[llvm] c060757 - [ARM] Correct v2i1 concat extract types.
David Green via llvm-commits
llvm-commits at lists.llvm.org
Wed Oct 18 05:40:44 PDT 2023
Author: David Green
Date: 2023-10-18T13:40:38+01:00
New Revision: c060757bcc8a0d4537bc2f98e1208f089aef79e6
URL: https://github.com/llvm/llvm-project/commit/c060757bcc8a0d4537bc2f98e1208f089aef79e6
DIFF: https://github.com/llvm/llvm-project/commit/c060757bcc8a0d4537bc2f98e1208f089aef79e6.diff
LOG: [ARM] Correct v2i1 concat extract types.
For two v2i1 concat into a v4i1, we cannot extract each i64 element as an i32.
This casts to a v4i32 instead and extracts the correct vector lanes.
Added:
llvm/test/CodeGen/Thumb2/mve-concat.ll
Modified:
llvm/lib/Target/ARM/ARMISelLowering.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 8cca167e32ee791..6e58cbaf2ac4a84 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -9082,6 +9082,8 @@ static SDValue LowerCONCAT_VECTORS_i1(SDValue Op, SelectionDAG &DAG,
EVT Op1VT = V1.getValueType();
EVT Op2VT = V2.getValueType();
assert(Op1VT == Op2VT && "Operand types don't match!");
+ assert((Op1VT == MVT::v2i1 || Op1VT == MVT::v4i1 || Op1VT == MVT::v8i1) &&
+ "Unexpected i1 concat operations!");
EVT VT = Op1VT.getDoubleNumVectorElementsVT(*DAG.getContext());
SDValue NewV1 = PromoteMVEPredVector(dl, V1, Op1VT, DAG);
@@ -9103,9 +9105,14 @@ static SDValue LowerCONCAT_VECTORS_i1(SDValue Op, SelectionDAG &DAG,
auto ExtractInto = [&DAG, &dl](SDValue NewV, SDValue ConVec, unsigned &j) {
EVT NewVT = NewV.getValueType();
EVT ConcatVT = ConVec.getValueType();
+ unsigned ExtScale = 1;
+ if (NewVT == MVT::v2f64) {
+ NewV = DAG.getNode(ARMISD::VECTOR_REG_CAST, dl, MVT::v4i32, NewV);
+ ExtScale = 2;
+ }
for (unsigned i = 0, e = NewVT.getVectorNumElements(); i < e; i++, j++) {
SDValue Elt = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::i32, NewV,
- DAG.getIntPtrConstant(i, dl));
+ DAG.getIntPtrConstant(i * ExtScale, dl));
ConVec = DAG.getNode(ISD::INSERT_VECTOR_ELT, dl, ConcatVT, ConVec, Elt,
DAG.getConstant(j, dl, MVT::i32));
}
@@ -9116,14 +9123,7 @@ static SDValue LowerCONCAT_VECTORS_i1(SDValue Op, SelectionDAG &DAG,
ConVec = ExtractInto(NewV2, ConVec, j);
// Now return the result of comparing the subvector with zero, which will
- // generate a real predicate, i.e. v4i1, v8i1 or v16i1. For a v2i1 we
- // convert to a v4i1 compare to fill in the two halves of the i64 as i32s.
- if (VT == MVT::v2i1) {
- SDValue BC = DAG.getNode(ARMISD::VECTOR_REG_CAST, dl, MVT::v4i32, ConVec);
- SDValue Cmp = DAG.getNode(ARMISD::VCMPZ, dl, MVT::v4i1, BC,
- DAG.getConstant(ARMCC::NE, dl, MVT::i32));
- return DAG.getNode(ARMISD::PREDICATE_CAST, dl, MVT::v2i1, Cmp);
- }
+ // generate a real predicate, i.e. v4i1, v8i1 or v16i1.
