aboutsummaryrefslogtreecommitdiffhomepage
path: root/ptx
diff options
context:
space:
mode:
authorAndrzej Janik <[email protected]>2024-11-02 16:07:44 +0100
committerGitHub <[email protected]>2024-11-02 16:07:44 +0100
commit970ba5aa25b8f21b97176650bd3b6c034407e843 (patch)
treeed4ceea07824d5aa9d3d4e16e6b1d05e439d4793 /ptx
parentb4cb3ade63af94ccb709f2c0858253b13125fcc6 (diff)
downloadZLUDA-970ba5aa25b8f21b97176650bd3b6c034407e843.tar.gz
ZLUDA-970ba5aa25b8f21b97176650bd3b6c034407e843.zip
Fix linking of AMD device libraries (#296)
It's weird that it fails without `-mno-link-builtin-bitcode-postopt`. I've tested it only on ROCm 6.2, might be broken on older or newer ROCm
Diffstat (limited to 'ptx')
-rw-r--r--ptx/lib/zluda_ptx_impl.cpp4
1 files changed, 2 insertions, 2 deletions
diff --git a/ptx/lib/zluda_ptx_impl.cpp b/ptx/lib/zluda_ptx_impl.cpp
index f1b416d..d0ec853 100644
--- a/ptx/lib/zluda_ptx_impl.cpp
+++ b/ptx/lib/zluda_ptx_impl.cpp
@@ -1,5 +1,5 @@
-// Every time this file changes it must te rebuilt, you need llvm-17:
-// /opt/rocm/llvm/bin/clang -Wall -Wextra -Wsign-compare -Wconversion -x hip zluda_ptx_impl.cpp -nogpulib -O3 -mno-wavefrontsize64 -o zluda_ptx_impl.bc -emit-llvm -c --offload-device-only --offload-arch=gfx1010 && llvm-dis-17 zluda_ptx_impl.bc -o - | sed '/@llvm.used/d' | sed '/wchar_size/d' | sed '/llvm.module.flags/d' | sed 's/define hidden/define linkonce_odr/g' | sed 's/\"target-cpu\"=\"gfx1010\"//g' | sed -E 's/\"target-features\"=\"[^\"]+\"//g' | llvm-as-17 - -o zluda_ptx_impl.bc && llvm-dis-17 zluda_ptx_impl.bc
+// Every time this file changes it must te rebuilt, you need `rocm-llvm-dev` and `llvm-17`:
+// /opt/rocm/llvm/bin/clang -Wall -Wextra -Wsign-compare -Wconversion -x hip zluda_ptx_impl.cpp -nogpulib -O3 -mno-wavefrontsize64 -o zluda_ptx_impl.bc -emit-llvm -c --offload-device-only --offload-arch=gfx1010 && /opt/rocm/llvm/bin/llvm-dis zluda_ptx_impl.bc -o - | sed '/@llvm.used/d' | sed '/wchar_size/d' | sed '/llvm.module.flags/d' | sed 's/define hidden/define linkonce_odr/g' | sed 's/\"target-cpu\"=\"gfx1010\"//g' | sed -E 's/\"target-features\"=\"[^\"]+\"//g' | sed 's/ nneg / /g' | sed 's/ disjoint / /g' | llvm-as-17 - -o zluda_ptx_impl.bc && /opt/rocm/llvm/bin/llvm-dis zluda_ptx_impl.bc
#include <cstddef>
#include <cstdint>