.indices(cast(SupportsIndex, size))