return DAG.getNode(ARMISD::VCMPZ, dl, VT, ConVec,
DAG.getConstant(ARMCC::NE, dl, MVT::i32));
};
diff --git a/llvm/test/CodeGen/Thumb2/mve-concat.ll b/llvm/test/CodeGen/Thumb2/mve-concat.ll
new file mode 100644
index 000000000000000..0a3d9be820e6add
--- /dev/null
+++ b/llvm/test/CodeGen/Thumb2/mve-concat.ll
@@ -0,0 +1,219 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=thumbv8.1m.main-none-none-eabi -mattr=+mve -verify-machineinstrs %s -o - | FileCheck %s
+
+define arm_aapcs_vfpcc <4 x i32> @concat_v2i1(i32 %a, i32 %b, <4 x i32> %c) {
+; CHECK-LABEL: concat_v2i1:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: vmsr p0, r1
+; CHECK-NEXT: vmov.i8 q1, #0x0
+; CHECK-NEXT: vmov.i8 q2, #0xff
+; CHECK-NEXT: vpsel q3, q2, q1
+; CHECK-NEXT: vmsr p0, r0
+; CHECK-NEXT: vpsel q1, q2, q1
+; CHECK-NEXT: vmov r1, s12
+; CHECK-NEXT: vmov r0, s4
+; CHECK-NEXT: vmov q2[2], q2[0], r0, r1
+; CHECK-NEXT: vmov r1, s6
+; CHECK-NEXT: vmov r0, s14
+; CHECK-NEXT: vmov.i32 q1, #0x0
+; CHECK-NEXT: vmov q2[3], q2[1], r1, r0
+; CHECK-NEXT: vcmp.i32 ne, q2, zr
+; CHECK-NEXT: vpsel q0, q0, q1
+; CHECK-NEXT: bx lr
+entry:
+ %ai = call <2 x i1> @llvm.arm.mve.pred.i2v.v2i1(i32 %a)
+ %bi = call <2 x i1> @llvm.arm.mve.pred.i2v.v2i1(i32 %b)
+ %s = shufflevector <2 x i1> %ai, <2 x i1> %bi, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+ %ci = select <4 x i1> %s, <4 x i32> %c, <4 x i32> zeroinitializer
+ ret <4 x i32> %ci
+}
+
+declare <2 x i1> @llvm.arm.mve.pred.i2v.v2i1(i32)
+
+
+define arm_aapcs_vfpcc <8 x i16> @concat_v4i1(<4 x i32> %a, <4 x i32> %b, <8 x i16> %c) {
+; CHECK-LABEL: concat_v4i1:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: .vsave {d8, d9, d10, d11}
+; CHECK-NEXT: vpush {d8, d9, d10, d11}
+; CHECK-NEXT: vmov.i8 q3, #0x0
+; CHECK-NEXT: vmov.i8 q4, #0xff
+; CHECK-NEXT: vcmp.s32 lt, q0, zr
+; CHECK-NEXT: vpsel q5, q4, q3
+; CHECK-NEXT: vcmp.s32 lt, q1, zr
+; CHECK-NEXT: vmov r0, r1, d10
+; CHECK-NEXT: vpsel q1, q4, q3
+; CHECK-NEXT: vmov.16 q0[0], r0
+; CHECK-NEXT: vmov.16 q0[1], r1
+; CHECK-NEXT: vmov r0, r1, d11
+; CHECK-NEXT: vmov.16 q0[2], r0
+; CHECK-NEXT: vmov.16 q0[3], r1
+; CHECK-NEXT: vmov r0, r1, d2
+; CHECK-NEXT: vmov.16 q0[4], r0
+; CHECK-NEXT: vmov.16 q0[5], r1
+; CHECK-NEXT: vmov r0, r1, d3
+; CHECK-NEXT: vmov.16 q0[6], r0
+; CHECK-NEXT: vmov.16 q0[7], r1
+; CHECK-NEXT: vcmp.i16 ne, q0, zr
+; CHECK-NEXT: vmov.i32 q0, #0x0
+; CHECK-NEXT: vpsel q0, q2, q0
+; CHECK-NEXT: vpop {d8, d9, d10, d11}
+; CHECK-NEXT: bx lr
+entry:
+ %ai = icmp slt <4 x i32> %a, zeroinitializer
+ %bi = icmp slt <4 x i32> %b, zeroinitializer
+ %s = shufflevector <4 x i1> %ai, <4 x i1> %bi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %ci = select <8 x i1> %s, <8 x i16> %c, <8 x i16> zeroinitializer
