diff options
author | Andrew Trick <atrick@apple.com> | 2013-01-09 03:36:49 +0000 |
---|---|---|
committer | Andrew Trick <atrick@apple.com> | 2013-01-09 03:36:49 +0000 |
commit | 47579cf390c42e0577519e0a2b6044baece9df00 (patch) | |
tree | 2744c5de5c5c825a168a20f90b9e099d8feaaf88 /test/CodeGen/PowerPC/misched-inorder-latency.ll | |
parent | 2af949ddddfaf2feb4a446c754e09d2d8c207ce4 (diff) | |
download | external_llvm-47579cf390c42e0577519e0a2b6044baece9df00.zip external_llvm-47579cf390c42e0577519e0a2b6044baece9df00.tar.gz external_llvm-47579cf390c42e0577519e0a2b6044baece9df00.tar.bz2 |
MIsched: add an ILP window property to machine model.
This was an experimental option, but needs to be defined
per-target. e.g. PPC A2 needs to aggressively hide latency.
I converted some in-order scheduling tests to A2. Hal is working on
more test cases.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@171946 91177308-0d34-0410-b5e6-96231b3b80d8
Diffstat (limited to 'test/CodeGen/PowerPC/misched-inorder-latency.ll')
-rw-r--r-- | test/CodeGen/PowerPC/misched-inorder-latency.ll | 55 |
1 files changed, 55 insertions, 0 deletions
diff --git a/test/CodeGen/PowerPC/misched-inorder-latency.ll b/test/CodeGen/PowerPC/misched-inorder-latency.ll new file mode 100644 index 0000000..8fae7ad --- /dev/null +++ b/test/CodeGen/PowerPC/misched-inorder-latency.ll @@ -0,0 +1,55 @@ +; RUN: llc < %s -enable-misched -pre-RA-sched=source -scheditins=false \ +; RUN: -disable-ifcvt-triangle-false -disable-post-ra | FileCheck %s +; +target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f128:128:128-v128:128:128-n32:64" +target triple = "powerpc64-bgq-linux" + +; %val1 is a load live out of %entry. It should be hoisted +; above the add. +; CHECK: testload: +; CHECK: %entry +; CHECK: lwz +; CHECK: addi +; CHECK: bne +; CHECK: %true +define i32 @testload(i32 *%ptr, i32 %sumin) { +entry: + %sum1 = add i32 %sumin, 1 + %val1 = load i32* %ptr + %p = icmp eq i32 %sumin, 0 + br i1 %p, label %true, label %end +true: + %sum2 = add i32 %sum1, 1 + %ptr2 = getelementptr i32* %ptr, i32 1 + %val = load i32* %ptr2 + %val2 = add i32 %val1, %val + br label %end +end: + %valmerge = phi i32 [ %val1, %entry], [ %val2, %true ] + %summerge = phi i32 [ %sum1, %entry], [ %sum2, %true ] + %sumout = add i32 %valmerge, %summerge + ret i32 %sumout +} + +; The prefetch gets a default latency of 3 cycles and should be hoisted +; above the add. +; +; CHECK: testprefetch: +; CHECK: %entry +; CHECK: dcbt +; CHECK: addi +; CHECK: blr +define i32 @testprefetch(i8 *%ptr, i32 %i) { +entry: + %val1 = add i32 %i, 1 + tail call void @llvm.prefetch( i8* %ptr, i32 0, i32 3, i32 1 ) + %p = icmp eq i32 %i, 0 + br i1 %p, label %true, label %end +true: + %val2 = add i32 %val1, 1 + br label %end +end: + %valmerge = phi i32 [ %val1, %entry], [ %val2, %true ] + ret i32 %valmerge +} +declare void @llvm.prefetch(i8*, i32, i32, i32) nounwind |