[PowerPC] Fix the unexpected modification caused by D62993 in LowerSELECT_CC for power9

[ZhangKang]

Summary

The patch D62993 : [PowerPC] Emit scalar min/max instructions with unsafe fp math has modified the functionality when Subtarget.hasP9Vector() && (!HasNoInfs || !HasNoNaNs), this modification is not expected.

For below case:

define double @test(double %a, double %b, double %c, double %d) {
entry:
  %cmp = fcmp fast oeq double %a, %b
  %cond = select fast i1 %cmp, double %c, double %d
  ret double %cond
}

Use the below command for pwr9 without the fast-math option,

llc -ppc-vsr-nums-as-vr -ppc-asm-full-reg-names test.ll -o p9.s -mcpu=pwr9 -O3

we will get below assembly, this is not right, because we haven't used the fast-math option :

asm
	xssubdp f0, f1, f2
	fsel f1, f0, f3, f4
	xsnegdp f0, f0
	fsel f1, f0, f1, f4
	blr

But in fact, if without the patch D62993, above case we will get below assembly, this is the right:

# %bb.0:                                # %entry
	xscmpudp cr0, f1, f2
	beq	cr0, .LBB0_2
# %bb.1:                                # %entry
	fmr f3, f4
.LBB0_2:                                # %entry
	fmr f1, f3
	blr

[ZhangKang]

@nemanjai , do you have any comments for this patch?

[nemanjai]

LGTM. Sorry I missed this for so long. Thank you for fixing it.

llvm/lib/Target/PowerPC/PPCISelLowering.cpp

default:
       return Op;
     case ISD::SETOGT:
     case ISD::SETGT:
       return DAG.getNode(PPCISD::XSMAXCDP, dl, Op.getValueType(), LHS, RHS);
     case ISD::SETOLT:
     case ISD::SETLT:
       return DAG.getNode(PPCISD::XSMINCDP, dl, Op.getValueType(), LHS, RHS);
     }
-  }
+  } else if (!HasNoInfs || !HasNoNaNs)
+    return Op;
 
   // TODO: Propagate flags from the select rather than global settings.
   SDNodeFlags Flags;
   Flags.setNoInfs(true);
   Flags.setNoNaNs(true);
 
   // If the RHS of the comparison is a 0.0, we don't need to do the
   // subtraction at all.

llvm/test/CodeGen/PowerPC/scalar-equal.ll

 ; FAST-P9-NEXT:    xssubdp f0, f1, f2
 ; FAST-P9-NEXT:    fsel f1, f0, f3, f4
 ; FAST-P9-NEXT:    xsnegdp f0, f0
 ; FAST-P9-NEXT:    fsel f1, f0, f1, f4
 ; FAST-P9-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: testoeq:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
-; NO-FAST-P9-NEXT:    xsnegdp f0, f0
-; NO-FAST-P9-NEXT:    fsel f1, f0, f1, f4
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    beq cr0, .LBB0_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB0_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 ;
 ; NO-FAST-P8-LABEL: testoeq:
 ; NO-FAST-P8:       # %bb.0: # %entry
 ; NO-FAST-P8-NEXT:    xscmpudp cr0, f1, f2
 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    beqlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 entry:
   %cmp = fcmp fast oeq double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }

llvm/test/CodeGen/PowerPC/scalar_cmp.ll

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    beqlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_oeq_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f1, f2
-; NO-FAST-P9-NEXT:    xsnegdp f1, f0
-; NO-FAST-P9-NEXT:    fsel f0, f0, f3, f4
-; NO-FAST-P9-NEXT:    fsel f1, f1, f0, f4
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    beq cr0, .LBB2_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB2_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast oeq float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_oeq_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    beqlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_oeq_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
-; NO-FAST-P9-NEXT:    xsnegdp f0, f0
-; NO-FAST-P9-NEXT:    fsel f1, f0, f1, f4
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    beq cr0, .LBB3_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB3_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast oeq double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bnelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_one_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f1, f2
-; NO-FAST-P9-NEXT:    xsnegdp f1, f0
-; NO-FAST-P9-NEXT:    fsel f0, f0, f4, f3
-; NO-FAST-P9-NEXT:    fsel f1, f1, f0, f3
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    bne cr0, .LBB6_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB6_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast one float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_one_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bnelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_one_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f4, f3
-; NO-FAST-P9-NEXT:    xsnegdp f0, f0
-; NO-FAST-P9-NEXT:    fsel f1, f0, f1, f3
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    bne cr0, .LBB7_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB7_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast one double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bgelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_oge_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    bge cr0, .LBB10_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB10_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast oge float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_oge_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bgelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_oge_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    bge cr0, .LBB11_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB11_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast oge double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bltlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_olt_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f4, f3
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    blt cr0, .LBB14_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB14_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast olt float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_olt_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bltlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_olt_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f1, f2
-; NO-FAST-P9-NEXT:    fsel f1, f0, f4, f3
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    blt cr0, .LBB15_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB15_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast olt double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bgtlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_ogt_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f2, f1
-; NO-FAST-P9-NEXT:    fsel f1, f0, f4, f3
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    bgt cr0, .LBB18_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB18_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast ogt float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_ogt_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    bgtlr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_ogt_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f2, f1
-; NO-FAST-P9-NEXT:    fsel f1, f0, f4, f3
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    bgt cr0, .LBB19_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB19_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast ogt double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    blelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_ole_float:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubsp f0, f2, f1
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
+; NO-FAST-P9-NEXT:    fcmpu cr0, f1, f2
+; NO-FAST-P9-NEXT:    ble cr0, .LBB22_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB22_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast ole float %a, %b
   %cond = select fast i1 %cmp, float %c, float %d
   ret float %cond
 }
 
 define double @select_fast_ole_double(double %a, double %b, double %c, double %d) {

 ; NO-FAST-P8-NEXT:    fmr f1, f3
 ; NO-FAST-P8-NEXT:    blelr cr0
 ; NO-FAST-P8-NEXT:  # %bb.1: # %entry
 ; NO-FAST-P8-NEXT:    fmr f1, f4
 ; NO-FAST-P8-NEXT:    blr
 ;
 ; NO-FAST-P9-LABEL: select_fast_ole_double:
 ; NO-FAST-P9:       # %bb.0: # %entry
-; NO-FAST-P9-NEXT:    xssubdp f0, f2, f1
-; NO-FAST-P9-NEXT:    fsel f1, f0, f3, f4
+; NO-FAST-P9-NEXT:    xscmpudp cr0, f1, f2
+; NO-FAST-P9-NEXT:    ble cr0, .LBB23_2
+; NO-FAST-P9-NEXT:  # %bb.1: # %entry
+; NO-FAST-P9-NEXT:    fmr f3, f4
+; NO-FAST-P9-NEXT:  .LBB23_2: # %entry
+; NO-FAST-P9-NEXT:    fmr f1, f3
 ; NO-FAST-P9-NEXT:    blr
 entry:
   %cmp = fcmp fast ole double %a, %b
   %cond = select fast i1 %cmp, double %c, double %d
   ret double %cond
 }
 
 ; Test RHS is 1.000000e+00