[ARM] MVE bitwise instruction patterns

[dmgreen]

Summary

This adds patterns for the simpler VAND, VORR and VEOR instructions. It also adjusts the top16Zero PatLeaf to not match on vector instructions, which can otherwise cause problems.

Code written by David Sherwood.

[SjoerdMeijer]

LGTM

llvm/trunk/lib/Target/ARM/ARMInstrMVE.td

foreach s=["s8", "s16", "s32", "u8", "u16", "u32", "i8", "i16", "i32", "f16", "f32"] in {
   def : MVEInstAlias<"vorn${vp}." # s # "\t$QdSrc, $QnSrc, $QmSrc",
         (MVE_VORN MQPR:$QdSrc, MQPR:$QnSrc, MQPR:$QmSrc, vpred_r:$vp)>;
   def : MVEInstAlias<"vorr${vp}." # s # "\t$QdSrc, $QnSrc, $QmSrc",
         (MVE_VORR MQPR:$QdSrc, MQPR:$QnSrc, MQPR:$QmSrc, vpred_r:$vp)>;
   def : MVEInstAlias<"vand${vp}." # s # "\t$QdSrc, $QnSrc, $QmSrc",
         (MVE_VAND MQPR:$QdSrc, MQPR:$QnSrc, MQPR:$QmSrc, vpred_r:$vp)>;
 }
 
+let Predicates = [HasMVEInt] in {
+  def : Pat<(v16i8 (and (v16i8 MQPR:$val1), (v16i8 MQPR:$val2))),
+            (v16i8 (MVE_VAND (v16i8 MQPR:$val1), (v16i8 MQPR:$val2)))>;
+  def : Pat<(v8i16 (and (v8i16 MQPR:$val1), (v8i16 MQPR:$val2))),
+            (v8i16 (MVE_VAND (v8i16 MQPR:$val1), (v8i16 MQPR:$val2)))>;
+  def : Pat<(v4i32 (and (v4i32 MQPR:$val1), (v4i32 MQPR:$val2))),
+            (v4i32 (MVE_VAND (v4i32 MQPR:$val1), (v4i32 MQPR:$val2)))>;
+
+  def : Pat<(v16i8 (or (v16i8 MQPR:$val1), (v16i8 MQPR:$val2))),
+            (v16i8 (MVE_VORR (v16i8 MQPR:$val1), (v16i8 MQPR:$val2)))>;
+  def : Pat<(v8i16 (or (v8i16 MQPR:$val1), (v8i16 MQPR:$val2))),
+            (v8i16 (MVE_VORR (v8i16 MQPR:$val1), (v8i16 MQPR:$val2)))>;
+  def : Pat<(v4i32 (or (v4i32 MQPR:$val1), (v4i32 MQPR:$val2))),
+            (v4i32 (MVE_VORR (v4i32 MQPR:$val1), (v4i32 MQPR:$val2)))>;
+
+  def : Pat<(v16i8 (xor (v16i8 MQPR:$val1), (v16i8 MQPR:$val2))),
+            (v16i8 (MVE_VEOR (v16i8 MQPR:$val1), (v16i8 MQPR:$val2)))>;
+  def : Pat<(v8i16 (xor (v8i16 MQPR:$val1), (v8i16 MQPR:$val2))),
+            (v8i16 (MVE_VEOR (v8i16 MQPR:$val1), (v8i16 MQPR:$val2)))>;
+  def : Pat<(v4i32 (xor (v4i32 MQPR:$val1), (v4i32 MQPR:$val2))),
+            (v4i32 (MVE_VEOR (v4i32 MQPR:$val1), (v4i32 MQPR:$val2)))>;
+}
+
 class MVE_bit_cmode<string iname, string suffix, bits<4> cmode, dag inOps>
   : MVE_p<(outs MQPR:$Qd), inOps, NoItinerary,
           iname, suffix, "$Qd, $imm", vpred_n, "$Qd = $Qd_src"> {
   bits<8> imm;
   bits<4> Qd;
 
   let Inst{28} = imm{7};
   let Inst{27-23} = 0b11111;

llvm/trunk/lib/Target/ARM/ARMInstrThumb2.td

defm t2MVN  : T2I_un_irs <0b0011, "mvn",
                           not, 1, 1, 1>;
 
 let AddedComplexity = 1 in
 def : T2Pat<(and     rGPR:$src, t2_so_imm_not:$imm),
             (t2BICri rGPR:$src, t2_so_imm_not:$imm)>;
 
 // top16Zero - answer true if the upper 16 bits of $src are 0, false otherwise
 def top16Zero: PatLeaf<(i32 rGPR:$src), [{
-  return CurDAG->MaskedValueIsZero(SDValue(N,0), APInt::getHighBitsSet(32, 16));
+  return !SDValue(N,0)->getValueType(0).isVector() &&
+         CurDAG->MaskedValueIsZero(SDValue(N,0), APInt::getHighBitsSet(32, 16));
   }]>;
 
 // so_imm_notSext is needed instead of so_imm_not, as the value of imm
 // will match the extended, not the original bitWidth for $src.
 def : T2Pat<(and top16Zero:$src, t2_so_imm_notSext:$imm),
             (t2BICri rGPR:$src, t2_so_imm_notSext:$imm)>;
 
