name         old time/op    new time/op    delta
IndexLoad-4    2.49ms ± 1%    2.18ms ± 3%  -12.39%  (p=0.000 n=9+10)

name         old alloc/op   new alloc/op   delta
IndexLoad-4    3.77MB ± 0%    3.49MB ± 0%   -7.50%  (p=0.000 n=8+9)

name         old allocs/op  new allocs/op  delta
IndexLoad-4     1.98k ± 0%     0.53k ± 0%  -72.96%  (p=0.000 n=10+10)

Code generated for

	copy(posb[2:], posBuf[i*6:])
	pos := int64(binary.BigEndian.Uint64(posb[:]))

was awful.

name         old time/op    new time/op    delta
IndexLoad-4    3.06ms ± 0%    2.49ms ± 1%  -18.71%  (p=0.000 n=9+10)

looks like slice modification inside loop is not well optimized yet.

name       old time/op  new time/op  delta
PyQuote-4   894ns ± 0%   861ns ± 0%  -3.78%  (p=0.008 n=5+5)

name       old time/op    new time/op    delta
PyQuote-4     950ns ± 0%     894ns ± 0%  -5.89%  (p=0.008 n=5+5)

name       old alloc/op   new alloc/op   delta
PyQuote-4     0.00B          0.00B         ~     (all equal)

name       old allocs/op  new allocs/op  delta
PyQuote-4      0.00           0.00         ~     (all equal)

name       old time/op    new time/op    delta
PyQuote-4     976ns ± 0%     950ns ± 0%  -2.64%  (p=0.008 n=5+5)

name       old alloc/op   new alloc/op   delta
PyQuote-4     0.00B          0.00B         ~     (all equal)

name       old allocs/op  new allocs/op  delta
PyQuote-4      0.00           0.00         ~     (all equal)

it is annoying to see this function high in fstail profiles taking up to 10%

name       old time/op    new time/op    delta
PyQuote-4    1.33µs ± 1%    0.98µs ± 0%  -26.77%  (p=0.008 n=5+5)

name       old alloc/op   new alloc/op   delta
PyQuote-4     0.00B          0.00B          ~     (all equal)

name       old allocs/op  new allocs/op  delta
PyQuote-4      0.00           0.00          ~     (all equal)