+ ret <8 x i16> %ci
+}
+
+define arm_aapcs_vfpcc <16 x i8> @concat_v8i1(<8 x i16> %a, <8 x i16> %b, <16 x i8> %c) {
+; CHECK-LABEL: concat_v8i1:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: .vsave {d8, d9, d10, d11}
+; CHECK-NEXT: vpush {d8, d9, d10, d11}
+; CHECK-NEXT: vmov.i8 q3, #0x0
+; CHECK-NEXT: vmov.i8 q4, #0xff
+; CHECK-NEXT: vcmp.s16 lt, q0, zr
+; CHECK-NEXT: vpsel q5, q4, q3
+; CHECK-NEXT: vcmp.s16 lt, q1, zr
+; CHECK-NEXT: vmov.u16 r0, q5[0]
+; CHECK-NEXT: vpsel q1, q4, q3
+; CHECK-NEXT: vmov.8 q0[0], r0
+; CHECK-NEXT: vmov.u16 r0, q5[1]
+; CHECK-NEXT: vmov.8 q0[1], r0
+; CHECK-NEXT: vmov.u16 r0, q5[2]
+; CHECK-NEXT: vmov.8 q0[2], r0
+; CHECK-NEXT: vmov.u16 r0, q5[3]
+; CHECK-NEXT: vmov.8 q0[3], r0
+; CHECK-NEXT: vmov.u16 r0, q5[4]
+; CHECK-NEXT: vmov.8 q0[4], r0
+; CHECK-NEXT: vmov.u16 r0, q5[5]
+; CHECK-NEXT: vmov.8 q0[5], r0
+; CHECK-NEXT: vmov.u16 r0, q5[6]
+; CHECK-NEXT: vmov.8 q0[6], r0
+; CHECK-NEXT: vmov.u16 r0, q5[7]
+; CHECK-NEXT: vmov.8 q0[7], r0
+; CHECK-NEXT: vmov.u16 r0, q1[0]
+; CHECK-NEXT: vmov.8 q0[8], r0
+; CHECK-NEXT: vmov.u16 r0, q1[1]
+; CHECK-NEXT: vmov.8 q0[9], r0
+; CHECK-NEXT: vmov.u16 r0, q1[2]
+; CHECK-NEXT: vmov.8 q0[10], r0
+; CHECK-NEXT: vmov.u16 r0, q1[3]
+; CHECK-NEXT: vmov.8 q0[11], r0
+; CHECK-NEXT: vmov.u16 r0, q1[4]
+; CHECK-NEXT: vmov.8 q0[12], r0
+; CHECK-NEXT: vmov.u16 r0, q1[5]
+; CHECK-NEXT: vmov.8 q0[13], r0
+; CHECK-NEXT: vmov.u16 r0, q1[6]
+; CHECK-NEXT: vmov.8 q0[14], r0
+; CHECK-NEXT: vmov.u16 r0, q1[7]
+; CHECK-NEXT: vmov.8 q0[15], r0
+; CHECK-NEXT: vcmp.i8 ne, q0, zr
+; CHECK-NEXT: vmov.i32 q0, #0x0
+; CHECK-NEXT: vpsel q0, q2, q0
+; CHECK-NEXT: vpop {d8, d9, d10, d11}
+; CHECK-NEXT: bx lr
+entry:
+ %ai = icmp slt <8 x i16> %a, zeroinitializer
+ %bi = icmp slt <8 x i16> %b, zeroinitializer
+ %s = shufflevector <8 x i1> %ai, <8 x i1> %bi, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %ci = select <16 x i1> %s, <16 x i8> %c, <16 x i8> zeroinitializer
+ ret <16 x i8> %ci
+}
+
+
+define arm_aapcs_vfpcc <16 x i8> @concat_v48i1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %d, <4 x i32> %e, <16 x i8> %c) {
+; CHECK-LABEL: concat_v48i1:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: .vsave {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT: vpush {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT: vmov.i8 q4, #0x0
+; CHECK-NEXT: vmov.i8 q5, #0xff
+; CHECK-NEXT: vcmp.s32 lt, q0, zr
+; CHECK-NEXT: vpsel q6, q5, q4
+; CHECK-NEXT: vcmp.s32 lt, q1, zr
+; CHECK-NEXT: vmov r0, r1, d12
+; CHECK-NEXT: vpsel q1, q5, q4
+; CHECK-NEXT: vmov.16 q0[0], r0
+; CHECK-NEXT: vmov.16 q0[1], r1
+; CHECK-NEXT: vmov r0, r1, d13
+; CHECK-NEXT: vmov.16 q0[2], r0
+; CHECK-NEXT: vmov.16 q0[3], r1
+; CHECK-NEXT: vmov r0, r1, d2
