[ARM] MVE trunc to i1 vectors

[dmgreen]

Summary

This adds patterns for selecting trunc from vectors to i1's.

[SjoerdMeijer]

LGTM

llvm/lib/Target/ARM/ARMInstrMVE.td

def : Pat<(v4i32 (sext  (v4i1  VCCR:$pred))),
             (v4i32 (MVE_VPSEL (MVE_VMOVimmi8 255), (MVE_VMOVimmi32 0), 0, VCCR:$pred))>;
 
   def : Pat<(v16i8 (anyext  (v16i1 VCCR:$pred))),
             (v16i8 (MVE_VPSEL (MVE_VMOVimmi8 1), (MVE_VMOVimmi8 0), 0, VCCR:$pred))>;
   def : Pat<(v8i16 (anyext  (v8i1  VCCR:$pred))),
             (v8i16 (MVE_VPSEL (MVE_VMOVimmi16 1), (MVE_VMOVimmi16 0), 0, VCCR:$pred))>;
   def : Pat<(v4i32 (anyext  (v4i1  VCCR:$pred))),
             (v4i32 (MVE_VPSEL (MVE_VMOVimmi32 1), (MVE_VMOVimmi32 0), 0, VCCR:$pred))>;
+
+  def : Pat<(v16i1 (trunc (v16i8 MQPR:$v1))),
+            (v16i1 (MVE_VCMPi32r (v16i8 MQPR:$v1), ZR, 1))>;
+  def : Pat<(v8i1 (trunc (v8i16  MQPR:$v1))),
+            (v8i1 (MVE_VCMPi32r (v8i16 MQPR:$v1), ZR, 1))>;
+  def : Pat<(v4i1 (trunc (v4i32  MQPR:$v1))),
+            (v4i1 (MVE_VCMPi32r (v4i32 MQPR:$v1), ZR, 1))>;
 }
 
 let Predicates = [HasMVEFloat] in {
   // Pred <-> Float
   // 112 is 1.0 in float
   def : Pat<(v4f32 (uint_to_fp (v4i1 VCCR:$pred))),
             (v4f32 (MVE_VPSEL (v4f32 (MVE_VMOVimmf32 112)), (v4f32 (MVE_VMOVimmi32 0)), 0, VCCR:$pred))>;
   // 2620 in 1.0 in half

llvm/test/CodeGen/Thumb2/mve-pred-ext.ll

 ; CHECK-NEXT:    .long 0 @ 0x0
 entry:
   %c = icmp sgt <2 x i64> %src, zeroinitializer
   %0 = zext <2 x i1> %c to <2 x i64>
   ret <2 x i64> %0
 }
 
 
+define arm_aapcs_vfpcc <4 x i32> @trunc_v4i1_v4i32(<4 x i32> %src) {
+; CHECK-LABEL: trunc_v4i1_v4i32:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vcmp.i32 ne, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = trunc <4 x i32> %src to <4 x i1>
+  %1 = select <4 x i1> %0, <4 x i32> %src, <4 x i32> zeroinitializer
+  ret <4 x i32> %1
+}
+
+define arm_aapcs_vfpcc <8 x i16> @trunc_v8i1_v8i16(<8 x i16> %src) {
+; CHECK-LABEL: trunc_v8i1_v8i16:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vcmp.i32 ne, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = trunc <8 x i16> %src to <8 x i1>
+  %1 = select <8 x i1> %0, <8 x i16> %src, <8 x i16> zeroinitializer
+  ret <8 x i16> %1
+}
+
+define arm_aapcs_vfpcc <16 x i8> @trunc_v16i1_v16i8(<16 x i8> %src) {
+; CHECK-LABEL: trunc_v16i1_v16i8:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vmov.i32 q1, #0x0
+; CHECK-NEXT:    vcmp.i32 ne, q0, zr
+; CHECK-NEXT:    vpsel q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = trunc <16 x i8> %src to <16 x i1>
+  %1 = select <16 x i1> %0, <16 x i8> %src, <16 x i8> zeroinitializer
+  ret <16 x i8> %1
+}
+
+define arm_aapcs_vfpcc <2 x i64> @trunc_v2i1_v2i64(<2 x i64> %src) {
+; CHECK-LABEL: trunc_v2i1_v2i64:
+; CHECK:       @ %bb.0: @ %entry
+; CHECK-NEXT:    vmov r1, s0
+; CHECK-NEXT:    vmov r0, s2
+; CHECK-NEXT:    and r1, r1, #1
+; CHECK-NEXT:    rsbs r1, r1, #0
+; CHECK-NEXT:    and r0, r0, #1
+; CHECK-NEXT:    vmov.32 q1[0], r1
+; CHECK-NEXT:    rsbs r0, r0, #0
+; CHECK-NEXT:    vmov.32 q1[1], r1
+; CHECK-NEXT:    vmov.32 q1[2], r0
+; CHECK-NEXT:    vmov.32 q1[3], r0
+; CHECK-NEXT:    vand q0, q0, q1
+; CHECK-NEXT:    bx lr
+entry:
+  %0 = trunc <2 x i64> %src to <2 x i1>
+  %1 = select <2 x i1> %0, <2 x i64> %src, <2 x i64> zeroinitializer
+  ret <2 x i64> %1
+}
+
+
 define arm_aapcs_vfpcc <4 x float> @uitofp_v4i1_v4f32(<4 x i32> %src) {
 ; CHECK-LABEL: uitofp_v4i1_v4f32:
 ; CHECK:       @ %bb.0: @ %entry
 ; CHECK-NEXT:    vmov.i32 q1, #0x0
 ; CHECK-NEXT:    vmov.f32 q2, #1.000000e+00
 ; CHECK-NEXT:    vcmp.s32 gt, q0, zr
 ; CHECK-NEXT:    vpsel q0, q2, q1
 ; CHECK-NEXT:    bx lr