Skip to content

[None][fix] Remove FP8 K/V buffer from TRTLLM sparse MLA attention kernel #9529

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
lfr-0531 merged 1 commit into from
Nov 28, 2025
Merged

[None][fix] Remove FP8 K/V buffer from TRTLLM sparse MLA attention kernel #9529

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 34 additions & 8 deletions cpp/tensorrt_llm/common/attentionOp.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -778,8 +778,19 @@ size_t AttentionOp::getWorkspaceSizeForContext(nvinfer1::DataType type, int32_t
if (mEnableContextFMHA && mFP8ContextMLA && mFmhaDispatcher->isSeparateQAndKvInput())
{
fp8_q_buf_size = max_num_tokens * static_cast<size_t>(total_q_dim_all_heads);
fp8_k_buf_size = mChunkPrefillBufferBatchSize * max_num_tokens * static_cast<size_t>(total_k_dim_all_heads);
fp8_v_buf_size = mChunkPrefillBufferBatchSize * max_num_tokens * static_cast<size_t>(total_v_dim_all_heads);

if (useSparseMLA())
{
// Sparse MLA (absorption mode): K and V are stored directly in KV cache during MLA RoPE kernel.
// No separate FP8 buffers needed for K/V since they're read from paged KV cache (Q_PAGED_KV layout).
fp8_k_buf_size = 0;
fp8_v_buf_size = 0;
}
else
{
fp8_k_buf_size = mChunkPrefillBufferBatchSize * max_num_tokens * static_cast<size_t>(total_k_dim_all_heads);
fp8_v_buf_size = mChunkPrefillBufferBatchSize * max_num_tokens * static_cast<size_t>(total_v_dim_all_heads);
}
}

size_t const padding_offset_size = mEnableContextFMHA ? 0 : sizeof(int) * max_num_tokens;
Expand Down Expand Up @@ -1436,8 +1447,19 @@ int AttentionOp::enqueueContext(EnqueueContextParams<T> const& params, cudaStrea
if (mEnableContextFMHA && mFP8ContextMLA && mFmhaDispatcher->isSeparateQAndKvInput())
{
fp8_q_buf_size = params.num_tokens * static_cast<size_t>(total_q_dim_all_heads);
fp8_k_buf_size = params.total_kv_len * static_cast<size_t>(total_k_dim_all_heads);
fp8_v_buf_size = params.total_kv_len * static_cast<size_t>(total_v_dim_all_heads);

if (useSparseMLA())
{
// Sparse MLA (absorption mode): K and V are stored directly in KV cache during MLA RoPE kernel.
// No separate FP8 buffers needed for K/V since they're read from paged KV cache (Q_PAGED_KV layout).
fp8_k_buf_size = 0;
fp8_v_buf_size = 0;
}
else
{
fp8_k_buf_size = params.total_kv_len * static_cast<size_t>(total_k_dim_all_heads);
fp8_v_buf_size = params.total_kv_len * static_cast<size_t>(total_v_dim_all_heads);
}
}
size_t const padding_offset_size
= mEnableContextFMHA ? 0 : sizeof(int) * params.batch_size * params.input_seq_length;
Expand Down Expand Up @@ -1805,11 +1827,15 @@ int AttentionOp::enqueueContext(EnqueueContextParams<T> const& params, cudaStrea
TLLM_CHECK_WITH_INFO(
mFmhaDispatcher->isSeparateQAndKvInput(), "Separate QKV input is required for fp8 context MLA");
TLLM_CHECK_WITH_INFO(fp8_q_buf != nullptr, "FP8 q buffer is required for fp8 context MLA");
TLLM_CHECK_WITH_INFO(fp8_k_buf != nullptr, "FP8 k buffer is required for fp8 context MLA");
TLLM_CHECK_WITH_INFO(fp8_v_buf != nullptr, "FP8 v buffer is required for fp8 context MLA");
// In sparse MLA (absorption mode), K and V are stored in KV cache, not as separate FP8 buffers
TLLM_CHECK_WITH_INFO(useSparseMLA() || fp8_k_buf != nullptr,
"FP8 k buffer is required for fp8 context MLA in non-sparse mode");
TLLM_CHECK_WITH_INFO(useSparseMLA() || fp8_v_buf != nullptr,
"FP8 v buffer is required for fp8 context MLA in non-sparse mode");

fmhaParams.qPtr = reinterpret_cast<void const*>(fp8_q_buf);
fmhaParams.kPtr = reinterpret_cast<void const*>(fp8_k_buf);
fmhaParams.vPtr = reinterpret_cast<void const*>(fp8_v_buf);
fmhaParams.kPtr = useSparseMLA() ? nullptr : reinterpret_cast<void const*>(fp8_k_buf);
fmhaParams.vPtr = useSparseMLA() ? nullptr : reinterpret_cast<void const*>(fp8_v_buf);
}
else
{
Expand Down