[ARM] Tighten restrictions on use of SP in v8.1-M CSEL.

In the `CSEL Rd,Rm,Rn` instruction family (also including CSINC, CSINV
and CSNEG), the architecture lists it as CONSTRAINED UNPREDICTABLE
(i.e. SoftFail) to use SP in the Rd or Rm slot, but outright illegal
to use it in the Rn slot, not least because some encodings of that
form are used by MVE instructions such as UQRSHLL.

MC was treating all three slots the same, as SoftFail. So the only
reason UQRSHLL was disassembled correctly at all was because the MVE
decode table is separate from the Thumb2 one and takes priority; if
you turned off MVE, then encodings such as `[0x5f,0xea,0x0d,0x83]`
would disassemble as spurious CSELs.

Fixed by inventing another version of the `GPRwithZR` register class,
which disallows SP completely instead of just SoftFailing it.

Reviewers: DavidSpickett, ostannard

Subscribers: javed.absar, kristof.beyls, hiraditya, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63862

llvm-svn: 364531

llvm/lib/Target/ARM/ARMInstrThumb2.td

@@ -5223,7 +5223,7 @@ def t2LoopEnd :
 } // end isNotDuplicable
 
 class CS<string iname, bits<4> opcode, list<dag> pattern=[]>
-  : V8_1MI<(outs rGPR:$Rd), (ins GPRwithZR:$Rn, GPRwithZR:$Rm, pred_noal:$fcond),
+  : V8_1MI<(outs rGPR:$Rd), (ins GPRwithZR:$Rn, GPRwithZRnosp:$Rm, pred_noal:$fcond),
            AddrModeNone, NoItinerary, iname, "$Rd, $Rn, $Rm, $fcond", "", pattern> {
   bits<4> Rd;
   bits<4> Rm;
@@ -5255,11 +5255,11 @@ let Predicates = [HasV8_1MMainline] in {
                  (t2CSINC rGPR:$Rd, ZR, ZR, pred_noal_inv:$fcond)>;
 
   def : InstAlias<"cinc\t$Rd, $Rn, $fcond",
-                 (t2CSINC rGPR:$Rd, GPRwithZR:$Rn, GPRwithZR:$Rn, pred_noal_inv:$fcond)>;
+                 (t2CSINC rGPR:$Rd, GPRwithZRnosp:$Rn, GPRwithZRnosp:$Rn, pred_noal_inv:$fcond)>;
 
   def : InstAlias<"cinv\t$Rd, $Rn, $fcond",
-                 (t2CSINV rGPR:$Rd, GPRwithZR:$Rn, GPRwithZR:$Rn, pred_noal_inv:$fcond)>;
+                 (t2CSINV rGPR:$Rd, GPRwithZRnosp:$Rn, GPRwithZRnosp:$Rn, pred_noal_inv:$fcond)>;
 
   def : InstAlias<"cneg\t$Rd, $Rn, $fcond",
-                 (t2CSNEG rGPR:$Rd, GPRwithZR:$Rn, GPRwithZR:$Rn, pred_noal_inv:$fcond)>;
+                 (t2CSNEG rGPR:$Rd, GPRwithZRnosp:$Rn, GPRwithZRnosp:$Rn, pred_noal_inv:$fcond)>;
 }

llvm/lib/Target/ARM/ARMRegisterInfo.td

@@ -261,13 +261,19 @@ def GPRwithAPSRnosp : RegisterClass<"ARM", [i32], 32, (add (sequence "R%u", 0, 1
   let isAllocatable = 0;
 }
 
-def GPRwithZR : RegisterClass<"ARM", [i32], 32, (add (sequence "R%u", 0, 12),
+def GPRwithZR : RegisterClass<"ARM", [i32], 32, (add (sub GPR, PC), ZR)> {
-                                              LR, ZR)> {
-
   let AltOrders = [(add LR, GPRwithZR), (trunc GPRwithZR, 8)];
   let AltOrderSelect = [{
       return 1 + MF.getSubtarget<ARMSubtarget>().isThumb1Only();
   }];
+  let DiagnosticString = "operand must be a register in range [r0, r14] or zr";
+}
+
+def GPRwithZRnosp : RegisterClass<"ARM", [i32], 32, (sub GPRwithZR, SP)> {
+  let AltOrders = [(add LR, GPRwithZRnosp), (trunc GPRwithZRnosp, 8)];
+  let AltOrderSelect = [{
+      return 1 + MF.getSubtarget<ARMSubtarget>().isThumb1Only();
+  }];
   let DiagnosticString = "operand must be a register in range [r0, r12] or r14 or zr";
 }
 

