[PATCH] D70947: Add a default address space for globals to DataLayout

[Alexander Richardson via Phabricator via llvm-commits]

arichardson created this revision.
arichardson added reviewers: dylanmckay, bjope, arsenm, asb, theraven, kparzysz, rengolin, hfinkel.
Herald added subscribers: llvm-commits, s.egerton, lenary, PkmX, atanasyan, simoncook, hiraditya, wdng, sdardis.
Herald added a project: LLVM.

This is similar to the existing alloca and program address spaces (D37052 <https://reviews.llvm.org/D37052>)
and should be used when creating/accessing global variables.
We need this in our CHERI fork of LLVM to place all globals in address space 200.
This ensures that values are accessed using CHERI load/store instructions
instead of the normal MIPS/RISC-V ones.

The problem this is trying to fix is that most of the time the type of
globals is created using a simple PointerType::getUnqual() (or ::get() with
the default address-space value of 0). This does not work for us and we get
assertion/compilation/instruction selection failures whenever a new call
is added that uses the default value of zero.

In our fork we have removed the default parameter value of zero for most
address space arguments and use DL.getProgramAddressSpace() or
DL.getGlobalsAddressSpace() whenever possible. If this change is accepted,
I will upstream follow-up patches to use DL.getGlobalsAddressSpace() instead
of relying on the default value of 0 for PointerType::get(), etc.

This patch and the follow-up changes will not have any functional changes
for existing backends with the default globals address space of zero.


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D70947

Files:
  llvm/docs/LangRef.rst
  llvm/include/llvm/IR/DataLayout.h
  llvm/lib/IR/DataLayout.cpp
  llvm/test/Assembler/invalid-datalayout-globals-addrspace.ll


Index: llvm/test/Assembler/invalid-datalayout-globals-addrspace.ll
===================================================================
--- /dev/null
+++ llvm/test/Assembler/invalid-datalayout-globals-addrspace.ll
@@ -0,0 +1,4 @@
+; RUN: not llvm-as < %s 2>&1 | FileCheck %s
+
+; CHECK: Invalid address space, must be a 24-bit integer
+target datalayout = "G16777216"
Index: llvm/lib/IR/DataLayout.cpp
===================================================================
--- llvm/lib/IR/DataLayout.cpp
+++ llvm/lib/IR/DataLayout.cpp
@@ -179,6 +179,7 @@
   AllocaAddrSpace = 0;
   StackNaturalAlign.reset();
   ProgramAddrSpace = 0;
+  GlobalsAddrSpace = 0;
   FunctionPtrAlign.reset();
   TheFunctionPtrAlignType = FunctionPtrAlignType::Independent;
   ManglingMode = MM_None;
@@ -419,6 +420,10 @@
       AllocaAddrSpace = getAddrSpace(Tok);
       break;
     }
+    case 'G': { // Default address space for global variables.
+      GlobalsAddrSpace = getAddrSpace(Tok);
+      break;
+    }
     case 'm':
       if (!Tok.empty())
         report_fatal_error("Unexpected trailing characters after mangling specifier in datalayout string");
@@ -464,6 +469,7 @@
              AllocaAddrSpace == Other.AllocaAddrSpace &&
              StackNaturalAlign == Other.StackNaturalAlign &&
              ProgramAddrSpace == Other.ProgramAddrSpace &&
+             GlobalsAddrSpace == Other.GlobalsAddrSpace &&
              FunctionPtrAlign == Other.FunctionPtrAlign &&
              TheFunctionPtrAlignType == Other.TheFunctionPtrAlignType &&
              ManglingMode == Other.ManglingMode &&
Index: llvm/include/llvm/IR/DataLayout.h
===================================================================
--- llvm/include/llvm/IR/DataLayout.h
+++ llvm/include/llvm/IR/DataLayout.h
@@ -123,6 +123,7 @@
   unsigned AllocaAddrSpace;
   MaybeAlign StackNaturalAlign;
   unsigned ProgramAddrSpace;
+  unsigned GlobalsAddrSpace;
 
   MaybeAlign FunctionPtrAlign;
   FunctionPtrAlignType TheFunctionPtrAlignType;
@@ -212,6 +213,7 @@
     FunctionPtrAlign = DL.FunctionPtrAlign;
     TheFunctionPtrAlignType = DL.TheFunctionPtrAlignType;
     ProgramAddrSpace = DL.ProgramAddrSpace;
+    GlobalsAddrSpace = DL.GlobalsAddrSpace;
     ManglingMode = DL.ManglingMode;
     LegalIntWidths = DL.LegalIntWidths;
     Alignments = DL.Alignments;
@@ -284,6 +286,7 @@
   }
 
   unsigned getProgramAddressSpace() const { return ProgramAddrSpace; }
+  unsigned getGlobalsAddressSpace() const { return GlobalsAddrSpace; }
 
   bool hasMicrosoftFastStdCallMangling() const {
     return ManglingMode == MM_WinCOFFX86;
Index: llvm/docs/LangRef.rst
===================================================================
--- llvm/docs/LangRef.rst
+++ llvm/docs/LangRef.rst
@@ -2055,6 +2055,10 @@
     program memory space defaults to the default address space of 0,
     which corresponds to a Von Neumann architecture that has code
     and data in the same space.
+``G<address space>``
+    Specifies the address space that should be used for global variables.
+    If omitted, the globals address space defaults to the default address
+    space of 0.
 ``A<address space>``
     Specifies the address space of objects created by '``alloca``'.
     Defaults to the default address space of 0.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D70947.231848.patch
Type: text/x-patch
Size: 3258 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20191203/511d2f43/attachment.bin>