more readable and faster

tinygrad/ops/ops_gpu.py

@@ -14,19 +14,19 @@ class UnaryOp(Function):
 
   def backward(ctx, grad_output):
     input, = ctx.saved_tensors
-    return binary_op(ctx, ctx.bop, grad_output, input)
+    return binary_op(ctx, ctx.bop, input, grad_output)
 
 class ReLU(UnaryOp):
   fop = 'max(a, (float)0.)'
-  bop = 'a * (b >= 0)'
+  bop = 'b * (a >= 0)'
 
 class Log(UnaryOp):
   fop = 'log(a)'
-  bop = 'a / b'
+  bop = 'b / a'
 
 class Exp(UnaryOp):
   fop = 'exp(a)'
-  bop = 'a * exp(b)'
+  bop = 'b * exp(a)'
 
 # ************* reduce ops *************
 
@@ -37,7 +37,8 @@ class Sum(Function):
 
   def backward(ctx, grad_output):
     shape_input, = ctx.saved_tensors
-    return binary_op(ctx, 'a+b', grad_output, buffer_new(ctx, shape_input, zero=True))
+    # NOTE: the b buffer_new isn't used, since this is just for broadcast
+    return binary_op(ctx, 'a', grad_output, buffer_new(ctx, shape_input))
 
 class Max(Function):
   def forward(ctx, input, axis=None):