Skip to content

GitLab

Commit 9d75245e authored by zhouxiang's avatar zhouxiang
Browse files

修改注释

parent a0b44ca8
Hide whitespace changes
Inline Side-by-side
Showing with 3 additions and 3 deletions
src/turbomind/utils/cublasMMWrapper.cc
View file @ 9d75245e
......@@ -40,7 +40,7 @@ cublasMMWrapper::cublasMMWrapper(cublasHandle_t cublas_handle,
if (allocator_ != nullptr) {
cublas_workspace_ = allocator_->reMalloc(cublas_workspace_, CUBLAS_WORKSPACE_SIZE, false);
}
# hgemm-switch 0:fp32r,1:fp16r-fp32r,2:fp16r ----xzhou 20240427
// hgemm-switch 0:fp32r,1:fp16r-fp32r,2:fp16r ----xzhou 20240427
m_ihgemm_switch = 0;
const char* env_var_value_str = std::getenv("LMDEPLOY_HGEMM_SWITCH");
if (env_var_value_str != nullptr) {
......@@ -125,7 +125,7 @@ void cublasMMWrapper::Gemm(cublasOperation_t transa,
{
TM_LOG_DEBUG(__PRETTY_FUNCTION__);
mu_->lock();
# hgemm-switch xzhou 20240427
// hgemm-switch ----xzhou 20240427
if(m_ihgemm_switch == 1 && (m == 5120 || m == 4096 || m == 12288 || m == 11008) && n <= m_ihgemm_switch_n && Atype == CUDA_R_16F){
computeType = CUDA_R_16F;
}
......@@ -189,7 +189,7 @@ void cublasMMWrapper::Gemm(cublasOperation_t transa,
mu_->lock();
// TODO: default cublas libs
cudaDataType_t computeType = computeType_;
# hgemm-switch xzhou 20240427
// hgemm-switch ------xzhou 20240427
if(m_ihgemm_switch == 1 && (m == 5120 || m == 4096 || m == 12288 || m == 11008) && n <= m_ihgemm_switch_n && Atype_ == CUDA_R_16F){
computeType = CUDA_R_16F;
}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment