[VP][DAGCombine] Use simplifySelect when combining vp.select.
#87342
Conversation
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write If you have received no comments on your PR for a week, you can request a review If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
|
@llvm/pr-subscribers-llvm-selectiondag Author: None (AinsleySnow) ChangesHi all, This patch is a follow-up of #79101. It migrates logic from I'm a total newbie to llvm and I'm sure there's room for improvements in this patch. Please let me know if you have any advice. Thank you in advance! Full diff: https://github.com/llvm/llvm-project/pull/87342.diff 2 Files Affected:
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 2f46b23a97c62c..b889e4fe9af9a4 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -12056,6 +12056,13 @@ SDValue DAGCombiner::foldVSelectOfConstants(SDNode *N) {
}
SDValue DAGCombiner::visitVP_SELECT(SDNode *N) {
+ SDValue N0 = N->getOperand(0);
+ SDValue N1 = N->getOperand(1);
+ SDValue N2 = N->getOperand(2);
+
+ if (SDValue V = DAG.simplifySelect(N0, N1, N2))
+ return V;
+
if (SDValue V = foldBoolSelectToLogic<VPMatchContext>(N, DAG))
return V;
diff --git a/llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll b/llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll
index 0d52dd794fd56f..0d7f0cdeabc337 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vselect-vp.ll
@@ -825,3 +825,56 @@ define <vscale x 2 x i1> @select_cond_x_cond(<vscale x 2 x i1> %x, <vscale x 2 x
%a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, <vscale x 2 x i1> %x, i32 %evl)
ret <vscale x 2 x i1> %a
}
+
+define <vscale x 2 x i1> @select_undef_T_F(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, i32 zeroext %evl) {
+; CHECK-LABEL: select_undef_T_F:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vmv1r.v v0, v8
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> undef, <vscale x 2 x i1> %x, <vscale x 2 x i1> %y, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
+
+define <vscale x 2 x i1> @select_undef_undef_F(<vscale x 2 x i1> %x, i32 zeroext %evl) {
+; CHECK-LABEL: select_undef_undef_F:
+; CHECK: # %bb.0:
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> undef, <vscale x 2 x i1> undef, <vscale x 2 x i1> %x, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
+
+define <vscale x 2 x i1> @select_unknown_undef_F(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, i32 zeroext %evl) {
+; CHECK-LABEL: select_unknown_undef_F:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vmv1r.v v0, v8
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> %x, <vscale x 2 x i1> undef, <vscale x 2 x i1> %y, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
+
+define <vscale x 2 x i1> @select_unknown_T_undef(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, i32 zeroext %evl) {
+; CHECK-LABEL: select_unknown_T_undef:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vmv1r.v v0, v8
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, <vscale x 2 x i1> undef, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
+
+define <vscale x 2 x i1> @select_false_T_F(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, <vscale x 2 x i1> %z, i32 zeroext %evl) {
+; CHECK-LABEL: select_false_T_F:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vmv1r.v v0, v9
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> zeroinitializer, <vscale x 2 x i1> %y, <vscale x 2 x i1> %z, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
+
+define <vscale x 2 x i1> @select_unknown_T_T(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, i32 zeroext %evl) {
+; CHECK-LABEL: select_unknown_T_T:
+; CHECK: # %bb.0:
+; CHECK-NEXT: vmv1r.v v0, v8
+; CHECK-NEXT: ret
+ %a = call <vscale x 2 x i1> @llvm.vp.select.nxv2i1(<vscale x 2 x i1> %x, <vscale x 2 x i1> %y, <vscale x 2 x i1> %y, i32 %evl)
+ ret <vscale x 2 x i1> %a
+}
|
|
Are you sure these tests are hitting the DAG combine, and not getting cleaned up in InstSimplify |
|
I addressed the comments in the newly-pushed commits. I added pre-commit test cases to demonstrate that the tests will hit the DAG combine. Before we use Thank you for your suggestions! |
|
Thank you. We can add logic to InstSimplify in later patches. By the way could you please help me commit these changes? I have no write access right now. |
|
@AinsleySnow Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested Please check whether problems have been caused by your change specifically, as How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
Hi all,
This patch is a follow-up of #79101. It migrates logic from
visitVSELECTtovisitVP_SELECTto simplifyvp.select. With this patch we can do the following combinations:I'm a total newbie to llvm and I'm sure there's room for improvements in this patch. Please let me know if you have any advice. Thank you in advance!