[llvm-branch-commits] [llvm] 8c2ad9e - [VE] Correct VMP allocation in calling conv

[Kazushi Marukawa via llvm-branch-commits]

Author: Kazushi (Jam) Marukawa
Date: 2020-12-21T22:42:24+09:00
New Revision: 8c2ad9e85f677546021880dc88e24f633ccacd93

URL: https://github.com/llvm/llvm-project/commit/8c2ad9e85f677546021880dc88e24f633ccacd93
DIFF: https://github.com/llvm/llvm-project/commit/8c2ad9e85f677546021880dc88e24f633ccacd93.diff

LOG: [VE] Correct VMP allocation in calling conv

VE used to allocate VM1, VM2, VMP2 (VM4+VM5), and VM3.  This patch
corrects to allocate VM1, VM2, VMP2 (VM4+VM5), and VM6.  Also add
a regression test.

Reviewed By: simoll

Differential Revision: https://reviews.llvm.org/D93570

Added: 
    

Modified: 
    llvm/lib/Target/VE/VECallingConv.td
    llvm/test/CodeGen/VE/Vector/fastcc_callee.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/VE/VECallingConv.td b/llvm/lib/Target/VE/VECallingConv.td
index 6741d1cc8eaf..93899c2cae3d 100644
--- a/llvm/lib/Target/VE/VECallingConv.td
+++ b/llvm/lib/Target/VE/VECallingConv.td
@@ -116,7 +116,7 @@ def CC_VE_Fast : CallingConv<[
   // pair of vector mask --> generic vector mask registers
   CCIfType<[v512i1],
            CCAssignToRegWithShadow<[VMP1, VMP2, VMP3],
-                                   [VM1, VM1, VM3]>>,
+                                   [VM1, VM3, VM5]>>,
 
   // Follow the standard C CC for scalars.
   CCDelegateTo<CC_VE_C>
@@ -137,7 +137,7 @@ def RetCC_VE_Fast : CallingConv<[
   // pair of vector mask --> generic vector mask registers
   CCIfType<[v512i1],
            CCAssignToRegWithShadow<[VMP1, VMP2, VMP3],
-                                   [VM1, VM1, VM3]>>,
+                                   [VM1, VM3, VM5]>>,
 
   // Follow the standard C CC for scalars.
   CCDelegateTo<RetCC_VE_C>

diff  --git a/llvm/test/CodeGen/VE/Vector/fastcc_callee.ll b/llvm/test/CodeGen/VE/Vector/fastcc_callee.ll
index c0ad247d0e74..aa7b4944e7e4 100644
--- a/llvm/test/CodeGen/VE/Vector/fastcc_callee.ll
+++ b/llvm/test/CodeGen/VE/Vector/fastcc_callee.ll
@@ -137,3 +137,11 @@ define fastcc <512 x i1> @vreg_arg_v512i1_vmp3(<512 x i1> %vmp1, <512 x i1> %vmp
 ; CHECK-NEXT:    b.l.t (, %s10)
   ret <512 x i1> %vmp3
 }
+
+define fastcc <256 x i1> @vmp_cc_bug(<256 x i1> %vm1, <256 x i1> %vm2, <512 x i1> %vmp2, <256 x i1> %vm6) {
+; CHECK-LABEL: vmp_cc_bug:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    andm %vm1, %vm0, %vm6
+; CHECK-NEXT:    b.l.t (, %s10)
+  ret <256 x i1> %vm6
+}