[llvm] X86 data v2 (PR #156287)
via llvm-commits
llvm-commits at lists.llvm.org
Mon Sep 1 00:24:22 PDT 2025
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-x86
Author: Danny Milosavljevic (daym)
<details>
<summary>Changes</summary>
See also https://github.com/llvm/llvm-project/issues/156286
---
Full diff: https://github.com/llvm/llvm-project/pull/156287.diff
4 Files Affected:
- (modified) llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp (+32-6)
- (added) llvm/test/MC/X86/x86-16-intel.s (+13)
- (modified) llvm/test/MC/X86/x86-16.s (+12)
- (added) llvm/test/MC/X86/x86-32-intel.s (+28)
``````````diff
diff --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index d7671ed19589b..211aa47e48fe3 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -3523,8 +3523,26 @@ bool X86AsmParser::parseInstruction(ParseInstructionInfo &Info, StringRef Name,
PatchedName = Name;
// Hacks to handle 'data16' and 'data32'
- if (PatchedName == "data16" && is16BitMode()) {
- return Error(NameLoc, "redundant data16 prefix");
+ if (PatchedName == "data16") {
+ if (is16BitMode())
+ return Error(NameLoc, "redundant data16 prefix");
+ if (is64BitMode())
+ return Error(NameLoc, "'data16' is not supported in 64-bit mode");
+ if (getLexer().isNot(AsmToken::EndOfStatement)) {
+ StringRef Next = Parser.getTok().getString();
+ getLexer().Lex();
+ // data16 effectively changes the instruction suffix.
+ // TODO Generalize.
+ if (Next == "call")
+ Next = "callw";
+ if (Next == "ljmp")
+ Next = "ljmpw";
+
+ Name = Next;
+ PatchedName = Name;
+ ForcedDataPrefix = X86::Is16Bit;
+ IsPrefix = false;
+ }
}
if (PatchedName == "data32") {
if (is32BitMode())
@@ -4538,14 +4556,22 @@ bool X86AsmParser::matchAndEmitIntelInstruction(
if (X86Op->isImm()) {
// If it's not a constant fall through and let remainder take care of it.
const auto *CE = dyn_cast<MCConstantExpr>(X86Op->getImm());
- unsigned Size = getPointerWidth();
+ // Determine the size. Prioritize the ForcedDataPrefix flag if it was set
+ // by a 'data32' prefix. Otherwise, fall back to the pointer width of the
+ // current mode.
+ unsigned Size = (ForcedDataPrefix == X86::Is32Bit) ? 32
+ : (ForcedDataPrefix == X86::Is16Bit) ? 16
+ : getPointerWidth();
+ ForcedDataPrefix = 0;
if (CE &&
(isIntN(Size, CE->getValue()) || isUIntN(Size, CE->getValue()))) {
SmallString<16> Tmp;
Tmp += Base;
- Tmp += (is64BitMode())
- ? "q"
- : (is32BitMode()) ? "l" : (is16BitMode()) ? "w" : " ";
+ // Append the suffix corresponding to the determined size.
+ if (Size == 64) Tmp += "q";
+ else if (Size == 32) Tmp += "l";
+ else if (Size == 16) Tmp += "w";
+ else Tmp += " ";
Op.setTokenValue(Tmp);
// Do match in ATT mode to allow explicit suffix usage.
Match.push_back(MatchInstruction(Operands, Inst, ErrorInfo,
diff --git a/llvm/test/MC/X86/x86-16-intel.s b/llvm/test/MC/X86/x86-16-intel.s
new file mode 100644
index 0000000000000..77ae4ae217218
--- /dev/null
+++ b/llvm/test/MC/X86/x86-16-intel.s
@@ -0,0 +1,13 @@
+// RUN: llvm-mc -triple i386-unknown-unknown-code16 --x86-asm-syntax=intel --show-encoding %s | FileCheck %s
+
+// CHECK: pushl $8
+// CHECK: encoding: [0x66,0x6a,0x08]
+ data32 push 8
+
+// CHECK: pushw $8
+// CHECK: encoding: [0x6a,0x08]
+ push 8
+
+// CHECK: lretl
+// CHECK: encoding: [0x66,0xcb]
+ data32 retf
diff --git a/llvm/test/MC/X86/x86-16.s b/llvm/test/MC/X86/x86-16.s
index b0a4bda56fcbf..b4e116ab1a0fb 100644
--- a/llvm/test/MC/X86/x86-16.s
+++ b/llvm/test/MC/X86/x86-16.s
@@ -1060,3 +1060,15 @@ xresldtrk
// CHECK: encoding: [0x66,0x8b,0x1e,A,A]
// CHECK: fixup A - offset: 3, value: nearer, kind: FK_Data_2
movl nearer, %ebx
+
+// CHECK: pushl $8
+// CHECK: encoding: [0x66,0x6a,0x08]
+data32 push $8
+
+// CHECK: pushl $8
+// CHECK: encoding: [0x66,0x6a,0x08]
+pushl $8
+
+// CHECK: pushw $8
+// CHECK: encoding: [0x6a,0x08]
+push $8
diff --git a/llvm/test/MC/X86/x86-32-intel.s b/llvm/test/MC/X86/x86-32-intel.s
new file mode 100644
index 0000000000000..44fc5104653c2
--- /dev/null
+++ b/llvm/test/MC/X86/x86-32-intel.s
@@ -0,0 +1,28 @@
+// RUN: llvm-mc -triple i386-unknown-unknown --x86-asm-syntax=intel --show-encoding %s | FileCheck %s
+
+// CHECK: encoding: [0x66,0x6a,0x08]
+ data16 push 8
+
+// CHECK: encoding: [0x6a,0x08]
+ push 8
+
+// CxHECK: encoding: [0x66,0xcb]
+ // data16 retf
+
+// CHECK: encoding: [0xcb]
+ retf
+
+// CHECK: encoding: [0x66,0x9a,0xcd,0xab,0xce,0x7a]
+ data16 call 0x7ace, 0xabcd
+
+// CHECK: encoding: [0x9a,0xcd,0xab,0x00,0x00,0xce,0x7a]
+ call 0x7ace, 0xabcd
+
+// CHECK: encoding: [0xe8,A,A,A,A]
+ call a
+
+// CxHECK: encoding: [0x66,0xea,0xcd,0xab,0xce,0x7a]
+ // data16 ljmp 0x7ace, 0xabcd
+
+// CHECK: encoding: [0xea,0xcd,0xab,0x00,0x00,0xce,0x7a]
+ ljmp 0x7ace, 0xabcd
``````````
</details>
https://github.com/llvm/llvm-project/pull/156287
More information about the llvm-commits
mailing list