[cfe-commits] r134946 - in /cfe/trunk: lib/CodeGen/TargetInfo.cpp test/CodeGen/x86_64-arguments.c

[Bruno Cardoso Lopes]

Taking a look...! If I don't find a solution soon will revert! Thank you!

On Tue, Jul 12, 2011 at 1:32 PM, Eli Friedman <eli.friedman at gmail.com> wrote:
> On Tue, Jul 12, 2011 at 1:26 PM, Benjamin Kramer
> <benny.kra at googlemail.com> wrote:
>> On Mon, Jul 11, 2011 at 17:30, Bruno Cardoso Lopes
>> <bruno.cardoso at gmail.com> wrote:
>>> Author: bruno
>>> Date: Mon Jul 11 19:30:27 2011
>>> New Revision: 134946
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=134946&view=rev
>>> Log:
>>> Fix one x86_64 abi issue and the test to actually look for the right thing,
>>> which is: { <4 x float>, <4 x float> } should continue to go through memory.
>>>
>>> Modified:
>>>    cfe/trunk/lib/CodeGen/TargetInfo.cpp
>>>    cfe/trunk/test/CodeGen/x86_64-arguments.c
>>>
>>> Modified: cfe/trunk/lib/CodeGen/TargetInfo.cpp
>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/CodeGen/TargetInfo.cpp?rev=134946&r1=134945&r2=134946&view=diff
>>> ==============================================================================
>>> --- cfe/trunk/lib/CodeGen/TargetInfo.cpp (original)
>>> +++ cfe/trunk/lib/CodeGen/TargetInfo.cpp Mon Jul 11 19:30:27 2011
>>> @@ -1228,6 +1228,13 @@
>>>
>>>     const RecordDecl *RD = RT->getDecl();
>>>
>>> +    // The only case a 256-bit wide vector could be used is when the struct
>>> +    // contains a single 256-bit element. Since Lo and Hi logic isn't extended
>>> +    // to work for sizes wider than 128, early check and fallback to memory.
>>> +    RecordDecl::field_iterator FirstElt = RD->field_begin();
>>> +    if (Size > 128 && getContext().getTypeSize(FirstElt->getType()) != 256)
>>> +      return;
>>> +
>>>     // Assume variable sized types are passed in memory.
>>>     if (RD->hasFlexibleArrayMember())
>>>       return;
>>> @@ -1263,7 +1270,7 @@
>>>
>>>     // Classify the fields one at a time, merging the results.
>>>     unsigned idx = 0;
>>> -    for (RecordDecl::field_iterator i = RD->field_begin(), e = RD->field_end();
>>> +    for (RecordDecl::field_iterator i = FirstElt, e = RD->field_end();
>>>            i != e; ++i, ++idx) {
>>>       uint64_t Offset = OffsetBase + Layout.getFieldOffset(idx);
>>>       bool BitField = i->isBitField();
>>>
>>> Modified: cfe/trunk/test/CodeGen/x86_64-arguments.c
>>> URL: http://llvm.org/viewvc/llvm-project/cfe/trunk/test/CodeGen/x86_64-arguments.c?rev=134946&r1=134945&r2=134946&view=diff
>>> ==============================================================================
>>> --- cfe/trunk/test/CodeGen/x86_64-arguments.c (original)
>>> +++ cfe/trunk/test/CodeGen/x86_64-arguments.c Mon Jul 11 19:30:27 2011
>>> @@ -280,9 +280,9 @@
>>>  // Make sure that the struct below is passed in the same way
>>>  // regardless of avx being used
>>>  //
>>> -// CHECK: define void @func41(<2 x double> %s.coerce)
>>> +// CHECK: declare void @func40(%struct.t128* byval align 16)
>>>  typedef float __m128 __attribute__ ((__vector_size__ (16)));
>>> -typedef struct {
>>> +typedef struct t128 {
>>>   __m128 m;
>>>   __m128 n;
>>>  } two128;
>>
>>
>> Hi Bruno,
>>
>> Looks like this change broke selfhost on x86_64, here's a test case:
>
> Also breaking a couple tests from the nightly tests (specifically,
> SingleSource/Benchmarks/Misc-C++-EH/spirit and
> MultiSource/Benchmarks/Prolangs-C++/city/city).
>
> -Eli
>



-- 
Bruno Cardoso Lopes
http://www.brunocardoso.cc