Flang: __ARM_ARCH and friends defined for AArch64 #30
Conversation
|
@vjayathirtha-nv, please review. Thanks! |
| UpperCmdArgs.push_back("-def"); UpperCmdArgs.push_back("__LONG_MAX__=9223372036854775807L"); | ||
| UpperCmdArgs.push_back("-def"); UpperCmdArgs.push_back("__SIZE_TYPE__=unsigned long int"); | ||
| UpperCmdArgs.push_back("-def"); UpperCmdArgs.push_back("__PTRDIFF_TYPE__=long int"); | ||
| switch (getToolChain().getEffectiveTriple().getArch()) { |
There was a problem hiding this comment.
The lines common to all the cases can be factored out.
test/Driver/arch-specific-macros.f
Outdated
| @@ -0,0 +1,15 @@ | |||
| ! RUN: %flang -target aarch64 -v -x f77 -### -c - 2>&1 >/dev/null | FileCheck --check-prefix=AARCH64 %s | |||
There was a problem hiding this comment.
This test file is useless anyway. It uses %flang as it assumes that pull request flang-compiler/clang#20 was pulled into frontend driver. Now as flang moved to release_60 and affected configuration file lit.cfg migrated to lit.cfg.py (with significant rewrite and different ideas about tool discovery), referenced pull request won't apply any longer.
test/Driver/arch-specific-macros.f
Outdated
| ! AARCH64: __aarch64__ | ||
| ! AARCH64: __ARM_ARCH=8 | ||
| ! AARCH64: __ARM_ARCH__=8 | ||
| ! AARCH64-NOT: __x86_64__ |
There was a problem hiding this comment.
Flang can be built on OpenPower too.
AARCH64-NOT should not imply __X86_64__.
Vice-versa for X86_64-NOT
There was a problem hiding this comment.
That's not what this test case tests. This test case tests only what this patch introduces: awareness of x86_64 and aarch64. Iff soneone adds OpenPOWER, this test will have to be amended.
…s defined for x86_64 only Signed-off-by: Paul Osmialowski <[email protected]>
|
Simplified version uploaded. |
|
LGTM. |
|
@gklimowicz You'll want to update the driver to support OpenPower. |
3 participants
No description provided.