[llvm] 5fe5c0a - [X86] Preserve fpexcept property when turning strict_fp_extend and strict_fp_round into stack operations.

[Craig Topper via llvm-commits]

Author: Craig Topper
Date: 2020-01-10T23:41:06-08:00
New Revision: 5fe5c0a60f9a5f32da4316ba0d1732a1e439703b

URL: https://github.com/llvm/llvm-project/commit/5fe5c0a60f9a5f32da4316ba0d1732a1e439703b
DIFF: https://github.com/llvm/llvm-project/commit/5fe5c0a60f9a5f32da4316ba0d1732a1e439703b.diff

LOG: [X86] Preserve fpexcept property when turning strict_fp_extend and strict_fp_round into stack operations.

We use the stack for X87 fp_round and for moving from SSE f32/f64 to
X87 f64/f80. Or from X87 f64/f80 to SSE f32/f64.

Note for the SSE<->X87 conversions the conversion always happens in the
X87 domain. The load/store ops in the X87 instructions are able
to signal exceptions.

Added: 
    

Modified: 
    llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
    llvm/lib/Target/X86/X86InstrFPStack.td

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
index 201082fac57f..a621d7ed7408 100644
--- a/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
+++ b/llvm/lib/Target/X86/X86ISelDAGToDAG.cpp
@@ -1128,10 +1128,40 @@ void X86DAGToDAGISel::PreprocessISelDAG() {
       // FIXME: optimize the case where the src/dest is a load or store?
 
       //Since the operation is StrictFP, use the preexisting chain.
-      SDValue Store = CurDAG->getTruncStore(N->getOperand(0), dl, N->getOperand(1),
-                                MemTmp, MachinePointerInfo(), MemVT);
-      SDValue Result = CurDAG->getExtLoad(ISD::EXTLOAD, dl, DstVT, Store, MemTmp,
-                                          MachinePointerInfo(), MemVT);
+      SDValue Store, Result;
+      if (!SrcIsSSE) {
+        SDVTList VTs = CurDAG->getVTList(MVT::Other);
+        SDValue Ops[] = {N->getOperand(0), N->getOperand(1), MemTmp};
+        Store = CurDAG->getMemIntrinsicNode(X86ISD::FST, dl, VTs, Ops, MemVT,
+                                            MachinePointerInfo(), 0,
+                                            MachineMemOperand::MOStore);
+        if (N->getFlags().hasNoFPExcept()) {
+          SDNodeFlags Flags = Store->getFlags();
+          Flags.setNoFPExcept(true);
+          Store->setFlags(Flags);
+        }
+      } else {
+        assert(SrcVT == MemVT && "Unexpected VT!");
+        Store = CurDAG->getStore(N->getOperand(0), dl, N->getOperand(1), MemTmp,
+                                 MachinePointerInfo());
+      }
+
+      if (!DstIsSSE) {
+        SDVTList VTs = CurDAG->getVTList(DstVT, MVT::Other);
+        SDValue Ops[] = {Store, MemTmp};
+        Result = CurDAG->getMemIntrinsicNode(X86ISD::FLD, dl, VTs, Ops, MemVT,
+                                             MachinePointerInfo(), 0,
+                                             MachineMemOperand::MOLoad);
+        if (N->getFlags().hasNoFPExcept()) {
+          SDNodeFlags Flags = Result->getFlags();
+          Flags.setNoFPExcept(true);
+          Result->setFlags(Flags);
+        }
+      } else {
+        assert(DstVT == MemVT && "Unexpected VT!");
+        Result =
+            CurDAG->getLoad(DstVT, dl, Store, MemTmp, MachinePointerInfo());
+      }
 
       // We're about to replace all uses of the FP_ROUND/FP_EXTEND with the
       // extload we created.  This will cause general havok on the dag because

diff  --git a/llvm/lib/Target/X86/X86InstrFPStack.td b/llvm/lib/Target/X86/X86InstrFPStack.td
index 1bd9ea47c2e7..2cdf2ae15ce4 100644
--- a/llvm/lib/Target/X86/X86InstrFPStack.td
+++ b/llvm/lib/Target/X86/X86InstrFPStack.td
@@ -777,7 +777,10 @@ def FXRSTOR64 : RI<0xAE, MRM1m, (outs), (ins opaquemem:$src),
 
 // Required for RET of f32 / f64 / f80 values.
 def : Pat<(X86fldf32 addr:$src), (LD_Fp32m addr:$src)>;
+def : Pat<(X86fldf32 addr:$src), (LD_Fp32m64 addr:$src)>;
 def : Pat<(X86fldf64 addr:$src), (LD_Fp64m addr:$src)>;
+def : Pat<(X86fldf32 addr:$src), (LD_Fp32m80 addr:$src)>;
+def : Pat<(X86fldf64 addr:$src), (LD_Fp64m80 addr:$src)>;
 def : Pat<(X86fldf80 addr:$src), (LD_Fp80m addr:$src)>;
 
 // Required for CALL which return f32 / f64 / f80 values.