youngkingdom/vllm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[MISC] Remove unused variableds in C++ (#19609)

Browse Source 
Signed-off-by: Lu Fang <lufang@fb.com>
This commit is contained in:
Lu Fang
2025-06-16 11:05:28 +08:00
committed by GitHub
parent a5e7242d5f
commit c6703d1e0d
6 changed files with 2 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
csrc/quantization/fp8/amd/quant_utils.cuh
View File

@ -446,8 +446,6 @@ scaled_vec_conversion<uint16_t, uint8_t>(const uint8_t& a, float scale) {
template <>
__inline__ __device__ uint32_t
scaled_vec_conversion<uint32_t, uint16_t>(const uint16_t& a, float scale) {
[[maybe_unused]] __half2_raw h2r =
__hip_cvt_fp8x2_to_halfraw2(a, fp8_type::__default_interpret);
union {
__half2_raw h2r;
uint32_t ui32;

8
csrc/quantization/gptq/q_gemm.cu
View File

@ -206,8 +206,6 @@ __global__ void gemm_half_q_half_gptq_4bit_kernel(
auto offset_m = blockIdx.y * m_count;
auto offset_k = blockIdx.z * BLOCK_KN_SIZE;
[[maybe_unused]] int end_n = min(offset_n + BLOCK_KN_SIZE * 4, size_n);
[[maybe_unused]] int end_m = min(offset_m + m_count, size_m);
int end_k = min(offset_k + BLOCK_KN_SIZE, size_k);
int n = offset_n + t * 4;
@ -344,8 +342,6 @@ __global__ void gemm_half_q_half_gptq_2bit_kernel(
auto offset_m = blockIdx.y * m_count;
auto offset_k = blockIdx.z * BLOCK_KN_SIZE;
[[maybe_unused]] int end_n = min(offset_n + BLOCK_KN_SIZE * 4, size_n);
[[maybe_unused]] int end_m = min(offset_m + m_count, size_m);
int end_k = min(offset_k + BLOCK_KN_SIZE, size_k);
int n = offset_n + t * 4;
@ -465,8 +461,6 @@ __global__ void gemm_half_q_half_gptq_3bit_kernel(
auto offset_m = blockIdx.y * m_count;
auto offset_k = blockIdx.z * BLOCK_KN_SIZE;
[[maybe_unused]] int end_n = min(offset_n + BLOCK_KN_SIZE * 4, size_n);
[[maybe_unused]] int end_m = min(offset_m + m_count, size_m);
int end_k = min(offset_k + BLOCK_KN_SIZE, size_k);
int n = offset_n + t * 4;
@ -593,8 +587,6 @@ __global__ void gemm_half_q_half_gptq_8bit_kernel(
auto offset_m = blockIdx.y * m_count;
auto offset_k = blockIdx.z * BLOCK_KN_SIZE;
[[maybe_unused]] int end_n = min(offset_n + BLOCK_KN_SIZE * 4, size_n);
[[maybe_unused]] int end_m = min(offset_m + m_count, size_m);
int end_k = min(offset_k + BLOCK_KN_SIZE, size_k);
int n = offset_n + t * 4;