From 16b9cd193965769089881bb8ec012fccca7b37b6 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Johannes=20G=C3=A4=C3=9Fler?= <johannesg@5d6.de>
Date: Mon, 19 Jun 2023 10:23:56 +0200
Subject: Convert vector to f16 for dequantize mul mat vec (#1913)

* Convert vector to f16 for dmmv

* compile option

* Added compilation option description to README

* Changed cmake CUDA_ARCHITECTURES from "OFF" to "native"
---
 Makefile | 3 +++
 1 file changed, 3 insertions(+)

(limited to 'Makefile')

diff --git a/Makefile b/Makefile
index afd06e0..5dd676f 100644
--- a/Makefile
+++ b/Makefile
@@ -169,6 +169,9 @@ ifdef LLAMA_CUDA_DMMV_Y
 else
 	NVCCFLAGS += -DGGML_CUDA_DMMV_Y=1
 endif # LLAMA_CUDA_DMMV_Y
+ifdef LLAMA_CUDA_DMMV_F16
+	NVCCFLAGS += -DGGML_CUDA_DMMV_F16
+endif # LLAMA_CUDA_DMMV_F16
 ifdef LLAMA_CUDA_KQUANTS_ITER
 	NVCCFLAGS += -DK_QUANTS_PER_ITERATION=$(LLAMA_CUDA_KQUANTS_ITER)
 else
-- 
cgit v1.2.3