[WebAssembly] Fix load/store name detection for atomic instructions

[aheejin]

Summary

Fixed a bug in the routine in AsmParser that determines whether the
current instruction is a load or a store. Atomic instructions' prefixes
are not atomic_ but atomic., and all atomic instructions are also
memory instructions. Also fixed the printing format of atomic
instructions to match other memory instructions and added encoding tests
for atomic instructions.

[tlively]

test/MC/WebAssembly/atomics-encodings.s

1 ↗

(On Diff #187183)

🎉

[aardappel]

lib/Target/WebAssembly/WebAssemblyInstrAtomics.td

340 ↗

(On Diff #187183)

Is this syntax in line with all other uses of p2align in .td?

[aheejin]

lib/Target/WebAssembly/WebAssemblyInstrAtomics.td

340 ↗

(On Diff #187183)

Yes: example

llvm/trunk/lib/Target/WebAssembly/AsmParser/WebAssemblyAsmParser.cpp

#include "WebAssemblyGenAsmMatcher.inc"
   }
 
   bool parseOperandStartingWithInteger(bool IsNegative, OperandVector &Operands,
                                        StringRef InstName) {
     parseSingleInteger(IsNegative, Operands);
     // FIXME: there is probably a cleaner way to do this.
     auto IsLoadStore = InstName.startswith("load") ||
                        InstName.startswith("store") ||
-                       InstName.startswith("atomic_load") ||
-                       InstName.startswith("atomic_store");
+                       InstName.startswith("atomic");
     if (IsLoadStore) {
       // Parse load/store operands of the form: offset align
       auto &Offset = Lexer.getTok();
       if (Offset.is(AsmToken::Integer)) {
         parseSingleInteger(false, Operands);
       } else {
         // Alignment not specified.
         // FIXME: correctly derive a default from the instruction.

llvm/trunk/lib/Target/WebAssembly/WebAssemblyInstrAtomics.td

 // Atomic binary read-modify-writes
 //===----------------------------------------------------------------------===//
 
 multiclass WebAssemblyBinRMW<WebAssemblyRegClass rc, string Name, int Opcode> {
   defm "" : I<(outs rc:$dst),
               (ins P2Align:$p2align, offset32_op:$off, I32:$addr, rc:$val),
               (outs), (ins P2Align:$p2align, offset32_op:$off), [],
               !strconcat(Name, "\t$dst, ${off}(${addr})${p2align}, $val"),
-              !strconcat(Name, "\t${off}, ${p2align}"), Opcode>;
+              !strconcat(Name, "\t${off}${p2align}"), Opcode>;
 }
 
 defm ATOMIC_RMW_ADD_I32 : WebAssemblyBinRMW<I32, "i32.atomic.rmw.add", 0xfe1e>;
 defm ATOMIC_RMW_ADD_I64 : WebAssemblyBinRMW<I64, "i64.atomic.rmw.add", 0xfe1f>;
 defm ATOMIC_RMW8_U_ADD_I32 :
   WebAssemblyBinRMW<I32, "i32.atomic.rmw8.add_u", 0xfe20>;
 defm ATOMIC_RMW16_U_ADD_I32 :
   WebAssemblyBinRMW<I32, "i32.atomic.rmw16.add_u", 0xfe21>;

 // if it is frequent.
 
 multiclass WebAssemblyTerRMW<WebAssemblyRegClass rc, string Name, int Opcode> {
   defm "" : I<(outs rc:$dst),
               (ins P2Align:$p2align, offset32_op:$off, I32:$addr, rc:$exp,
                    rc:$new),
               (outs), (ins P2Align:$p2align, offset32_op:$off), [],
               !strconcat(Name, "\t$dst, ${off}(${addr})${p2align}, $exp, $new"),
-              !strconcat(Name, "\t${off}, ${p2align}"), Opcode>;
+              !strconcat(Name, "\t${off}${p2align}"), Opcode>;
 }
 
 defm ATOMIC_RMW_CMPXCHG_I32 :
   WebAssemblyTerRMW<I32, "i32.atomic.rmw.cmpxchg", 0xfe48>;
 defm ATOMIC_RMW_CMPXCHG_I64 :
   WebAssemblyTerRMW<I64, "i64.atomic.rmw.cmpxchg", 0xfe49>;
 defm ATOMIC_RMW8_U_CMPXCHG_I32 :
   WebAssemblyTerRMW<I32, "i32.atomic.rmw8.cmpxchg_u", 0xfe4a>;

