Use LLVM args to handle -malign-double option

Use LLVM args to handle -malign-double option

We previously used CPU features to pass the align-double option from
the driver to the frontend, but these passed to backend causing
spurious warnings. The implementation now uses LLVM args instead
which doesn't have the same side-effect.

BUG=none
R=dschuff@chromium.org

Committed: 605ea9bc25022747e435e57ee7b578e1fc220a63

[Petr Hosek, 2015-09-09 16:03:05]

Patchset #1 (id:1) has been deleted

[Derek Schuff, 2015-09-09 22:45:56]

https://chromiumcodereview.appspot.com/1323073010/diff/20001/lib/Basic/Target...
File lib/Basic/Targets.cpp (right):

https://chromiumcodereview.appspot.com/1323073010/diff/20001/lib/Basic/Target...
lib/Basic/Targets.cpp:7305: if (!Target->handleLLVMArgs(Opts->LLVMArgs))
LOCALMOD

https://chromiumcodereview.appspot.com/1323073010/diff/20001/test/Driver/mali...
File test/Driver/malign-double.c (right):

https://chromiumcodereview.appspot.com/1323073010/diff/20001/test/Driver/mali...
test/Driver/malign-double.c:6: // CHECK-I686-NOT: backend data layout
'e-m:e-p:32:32-i64:64-f80:32-n8:16:32-S128' does not match expected target
description 'e-m:e-p:32:32-f64:32:64-f80:32-n8:16:32-S128'
could we maybe reduce this so it still fires even if different datalayout
strings are printed?

[Petr Hosek, 2015-09-09 23:00:24]

https://chromiumcodereview.appspot.com/1323073010/diff/20001/lib/Basic/Target...
File lib/Basic/Targets.cpp (right):

https://chromiumcodereview.appspot.com/1323073010/diff/20001/lib/Basic/Target...
lib/Basic/Targets.cpp:7305: if (!Target->handleLLVMArgs(Opts->LLVMArgs))
On 2015/09/09 22:45:56, Derek Schuff wrote:
> LOCALMOD

Done.

https://chromiumcodereview.appspot.com/1323073010/diff/20001/test/Driver/mali...
File test/Driver/malign-double.c (right):

https://chromiumcodereview.appspot.com/1323073010/diff/20001/test/Driver/mali...
test/Driver/malign-double.c:6: // CHECK-I686-NOT: backend data layout
'e-m:e-p:32:32-i64:64-f80:32-n8:16:32-S128' does not match expected target
description 'e-m:e-p:32:32-f64:32:64-f80:32-n8:16:32-S128'
On 2015/09/09 22:45:56, Derek Schuff wrote:
> could we maybe reduce this so it still fires even if different datalayout
> strings are printed?

Regex to the rescue!

[Derek Schuff, 2015-09-09 23:02:28]

https://chromiumcodereview.appspot.com/1323073010/diff/40001/lib/Frontend/Com...
File lib/Frontend/CompilerInvocation.cpp (right):

https://chromiumcodereview.appspot.com/1323073010/diff/40001/lib/Frontend/Com...
lib/Frontend/CompilerInvocation.cpp:1829: Opts.LLVMArgs =
Args.getAllArgValues(OPT_mllvm);
LOCALMOD here too

https://chromiumcodereview.appspot.com/1323073010/diff/40001/test/Driver/mali...
File test/Driver/malign-double.c (right):

https://chromiumcodereview.appspot.com/1323073010/diff/40001/test/Driver/mali...
test/Driver/malign-double.c:8: // RUN: %clang -v -c -malign-double %s -target
x86_64-unknown-linux 2>&1 \
-malign-double will have no effect on x86-64. Maybe we should ignore it rather
than passing it through?

[Petr Hosek, 2015-09-09 23:45:24]

https://chromiumcodereview.appspot.com/1323073010/diff/40001/lib/Frontend/Com...
File lib/Frontend/CompilerInvocation.cpp (right):

https://chromiumcodereview.appspot.com/1323073010/diff/40001/lib/Frontend/Com...
lib/Frontend/CompilerInvocation.cpp:1829: Opts.LLVMArgs =
Args.getAllArgValues(OPT_mllvm);
On 2015/09/09 23:02:28, Derek Schuff wrote:
> LOCALMOD here too

Done.

https://chromiumcodereview.appspot.com/1323073010/diff/40001/test/Driver/mali...
File test/Driver/malign-double.c (right):

https://chromiumcodereview.appspot.com/1323073010/diff/40001/test/Driver/mali...
test/Driver/malign-double.c:8: // RUN: %clang -v -c -malign-double %s -target
x86_64-unknown-linux 2>&1 \
On 2015/09/09 23:02:28, Derek Schuff wrote:
> -malign-double will have no effect on x86-64. Maybe we should ignore it rather
> than passing it through?

Done by rejecting -mno-align-double as per our discussion in person.

[Derek Schuff, 2015-09-10 16:16:38]

lgtm

[Petr Hosek, 2015-09-12 20:07:46]

Committed patchset #3 (id:60001) manually as
605ea9bc25022747e435e57ee7b578e1fc220a63 (presubmit successful).