EngineX/xc-llm-ascend
3
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[BugFix]hccl bufferSize check for dispatch_ffn_combine (#6130)

Browse Source 
### What this PR does / why we need it?
dispatch_ffn_combine use hccl buffer as shared buffer, if hccl buffer
not enough，operator will error with "MTE out of range"
now add check for hccl buffer size, if not enough, will prompt "hccl
buffer is too small" and indicate what the expectation is.

### Does this PR introduce _any_ user-facing change?
N/A

### How was this patch tested?

- vLLM version: v0.13.0
- vLLM main:
d68209402d

---------

Signed-off-by: lhchg <lhao_cheng@163.com>
This commit is contained in:
lhchg
2026-01-23 08:41:40 +08:00
committed by GitHub
parent 7725314b26
commit 27a513b672
1 changed files with 35 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
csrc/dispatch_ffn_combine/op_host/dispatch_ffn_combine_tiling.cpp
View File

@@ -26,6 +26,8 @@
using namespace AscendC;
using namespace ge;
#define HCCL_BUFFSIZE "HCCL_BUFFSIZE"
namespace {
// 1. Constant definitions
const char *K_INNER_DEBUG = "DispatchFFNCombine Tiling Debug";
@@ -42,6 +44,7 @@ namespace {
constexpr uint32_t EXPERTID_INDEX = 3;
constexpr uint32_t BLOCK_NUM = 20;
constexpr uint32_t SYSTEM_NEED_WORKSPACE = 16 * 1024 * 1024;
constexpr uint64_t MB_SIZE = 1024 * 1024UL;
}
namespace optiling {
@@ -54,6 +57,30 @@ static int32_t CeilDev(int32_t num, int32_t div)
return (num + div - 1) / div;
}
static uint64_t GetMaxWindowSize()
{
uint16_t defaultWindowSize = 200;
const char* hccl_buffsize_env = getenv(HCCL_BUFFSIZE);
if (hccl_buffsize_env != nullptr) {
try {
std::string envStr(hccl_buffsize_env);
unsigned long val = std::stoul(envStr);
if (val <= std::numeric_limits<uint16_t>::max()) {
defaultWindowSize = static_cast<uint16_t>(val);
} else {
OP_LOGW(K_INNER_DEBUG, "HCCL_BUFFSIZE value %lu is out of range, using default.", val);
}
} catch (const std::exception& e) {
OP_LOGE(K_INNER_DEBUG, "Exception encountered when parsing env HCCL_BUFFSIZE: %s", e.what());
}
} else {
OP_LOGD(K_INNER_DEBUG, "Env HCCL_BUFFSIZE not set");
}
const uint64_t maxWindowSize = static_cast<uint64_t>(defaultWindowSize) * MB_SIZE;
OP_LOGD(K_INNER_DEBUG, "Get maxWindowSize is %lu", maxWindowSize);
return maxWindowSize;
}
// Parse and validate rankId, group, worldSize, and isTransB attributes
static ge::graphStatus DispatchFFNCombineCheckAttrAndSetTiling(gert::TilingContext *context, DispatchFFNCombineInfo& info)
{
@@ -232,6 +259,14 @@ static ge::graphStatus DispatchFFNCombineTilingFuncImpl(gert::TilingContext *con
tilingData->cocTiling.moeInitRoutingQuantV2TilingData.gatherOutComputeParamsOp = moeInitRoutingQuantV2TilingBase.quantTilingData.gatherOutComputeParamsOp;
tilingData->cocTiling.initRoutingQuantTilingKey = initRoutingQuantTilingKey;
uint64_t maxWindowSize = GetMaxWindowSize();
uint64_t actualSize = static_cast<uint64_t>(info.M) * info.topK * info.K * sizeof(int8_t) * 3 + 10 * MB_SIZE ;
OP_TILING_CHECK((actualSize > maxWindowSize),
OP_LOGE(nodeName, "HCCL_BUFFSIZE is too SMALL, m = %lu, k = %lu, topK = %lu"
" expected HCCL_BUFFSIZE is ((m * k * topK * sizeof(int8_t)) * 3 + 3MB)= %luMB, HCCL_BUFFSIZE=%luMB.",
info.M, info.K, info.topK, (actualSize + MB_SIZE - 1) / MB_SIZE, maxWindowSize / MB_SIZE),
return ge::GRAPH_FAILED);
// 4. workspace
size_t *workSpaces = context->GetWorkspaceSizes(1);
OP_TILING_CHECK(workSpaces == nullptr, OP_LOGE(nodeName, "workSpaces is nullptr."),