[NVPTX] enable NaryReassociate in NVPTX

Summary:
We run NaryReassociate right after SLSR because SLSR enables many
opportunities for NaryReassociate. For example, in nary-slsr.ll

  foo((a + b) + c);
  foo((a + b * 2) + c);
  foo((a + b * 3) + c);   // 2 muls and 6 adds

after SLSR:

  ab = a + b;
  foo(ab + c);
  ab2 = ab + b;
  foo(ab2 + c);
  ab3 = ab2 + b;
  foo(ab3 + c);           // 6 adds

after NaryReassociate:

  abc = (a + b) + c;
  foo(abc);
  ab2c = abc + b;
  foo(ab2c);
  ab3c = ab2c + b;
  foo(ab3c);              // 4 adds

Test Plan: nary-slsr.ll

Reviewers: jholewinski, eliben

Reviewed By: eliben

Subscribers: jholewinski, llvm-commits

Differential Revision: http://reviews.llvm.org/D9066

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@235688 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/NVPTX/NVPTXTargetMachine.cpp b/lib/Target/NVPTX/NVPTXTargetMachine.cpp
index dc8e8a2952826db9bc39849a6faaa0e6285d547f..ac27c30aabab0dc66f81cbcf177755a123d328a6 100644 (file)
--- a/lib/Target/NVPTX/NVPTXTargetMachine.cpp
+++ b/lib/Target/NVPTX/NVPTXTargetMachine.cpp
@@ -179,6 +179,8 @@ void NVPTXPassConfig::addIRPasses() {
     addPass(createGVNPass());
   else
     addPass(createEarlyCSEPass());
+  // Run NaryReassociate after EarlyCSE/GVN to be more effective.
+  addPass(createNaryReassociatePass());
 }
 
 bool NVPTXPassConfig::addInstSelector() {

diff --git a/test/Transforms/NaryReassociate/NVPTX/lit.local.cfg b/test/Transforms/NaryReassociate/NVPTX/lit.local.cfg
new file mode 100644 (file)
index 0000000..2cb98eb
--- /dev/null
+++ b/test/Transforms/NaryReassociate/NVPTX/lit.local.cfg
@@ -0,0 +1,2 @@
+if not 'NVPTX' in config.root.targets:
+    config.unsupported = True

diff --git a/test/Transforms/NaryReassociate/NVPTX/nary-slsr.ll b/test/Transforms/NaryReassociate/NVPTX/nary-slsr.ll
new file mode 100644 (file)
index 0000000..2360fe7
--- /dev/null
+++ b/test/Transforms/NaryReassociate/NVPTX/nary-slsr.ll
@@ -0,0 +1,47 @@
+; RUN: opt < %s -slsr -nary-reassociate -S | FileCheck %s
+; RUN: llc < %s -march=nvptx64 -mcpu=sm_20 | FileCheck %s --check-prefix=PTX
+
+target datalayout = "e-i64:64-v16:16-v32:32-n16:32:64"
+
+; foo((a + b) + c);
+; foo((a + b * 2) + c);
+; foo((a + b * 3) + c);
+;   =>
+; abc = (a + b) + c;
+; foo(abc);
+; ab2c = abc + b;
+; foo(ab2c);
+; ab3c = ab2c + b;
+; foo(ab3c);
+define void @nary_reassociate_after_slsr(i32 %a, i32 %b, i32 %c) {
+; CHECK-LABEL: @nary_reassociate_after_slsr(
+; PTX-LABEL: .visible .func nary_reassociate_after_slsr(
+; PTX: ld.param.u32 [[b:%r[0-9]+]], [nary_reassociate_after_slsr_param_1];
+  %ab = add i32 %a, %b
+  %abc = add i32 %ab, %c
+  call void @foo(i32 %abc)
+; CHECK: call void @foo(i32 %abc)
+; PTX: st.param.b32 [param0+0], [[abc:%r[0-9]+]];
+
+  %b2 = shl i32 %b, 1
+  %ab2 = add i32 %a, %b2
+  %ab2c = add i32 %ab2, %c
+; CHECK-NEXT: %ab2c = add i32 %abc, %b
+; PTX: add.s32 [[ab2c:%r[0-9]+]], [[abc]], [[b]]
+  call void @foo(i32 %ab2c)
+; CHECK-NEXT: call void @foo(i32 %ab2c)
+; PTX: st.param.b32 [param0+0], [[ab2c]];
+
+  %b3 = mul i32 %b, 3
+  %ab3 = add i32 %a, %b3
+  %ab3c = add i32 %ab3, %c
+; CHECK-NEXT: %ab3c = add i32 %ab2c, %b
+; PTX: add.s32 [[ab3c:%r[0-9]+]], [[ab2c]], [[b]]
+  call void @foo(i32 %ab3c)
+; CHECK-NEXT: call void @foo(i32 %ab3c)
+; PTX: st.param.b32 [param0+0], [[ab3c]];
+
+  ret void
+}
+
+declare void @foo(i32)