Feature × Feature

Feature × Feature#

The tables below show mutually exclusive features and the support on Ascend hardware, extended from vLLM table.

The symbols used have the following meanings:

✅ = Full compatibility
🟠 = Partial compatibility
❌ = No compatibility
❔ = Unknown or TBD

Feature	ACLGraph Full_Decode_Only	ACLGraph Piecewise	Async Scheduling	APC	Chunked Prefill	Context Parallel	Cpu Binding	DP	Disaggregated Prefill	Eagle3	Eplb	EP	Flashcomm1	KV Cache Pool	Layer Sharding	Lmhead TP	Mlapo	mm	Multistream Moe	Shared Expert DP	Quantization W4A4	Quantization W4A8	Quantization W8A8	TP	Weight nz
ACLGraph Full_Decode_Only	✅
ACLGraph Piecewise	❌	✅
Async Scheduling	✅	✅	✅
APC	✅	✅	✅	✅
Chunked Prefill	✅	✅	✅	✅	✅
Context Parallel	✅	✅	✅	✅	✅	✅
Cpu Binding	✅	✅	✅	✅	✅	✅	✅
DP	✅	✅	✅	✅	✅	🟠¹	✅	✅
Disaggregated Prefill	✅	✅	✅	✅	✅	✅	✅	✅	✅
Eagle3	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
Eplb	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
EP	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
Flashcomm1	✅	✅	✅	✅	✅	✅	✅	✅	🟠²	✅	✅	✅	✅
KV Cache Pool	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
Layer Sharding	✅	✅	✅	✅	✅	🟠	✅	✅	🟠³	✅	✅	✅	✅	❔	✅
Lmhead TP	✅	✅	✅	✅	✅	❔	✅	🟠⁴	✅	✅	✅	✅	❌	❔	✅	✅
Mlapo	✅	✅	✅	✅	✅	✅	✅	✅	🟠⁵	✅	✅	✅	❌	❔	❌	✅	✅
mm	✅	✅	✅	✅	✅	🟠	✅	✅	✅	✅	✅	✅	✅	❌	✅	✅	✅	✅
Multistream Moe	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	❔	✅	✅	✅	✅
Shared Expert DP	✅	✅	✅	✅	✅	🟠¹	✅	✅	✅	✅	✅	✅	✅	✅	✅	❔	✅	✅	❔	✅
Quantization W4A4	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	❌	❔	❔	✅	❔	✅	❔	❌	❔	❔	✅
Quantization W4A8	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	❌	✅	✅	✅	❔	✅	❔	❌	✅	✅	❔	✅
Quantization W8A8	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	❔	✅	✅
TP	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅
Weight nz	✅	✅	✅	✅	✅	❔	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	✅	❌	🟠	✅	✅	✅

¹ Only dcp supports dp while pcp does not support dp.
² Falshcomm is only enabled on the prefill stage.
³ Layer sharding is only enabled on the prefill stage.
⁴ Lmhead TP is only enabled in the pure dp scenarios.
⁵ MLAPO is only supported on the decode stage.