From 8a25bd41b3c1c9b1bdb983e67a22c5b1be32256d Mon Sep 17 00:00:00 2001
From: Kylin <y13308074856@gmail.com>
Date: Sun, 20 Aug 2023 00:07:50 +0800
Subject: [PATCH] ggml: support CUDA's half type for aarch64(#1455) support
 CUDA's half type for aarch64 in ggml_fp16_t definition

---
 ggml.h | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/ggml.h b/ggml.h
index bdbd12800..413bc1de5 100644
--- a/ggml.h
+++ b/ggml.h
@@ -255,9 +255,11 @@
 extern "C" {
 #endif
 
-#ifdef __ARM_NEON
+#if defined(__ARM_NEON) && !defined(GGML_CUDA_F16)
     // we use the built-in 16-bit float type
     typedef __fp16 ggml_fp16_t;
+#elif defined(GGML_CUDA_F16)
+    typedef half ggml_fp16_t;
 #else
     typedef uint16_t ggml_fp16_t;
 #endif