 //===----------------------------------------------------------------------===//
 
 let hasSideEffects = 1 in {
 defm ATOMIC_NOTIFY :
   I<(outs I32:$dst),
     (ins P2Align:$p2align, offset32_op:$off, I32:$addr, I32:$count),
     (outs), (ins P2Align:$p2align, offset32_op:$off), [],
     "atomic.notify \t$dst, ${off}(${addr})${p2align}, $count",
-    "atomic.notify \t${off}, ${p2align}", 0xfe00>;
+    "atomic.notify \t${off}${p2align}", 0xfe00>;
 let mayLoad = 1 in {
 defm ATOMIC_WAIT_I32 :
   I<(outs I32:$dst),
     (ins P2Align:$p2align, offset32_op:$off, I32:$addr, I32:$exp, I64:$timeout),
     (outs), (ins P2Align:$p2align, offset32_op:$off), [],
     "i32.atomic.wait \t$dst, ${off}(${addr})${p2align}, $exp, $timeout",
-    "i32.atomic.wait \t${off}, ${p2align}", 0xfe01>;
+    "i32.atomic.wait \t${off}${p2align}", 0xfe01>;
 defm ATOMIC_WAIT_I64 :
   I<(outs I32:$dst),
     (ins P2Align:$p2align, offset32_op:$off, I32:$addr, I64:$exp, I64:$timeout),
     (outs), (ins P2Align:$p2align, offset32_op:$off), [],
     "i64.atomic.wait \t$dst, ${off}(${addr})${p2align}, $exp, $timeout",
-    "i64.atomic.wait \t${off}, ${p2align}", 0xfe02>;
+    "i64.atomic.wait \t${off}${p2align}", 0xfe02>;
 } // mayLoad = 1
 } // hasSideEffects = 1
 
 let Predicates = [HasAtomics] in {
 // Select notifys with no constant offset.
 class NotifyPatNoOffset<Intrinsic kind> :
   Pat<(i32 (kind I32:$addr, I32:$count)),
       (ATOMIC_NOTIFY 0, 0, I32:$addr, I32:$count)>;

