amdgcn: Add waitcnt after LDS write instructions

amdgcn: Add waitcnt after LDS write instructions

Data-share write (ds_write) instructions do not necessarily complete
the write to LDS immediately. When a write completes, LGKM_CNT is
decremented. For now, we wait until LGKM_CNT reaches zero after each
ds_write instruction.

This fixes a race condition in the case where LDS is read immediately
after being written. This can happen with broadcast operations.

2020-07-15  Julian Brown  <julian@codesourcery.com>

gcc/
	* config/gcn/gcn-valu.md (scatter<mode>_insn_1offset_ds<exec_scatter>):
	Add waitcnt.
	(*mov<mode>_insn, *movti_insn): Add waitcnt to ds_write alternatives.