In order to make metadata useful to the backend, I'd like to permit
intrinsics to be written that take metadata-typed objects. To allow
this, I decided to try adding a new machine value type right next to isVoid.
I updated both ValueTypes.td and ValueTypes.h as well as
TableGen/IntrinsicEmitter.cpp and TableGen/CodeGenTarget.cpp where they
have switch(mvt). The actual patch is attached.
This runs into trouble with the X86 backend where the SSE2 forms of ADD
are defined:
llvm[3]: Building X86.td DAG instruction selector implementation with tblgen
ADDPDrm: (set:isVoid VR128:isFP:$dst, (fadd:isFP
VR128:isFP:$src1, (ld:isFP
addr:iPTR:$src2)<<P:Predicate_unindexedload>><<P:Predicate_load>><<P:Predicate_memop>>))
/home/nicholas/llvm-commit/Debug/bin/tblgen: error:
Included from X86.td:128:
Included from X86InstrInfo.td:3964:
Parsing X86InstrSSE.td:1320: In ADDPDrm: Could not infer all types in
pattern!
defm ADD : basic_sse2_fp_binop_rm<0x58, "add", fadd,
int_x86_sse2_add_sd, 1>;
^
I'm hoping someone with backend experience understands why this would
happen. I've looked over the definition of basic_sse2_fp_binop_rm and
can't see any way in which adding MVT::isMetadata should impact it or
the type inference step.
I'm stumped. Does anyone have any suggestions?
Nick
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mvt-metadata.patch
Type: text/x-patch
Size: 7893 bytes
Desc: not available
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20090616/e5cccc31/attachment.bin>
