more ops on the SparseCore. Ops are broken up into:
  1. SC forward pass
  2. TC forward/backward pass
  3. SC backward pass
  4. non-TPU loop counter updates
These 4 functions are then staggered so as to enable parallel execution.

In pseudocode, the algorithm is as follows:

// Start step 0
C_0 = cond(args_0)
N_0 = non_tpu(args_0)
if (C_0) {
   F_0 = forward(args_0, N_0)
   T_0 = core_tpu(args_0, N_0, F_0)

		// and we want the concatenation of all of those logs.
		// This means, of course, that users need to rm the old log
		// to get fresh data.
		// TODO: all.bash runs compilers in parallel. Need to synchronize logging somehow?
		w, err := os.OpenFile(filepath.Join(os.Getenv("GOROOT"), "src", "rulelog"),
			os.O_CREATE|os.O_WRONLY|os.O_APPEND, 0666)
		if err != nil {
			panic(err)
		}

// (Is the function atomic?)
//
// For each function, we write a "hammer" function that repeatedly
// uses the atomic operation to add 1 to a value. After running
// multiple hammers in parallel, check that we end with the correct
// total.
// Swap can't add 1, so it uses a different scheme.
// The functions repeatedly generate a pseudo-random number such that

github.com/modern-go/reflect2 v1.0.1/go.mod h1:bx2lNnkwVCuqBIxFjflWJWanXIb3RllmbCylyMrvgv0=
github.com/modern-go/reflect2 v1.0.2/go.mod h1:yWuevngMOJpCy52FWWMvUC8ws7m/LJsjYzDa0/r8luk=
github.com/moricho/tparallel v0.3.1 h1:fQKD4U1wRMAYNngDonW5XupoB/ZGJHdpzrWqgyg9krA=
github.com/moricho/tparallel v0.3.1/go.mod h1:leENX2cUv7Sv2qDgdi0D0fCftN8fRC67Bcn8pqzeYNI=
github.com/mwitkow/go-conntrack v0.0.0-20161129095857-cc309e4a2223/go.mod h1:qRWi+5nqEBWmkhHvq77mSJWrCKwh8bxhgT7d/eI7P4U=