[ARM] Add scheduling classes for VFNM[AS]

[olista01]

Summary

The VFNM[AS] instructions did not have scheduling information attached, which was causing assertion failures with the Cortex-A57 scheduling model and -fp-contract=fast, because the Cortex-A57 sched model claims to be complete.

[javed.absar]

Thanks Oliver for fixing it.

lib/Target/ARM/ARMInstrVFP.td

def : Pat<(f32 (fma SPR:$Sn, (fneg SPR:$Sm), SPR:$Sdin)),
       Requires<[HasVFP4]>;
 
 def VFNMAD : ADbI<0b11101, 0b01, 1, 0,
                   (outs DPR:$Dd), (ins DPR:$Ddin, DPR:$Dn, DPR:$Dm),
                   IIC_fpFMAC64, "vfnma", ".f64\t$Dd, $Dn, $Dm",
                   [(set DPR:$Dd,(fsub_mlx (fneg (fmul_su DPR:$Dn,DPR:$Dm)),
                                           (f64 DPR:$Ddin)))]>,
                 RegConstraint<"$Ddin = $Dd">,
-                Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>;
+                Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>,
+                Sched<[WriteFPMAC64, ReadFPMAC, ReadFPMUL, ReadFPMUL]>;
 
 def VFNMAS : ASbI<0b11101, 0b01, 1, 0,
                   (outs SPR:$Sd), (ins SPR:$Sdin, SPR:$Sn, SPR:$Sm),
                   IIC_fpFMAC32, "vfnma", ".f32\t$Sd, $Sn, $Sm",
                   [(set SPR:$Sd, (fsub_mlx (fneg (fmul_su SPR:$Sn, SPR:$Sm)),
                                            SPR:$Sdin))]>,
                 RegConstraint<"$Sdin = $Sd">,
-                Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]> {
+                Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>,
+                Sched<[WriteFPMAC32, ReadFPMAC, ReadFPMUL, ReadFPMUL]> {
   // Some single precision VFP instructions may be executed on both NEON and
   // VFP pipelines.
 }
 
 def VFNMAH : AHbI<0b11101, 0b01, 1, 0,
                   (outs SPR:$Sd), (ins SPR:$Sdin, SPR:$Sn, SPR:$Sm),
                   IIC_fpFMAC16, "vfnma", ".f16\t$Sd, $Sn, $Sm",
                   []>,
                 RegConstraint<"$Sdin = $Sd">,
-                Requires<[HasFullFP16,UseFusedMAC]>;
+                Requires<[HasFullFP16,UseFusedMAC]>,
+                Sched<[WriteFPMAC32, ReadFPMAC, ReadFPMUL, ReadFPMUL]>;
 
 def : Pat<(fsub_mlx (fneg (fmul_su DPR:$a, (f64 DPR:$b))), DPR:$dstin),
           (VFNMAD DPR:$dstin, DPR:$a, DPR:$b)>,
           Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>;
 def : Pat<(fsub_mlx (fneg (fmul_su SPR:$a, SPR:$b)), SPR:$dstin),
           (VFNMAS SPR:$dstin, SPR:$a, SPR:$b)>,
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 
def : Pat<(f32 (fma (fneg SPR:$Sn), SPR:$Sm, (fneg SPR:$Sdin))),
       Requires<[HasVFP4]>;
 
 def VFNMSD : ADbI<0b11101, 0b01, 0, 0,
                   (outs DPR:$Dd), (ins DPR:$Ddin, DPR:$Dn, DPR:$Dm),
                   IIC_fpFMAC64, "vfnms", ".f64\t$Dd, $Dn, $Dm",
                   [(set DPR:$Dd, (fsub_mlx (fmul_su DPR:$Dn, DPR:$Dm),
                                            (f64 DPR:$Ddin)))]>,
                RegConstraint<"$Ddin = $Dd">,
-               Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>;
+               Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>,
+               Sched<[WriteFPMAC64, ReadFPMAC, ReadFPMUL, ReadFPMUL]>;
 
