[RISCV][MC] Find matching pcrel_hi fixup in more cases.

[efriedma]

Summary

If a symbol points to the end of a fragment, instead of searching for fixups in that fragment, search in the next fragment.

Fixes spurious assembler error with subtarget change next to "la" pseudo-instruction, or expanded equivalent.

Alternate proposal to fix the problem discussed in https://reviews.llvm.org/D58759.

[lewis-revill]

On balance this patch seems a more appropriate solution given it doesn't touch generic code. I hope there isn't a case that isn't covered by this because I honestly can't think of one.

lib/Target/RISCV/MCTargetDesc/RISCVMCExpr.cpp

53

Perhaps I'm missing something, but I don't quite follow why this change (AUIPCSRE->findAssociatedFragment() -> AUIPCSymbol->getFragment()) was made?

59

Nitpick: could we swap this to Offset == DF->getContents().size()?

[efriedma]

lib/Target/RISCV/MCTargetDesc/RISCVMCExpr.cpp

53

It's a no-op. I just wanted to make the relationship between the fragment and the offset a bit more obvious.

59

Sure.

[apazos]

LGTM, I have verified the patch with some workloads and found no new issue.

[asb]

It took me a bit of time to go through the alternate patches and related discussion, but I've now reviewed and this seems like the most straight-forward solution and the right way to go. Many thanks Eli. LGTM.

[arunthomas]

Looking forward to seeing this patch merged. With this patch applied, I am nearly able to build FreeRTOS 10.2 with clang/RISC-V.

lib/Target/RISCV/MCTargetDesc/RISCVMCExpr.cpp

const MCFixup *RISCVMCExpr::getPCRelHiFixup() const {
   MCValue AUIPCLoc;
   if (!getSubExpr()->evaluateAsRelocatable(AUIPCLoc, nullptr, nullptr))
     return nullptr;
 
   const MCSymbolRefExpr *AUIPCSRE = AUIPCLoc.getSymA();
   if (!AUIPCSRE)
     return nullptr;
 
-  const auto *DF =
-      dyn_cast_or_null<MCDataFragment>(AUIPCSRE->findAssociatedFragment());
+  const MCSymbol *AUIPCSymbol = &AUIPCSRE->getSymbol();
+  const auto *DF = dyn_cast_or_null<MCDataFragment>(AUIPCSymbol->getFragment());
+
   if (!DF)
     return nullptr;
 
-  const MCSymbol *AUIPCSymbol = &AUIPCSRE->getSymbol();
+  uint64_t Offset = AUIPCSymbol->getOffset();
+  if (DF->getContents().size() == Offset) {
+    DF = dyn_cast_or_null<MCDataFragment>(DF->getNextNode());
+    if (!DF)
+      return nullptr;
+    Offset = 0;
+  }
+
   for (const MCFixup &F : DF->getFixups()) {
-    if (F.getOffset() != AUIPCSymbol->getOffset())
+    if (F.getOffset() != Offset)
       continue;
 
     switch ((unsigned)F.getKind()) {
     default:
       continue;
     case RISCV::fixup_riscv_got_hi20:
     case RISCV::fixup_riscv_pcrel_hi20:
       return &F;

test/MC/RISCV/option-mix.s

+# RUN: llvm-mc %s -triple=riscv32 | FileCheck -check-prefix=ASM %s
+# RUN: llvm-mc %s -triple=riscv64 | FileCheck -check-prefix=ASM %s
+# RUN: llvm-mc -filetype=obj -triple riscv32 < %s \
+# RUN:   | llvm-objdump -d -riscv-no-aliases - | FileCheck -check-prefix=DISASM %s
+# RUN: llvm-mc -filetype=obj -triple riscv64 < %s \
+# RUN:   | llvm-objdump -d -riscv-no-aliases - | FileCheck -check-prefix=DISASM %s
+
+# Checks change of options does not cause error: could not find corresponding %pcrel_hi 
+# when assembling pseudoinstruction and its extended form.
+
+.option push
+.option norelax
+  la a0, a_symbol
+.option pop
+  la a1, another_symbol
+
+# ASM: .Lpcrel_hi0:
+# ASM: auipc   a0, %pcrel_hi(a_symbol)
+# ASM: addi    a0, a0, %pcrel_lo(.Lpcrel_hi0)
+# ASM: .Lpcrel_hi1:
+# ASM: auipc   a1, %pcrel_hi(another_symbol)
+# ASM: addi    a1, a1, %pcrel_lo(.Lpcrel_hi1)
+
+# DISASM: .Lpcrel_hi0:
+# DISASM: auipc   a0, 0
+# DISASM: addi    a0, a0, 0
+# DISASM:.Lpcrel_hi1:
+# DISASM: auipc   a1, 0
+# DISASM: addi    a1, a1, 0
+
+.option push
+.option norelax
+1:auipc   a0, %pcrel_hi(a_symbol)
+  addi    a0, a0, %pcrel_lo(1b)
+.option pop
+2:auipc   a1, %pcrel_hi(another_symbol)
+  addi    a1, a1, %pcrel_lo(2b)
+
+# ASM: .Ltmp0:
+# ASM: auipc   a0, %pcrel_hi(a_symbol)
+# ASM: addi    a0, a0, %pcrel_lo(.Ltmp0)
+# ASM: .Ltmp1:
+# ASM: auipc   a1, %pcrel_hi(another_symbol)
+# ASM: addi    a1, a1, %pcrel_lo(.Ltmp1)
+
+# DISASM: .Ltmp0:
+# DISASM: auipc   a0, 0
+# DISASM: addi    a0, a0, 0
+# DISASM: .Ltmp1:
+# DISASM: auipc   a1, 0
+# DISASM: addi    a1, a1, 0