 

llvm/trunk/test/CodeGen/Thumb2/mve-bitarith.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s
+
+define arm_aapcs_vfpcc <16 x i8> @and_int8_t(<16 x i8> %src1, <16 x i8> %src2) {
+; CHECK-LABEL: and_int8_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = and <16 x i8> %src1, %src2
+  ret <16 x i8> %0
+}
+
+define arm_aapcs_vfpcc <8 x i16> @and_int16_t(<8 x i16> %src1, <8 x i16> %src2) {
+; CHECK-LABEL: and_int16_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = and <8 x i16> %src1, %src2
+  ret <8 x i16> %0
+}
+
+define arm_aapcs_vfpcc <4 x i32> @and_int32_t(<4 x i32> %src1, <4 x i32> %src2) {
+; CHECK-LABEL: and_int32_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = and <4 x i32> %src1, %src2
+  ret <4 x i32> %0
+}
+
+
+define arm_aapcs_vfpcc <16 x i8> @or_int8_t(<16 x i8> %src1, <16 x i8> %src2) {
+; CHECK-LABEL: or_int8_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = or <16 x i8> %src1, %src2
+  ret <16 x i8> %0
+}
+
+define arm_aapcs_vfpcc <8 x i16> @or_int16_t(<8 x i16> %src1, <8 x i16> %src2) {
+; CHECK-LABEL: or_int16_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = or <8 x i16> %src1, %src2
+  ret <8 x i16> %0
+}
+
+define arm_aapcs_vfpcc <4 x i32> @or_int32_t(<4 x i32> %src1, <4 x i32> %src2) {
+; CHECK-LABEL: or_int32_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vorr q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = or <4 x i32> %src1, %src2
+  ret <4 x i32> %0
+}
+
+
+define arm_aapcs_vfpcc <16 x i8> @xor_int8_t(<16 x i8> %src1, <16 x i8> %src2) {
+; CHECK-LABEL: xor_int8_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = xor <16 x i8> %src1, %src2
+  ret <16 x i8> %0
+}
+
+define arm_aapcs_vfpcc <8 x i16> @xor_int16_t(<8 x i16> %src1, <8 x i16> %src2) {
+; CHECK-LABEL: xor_int16_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = xor <8 x i16> %src1, %src2
+  ret <8 x i16> %0
+}
+
+define arm_aapcs_vfpcc <4 x i32> @xor_int32_t(<4 x i32> %src1, <4 x i32> %src2) {
+; CHECK-LABEL: xor_int32_t:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    veor q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = xor <4 x i32> %src1, %src2
+  ret <4 x i32> %0
+}
+

llvm/trunk/test/CodeGen/Thumb2/mve-div-expand.ll

 ; CHECK-LABEL: frem_f16:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    .pad #64
 ; CHECK-NEXT:    sub sp, #64
-; CHECK-NEXT:    vmov q5, q1
 ; CHECK-NEXT:    vmov q4, q0
 ; CHECK-NEXT:    vmov.u16 r0, q1[0]
 ; CHECK-NEXT:    vmov s0, r0
 ; CHECK-NEXT:    vmov.u16 r0, q4[0]
 ; CHECK-NEXT:    vmov s2, r0
+; CHECK-NEXT:    vmov q5, q1
 ; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
 ; CHECK-NEXT:    vstr s2, [sp, #56]
 ; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
 ; CHECK-NEXT:    vstr s0, [sp, #60]
 ; CHECK-NEXT:    ldrd r0, r1, [sp, #56]
 ; CHECK-NEXT:    bl fmodf
 ; CHECK-NEXT:    vmov s0, r0
 ; CHECK-NEXT:    vmov.u16 r0, q5[1]

llvm/trunk/test/CodeGen/Thumb2/mve-fmath.ll

 ; CHECK-LABEL: pow_float16_t:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    .save {r4, lr}
 ; CHECK-NEXT:    push {r4, lr}
 ; CHECK-NEXT:    .vsave {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    vpush {d8, d9, d10, d11, d12, d13}
 ; CHECK-NEXT:    .pad #64
 ; CHECK-NEXT:    sub sp, #64
-; CHECK-NEXT:    vmov q5, q1
 ; CHECK-NEXT:    vmov q4, q0
 ; CHECK-NEXT:    vmov.u16 r0, q1[0]
 ; CHECK-NEXT:    vmov s0, r0
 ; CHECK-NEXT:    vmov.u16 r0, q4[0]
 ; CHECK-NEXT:    vmov s2, r0
+; CHECK-NEXT:    vmov q5, q1
 ; CHECK-NEXT:    vcvtb.f32.f16 s2, s2
 ; CHECK-NEXT:    vstr s2, [sp, #56]
 ; CHECK-NEXT:    vcvtb.f32.f16 s0, s0
 ; CHECK-NEXT:    vstr s0, [sp, #60]
 ; CHECK-NEXT:    ldrd r0, r1, [sp, #56]
 ; CHECK-NEXT:    bl powf
 ; CHECK-NEXT:    vmov s0, r0
 ; CHECK-NEXT:    vmov.u16 r0, q5[1]