[AArch64][GlobalISel] Support lowering <1 x i8> arguments.

We don't yet have working codegen for the resulting unmerges, and if
we did it would probably be horrible.

Differential Revision: https://reviews.llvm.org/D97035

diff --git a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
index 9b7954d..4ef66fc 100644 (file)
--- a/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
+++ b/llvm/lib/CodeGen/GlobalISel/CallLowering.cpp
@@ -545,14 +545,21 @@ bool CallLowering::handleAssignments(CCState &CCInfo,
       // or do an unmerge to get the lower block of elements.
       if (VATy.isVector() &&
           VATy.getNumElements() > OrigVT.getVectorNumElements()) {
-        // Just handle the case where the VA type is 2 * original type.
-        if (VATy.getNumElements() != OrigVT.getVectorNumElements() * 2) {
-          LLVM_DEBUG(dbgs()
-                     << "Incoming promoted vector arg has too many elts");
+        // Just handle the case where the VA type is a multiple of original
+        // type.
+        if (VATy.getNumElements() % OrigVT.getVectorNumElements() != 0) {
+          LLVM_DEBUG(dbgs() << "Incoming promoted vector arg elts is not a "
+                               "multiple of orig type elt: "
+                            << VATy << " vs " << OrigTy);
           return false;
         }
-        auto Unmerge = MIRBuilder.buildUnmerge({OrigTy, OrigTy}, {NewReg});
-        MIRBuilder.buildCopy(ArgReg, Unmerge.getReg(0));
+        SmallVector<Register, 4> DstRegs = {ArgReg};
+        unsigned NumParts =
+            VATy.getNumElements() / OrigVT.getVectorNumElements() - 1;
+        for (unsigned Idx = 0; Idx < NumParts; ++Idx)
+          DstRegs.push_back(
+              MIRBuilder.getMRI()->createGenericVirtualRegister(OrigTy));
+        MIRBuilder.buildUnmerge(DstRegs, {NewReg});
       } else {
         MIRBuilder.buildTrunc(ArgReg, {NewReg}).getReg(0);
       }

diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/call-lowering-vectors.ll b/llvm/test/CodeGen/AArch64/GlobalISel/call-lowering-vectors.ll
new file mode 100644 (file)
index 0000000..2ade4ce
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/call-lowering-vectors.ll
@@ -0,0 +1,15 @@
+; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+; RUN: llc -mtriple=aarch64-linux-gnu -O0 -stop-after=irtranslator -global-isel -verify-machineinstrs %s -o - 2>&1 | FileCheck %s
+
+define i8 @v1s8_add(<1 x i8> %a0) nounwind {
+  ; CHECK-LABEL: name: v1s8_add
+  ; CHECK: bb.1 (%ir-block.0):
+  ; CHECK:   liveins: $d0
+  ; CHECK:   [[COPY:%[0-9]+]]:_(<8 x s8>) = COPY $d0
+  ; CHECK:   [[UV:%[0-9]+]]:_(s8), [[UV1:%[0-9]+]]:_(s8), [[UV2:%[0-9]+]]:_(s8), [[UV3:%[0-9]+]]:_(s8), [[UV4:%[0-9]+]]:_(s8), [[UV5:%[0-9]+]]:_(s8), [[UV6:%[0-9]+]]:_(s8), [[UV7:%[0-9]+]]:_(s8) = G_UNMERGE_VALUES [[COPY]](<8 x s8>)
+  ; CHECK:   [[ANYEXT:%[0-9]+]]:_(s32) = G_ANYEXT [[UV]](s8)
+  ; CHECK:   $w0 = COPY [[ANYEXT]](s32)
+  ; CHECK:   RET_ReallyLR implicit $w0
+  %res = bitcast <1 x i8> %a0 to i8
+  ret i8 %res
+}

diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/ret-1x-vec.ll b/llvm/test/CodeGen/AArch64/GlobalISel/ret-1x-vec.ll
index 529f69d..8d5ded8 100644 (file)
--- a/llvm/test/CodeGen/AArch64/GlobalISel/ret-1x-vec.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/ret-1x-vec.ll
@@ -7,9 +7,8 @@ define <1 x float> @foo(<1 x float> %v) {
   ; CHECK:   liveins: $d0
   ; CHECK:   [[COPY:%[0-9]+]]:_(<2 x s32>) = COPY $d0
   ; CHECK:   [[UV:%[0-9]+]]:_(s32), [[UV1:%[0-9]+]]:_(s32) = G_UNMERGE_VALUES [[COPY]](<2 x s32>)
-  ; CHECK:   [[COPY1:%[0-9]+]]:_(s32) = COPY [[UV]](s32)
   ; CHECK:   [[DEF:%[0-9]+]]:_(s32) = G_IMPLICIT_DEF
-  ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[COPY1]](s32), [[DEF]](s32)
+  ; CHECK:   [[BUILD_VECTOR:%[0-9]+]]:_(<2 x s32>) = G_BUILD_VECTOR [[UV]](s32), [[DEF]](s32)
   ; CHECK:   $d0 = COPY [[BUILD_VECTOR]](<2 x s32>)
   ; CHECK:   RET_ReallyLR implicit $d0
   ret <1 x float> %v

diff --git a/llvm/test/CodeGen/AArch64/GlobalISel/vec-s16-param.ll b/llvm/test/CodeGen/AArch64/GlobalISel/vec-s16-param.ll
index f8319a9..2684756 100644 (file)
--- a/llvm/test/CodeGen/AArch64/GlobalISel/vec-s16-param.ll
+++ b/llvm/test/CodeGen/AArch64/GlobalISel/vec-s16-param.ll
@@ -7,9 +7,8 @@ define <2 x half> @f16_vec_param(<2 x half> %v) {
   ; CHECK:   liveins: $d0
   ; CHECK:   [[COPY:%[0-9]+]]:_(<4 x s16>) = COPY $d0
   ; CHECK:   [[UV:%[0-9]+]]:_(<2 x s16>), [[UV1:%[0-9]+]]:_(<2 x s16>) = G_UNMERGE_VALUES [[COPY]](<4 x s16>)
-  ; CHECK:   [[COPY1:%[0-9]+]]:_(<2 x s16>) = COPY [[UV]](<2 x s16>)
   ; CHECK:   [[DEF:%[0-9]+]]:_(<2 x s16>) = G_IMPLICIT_DEF
-  ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[COPY1]](<2 x s16>), [[DEF]](<2 x s16>)
+  ; CHECK:   [[CONCAT_VECTORS:%[0-9]+]]:_(<4 x s16>) = G_CONCAT_VECTORS [[UV]](<2 x s16>), [[DEF]](<2 x s16>)
   ; CHECK:   $d0 = COPY [[CONCAT_VECTORS]](<4 x s16>)
   ; CHECK:   RET_ReallyLR implicit $d0
   ret <2 x half> %v