diff options
| author | Mark Mendell <mark.p.mendell@intel.com> | 2014-10-21 17:44:32 -0400 |
|---|---|---|
| committer | Mark Mendell <mark.p.mendell@intel.com> | 2014-10-22 09:20:32 -0400 |
| commit | a65c1dbb8d3511da6c0804f8063c453f744629c2 (patch) | |
| tree | bf8d22524d37c9150dcb66049d029d6ec8223310 /compiler | |
| parent | 6207b2b0c299798fd5213bb3d6c05e5b978747e8 (diff) | |
| download | art-a65c1dbb8d3511da6c0804f8063c453f744629c2.zip art-a65c1dbb8d3511da6c0804f8063c453f744629c2.tar.gz art-a65c1dbb8d3511da6c0804f8063c453f744629c2.tar.bz2 | |
X86 Long Min/Max: Avoid calling SRegToVReg with -1
It is possible that the result of a call to min/max can't be combined
with the following move. In that case, the destination will use the
default long return value (EAX/EDX), with a s_reg_low value of -1.
A debug compiler will assert fail in that case.
Fix: A result with no s_reg_low must be unused. Just return with no
code generated, like X86 GenInlinedAbsFloat().
Seen compiling GmsCore.apk on the host with a debug backend.
Change-Id: I8006e822e8dcb2112d86e4047bb2e3037ba6fece
Signed-off-by: Mark Mendell <mark.p.mendell@intel.com>
Diffstat (limited to 'compiler')
| -rwxr-xr-x | compiler/dex/quick/x86/int_x86.cc | 5 |
1 files changed, 5 insertions, 0 deletions
diff --git a/compiler/dex/quick/x86/int_x86.cc b/compiler/dex/quick/x86/int_x86.cc index 07034cb..acf5599 100755 --- a/compiler/dex/quick/x86/int_x86.cc +++ b/compiler/dex/quick/x86/int_x86.cc @@ -858,6 +858,11 @@ bool X86Mir2Lir::GenInlinedMinMax(CallInfo* info, bool is_min, bool is_long) { RegLocation rl_dest = InlineTargetWide(info); int res_vreg, src1_vreg, src2_vreg; + if (rl_dest.s_reg_low == INVALID_SREG) { + // Result is unused, the code is dead. Inlining successful, no code generated. + return true; + } + /* * If the result register is the same as the second element, then we * need to be careful. The reason is that the first copy will |