Skip to content

selective scan cuda kernel x buffer overflow fix #883

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

Open
darxradi3nt wants to merge 4 commits into
base: main
Choose a base branch
from
+5 −24
Open

selective scan cuda kernel x buffer overflow fix #883

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9b4a52b
fixes issue #880
darxradi3nt Mar 25, 2026
1871459
fixes issue #880
darxradi3nt Mar 25, 2026
1392be9
fixes issue #880
darxradi3nt Mar 26, 2026
d9a121d
fix: x buffer offset corrupts gradients silently for large batch/dsta…
darxradi3nt Mar 26, 2026
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
21 changes: 1 addition & 20 deletions csrc/selective_scan/selective_scan.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,27 +4,8 @@

#pragma once

////////////////////////////////////////////////////////////////////////////////////////////////////

struct SSMScanParamsBase {
using index_t = uint64_t;

int batch, seqlen, n_chunks;
index_t a_batch_stride;
index_t b_batch_stride;
index_t out_batch_stride;

// Common data pointers.
void *__restrict__ a_ptr;
void *__restrict__ b_ptr;
void *__restrict__ out_ptr;
void *__restrict__ x_ptr;
};

////////////////////////////////////////////////////////////////////////////////////////////////////

struct SSMParamsBase {
using index_t = uint32_t;
using index_t = uint64_t;

int batch, dim, seqlen, dstate, n_groups, n_chunks;
int dim_ngroups_ratio;
Expand Down
2 changes: 1 addition & 1 deletion csrc/selective_scan/selective_scan_bwd_kernel.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -139,7 +139,7 @@ void selective_scan_bwd_kernel(SSMParamsBwd params) {
float delta_bias = params.delta_bias_ptr == nullptr ? 0 : reinterpret_cast<float *>(params.delta_bias_ptr)[dim_id];
scan_t *x = params.x_ptr == nullptr
? nullptr
: reinterpret_cast<scan_t *>(params.x_ptr) + (batch_id * params.dim + dim_id) * (params.n_chunks) * params.dstate;
: reinterpret_cast<scan_t *>(params.x_ptr) + (int64_t)(batch_id * params.dim + dim_id) * params.n_chunks * params.dstate;
float dD_val = 0;
float ddelta_bias_val = 0;

Expand Down
6 changes: 3 additions & 3 deletions csrc/selective_scan/selective_scan_fwd_kernel.cuh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -102,16 +102,16 @@ void selective_scan_fwd_kernel(SSMParamsBase params) {
const int batch_id = blockIdx.x;
const int dim_id = blockIdx.y;
const int group_id = dim_id / (params.dim_ngroups_ratio);
input_t *u = reinterpret_cast<input_t *>(params.u_ptr) + (int64_t)batch_id * params.u_batch_stride
+ (int64_t)dim_id * kNRows * params.u_d_stride;
input_t *u = reinterpret_cast<input_t *>(params.u_ptr) + batch_id * params.u_batch_stride
+ dim_id * kNRows * params.u_d_stride;
input_t *delta = reinterpret_cast<input_t *>(params.delta_ptr) + batch_id * params.delta_batch_stride
+ dim_id * kNRows * params.delta_d_stride;
weight_t *A = reinterpret_cast<weight_t *>(params.A_ptr) + dim_id * kNRows * params.A_d_stride;
weight_t *B = reinterpret_cast<weight_t *>(params.B_ptr) + dim_id * kNRows * params.B_d_stride;
input_t *Bvar = reinterpret_cast<input_t *>(params.B_ptr) + batch_id * params.B_batch_stride + group_id * params.B_group_stride;
weight_t *C = reinterpret_cast<weight_t *>(params.C_ptr) + dim_id * kNRows * params.C_d_stride;
input_t *Cvar = reinterpret_cast<input_t *>(params.C_ptr) + batch_id * params.C_batch_stride + group_id * params.C_group_stride;
scan_t *x = reinterpret_cast<scan_t *>(params.x_ptr) + (batch_id * params.dim + dim_id * kNRows) * params.n_chunks * params.dstate;
scan_t *x = reinterpret_cast<scan_t *>(params.x_ptr) + (int64_t)(batch_id * params.dim + dim_id * kNRows) * params.n_chunks * params.dstate;

float D_val[kNRows] = {0};
if (params.D_ptr != nullptr) {
Expand Down