CUDA/HIP: add warp_size to cuda_device_info

ggml/src/ggml-cuda/common.cuh

@@ -520,6 +520,7 @@ struct ggml_cuda_device_info {
         bool    vmm;                // virtual memory support
         size_t  vmm_granularity;    // granularity of virtual memory
         size_t  total_vram;
+        int     warp_size;          // Number of threads in a dispatch
     };
 
     cuda_device_info devices[GGML_CUDA_MAX_DEVICES] = {};

ggml/src/ggml-cuda/ggml-cuda.cu

@@ -242,6 +242,7 @@ static ggml_cuda_device_info ggml_cuda_init() {
 
         info.devices[id].nsm   = prop.multiProcessorCount;
         info.devices[id].smpb  = prop.sharedMemPerBlock;
+        info.devices[id].warp_size = prop.warpSize;
 #if defined(GGML_USE_HIP) && defined(__HIP_PLATFORM_AMD__)
         info.devices[id].smpbo = prop.sharedMemPerBlock;
 
@@ -256,8 +257,9 @@ static ggml_cuda_device_info ggml_cuda_init() {
                 info.devices[id].cc += prop.minor * 0x10;
             }
         }
-        GGML_LOG_INFO("  Device %d: %s, %s (0x%x), VMM: %s\n",
-                        id, prop.name, prop.gcnArchName, info.devices[id].cc & 0xffff, device_vmm ? "yes" : "no");
+        GGML_LOG_INFO("  Device %d: %s, %s (0x%x), VMM: %s, Wave Size: %d\n",
+                      id, prop.name, prop.gcnArchName, info.devices[id].cc & 0xffff,
+                      device_vmm ? "yes" : "no", prop.warpSize);
 #else
         info.devices[id].smpbo = prop.sharedMemPerBlockOptin;
         info.devices[id].cc = 100*prop.major + 10*prop.minor;