 def VFNMSS : ASbI<0b11101, 0b01, 0, 0,
                   (outs SPR:$Sd), (ins SPR:$Sdin, SPR:$Sn, SPR:$Sm),
                   IIC_fpFMAC32, "vfnms", ".f32\t$Sd, $Sn, $Sm",
              [(set SPR:$Sd, (fsub_mlx (fmul_su SPR:$Sn, SPR:$Sm), SPR:$Sdin))]>,
                          RegConstraint<"$Sdin = $Sd">,
-                  Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]> {
+                  Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>,
+                  Sched<[WriteFPMAC32, ReadFPMAC, ReadFPMUL, ReadFPMUL]> {
   // Some single precision VFP instructions may be executed on both NEON and
   // VFP pipelines.
 }
 
 def VFNMSH : AHbI<0b11101, 0b01, 0, 0,
                   (outs SPR:$Sd), (ins SPR:$Sdin, SPR:$Sn, SPR:$Sm),
                   IIC_fpFMAC16, "vfnms", ".f16\t$Sd, $Sn, $Sm",
              []>,
                          RegConstraint<"$Sdin = $Sd">,
-                  Requires<[HasFullFP16,UseFusedMAC]>;
+                  Requires<[HasFullFP16,UseFusedMAC]>,
+                  Sched<[WriteFPMAC32, ReadFPMAC, ReadFPMUL, ReadFPMUL]>;
 
 def : Pat<(fsub_mlx (fmul_su DPR:$a, (f64 DPR:$b)), DPR:$dstin),
           (VFNMSD DPR:$dstin, DPR:$a, DPR:$b)>,
           Requires<[HasVFP4,HasDPVFP,UseFusedMAC]>;
 def : Pat<(fsub_mlx (fmul_su SPR:$a, SPR:$b), SPR:$dstin),
           (VFNMSS SPR:$dstin, SPR:$a, SPR:$b)>,
           Requires<[HasVFP4,DontUseNEONForFP,UseFusedMAC]>;
 

test/CodeGen/ARM/cortex-a57-misched-vfma.ll

 ; f1 * f2 + f3 * f4 + f5 * f6  ==>  VMULS, VMLSS, VMLSS
   %mul1 = fmul <2 x float> %f1, %f2
   %mul2 = fmul <2 x float> %f3, %f4
   %mul3 = fmul <2 x float> %f5, %f6
   %sub1 = fsub <2 x float> %mul1, %mul2
   %sub2 = fsub <2 x float> %sub1, %mul3
   ret <2 x float> %sub2
 }
+
+define float @Test5(float %f1, float %f2, float %f3) {
+; CHECK:       ********** MI Scheduling **********
+; CHECK:       Test5:BB#0
+
+; CHECK-DEFAULT: VNMLS
+; CHECK-FAST:    VFNMS
+; CHECK:       Latency            : 9
+; CHECK:       Successors:
+; CHECK:       data
+; > VMLAS not-optimized latency to VMOVRS = 9
+; CHECK-SAME:  Latency=9
+
+; f1 * f2 - f3  ==>  VNMLS/VFNMS
+  %mul = fmul float %f1, %f2
+  %sub = fsub float %mul, %f3
+  ret float %sub
+}
+
+
+define float @Test6(float %f1, float %f2, float %f3) {
+; CHECK:       ********** MI Scheduling **********
+; CHECK:       Test6:BB#0
+
+; CHECK-DEFAULT: VNMLA
+; CHECK-FAST:    VFNMA
+; CHECK:       Latency            : 9
+; CHECK:       Successors:
+; CHECK:       data
+; > VMLAS not-optimized latency to VMOVRS = 9
+; CHECK-SAME:  Latency=9
+
+; f1 * f2 - f3  ==>  VNMLA/VFNMA
+  %mul = fmul float %f1, %f2
+  %sub1 = fsub float -0.0, %mul
+  %sub2 = fsub float %sub1, %f2
+  ret float %sub2
+}