Tweak ARM fastcc by adopting these two AAPCS rules:

* CPRCs may be allocated to co-processor registers or the stack – they may never be allocated to core registers
* When a CPRC is allocated to the stack, all other VFP registers should be marked as unavailable

The difference is only noticeable in rare cases where there are a large number of floating point arguments (e.g.
7 doubles + additional float, double arguments). Although it's probably still better to avoid vmov as it can cause
stalls in some older ARM cores. The other, more subtle benefit, is to minimize difference between the various
calling conventions.

rdar://16039676


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@201193 91177308-0d34-0410-b5e6-96231b3b80d8

2 files changed, 47 insertions, 0 deletions

diff --git a/lib/Target/ARM/ARMCallingConv.td b/lib/Target/ARM/ARMCallingConv.td
index 9ca2e465d8..b1c1f3eab1 100644
--- a/lib/Target/ARM/ARMCallingConv.td
+++ b/lib/Target/ARM/ARMCallingConv.td
@@ -64,6 +64,13 @@ def FastCC_ARM_APCS : CallingConv<[
   CCIfType<[f64], CCAssignToReg<[D0, D1, D2, D3, D4, D5, D6, D7]>>,
   CCIfType<[f32], CCAssignToReg<[S0, S1, S2, S3, S4, S5, S6, S7, S8,
                                  S9, S10, S11, S12, S13, S14, S15]>>,
+
+  // CPRCs may be allocated to co-processor registers or the stack – they
+  // may never be allocated to core registers. 
+  CCIfType<[f32], CCAssignToStackWithShadow<4, 4, [Q0, Q1, Q2, Q3]>>,
+  CCIfType<[f64], CCAssignToStackWithShadow<8, 4, [Q0, Q1, Q2, Q3]>>,
+  CCIfType<[v2f64], CCAssignToStackWithShadow<16, 4, [Q0, Q1, Q2, Q3]>>,
+
   CCDelegateTo<CC_ARM_APCS>
 ]>;
 
diff --git a/test/CodeGen/ARM/fastcc-vfp.ll b/test/CodeGen/ARM/fastcc-vfp.ll
new file mode 100644
index 0000000000..4c98150c70
--- /dev/null
+++ b/test/CodeGen/ARM/fastcc-vfp.ll
@@ -0,0 +1,40 @@
+; RUN: llc < %s -mtriple=armv7-apple-ios -mattr=+vfp2 | FileCheck %s
+
+define fastcc double @t1(double %d0, double %d1, double %d2, double %d3, double %d4, double %d5, double %d6, double %d7, float %a, float %b) {
+entry:
+; CHECK-LABEL: t1:
+; CHECK-NOT: vmov
+; CHECK: vldr
+  %add = fadd float %a, %b
+  %conv = fpext float %add to double
+  ret double %conv
+}
+
+define fastcc double @t2(double %d0, double %d1, double %d2, double %d3, double %d4, double %d5, double %a, float %b, double %c) {
+entry:
+; CHECK-LABEL: t2:
+; CHECK-NOT: vmov
+; CHECK: vldr
+  %add = fadd double %a, %c
+  ret double %add
+}
+
+define fastcc float @t3(double %d0, double %d1, double %d2, double %d3, double %d4, double %d5, double %d6, float %a, double %b, float %c) {
+entry:
+; CHECK-LABEL: t3:
+; CHECK: vldr
+  %add = fadd float %a, %c
+  ret float %add
+}
+
+define fastcc double @t4(double %a, double %b) #0 {
+entry:
+; CHECK-LABEL: t4:
+; CHECK: vstr
+  %add = fadd double %a, %b
+  %sub = fsub double %a, %b
+  %call = tail call fastcc double @x(double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double %add, float 0.000000e+00, double %sub) #2
+  ret double %call
+}
+
+declare fastcc double @x(double, double, double, double, double, double, double, float, double)