llvm/trunk/test/MC/WebAssembly/atomics-encodings.s

+# RUN: llvm-mc -show-encoding -triple=wasm32-unkown-unknown -mattr=+atomics < %s | FileCheck %s
+
+main:
+  .functype main () -> ()
+
+  # FIXME This doesn't work because of PR40728. Enable this once it's fixed.
+  # C HECK:  atomic.notify 0:p2align=0 # encoding: [0xfe,0x00,0x00,0x00]
+  # atomic.notify 0
+  # CHECK:  i32.atomic.wait 0:p2align=0 # encoding: [0xfe,0x01,0x00,0x00]
+  i32.atomic.wait 0
+  # CHECK:  i64.atomic.wait 0:p2align=0 # encoding: [0xfe,0x02,0x00,0x00]
+  i64.atomic.wait 0
+
+  # CHECK: i32.atomic.load 0:p2align=0 # encoding: [0xfe,0x10,0x00,0x00]
+  i32.atomic.load 0
+  # CHECK: i64.atomic.load 4:p2align=0 # encoding: [0xfe,0x11,0x00,0x04]
+  i64.atomic.load 4
+  # CHECK:  i32.atomic.load8_u 48 # encoding: [0xfe,0x12,0x00,0x30]
+  i32.atomic.load8_u 48
+  # CHECK:  i32.atomic.load16_u 0:p2align=0 # encoding: [0xfe,0x13,0x00,0x00]
+  i32.atomic.load16_u 0
+  # CHECK:  i64.atomic.load8_u 0 # encoding: [0xfe,0x14,0x00,0x00]
+  i64.atomic.load8_u 0
+  # CHECK:  i64.atomic.load16_u 0:p2align=0 # encoding: [0xfe,0x15,0x00,0x00]
+  i64.atomic.load16_u 0
+  # CHECK:  i64.atomic.load32_u 0:p2align=0 # encoding: [0xfe,0x16,0x00,0x00]
+  i64.atomic.load32_u 0
+
+  # CHECK:  i32.atomic.store 0:p2align=0 # encoding: [0xfe,0x17,0x00,0x00]
+  i32.atomic.store 0
+  # CHECK:  i64.atomic.store 8:p2align=0 # encoding: [0xfe,0x18,0x00,0x08]
+  i64.atomic.store 8
+  # CHECK:  i32.atomic.store8 0 # encoding: [0xfe,0x19,0x00,0x00]
+  i32.atomic.store8 0
+  # CHECK:  i32.atomic.store16 0:p2align=0 # encoding: [0xfe,0x1a,0x00,0x00]
+  i32.atomic.store16 0
+  # CHECK:  i64.atomic.store8 16 # encoding: [0xfe,0x1b,0x00,0x10]
+  i64.atomic.store8 16
+  # CHECK:  i64.atomic.store16 0:p2align=0 # encoding: [0xfe,0x1c,0x00,0x00]
+  i64.atomic.store16 0
+  # CHECK:  i64.atomic.store32 0:p2align=0 # encoding: [0xfe,0x1d,0x00,0x00]
+  i64.atomic.store32 0
+
+  # CHECK:  i32.atomic.rmw.add 0:p2align=0 # encoding: [0xfe,0x1e,0x00,0x00]
+  i32.atomic.rmw.add 0
+  # CHECK:  i64.atomic.rmw.add 0:p2align=0 # encoding: [0xfe,0x1f,0x00,0x00]
+  i64.atomic.rmw.add 0
+  # CHECK:  i32.atomic.rmw8.add_u 0 # encoding: [0xfe,0x20,0x00,0x00]
+  i32.atomic.rmw8.add_u 0
+  # CHECK:  i32.atomic.rmw16.add_u 0:p2align=0 # encoding: [0xfe,0x21,0x00,0x00]
+  i32.atomic.rmw16.add_u 0
+  # CHECK:  i64.atomic.rmw8.add_u 0 # encoding: [0xfe,0x22,0x00,0x00]
+  i64.atomic.rmw8.add_u 0
+  # CHECK:  i64.atomic.rmw16.add_u 0:p2align=0 # encoding: [0xfe,0x23,0x00,0x00]
+  i64.atomic.rmw16.add_u 0
+  # CHECK:  i64.atomic.rmw32.add_u 16:p2align=0 # encoding: [0xfe,0x24,0x00,0x10]
+  i64.atomic.rmw32.add_u 16
+
+  # CHECK:  i32.atomic.rmw.sub 0:p2align=0 # encoding: [0xfe,0x25,0x00,0x00]
+  i32.atomic.rmw.sub 0
+  # CHECK:  i64.atomic.rmw.sub 0:p2align=0 # encoding: [0xfe,0x26,0x00,0x00]
+  i64.atomic.rmw.sub 0
+  # CHECK:  i32.atomic.rmw8.sub_u 0 # encoding: [0xfe,0x27,0x00,0x00]
+  i32.atomic.rmw8.sub_u 0
+  # CHECK:  i32.atomic.rmw16.sub_u 0:p2align=0 # encoding: [0xfe,0x28,0x00,0x00]
+  i32.atomic.rmw16.sub_u 0
+  # CHECK:  i64.atomic.rmw8.sub_u 8 # encoding: [0xfe,0x29,0x00,0x08]
+  i64.atomic.rmw8.sub_u 8
+  # CHECK:  i64.atomic.rmw16.sub_u 0:p2align=0 # encoding: [0xfe,0x2a,0x00,0x00]
+  i64.atomic.rmw16.sub_u 0
+  # CHECK:  i64.atomic.rmw32.sub_u 0:p2align=0 # encoding: [0xfe,0x2b,0x00,0x00]
+  i64.atomic.rmw32.sub_u 0
+
+  # CHECK:  i32.atomic.rmw.and 0:p2align=0 # encoding: [0xfe,0x2c,0x00,0x00]
+  i32.atomic.rmw.and 0
+  # CHECK:  i64.atomic.rmw.and 0:p2align=0 # encoding: [0xfe,0x2d,0x00,0x00]
