return
	}

	// Increment mp.locks when allocating. We are called within gcStart,
	// and thus must not trigger another gcStart via an allocation. gcStart
	// bails when allocating with locks held, so simulate that for these
	// allocations.
	//
	// TODO(prattmic): cleanup gcStart to use a more explicit "in gcStart"
	// check for bailing.
	mp := acquirem()
	ready := make(chan struct{}, 1)
	releasem(mp)

	if blocked {
		ev = traceEvGoCreateBlocked
	}
	tl.eventWriter(traceGoRunning, traceProcRunning).commit(ev, traceArg(newg.goid), tl.startPC(pc), tl.stack(2))
}

// GoStart emits a GoStart event.
//
// Must be called with a valid P.
func (tl traceLocker) GoStart() {
	gp := getg().m.curg
	pp := gp.m.p
	w := tl.eventWriter(traceGoRunnable, traceProcRunning)
	w = w.write(traceEvGoStart, traceArg(gp.goid), gp.trace.nextSeq(tl.gen))

		println("bad flushGen", flushGen, "in prepareForSweep; sweepgen", sg)
		throw("bad flushGen")
	}
	c.releaseAll()
	stackcache_clear(c)
	c.flushGen.Store(mheap_.sweepgen) // Synchronizes with gcStart

		}
	}
	mp.mallocing = 0
	releasem(mp)

	// Again, because this chunk counts toward heapLive, potentially trigger a GC.
	if t := (gcTrigger{kind: gcTriggerHeap}); t.test() {
		gcStart(t)
	}

	if debug.malloc {
		if inittrace.active && inittrace.id == getg().goid {
			// Init functions are executed sequentially in a single goroutine.
			inittrace.bytes += uint64(userArenaChunkBytes)
		}

	EvProcStart:         {"ProcStart", 5, false, []string{"thread"}, nil},
	EvProcStop:          {"ProcStop", 5, false, []string{}, nil},
	EvGCStart:           {"GCStart", 5, true, []string{"seq"}, nil}, // in 1.5 format it was {}
	EvGCDone:            {"GCDone", 5, false, []string{}, nil},

		// of the GC accounts for bytes marked.
		assistG.gcAssistBytes -= int64(fullSize - dataSize)
	}

	if shouldhelpgc {
		if t := (gcTrigger{kind: gcTriggerHeap}); t.test() {
			gcStart(t)
		}
	}

	if raceenabled && noscan && dataSize < maxTinySize {
		// Pad tinysize allocations so they are aligned with the end
		// of the tinyalloc region. This ensures that any arithmetic

GoStart dt=5 g=53 g_seq=4
GoLabel dt=1 label_string=2
GoBlock dt=1751 reason_string=15 stack=5
GoUnblock dt=12 g=53 g_seq=5 stack=0
GoStart dt=6 g=53 g_seq=6
GoLabel dt=3 label_string=2
GoBlock dt=119 reason_string=15 stack=5
GoStart dt=15 g=88 g_seq=4
GoBlock dt=50 reason_string=13 stack=11
GoUnblock dt=1212 g=54 g_seq=15 stack=0
GoStart dt=6 g=54 g_seq=16
GoLabel dt=1 label_string=4

Frequency freq=15625000
EventBatch gen=1 m=1709048 time=7689670869319 size=423
ProcStart dt=409 p=7 p_seq=1
GoStart dt=31 g=34 g_seq=1
GoStop dt=291990 reason_string=16 stack=50
GoStart dt=21 g=34 g_seq=2
GoStop dt=315853 reason_string=16 stack=50
GoStart dt=30 g=34 g_seq=3
GoUnblock dt=173432 g=1 g_seq=73 stack=52
GoDestroy dt=96
GoStart dt=22 g=1 g_seq=74
HeapAlloc dt=79 heapalloc_value=26397576
HeapAlloc dt=51 heapalloc_value=26405640

EventBatch gen=1 m=1986496 time=2753925246280 size=267
ProcStart dt=549 p=0 p_seq=1
GoStart dt=211 g=18 g_seq=1
GoBlock dt=3533 reason_string=12 stack=21
GoStart dt=41 g=21 g_seq=1
GoBlock dt=150 reason_string=10 stack=22
GoStart dt=93 g=20 g_seq=1
GoSyscallBegin dt=51 p_seq=2 stack=23
GoSyscallEnd dt=400
GoBlock dt=582 reason_string=15 stack=25
GoStart dt=26 g=23 g_seq=1
HeapAlloc dt=50 heapalloc_value=1646592

	// Use Go to start 4 goroutines each of which handles 1/4 of the
	// board.
	r = GoStart(0, x, y, 0, x / 2, 0, y / 2, a, n);
	assert(r.r0 == 0 && r.r1 == 100);	// test multiple returns
	r = GoStart(1, x, y, x / 2, x, 0, y / 2, a, n);
	assert(r.r0 == 1 && r.r1 == 101);	// test multiple returns
	GoStart(2, x, y, 0, x / 2, y / 2, y, a, n);
	GoStart(3, x, y, x / 2, x, y / 2, y, a, n);
	GoWait(0);
	GoWait(1);
	GoWait(2);
	GoWait(3);
}