diff options
author | Evan Cheng <evan.cheng@apple.com> | 2010-11-03 00:45:17 +0000 |
---|---|---|
committer | Evan Cheng <evan.cheng@apple.com> | 2010-11-03 00:45:17 +0000 |
commit | 8239daf7c83a65a189c352cce3191cdc3bbfe151 (patch) | |
tree | 84b82c2cf503208d1f67007255f2f56fdb383c55 /test/CodeGen/ARM/ifcvt11.ll | |
parent | 41957f6eb2271e5f1981b32a873d1b58217c6411 (diff) | |
download | external_llvm-8239daf7c83a65a189c352cce3191cdc3bbfe151.zip external_llvm-8239daf7c83a65a189c352cce3191cdc3bbfe151.tar.gz external_llvm-8239daf7c83a65a189c352cce3191cdc3bbfe151.tar.bz2 |
Two sets of changes. Sorry they are intermingled.
1. Fix pre-ra scheduler so it doesn't try to push instructions above calls to
"optimize for latency". Call instructions don't have the right latency and
this is more likely to use introduce spills.
2. Fix if-converter cost function. For ARM, it should use instruction latencies,
not # of micro-ops since multi-latency instructions is completely executed
even when the predicate is false. Also, some instruction will be "slower"
when they are predicated due to the register def becoming implicit input.
rdar://8598427
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@118135 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/ARM/ifcvt11.ll')
-rw-r--r-- | test/CodeGen/ARM/ifcvt11.ll | 59 |
1 files changed, 59 insertions, 0 deletions
diff --git a/test/CodeGen/ARM/ifcvt11.ll b/test/CodeGen/ARM/ifcvt11.ll new file mode 100644 index 0000000..63f8557 --- /dev/null +++ b/test/CodeGen/ARM/ifcvt11.ll @@ -0,0 +1,59 @@ +; RUN: llc < %s -mtriple=arm-apple-darwin -mcpu=cortex-a8 | FileCheck %s +; rdar://8598427 +; Adjust if-converter heuristics to avoid predicating vmrs which can cause +; significant regression. + +%struct.xyz_t = type { double, double, double } + +define i32 @effie(i32 %tsets, %struct.xyz_t* nocapture %p, i32 %a, i32 %b, i32 %c) nounwind readonly noinline { +; CHECK: effie: +entry: + %0 = icmp sgt i32 %tsets, 0 + br i1 %0, label %bb.nph, label %bb6 + +bb.nph: ; preds = %entry + %1 = add nsw i32 %b, %a + %2 = add nsw i32 %1, %c + br label %bb + +bb: ; preds = %bb4, %bb.nph +; CHECK: vcmpe.f64 +; CHECK: vmrs apsr_nzcv, fpscr + %r.19 = phi i32 [ 0, %bb.nph ], [ %r.0, %bb4 ] + %n.08 = phi i32 [ 0, %bb.nph ], [ %10, %bb4 ] + %scevgep10 = getelementptr inbounds %struct.xyz_t* %p, i32 %n.08, i32 0 + %scevgep11 = getelementptr %struct.xyz_t* %p, i32 %n.08, i32 1 + %3 = load double* %scevgep10, align 4 + %4 = load double* %scevgep11, align 4 + %5 = fcmp uge double %3, %4 + br i1 %5, label %bb3, label %bb1 + +bb1: ; preds = %bb +; CHECK-NOT: it +; CHECK-NOT: vcmpemi +; CHECK-NOT: vmrsmi +; CHECK: vcmpe.f64 +; CHECK: vmrs apsr_nzcv, fpscr + %scevgep12 = getelementptr %struct.xyz_t* %p, i32 %n.08, i32 2 + %6 = load double* %scevgep12, align 4 + %7 = fcmp uge double %3, %6 + br i1 %7, label %bb3, label %bb2 + +bb2: ; preds = %bb1 + %8 = add nsw i32 %2, %r.19 + br label %bb4 + +bb3: ; preds = %bb1, %bb + %9 = add nsw i32 %r.19, 1 + br label %bb4 + +bb4: ; preds = %bb3, %bb2 + %r.0 = phi i32 [ %9, %bb3 ], [ %8, %bb2 ] + %10 = add nsw i32 %n.08, 1 + %exitcond = icmp eq i32 %10, %tsets + br i1 %exitcond, label %bb6, label %bb + +bb6: ; preds = %bb4, %entry + %r.1.lcssa = phi i32 [ 0, %entry ], [ %r.0, %bb4 ] + ret i32 %r.1.lcssa +} |