simpler webgpu copyin [pr]

can directly construct the buffer instead of init then assign.
also put standard lib import together

tinygrad/runtime/ops_webgpu.py

@@ -1,9 +1,8 @@
-import functools
+import functools, struct
 from tinygrad.device import  Compiled, Allocator, Compiler
 from tinygrad.renderer.wgsl import WGSLRenderer
 from tinygrad.helpers import round_up
 import wgpu
-import struct
 
 def create_uniform(wgpu_device, val) -> wgpu.GPUBuffer:
   buf = wgpu_device.create_buffer(size=4, usage=wgpu.BufferUsage.UNIFORM | wgpu.BufferUsage.COPY_DST)
@@ -47,10 +46,8 @@ def __init__(self, dev): self.dev = dev
   def _alloc(self, size: int, options):
     return self.dev.create_buffer(size=round_up(size, 4), usage=wgpu.BufferUsage.STORAGE | wgpu.BufferUsage.COPY_DST | wgpu.BufferUsage.COPY_SRC)
   def _copyin(self, dest, src: memoryview):
-    if src.nbytes % 4:
-      padded_src = bytearray(round_up(src.nbytes, 4))
-      padded_src[:src.nbytes] = src
-    self.dev.queue.write_buffer(dest, 0, padded_src if src.nbytes % 4 else src)
+    if (pad := src.nbytes % 4): src = memoryview(bytearray(src) + bytearray(pad))
+    self.dev.queue.write_buffer(dest, 0, src)
   def _copyout(self, dest: memoryview, src):
     buffer_data = self.dev.queue.read_buffer(src, 0)
     dest[:] = buffer_data[:dest.nbytes] if src._nbytes > dest.nbytes else buffer_data