[lld] [LLD] Implement --enable-non-contiguous-regions (PR #90007)

[Daniel Thornburgh via llvm-commits]

================
@@ -1364,6 +1436,117 @@ const Defined *LinkerScript::assignAddresses() {
   return getChangedSymbolAssignment(oldValues);
 }
 
+static bool hasRegionOverflowed(MemoryRegion *mr) {
+  if (!mr)
+    return false;
+  return mr->curPos - mr->getOrigin() > mr->getLength();
+}
+
+// Spill input sections in reverse order of address assignment to (potentially)
+// bring memory regions out of overflow. The size savings of a spill can only be
+// estimated, since general linker script arithmetic may occur afterwards.
+// Under-estimates may cause unnecessary spills, but over-estimates can always
+// be corrected on the next pass.
+bool LinkerScript::spillSections() {
+  if (!config->enableNonContiguousRegions)
+    return false;
+
+  bool spilled = false;
+  for (SectionCommand *cmd : reverse(sectionCommands)) {
+    auto *od = dyn_cast<OutputDesc>(cmd);
+    if (!od)
+      continue;
+    OutputSection *osec = &od->osec;
+    if (!osec->size || !osec->memRegion)
+      continue;
+
+    // Input sections that have replaced a potential spill and should be removed
+    // from their input section description.
+    DenseSet<InputSection *> spilledInputSections;
+
+    for (SectionCommand *cmd : reverse(osec->commands)) {
+      if (!hasRegionOverflowed(osec->memRegion) &&
+          !hasRegionOverflowed(osec->lmaRegion))
+        break;
+
+      auto *isd = dyn_cast<InputSectionDescription>(cmd);
+      if (!isd)
+        continue;
+      for (InputSection *isec : reverse(isd->sections)) {
+        // Potential spill locations cannot be spilled.
+        if (isa<PotentialSpillSection>(isec))
+          continue;
+
+        // Find the next spill location.
+        auto it = potentialSpillLists.find(isec);
+        if (it == potentialSpillLists.end())
+          continue;
+
+        spilled = true;
+        PotentialSpillList &list = it->second;
+
----------------
mysterymath wrote:

If I've understood you correctly, it's desirable to have code in LLD as dense as is readable?

I've reordered the blocks based on topic: potential spill list manipulation and manipulating the input section. That should help make things denser, since the existing comments can be altered to cover more code.

https://github.com/llvm/llvm-project/pull/90007