=== Performance improvements

Apart from skipping the configuration phase, the configuration cache provides some additional performance improvements:

- All tasks run in parallel by default, subject to dependency constraints.
- Dependency resolution is cached.
- Configuration state and dependency resolution state is discarded from heap after writing the task graph. This reduces the peak heap usage required for a given set of tasks.

		rk := p.From.Reg
		rj := p.To.Reg
		rd := p.RegTo2

		// See section 2.2.7.1 of https://loongson.github.io/LoongArch-Documentation/LoongArch-Vol1-EN.html
		// for the register usage constraints.
		if rd == rj || rd == rk {
			c.ctxt.Diag("illegal register combination: %v\n", p)
		}
		o1 = OP_RRR(atomicInst[p.As], uint32(rk), uint32(rj), uint32(rd))
	}

	out[0] = o1
	out[1] = o2

// they're fresh from the operating system. It updates heapArena metadata that is
// critical for future page allocations.
//
// There are no locking constraints on this method.
func (h *mheap) allocNeedsZero(base, npage uintptr) (needZero bool) {
	for npage > 0 {
		ai := arenaIndex(base)
		ha := h.arenas[ai.l1()][ai.l2()]