Clean up the handling of the x87 fp stack to make it more robust.
Drop the FpMov instructions, use plain COPY instead.
Drop the FpSET/GET instruction for accessing fixed stack positions.
Instead use normal COPY to/from ST registers around inline assembly, and
provide a single new FpPOP_RETVAL instruction that can access the return
value(s) from a call. This is still necessary since you cannot tell from
the CALL instruction alone if it returns anything on the FP stack. Teach
fast isel to use this.
This provides a much more robust way of handling fixed stack registers -
we can tolerate arbitrary FP stack instructions inserted around calls
and inline assembly. Live range splitting could sometimes break x87 code
by inserting spill code in unfortunate places.
As a bonus we handle floating point inline assembly correctly now.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@134018 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp
index 82f5d3a..a4d1c67 100644
--- a/lib/Target/X86/X86FastISel.cpp
+++ b/lib/Target/X86/X86FastISel.cpp
@@ -1848,16 +1848,18 @@
// stack, but where we prefer to use the value in xmm registers, copy it
// out as F80 and use a truncate to move it from fp stack reg to xmm reg.
if ((RVLocs[i].getLocReg() == X86::ST0 ||
- RVLocs[i].getLocReg() == X86::ST1) &&
- isScalarFPTypeInSSEReg(RVLocs[0].getValVT())) {
- CopyVT = MVT::f80;
+ RVLocs[i].getLocReg() == X86::ST1)) {
+ if (isScalarFPTypeInSSEReg(RVLocs[i].getValVT()))
+ CopyVT = MVT::f80;
CopyReg = createResultReg(X86::RFP80RegisterClass);
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(X86::FpPOP_RETVAL),
+ CopyReg);
+ } else {
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(TargetOpcode::COPY),
+ CopyReg).addReg(RVLocs[i].getLocReg());
+ UsedRegs.push_back(RVLocs[i].getLocReg());
}
- BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DL, TII.get(TargetOpcode::COPY),
- CopyReg).addReg(RVLocs[i].getLocReg());
- UsedRegs.push_back(RVLocs[i].getLocReg());
-
if (CopyVT != RVLocs[i].getValVT()) {
// Round the F80 the right size, which also moves to the appropriate xmm
// register. This is accomplished by storing the F80 value in memory and