[libc] Disable atomic optimizations for `libc` AMDGPU builds

[jhuber6]

Summary

Recently the AMDGPU backend automatically enables a pass to optimize
atomics. This results in the LTO build taking about 10x longer in all
cases. For now we disable this by default as was the case before the
patch in D152649.

[pravinjagtap]

The D153261 should be fixing the issue. Also, We need to reverse this changes since the command line option amdgpu-atomic-optimizations will be removed completely through D153007. Now, -amdgpu-atomic-optimizer-strategy=None disables the atomic optimizer pass.

[jhuber6]

In D153232#4437296, @pravinjagtap wrote:

The D153261 should be fixing the issue. Also, We need to reverse this changes since the command line option amdgpu-atomic-optimizations will be removed completely through D153007. Now, -amdgpu-atomic-optimizer-strategy=None disables the atomic optimizer pass.

Thanks for the fix, I've removed this upstream.

libc/startup/gpu/amdgpu/CMakeLists.txt

 # Ensure that clang uses the correct linker for this object type.
 target_link_libraries(
   ${fq_name}
   PUBLIC
   "-mcpu=${LIBC_GPU_TARGET_ARCHITECTURE}"
   "--target=${LIBC_GPU_TARGET_TRIPLE}"
   "-flto"
   "-Wl,-mllvm,-amdgpu-lower-global-ctor-dtor=0"
+  # FIXME: This pass causes builds to take ~10x longer. Disable until the
+  # performance problems have been addressed.
+  "-Wl,-mllvm,-amdgpu-atomic-optimizations=0"
 )