ggml : support CUDA's half type for aarch64(#1455) (#2670)

* ggml: support CUDA's half type for aarch64(#1455) support CUDA's half type for aarch64 in ggml_fp16_t definition * ggml: use __CUDACC__ to recognise nvcc compiler
ggerganov · Aug 22, 2023 · 1e3bc52 · 1e3bc52
1 parent 14b1d7e
commit 1e3bc52
Showing 1 changed file with 3 additions and 2 deletions.
diff --git a/ggml.h b/ggml.h
@@ -259,8 +259,9 @@
 extern "C" {
 #endif
 
-#ifdef __ARM_NEON
-    // we use the built-in 16-bit float type
+#if defined(__ARM_NEON) && defined(__CUDACC__)
+    typedef half ggml_fp16_t;
+#elif defined(__ARM_NEON)
     typedef __fp16 ggml_fp16_t;
 #else
     typedef uint16_t ggml_fp16_t;