llvm/lib/Target/ARM/Disassembler/ARMDisassembler.cpp

@@ -194,6 +194,8 @@ static DecodeStatus DecodeGPRwithAPSRRegisterClass(MCInst &Inst,
 static DecodeStatus DecodeGPRwithZRRegisterClass(MCInst &Inst,
                                                unsigned RegNo, uint64_t Address,
                                                const void *Decoder);
+static DecodeStatus DecodeGPRwithZRnospRegisterClass(
+    MCInst &Inst, unsigned RegNo, uint64_t Address, const void *Decoder);
 static DecodeStatus DecodetGPRRegisterClass(MCInst &Inst, unsigned RegNo,
                                    uint64_t Address, const void *Decoder);
 static DecodeStatus DecodetcGPRRegisterClass(MCInst &Inst, unsigned RegNo,
@@ -1184,12 +1186,22 @@ DecodeGPRwithZRRegisterClass(MCInst &Inst, unsigned RegNo,
   }
 
   if (RegNo == 13)
-    S = MCDisassembler::SoftFail;
+    Check(S, MCDisassembler::SoftFail);
 
   Check(S, DecodeGPRRegisterClass(Inst, RegNo, Address, Decoder));
   return S;
 }
 
+static DecodeStatus
+DecodeGPRwithZRnospRegisterClass(MCInst &Inst, unsigned RegNo,
+                                 uint64_t Address, const void *Decoder) {
+  DecodeStatus S = MCDisassembler::Success;
+  if (RegNo == 13)
+    return MCDisassembler::Fail;
+  Check(S, DecodeGPRwithZRRegisterClass(Inst, RegNo, Address, Decoder));
+  return S;
+}
+
 static DecodeStatus DecodetGPRRegisterClass(MCInst &Inst, unsigned RegNo,
                                    uint64_t Address, const void *Decoder) {
   if (RegNo > 7)

llvm/test/MC/Disassembler/ARM/mve-scalar-shift.txt

@@ -28,7 +28,7 @@
 
 [0x5f 0xea 0x2d 0x83]
 # CHECK: sqrshrl lr, r3, r8  @ encoding: [0x5f,0xea,0x2d,0x83]
-# CHECK-NOMVE: [[@LINE-2]]:2: warning: potentially undefined instruction encoding
+# CHECK-NOMVE: [[@LINE-2]]:2: warning: invalid instruction encoding
 
 [0x5e 0xea 0x7f 0x4f]
 # CHECK: sqshl   lr, #17  @ encoding: [0x5e,0xea,0x7f,0x4f]

llvm/test/MC/Disassembler/ARM/thumbv8.1m.s

@@ -1,4 +1,4 @@
-# RUN: llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -show-encoding %s 2> %t | FileCheck %s
+# RUN: not llvm-mc -disassemble -triple=thumbv8.1m.main-none-eabi -show-encoding %s 2> %t | FileCheck %s
 # RUN: FileCheck --check-prefix=ERROR < %t %s
 
 [0x52 0xea 0x22 0x9e]
@@ -37,6 +37,20 @@
 [0x50,0xea,0x01,0x80]
 # CHECK: csel r0, r0, r1, eq @ encoding: [0x50,0xea,0x01,0x80]
 
+[0x51,0xea,0x02,0x8d]
+# CHECK: csel sp, r1, r2, eq @ encoding: [0x51,0xea,0x02,0x8d]
+# ERROR: [[@LINE-2]]:2: warning: potentially undefined instruction encoding
+
+[0x5d,0xea,0x02,0x80]
+# CHECK: csel r0, sp, r2, eq @ encoding: [0x5d,0xea,0x02,0x80]
+# ERROR: [[@LINE-2]]:2: warning: potentially undefined instruction encoding
+
+[0x51,0xea,0x0d,0x80]
+# ERROR: [[@LINE-1]]:2: warning: invalid instruction encoding
+
+[0x5f,0xea,0x0d,0x83]
+# ERROR: [[@LINE-1]]:2: warning: invalid instruction encoding
+
 [0x5d 0xea 0x22 0x9e]
 # ERROR: [[@LINE-1]]:2: warning: potentially undefined instruction encoding