diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -2228,6 +2228,35 @@
                          return isa<Instruction>(U) &&
                                 cast<Instruction>(U)->getParent() == BB;
                        }) > 2;
+
+      // For Thumb1, we impose additional constraints
+      // due to low number of registers:
+      // 1. Number of args passed + caller's num of args <
+      //    Total number of available regs
+      // 2. Each arg to callee, is either a "pass thru" arg, OR 8-bit imm
+      //    OR a constant load. The intent is to allow only those args,
+      //    that need a single register for computation.
+
+      if (PreferIndirect && Subtarget->isThumb1Only()) {
+        const Instruction *I = cast<Instruction>(*GV->users().begin());
+        Function &F = MF.getFunction();
+        PreferIndirect = false;
+        // FIXME: What API to use to get number of available regs
+        // instead of hardcoding 7 ?
+        if (F.arg_size() + I->getNumOperands() < 7) {
+          unsigned i;
+          for (i = 0; i < I->getNumOperands() - 1; i++) {
+            Value *O = I->getOperand(i);
+            if (!(isa<GlobalValue>(O) ||
+                  (i < F.arg_size() && O == F.getArg(i)) ||
+                  (isa<ConstantInt>(O) &&
+                   cast<ConstantInt>(O)->getZExtValue() < 256)))
+              break;
+          }
+          if (i == I->getNumOperands() - 1)
+            PreferIndirect = true;
+        }
+      }
     }
   }
   if (isTailCall) {
diff --git a/llvm/test/CodeGen/ARM/minsize-call-cse-2.ll b/llvm/test/CodeGen/ARM/minsize-call-cse-2.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/CodeGen/ARM/minsize-call-cse-2.ll
@@ -0,0 +1,20 @@
+; RUN: llc < %s | FileCheck %s
+
+target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
+target triple = "thumbv6m-arm-none-eabi"
+
+; CHECK-LABEL: f:
+; CHECK: bl g
+; CHECK: bl g
+; CHECK: bl g
+; CHECK: bl g
+define void @f(i32* %p, i32 %x, i32 %y, i32 %z) minsize optsize {
+entry:
+  call void @g(i32* %p, i32 %x, i32 %y, i32 %z)
+  call void @g(i32* %p, i32 %x, i32 %y, i32 %z)
+  call void @g(i32* %p, i32 %x, i32 %y, i32 %z)
+  call void @g(i32* %p, i32 %x, i32 %y, i32 %z)
+  ret void
+}
+
+declare void @g(i32*,i32,i32,i32)