From 9414f93f53c3fb31e00546bbdb179a79737620fc Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Szymon=20=C5=81opaciuk?= Date: Mon, 27 Apr 2026 16:28:47 +0200 Subject: [PATCH] Enable -Oft=min on CUDA to decrease compile-time mem and time pressure --- xobjects/context_cupy.py | 4 ++++ 1 file changed, 4 insertions(+) diff --git a/xobjects/context_cupy.py b/xobjects/context_cupy.py index bd65115..715040b 100644 --- a/xobjects/context_cupy.py +++ b/xobjects/context_cupy.py @@ -466,6 +466,10 @@ def build_kernels( *extra_compile_args, *include_flags, "-DXO_CONTEXT_CUDA", + # Skip heavy optimizations (e.g. involving cloning), + # which for us don't translate to a lot of runtime gains, + # but consume a lot of compile time and memory: + "--Ofast-compile=min", ) module = cupy.RawModule(