This implements a more optimal algorithm for selecting a base constant in

constant hoisting. It not only takes into account the number of uses and the
cost of expressions in which constants appear, but now also the resulting
integer range of the offsets. Thus, the algorithm maximizes the number of uses
within an integer range that will enable more efficient code generation. On
ARM, for example, this will enable code size optimisations because less
negative offsets will be created. Negative offsets/immediates are not supported
by Thumb1 thus preventing more compact instruction encoding.

Differential Revision: http://reviews.llvm.org/D21183

llvm-svn: 275382

1 files changed, 11 insertions, 0 deletions

diff --git a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
index a673619b922..13c5dc61acd 100644
--- a/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMTargetTransformInfo.cpp
@@ -47,6 +47,17 @@ int ARMTTIImpl::getIntImmCost(const APInt &Imm, Type *Ty) {
   return 3;
 }
 
+
+// Constants smaller than 256 fit in the immediate field of
+// Thumb1 instructions so we return a zero cost and 1 otherwise.
+int ARMTTIImpl::getIntImmCodeSizeCost(unsigned Opcode, unsigned Idx,
+                                      const APInt &Imm, Type *Ty) {
+  if (Imm.isNonNegative() && Imm.getLimitedValue() < 256)
+    return 0;
+
+  return 1;
+}
+
 int ARMTTIImpl::getIntImmCost(unsigned Opcode, unsigned Idx, const APInt &Imm,
                               Type *Ty) {
   // Division by a constant can be turned into multiplication, but only if we