summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorAMD Toolchain Support <73240730+amd-toolchain-support@users.noreply.github.com>2024-08-23 01:54:58 +0100
committerGitHub <noreply@github.com>2024-08-22 17:54:58 -0700
commit8b811171c7aa5e63ee537aa29037d5b0541b18c7 (patch)
tree34bebfe1223e16a2432d2ea5ce473067c98350a0
parent823a2c1e4bf6ad2e2f7673ec42925f7868a9cf2c (diff)
downloadspack-8b811171c7aa5e63ee537aa29037d5b0541b18c7.tar.gz
spack-8b811171c7aa5e63ee537aa29037d5b0541b18c7.tar.bz2
spack-8b811171c7aa5e63ee537aa29037d5b0541b18c7.tar.xz
spack-8b811171c7aa5e63ee537aa29037d5b0541b18c7.zip
removing -Ofast with aocc (#45880)
Co-authored-by: shbhaska <shbhaska@amd.com>
-rw-r--r--var/spack/repos/builtin/packages/lammps/package.py9
1 files changed, 7 insertions, 2 deletions
diff --git a/var/spack/repos/builtin/packages/lammps/package.py b/var/spack/repos/builtin/packages/lammps/package.py
index 663adedcbd..c3d9e76109 100644
--- a/var/spack/repos/builtin/packages/lammps/package.py
+++ b/var/spack/repos/builtin/packages/lammps/package.py
@@ -824,15 +824,20 @@ class Lammps(CMakePackage, CudaPackage, ROCmPackage, PythonExtension):
if spec.satisfies("%aocc"):
if spec.satisfies("+intel"):
cxx_flags = (
- "-Ofast -fno-math-errno -fno-unroll-loops "
+ "-O3 -fno-math-errno -fno-unroll-loops "
"-fveclib=AMDLIBM -muse-unaligned-vector-move"
)
+ if spec.satisfies("%aocc@4.1:"):
+ cxx_flags += (
+ " -mllvm -force-gather-overhead-cost=50"
+ " -mllvm -enable-masked-gather-sequence=false"
+ )
# add -fopenmp-simd if OpenMP not already turned on
if spec.satisfies("~openmp"):
cxx_flags += " -fopenmp-simd"
cxx_flags += " -DLMP_SIMD_COMPILER -DUSE_OMP_SIMD -DLMP_INTEL_USELRT"
else:
- cxx_flags = "-Ofast -mfma -fvectorize -funroll-loops"
+ cxx_flags = "-O3 -mfma -fvectorize -funroll-loops"
args.append(self.define("CMAKE_CXX_FLAGS_RELEASE", cxx_flags))
args.append(self.define("CMAKE_CXX_FLAGS_RELWITHDEBINFO", cxx_flags))