+  i64.atomic.rmw.and 0
+  # CHECK:  i32.atomic.rmw8.and_u 0 # encoding: [0xfe,0x2e,0x00,0x00]
+  i32.atomic.rmw8.and_u 0
+  # CHECK:  i32.atomic.rmw16.and_u 0:p2align=0 # encoding: [0xfe,0x2f,0x00,0x00]
+  i32.atomic.rmw16.and_u 0
+  # CHECK:  i64.atomic.rmw8.and_u 96 # encoding: [0xfe,0x30,0x00,0x60]
+  i64.atomic.rmw8.and_u 96
+  # CHECK:  i64.atomic.rmw16.and_u 0:p2align=0 # encoding: [0xfe,0x31,0x00,0x00]
+  i64.atomic.rmw16.and_u 0
+  # CHECK:  i64.atomic.rmw32.and_u 0:p2align=0 # encoding: [0xfe,0x32,0x00,0x00]
+  i64.atomic.rmw32.and_u 0
+
+  # CHECK:  i32.atomic.rmw.or 0:p2align=0 # encoding: [0xfe,0x33,0x00,0x00]
+  i32.atomic.rmw.or 0
+  # CHECK:  i64.atomic.rmw.or 0:p2align=0 # encoding: [0xfe,0x34,0x00,0x00]
+  i64.atomic.rmw.or 0
+  # CHECK:  i32.atomic.rmw8.or_u 0 # encoding: [0xfe,0x35,0x00,0x00]
+  i32.atomic.rmw8.or_u 0
+  # CHECK:  i32.atomic.rmw16.or_u 0:p2align=0 # encoding: [0xfe,0x36,0x00,0x00]
+  i32.atomic.rmw16.or_u 0
+  # CHECK:  i64.atomic.rmw8.or_u 0 # encoding: [0xfe,0x37,0x00,0x00]
+  i64.atomic.rmw8.or_u 0
+  # CHECK:  i64.atomic.rmw16.or_u 48:p2align=0 # encoding: [0xfe,0x38,0x00,0x30]
+  i64.atomic.rmw16.or_u 48
+  # CHECK:  i64.atomic.rmw32.or_u 0:p2align=0 # encoding: [0xfe,0x39,0x00,0x00]
+  i64.atomic.rmw32.or_u 0
+
+  # CHECK:  i32.atomic.rmw.xor 0:p2align=0 # encoding: [0xfe,0x3a,0x00,0x00]
+  i32.atomic.rmw.xor 0
+  # CHECK:  i64.atomic.rmw.xor 0:p2align=0 # encoding: [0xfe,0x3b,0x00,0x00]
+  i64.atomic.rmw.xor 0
+  # CHECK:  i32.atomic.rmw8.xor_u 4 # encoding: [0xfe,0x3c,0x00,0x04]
+  i32.atomic.rmw8.xor_u 4
+  # CHECK:  i32.atomic.rmw16.xor_u 0:p2align=0 # encoding: [0xfe,0x3d,0x00,0x00]
+  i32.atomic.rmw16.xor_u 0
+  # CHECK:  i64.atomic.rmw8.xor_u 0 # encoding: [0xfe,0x3e,0x00,0x00]
+  i64.atomic.rmw8.xor_u 0
+  # CHECK:  i64.atomic.rmw16.xor_u 0:p2align=0 # encoding: [0xfe,0x3f,0x00,0x00]
+  i64.atomic.rmw16.xor_u 0
+  # CHECK:  i64.atomic.rmw32.xor_u 0:p2align=0 # encoding: [0xfe,0x40,0x00,0x00]
+  i64.atomic.rmw32.xor_u 0
+
+  # CHECK:  i32.atomic.rmw.xchg 0:p2align=0 # encoding: [0xfe,0x41,0x00,0x00]
+  i32.atomic.rmw.xchg 0
+  # CHECK:  i64.atomic.rmw.xchg 0:p2align=0 # encoding: [0xfe,0x42,0x00,0x00]
+  i64.atomic.rmw.xchg 0
+  # CHECK:  i32.atomic.rmw8.xchg_u 0 # encoding: [0xfe,0x43,0x00,0x00]
+  i32.atomic.rmw8.xchg_u 0
+  # CHECK:  i32.atomic.rmw16.xchg_u 0:p2align=0 # encoding: [0xfe,0x44,0x00,0x00]
+  i32.atomic.rmw16.xchg_u 0
+  # CHECK:  i64.atomic.rmw8.xchg_u 0 # encoding: [0xfe,0x45,0x00,0x00]
+  i64.atomic.rmw8.xchg_u 0
+  # CHECK:  i64.atomic.rmw16.xchg_u 8:p2align=0 # encoding: [0xfe,0x46,0x00,0x08]
+  i64.atomic.rmw16.xchg_u 8
+  # CHECK:  i64.atomic.rmw32.xchg_u 0:p2align=0 # encoding: [0xfe,0x47,0x00,0x00]
+  i64.atomic.rmw32.xchg_u 0
+
+  # CHECK:  i32.atomic.rmw.cmpxchg 32:p2align=0 # encoding: [0xfe,0x48,0x00,0x20]
+  i32.atomic.rmw.cmpxchg 32
+  # CHECK:  i64.atomic.rmw.cmpxchg 0:p2align=0 # encoding: [0xfe,0x49,0x00,0x00]
+  i64.atomic.rmw.cmpxchg 0
+  # CHECK:  i32.atomic.rmw8.cmpxchg_u 0 # encoding: [0xfe,0x4a,0x00,0x00]
+  i32.atomic.rmw8.cmpxchg_u 0
+  # CHECK:  i32.atomic.rmw16.cmpxchg_u 0:p2align=0 # encoding: [0xfe,0x4b,0x00,0x00]
+  i32.atomic.rmw16.cmpxchg_u 0
+  # CHECK:  i64.atomic.rmw8.cmpxchg_u 16 # encoding: [0xfe,0x4c,0x00,0x10]
+  i64.atomic.rmw8.cmpxchg_u 16
+  # CHECK:  i64.atomic.rmw16.cmpxchg_u 0:p2align=0 # encoding: [0xfe,0x4d,0x00,0x00]
+  i64.atomic.rmw16.cmpxchg_u 0
+  # CHECK:  i64.atomic.rmw32.cmpxchg_u 0:p2align=0 # encoding: [0xfe,0x4e,0x00,0x00]
+  i64.atomic.rmw32.cmpxchg_u 0
+
+  end_function