[VE] Support fneg and frem

[kaz7]

Summary

VE doesn't have fneg or frem instruction, so change them to expand. Add
regression tests also.

llvm/lib/Target/VE/VEISelLowering.cpp

VETargetLowering::VETargetLowering(const TargetMachine &TM,
   // fp16 not supported
   for (MVT FPVT : MVT::fp_valuetypes()) {
     setOperationAction(ISD::FP16_TO_FP, FPVT, Expand);
     setOperationAction(ISD::FP_TO_FP16, FPVT, Expand);
   }
   /// } Conversion
 
   /// Floating-point Ops {
+  /// Note: Floating-point operations are fneg, fadd, fsub, fmul, fdiv, frem,
+  ///       and fcmp.
+
+  // VE doesn't have following floating point operations.
+  for (MVT VT : MVT::fp_valuetypes()) {
+    setOperationAction(ISD::FNEG, VT, Expand);
+    setOperationAction(ISD::FREM, VT, Expand);
+  }
 
   // VE doesn't have fdiv of f128.
   setOperationAction(ISD::FDIV, MVT::f128, Expand);
 
   for (MVT FPVT : {MVT::f32, MVT::f64}) {
     // f32 and f64 uses ConstantFP.  f128 uses ConstantPool.
     setOperationAction(ISD::ConstantFP, FPVT, Legal);
   }

llvm/test/CodeGen/VE/fp_fneg.ll

+; RUN: llc < %s -mtriple=ve | FileCheck %s
+
+;;; Test ‘fneg’ Instruction
+;;;
+;;; Syntax:
+;;;   <result> = fneg [fast-math flags]* <ty> <op1>   ; yields ty:result
+;;;
+;;; Overview:
+;;;    The ‘fneg’ instruction returns the negation of its operand.
+;;;
+;;; Arguments:
+;;;   The argument to the ‘fneg’ instruction must be a floating-point or
+;;;   vector of floating-point values.
+;;;
+;;; Semantics:
+;;;
+;;;   The value produced is a copy of the operand with its sign bit flipped.
+;;;   This instruction can also take any number of fast-math flags, which are
+;;;   optimization hints to enable otherwise unsafe floating-point
+;;;   optimizations.
+;;;
+;;; Example:
+;;;   <result> = fneg float %val          ; yields float:result = -%var
+;;;
+;;; Note:
+;;;   We test only float/double/fp128.
+
+; Function Attrs: norecurse nounwind readnone
+define float @fneg_float(float %0) {
+; CHECK-LABEL: fneg_float:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    sra.l %s0, %s0, 32
+; CHECK-NEXT:    lea %s1, -2147483648
+; CHECK-NEXT:    and %s1, %s1, (32)0
+; CHECK-NEXT:    xor %s0, %s0, %s1
+; CHECK-NEXT:    sll %s0, %s0, 32
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = fneg float %0
+  ret float %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define double @fneg_double(double %0) {
+; CHECK-LABEL: fneg_double:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    xor %s0, %s0, (1)1
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = fneg double %0
+  ret double %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @fneg_quad(fp128 %0) {
+; CHECK-LABEL: fneg_quad:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    st %s1, 176(, %s11)
+; CHECK-NEXT:    st %s0, 184(, %s11)
+; CHECK-NEXT:    ld1b.zx %s0, 191(, %s11)
+; CHECK-NEXT:    lea %s1, 128
+; CHECK-NEXT:    xor %s0, %s0, %s1
+; CHECK-NEXT:    st1b %s0, 191(, %s11)
+; CHECK-NEXT:    ld %s1, 176(, %s11)
+; CHECK-NEXT:    ld %s0, 184(, %s11)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = fneg fp128 %0
+  ret fp128 %2
+}