+; CHECK-NEXT: vmov.16 q0[4], r0
+; CHECK-NEXT: vmov.16 q0[5], r1
+; CHECK-NEXT: vmov r0, r1, d3
+; CHECK-NEXT: vmov.16 q0[6], r0
+; CHECK-NEXT: vmov.16 q0[7], r1
+; CHECK-NEXT: vcmp.i16 ne, q0, zr
+; CHECK-NEXT: vpsel q1, q5, q4
+; CHECK-NEXT: vcmp.s32 lt, q2, zr
+; CHECK-NEXT: vmov.u16 r0, q1[0]
+; CHECK-NEXT: vpsel q2, q5, q4
+; CHECK-NEXT: vmov.8 q0[0], r0
+; CHECK-NEXT: vmov.u16 r0, q1[1]
+; CHECK-NEXT: vmov.8 q0[1], r0
+; CHECK-NEXT: vmov.u16 r0, q1[2]
+; CHECK-NEXT: vmov.8 q0[2], r0
+; CHECK-NEXT: vmov.u16 r0, q1[3]
+; CHECK-NEXT: vmov.8 q0[3], r0
+; CHECK-NEXT: vmov.u16 r0, q1[4]
+; CHECK-NEXT: vmov.8 q0[4], r0
+; CHECK-NEXT: vmov.u16 r0, q1[5]
+; CHECK-NEXT: vmov.8 q0[5], r0
+; CHECK-NEXT: vmov.u16 r0, q1[6]
+; CHECK-NEXT: vmov.8 q0[6], r0
+; CHECK-NEXT: vmov.u16 r0, q1[7]
+; CHECK-NEXT: vmov.8 q0[7], r0
+; CHECK-NEXT: vmov r0, r1, d4
+; CHECK-NEXT: vmov.16 q1[0], r0
+; CHECK-NEXT: vcmp.s32 lt, q3, zr
+; CHECK-NEXT: vmov.16 q1[1], r1
+; CHECK-NEXT: vmov r0, r1, d5
+; CHECK-NEXT: vmov.16 q1[2], r0
+; CHECK-NEXT: vpsel q2, q5, q4
+; CHECK-NEXT: vmov.16 q1[3], r1
+; CHECK-NEXT: vmov r0, r1, d4
+; CHECK-NEXT: vmov.16 q1[4], r0
+; CHECK-NEXT: vmov.16 q1[5], r1
+; CHECK-NEXT: vmov r0, r1, d5
+; CHECK-NEXT: vmov.16 q1[6], r0
+; CHECK-NEXT: vmov.16 q1[7], r1
+; CHECK-NEXT: vcmp.i16 ne, q1, zr
+; CHECK-NEXT: vpsel q1, q5, q4
+; CHECK-NEXT: vmov.u16 r0, q1[0]
+; CHECK-NEXT: vmov.8 q0[8], r0
+; CHECK-NEXT: vmov.u16 r0, q1[1]
+; CHECK-NEXT: vmov.8 q0[9], r0
+; CHECK-NEXT: vmov.u16 r0, q1[2]
+; CHECK-NEXT: vmov.8 q0[10], r0
+; CHECK-NEXT: vmov.u16 r0, q1[3]
+; CHECK-NEXT: vmov.8 q0[11], r0
+; CHECK-NEXT: vmov.u16 r0, q1[4]
+; CHECK-NEXT: vmov.8 q0[12], r0
+; CHECK-NEXT: vmov.u16 r0, q1[5]
+; CHECK-NEXT: vmov.8 q0[13], r0
+; CHECK-NEXT: vmov.u16 r0, q1[6]
+; CHECK-NEXT: vmov.8 q0[14], r0
+; CHECK-NEXT: vmov.u16 r0, q1[7]
+; CHECK-NEXT: vmov.8 q0[15], r0
+; CHECK-NEXT: add r0, sp, #48
+; CHECK-NEXT: vldrw.u32 q1, [r0]
+; CHECK-NEXT: vcmp.i8 ne, q0, zr
+; CHECK-NEXT: vmov.i32 q0, #0x0
+; CHECK-NEXT: vpsel q0, q1, q0
+; CHECK-NEXT: vpop {d8, d9, d10, d11, d12, d13}
+; CHECK-NEXT: bx lr
+entry:
+ %ai = icmp slt <4 x i32> %a, zeroinitializer
+ %bi = icmp slt <4 x i32> %b, zeroinitializer
+ %di = icmp slt <4 x i32> %d, zeroinitializer
+ %ei = icmp slt <4 x i32> %e, zeroinitializer
+ %s1 = shufflevector <4 x i1> %ai, <4 x i1> %bi, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %s2 = shufflevector <4 x i1> %di, <4 x i1> %ei, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+ %s = shufflevector <8 x i1> %s1, <8 x i1> %s2, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+ %ci = select <16 x i1> %s, <16 x i8> %c, <16 x i8> zeroinitializer
+ ret <16 x i8> %ci
+}
More information about the llvm-commits
mailing list