From: Johannes Gäßler <redacted>
Date: Mon, 11 Sep 2023 11:00:24 +0000 (+0200)
Subject: CUDA: add device number to error messages (#3112)
X-Git-Tag: gguf-v0.4.0~94
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=8a4ca9af569853023ce87f047eb5165df13f2ff1;p=pkg%2Fggml%2Fsources%2Fllama.cpp

CUDA: add device number to error messages (#3112)
---

diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index 00e9bbea..50344ae8 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -144,8 +144,11 @@ static_assert(sizeof(half) == sizeof(ggml_fp16_t), "wrong fp16 size");
     do {                                                                                \
         cudaError_t err_ = (err);                                                       \
         if (err_ != cudaSuccess) {                                                      \
-            fprintf(stderr, "CUDA error %d at %s:%d: %s\n", err_, __FILE__, __LINE__,   \
+            int id;                                                                     \
+            cudaGetDevice(&id);                                                         \
+            fprintf(stderr, "\nCUDA error %d at %s:%d: %s\n", err_, __FILE__, __LINE__, \
                 cudaGetErrorString(err_));                                              \
+            fprintf(stderr, "current device: %d\n", id);                                \
             exit(1);                                                                    \
         }                                                                               \
     } while (0)
@@ -155,8 +158,11 @@ static_assert(sizeof(half) == sizeof(ggml_fp16_t), "wrong fp16 size");
     do {                                                                                \
         cublasStatus_t err_ = (err);                                                    \
         if (err_ != CUBLAS_STATUS_SUCCESS) {                                            \
+            int id;                                                                     \
+            cudaGetDevice(&id);                                                         \
             fprintf(stderr, "\ncuBLAS error %d at %s:%d: %s\n",                         \
                     err_, __FILE__, __LINE__, cublasGetStatusString(err_));             \
+            fprintf(stderr, "current device: %d\n", id);                                \
             exit(1);                                                                    \
         }                                                                               \
     } while (0)
@@ -165,7 +171,10 @@ static_assert(sizeof(half) == sizeof(ggml_fp16_t), "wrong fp16 size");
     do {                                                                                \
         cublasStatus_t err_ = (err);                                                    \
         if (err_ != CUBLAS_STATUS_SUCCESS) {                                            \
+            int id;                                                                     \
+            cudaGetDevice(&id);                                                         \
             fprintf(stderr, "\ncuBLAS error %d at %s:%d\n", err_, __FILE__, __LINE__);  \
+            fprintf(stderr, "current device: %d\n", id);                                \
             exit(1);                                                                    \
         }                                                                               \
     } while (0)