on a double-word boundary instead of a single-word boundary.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@14177
91177308-0d34-0410-b5e6-
96231b3b80d8
// ----------
// 23 words * 4 bytes per word = 92 bytes
NumBytes += 92;
- NumBytes = (NumBytes + 3) & ~3; // Round up to next word boundary
+ NumBytes = (NumBytes + 7) & ~7; // Round up to next doubleword boundary
+ // (Technically, a word boundary should be sufficient, but SPARC as complains)
BuildMI(MBB, MBB.begin(), V8::SAVEri, 2,
V8::SP).addImm(-NumBytes).addReg(V8::SP);
}
// ----------
// 23 words * 4 bytes per word = 92 bytes
NumBytes += 92;
- NumBytes = (NumBytes + 3) & ~3; // Round up to next word boundary
+ NumBytes = (NumBytes + 7) & ~7; // Round up to next doubleword boundary
+ // (Technically, a word boundary should be sufficient, but SPARC as complains)
BuildMI(MBB, MBB.begin(), V8::SAVEri, 2,
V8::SP).addImm(-NumBytes).addReg(V8::SP);
}