[PATCH] [CodeGen] Emit Constants, not ScalarExprs, for "immediate"-constrained inlineasm arguments.

[Ahmed Bougacha]

In http://reviews.llvm.org/D10255#184520, @ahatanak wrote:

> Do we have to handle an expression like (long)&object - (long)&base when it's used with constraint "i"? I believe clang currently cannot handle this expression.


Yes, I don't think 'i' supports that.

In http://reviews.llvm.org/D10255#184893, @dsanders wrote:

> Just to mention it, there's something odd about the 'i' constraint in the backends but I haven't had chance to investigate it yet. For some reason, it has to be handled in *DAGToDAGISel::SelectInlineAsmMemoryOperand() despite not being a memory operand.


I'm not too familiar with inline-asm handling, but could it be because it supports symbols, and symbols aren't really immediates (e.g. aren't usually supported in non-addressing mode contexts) ?


================
Comment at: lib/CodeGen/CGStmt.cpp:1759-1760
@@ +1758,4 @@
+      return llvm::ConstantInt::get(getLLVMContext(), Result);
+    assert(!Info.requiresImmediateConstant() &&
+           "Required-immediate inlineasm arg isn't constant?");
+  }
----------------
dsanders wrote:
> Should this cause an error to be emitted on release builds too?
I'm not sure, my understanding of clang is, by the time we get to IR gen, source errors have already been caught and reported.

================
Comment at: test/CodeGen/inline-asm-immediate-ubsan.c:2
@@ +1,3 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -emit-llvm -o - %s \
+// RUN:     -fsanitize=signed-integer-overflow \
+// RUN:   | FileCheck %s --check-prefix=CHECK
----------------
dsanders wrote:
> Sanitizers aren't always available. It's possible we need a 'REQUIRES:' line to disable the test for some buildbots.
That makes sense for runtime tests, but is it necessary for codegen tests?  I thought all of IR gen was always available, much like targets.

http://reviews.llvm.org/D10255

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/