llvm/test/CodeGen/VE/fp_frem.ll

+; RUN: llc < %s -mtriple=ve | FileCheck %s
+
+;;; Test ‘frem’ Instruction
+;;;
+;;; Syntax:
+;;;   <result> = frem [fast-math flags]* <ty> <op1>, <op2> ; yields ty:result
+;;;
+;;; Overview:
+;;;   The ‘frem’ instruction returns the remainder from the division of its two
+;;;   operands.
+;;;
+;;; Arguments:
+;;;   The two arguments to the ‘frem’ instruction must be floating-point or
+;;;   vector of floating-point values. Both arguments must have identical types.
+;;;
+;;; Semantics:
+;;;   The value produced is the floating-point remainder of the two operands.
+;;;   This is the same output as a libm ‘fmod’ function, but without any
+;;;   possibility of setting errno. The remainder has the same sign as the
+;;;   dividend. This instruction is assumed to execute in the default
+;;;   floating-point environment. This instruction can also take any number
+;;;   of fast-math flags, which are optimization hints to enable otherwise
+;;;   unsafe floating-point optimizations:
+;;;
+;;; Example:
+;;;
+;;;   <result> = frem float 4.0, %var ; yields float:result = 4.0 % %var
+;;;
+;;; Note:
+;;;   We test only float/double/fp128.
+;;;   We have no way to generated frem from C source code, so convert fdiv
+;;;   to frem by using sed program.
+
+; Function Attrs: norecurse nounwind readnone
+define float @frem_float_var(float %0, float %1) {
+; CHECK-LABEL: frem_float_var:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s2, fmodf@lo
+; CHECK-NEXT:    and %s2, %s2, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodf@hi(, %s2)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %3 = frem float %0, %1
+  ret float %3
+}
+
+; Function Attrs: norecurse nounwind readnone
+define double @frem_double_var(double %0, double %1) {
+; CHECK-LABEL: frem_double_var:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s2, fmod@lo
+; CHECK-NEXT:    and %s2, %s2, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmod@hi(, %s2)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %3 = frem double %0, %1
+  ret double %3
+}
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @frem_quad_var(fp128 %0, fp128 %1) {
+; CHECK-LABEL: frem_quad_var:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    lea %s4, fmodl@lo
+; CHECK-NEXT:    and %s4, %s4, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodl@hi(, %s4)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %3 = frem fp128 %0, %1
+  ret fp128 %3
+}
+
+; Function Attrs: norecurse nounwind readnone
+define float @frem_float_zero(float %0) {
+; CHECK-LABEL: frem_float_zero:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s1, 0, %s0
+; CHECK-NEXT:    lea.sl %s0, 0
+; CHECK-NEXT:    lea %s2, fmodf@lo
+; CHECK-NEXT:    and %s2, %s2, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodf@hi(, %s2)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem float 0.000000e+00, %0
+  ret float %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define double @frem_double_zero(double %0) {
+; CHECK-LABEL: frem_double_zero:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s1, 0, %s0
+; CHECK-NEXT:    lea %s0, fmod@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmod@hi(, %s0)
+; CHECK-NEXT:    lea.sl %s0, 0
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem double 0.000000e+00, %0
+  ret double %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @frem_quad_zero(fp128 %0) {
+; CHECK-LABEL: frem_quad_zero:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s2, 0, %s0
+; CHECK-NEXT:    or %s3, 0, %s1
+; CHECK-NEXT:    lea %s0, .LCPI{{[0-9]+}}_0@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s4, .LCPI{{[0-9]+}}_0@hi(, %s0)
+; CHECK-NEXT:    ld %s0, 8(, %s4)
+; CHECK-NEXT:    ld %s1, (, %s4)
+; CHECK-NEXT:    lea %s4, fmodl@lo
+; CHECK-NEXT:    and %s4, %s4, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodl@hi(, %s4)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem fp128 0xL00000000000000000000000000000000, %0
+  ret fp128 %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define float @frem_float_cont(float %0) {
+; CHECK-LABEL: frem_float_cont:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s1, 0, %s0
+; CHECK-NEXT:    lea.sl %s0, -1073741824
+; CHECK-NEXT:    lea %s2, fmodf@lo
+; CHECK-NEXT:    and %s2, %s2, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodf@hi(, %s2)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem float -2.000000e+00, %0
+  ret float %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define double @frem_double_cont(double %0) {
+; CHECK-LABEL: frem_double_cont:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s1, 0, %s0
+; CHECK-NEXT:    lea %s0, fmod@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmod@hi(, %s0)
+; CHECK-NEXT:    lea.sl %s0, -1073741824
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem double -2.000000e+00, %0
+  ret double %2
+}
+
+; Function Attrs: norecurse nounwind readnone
+define fp128 @frem_quad_cont(fp128 %0) {
+; CHECK-LABEL: frem_quad_cont:
+; CHECK:       .LBB{{[0-9]+}}_2:
+; CHECK-NEXT:    or %s2, 0, %s0
+; CHECK-NEXT:    or %s3, 0, %s1
+; CHECK-NEXT:    lea %s0, .LCPI{{[0-9]+}}_0@lo
+; CHECK-NEXT:    and %s0, %s0, (32)0
+; CHECK-NEXT:    lea.sl %s4, .LCPI{{[0-9]+}}_0@hi(, %s0)
+; CHECK-NEXT:    ld %s0, 8(, %s4)
+; CHECK-NEXT:    ld %s1, (, %s4)
+; CHECK-NEXT:    lea %s4, fmodl@lo
+; CHECK-NEXT:    and %s4, %s4, (32)0
+; CHECK-NEXT:    lea.sl %s12, fmodl@hi(, %s4)
+; CHECK-NEXT:    bsic %s10, (, %s12)
+; CHECK-NEXT:    or %s11, 0, %s9
+  %2 = frem fp128 0xL0000000000000000C000000000000000, %0
+  ret fp128 %2
+}