From 55c5708b761e2b99e426aaef87e608c408d8c916 Mon Sep 17 00:00:00 2001 From: Spencer Bryngelson Date: Thu, 1 Oct 2026 10:20:16 -0500 Subject: [PATCH] Experiment: fixed device-array bounds for every compiler (USING_AMD = True) Measures in CI whether the compile-time bounds behind the USING_AMD guards speed up non-case-optimized NVHPC, CCE, and GNU builds as they do amdflang (3.8-5.3x on AFAR 24.3). Not for merge. --- src/common/include/shared_parallel_macros.fpp | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/src/common/include/shared_parallel_macros.fpp b/src/common/include/shared_parallel_macros.fpp index 0882db1718..155837ce57 100644 --- a/src/common/include/shared_parallel_macros.fpp +++ b/src/common/include/shared_parallel_macros.fpp @@ -6,7 +6,7 @@ #:set USING_NVHPC = (MFC_COMPILER == NVIDIA_COMPILER_ID or MFC_COMPILER == PGI_COMPILER_ID) #:set USING_CCE = (MFC_COMPILER == CCE_COMPILER_ID) -#:set USING_AMD = (MFC_COMPILER == AMD_COMPILER_ID) +#:set USING_AMD = True #! experiment: fixed device-array bounds for every compiler #! Fallback extents the USING_AMD guards substitute for device-global array bounds when case #! optimization is off. They are not independent: sys_size counts the species, so AMD_SYS_SIZE_MAX