From ec0ab2d2d0eb5b7be3a7fa4bb93953cf57a9e364 Mon Sep 17 00:00:00 2001
From: ademeure <arundemeure@gmail.com>
Date: Sun, 5 May 2024 00:00:58 +0100
Subject: [PATCH] Remove arch=native as it only available on recent CUDA
 versions

---
 Makefile | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/Makefile b/Makefile
index a755c30..05a02c3 100644
--- a/Makefile
+++ b/Makefile
@@ -15,7 +15,7 @@ CUDA_OUTPUT_FILE = -o $@
 # -t=0 is short for --threads, 0 = number of CPUs on the machine
 # include PTX for both SM52 (Maxwell) and SM80 (Ampere, our main optimisation target at the moment) + native SASS for current GPU
 # this increases compile time by ~5%, but we need >=SM70 PTX for some optimisations, and it allows "cuobjdump --dump-sass" to work
-NVCC_FLAGS = -O3 -t=0 --use_fast_math -gencode=arch=compute_52,code=compute_52 -gencode=arch=compute_80,code=compute_80 -arch=native
+NVCC_FLAGS = -O3 -t=0 --use_fast_math -gencode=arch=compute_52,code=compute_52 -gencode=arch=compute_80,code=sm_80
 NVCC_LDFLAGS = -lcublas -lcublasLt
 NVCC_INCLUDES =
 NVCC_LDLIBS =