summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorJustin Lebar <jlebar@google.com>2016-11-21 22:49:15 +0000
committerJustin Lebar <jlebar@google.com>2016-11-21 22:49:15 +0000
commit3e50a5be8f5259bc256f39830e72525dd9f90626 (patch)
treeb99385ff21dbca7fb6474649ed54e31953b3cb28
parent838c7f5a85a96f642fea0e2167c92e32682222b6 (diff)
downloadbcm5719-llvm-3e50a5be8f5259bc256f39830e72525dd9f90626.tar.gz
bcm5719-llvm-3e50a5be8f5259bc256f39830e72525dd9f90626.zip
[CodeGenPrepare] Don't sink non-cheap addrspacecasts.
Summary: Previously, CGP would unconditionally sink addrspacecast instructions, even going so far as to sink them into a loop. Now we check that the cast is "cheap", as defined by TLI. We introduce a new "is-cheap" function to TLI rather than using isNopAddrSpaceCast because some GPU platforms want the ability to ask for non-nop casts to be sunk. Reviewers: arsenm, tra Subscribers: jholewinski, wdng, llvm-commits Differential Revision: https://reviews.llvm.org/D26923 llvm-svn: 287591
-rw-r--r--llvm/include/llvm/Target/TargetLowering.h6
-rw-r--r--llvm/lib/CodeGen/CodeGenPrepare.cpp8
-rw-r--r--llvm/test/Transforms/CodeGenPrepare/NVPTX/dont-sink-nop-addrspacecast.ll21
3 files changed, 35 insertions, 0 deletions
diff --git a/llvm/include/llvm/Target/TargetLowering.h b/llvm/include/llvm/Target/TargetLowering.h
index 8beabc7b03e..3c7dc885ba8 100644
--- a/llvm/include/llvm/Target/TargetLowering.h
+++ b/llvm/include/llvm/Target/TargetLowering.h
@@ -1153,6 +1153,12 @@ public:
return false;
}
+ /// Returns true if a cast from SrcAS to DestAS is "cheap", such that e.g. we
+ /// are happy to sink it into basic blocks.
+ virtual bool isCheapAddrSpaceCast(unsigned SrcAS, unsigned DestAS) const {
+ return isNoopAddrSpaceCast(SrcAS, DestAS);
+ }
+
/// Return true if the pointer arguments to CI should be aligned by aligning
/// the object whose address is being passed. If so then MinSize is set to the
/// minimum size the object must be to be aligned and PrefAlign is set to the
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index 87ad3075873..68f090ccee2 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -927,6 +927,14 @@ static bool SinkCast(CastInst *CI) {
///
static bool OptimizeNoopCopyExpression(CastInst *CI, const TargetLowering &TLI,
const DataLayout &DL) {
+ // Sink only "cheap" (or nop) address-space casts. This is a weaker condition
+ // than sinking only nop casts, but is helpful on some platforms.
+ if (auto *ASC = dyn_cast<AddrSpaceCastInst>(CI)) {
+ if (!TLI.isCheapAddrSpaceCast(ASC->getSrcAddressSpace(),
+ ASC->getDestAddressSpace()))
+ return false;
+ }
+
// If this is a noop copy,
EVT SrcVT = TLI.getValueType(DL, CI->getOperand(0)->getType());
EVT DstVT = TLI.getValueType(DL, CI->getType());
diff --git a/llvm/test/Transforms/CodeGenPrepare/NVPTX/dont-sink-nop-addrspacecast.ll b/llvm/test/Transforms/CodeGenPrepare/NVPTX/dont-sink-nop-addrspacecast.ll
new file mode 100644
index 00000000000..97b24906d16
--- /dev/null
+++ b/llvm/test/Transforms/CodeGenPrepare/NVPTX/dont-sink-nop-addrspacecast.ll
@@ -0,0 +1,21 @@
+; RUN: opt -S -codegenprepare < %s | FileCheck %s
+
+target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
+target triple = "nvptx64-nvidia-cuda"
+
+; CHECK-LABEL: @test
+define i64 @test(i1 %pred, i64* %ptr) {
+; CHECK: addrspacecast
+ %ptr_as1 = addrspacecast i64* %ptr to i64 addrspace(1)*
+ br i1 %pred, label %l1, label %l2
+l1:
+; CHECK-LABEL: l1:
+; CHECK-NOT: addrspacecast
+ %v1 = load i64, i64* %ptr
+ ret i64 %v1
+l2:
+ ; CHECK-LABEL: l2:
+ ; CHECK-NOT: addrspacecast
+ %v2 = load i64, i64 addrspace(1)* %ptr_as1
+ ret i64 %v2
+}
OpenPOWER on IntegriCloud