refactor: remove debug printf

Author: ZhangZhiPku

rtp_llm/cpp/cuda/allocator_cuda.cc

@@ -101,7 +101,6 @@ Allocator<AllocatorType::CUDA>::~Allocator() {
 
 void* Allocator<AllocatorType::CUDA>::mallocPhysical(size_t size) {
     RTP_LLM_LOG_DEBUG("malloc physical memory with size %lu\n", size);
-    printf("malloc physical %lu\n", size);
     auto address = doMallocSync(size);
     if (!address) {
         return nullptr;

rtp_llm/cpp/cuda/cutlass/cutlass_extensions/include/cutlass_extensions/gemm/kernel/fused_moe_kernel.cuh

@@ -176,7 +176,7 @@ __global__ void run_global(__grid_constant__ typename GemmType::Params const par
 
 /// Computes the maximum number of active blocks per multiprocessor
 template <typename GemmType>
-static int fused_gemm_maximum_active_blocks(int smem_capacity = -1)
+static int  (int smem_capacity = -1)
 {
 
     CUTLASS_TRACE_HOST("BaseGrouped::maximum_active_blocks()");

rtp_llm/cpp/devices/BufferManager.cc

@@ -70,9 +70,7 @@ BufferPtr BufferManager::doAllocate(const BufferParams& params, const BufferHint
         accumulate(shape.begin(), shape.end(), (size_t)1, std::multiplies<size_t>()) * getTypeSize(params.type);
 
     if (params.vmem_ctl == VmemCtl::ForcePhysical) {
-        printf("Calling here\n");
         if (auto vmem_allocator = dynamic_cast<IVirtualMemAllocator*>(allocator)) {
-            printf("Calling here 2\n");
             const auto data    = vmem_allocator->mallocPhysical(alloc_bytes);
             const auto deleter = [this, vmem_allocator](Buffer* buffer) { this->recycle(buffer, vmem_allocator); };
             const auto buffer  = new Buffer(vmem_allocator->memoryType(), params.type, shape, data, deleter);