[llvm] r353872 - [AArch64] Expand v8i8 cttz (PR39729)
Nikita Popov via llvm-commits
llvm-commits at lists.llvm.org
Tue Feb 12 10:55:54 PST 2019
Author: nikic
Date: Tue Feb 12 10:55:53 2019
New Revision: 353872
URL: http://llvm.org/viewvc/llvm-project?rev=353872&view=rev
Log:
[AArch64] Expand v8i8 cttz (PR39729)
Fix for https://bugs.llvm.org/show_bug.cgi?id=39729.
Rather than adding just a case for v8i8 I'm setting cttz to expand
for all vector types.
Differential Revision: https://reviews.llvm.org/D58008
Added:
llvm/trunk/test/CodeGen/AArch64/vec_cttz.ll
Modified:
llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp
Modified: llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp?rev=353872&r1=353871&r2=353872&view=diff
==============================================================================
--- llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp (original)
+++ llvm/trunk/lib/Target/AArch64/AArch64ISelLowering.cpp Tue Feb 12 10:55:53 2019
@@ -681,15 +681,6 @@ AArch64TargetLowering::AArch64TargetLowe
setOperationAction(ISD::CTLZ, MVT::v1i64, Expand);
setOperationAction(ISD::CTLZ, MVT::v2i64, Expand);
- setOperationAction(ISD::CTTZ, MVT::v2i8, Expand);
- setOperationAction(ISD::CTTZ, MVT::v4i16, Expand);
- setOperationAction(ISD::CTTZ, MVT::v2i32, Expand);
- setOperationAction(ISD::CTTZ, MVT::v1i64, Expand);
- setOperationAction(ISD::CTTZ, MVT::v16i8, Expand);
- setOperationAction(ISD::CTTZ, MVT::v8i16, Expand);
- setOperationAction(ISD::CTTZ, MVT::v4i32, Expand);
- setOperationAction(ISD::CTTZ, MVT::v2i64, Expand);
-
// AArch64 doesn't have MUL.2d:
setOperationAction(ISD::MUL, MVT::v2i64, Expand);
// Custom handling for some quad-vector types to detect MULL.
@@ -728,6 +719,7 @@ AArch64TargetLowering::AArch64TargetLowe
setOperationAction(ISD::UMUL_LOHI, VT, Expand);
setOperationAction(ISD::BSWAP, VT, Expand);
+ setOperationAction(ISD::CTTZ, VT, Expand);
for (MVT InnerVT : MVT::vector_valuetypes()) {
setTruncStoreAction(VT, InnerVT, Expand);
Added: llvm/trunk/test/CodeGen/AArch64/vec_cttz.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/AArch64/vec_cttz.ll?rev=353872&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/AArch64/vec_cttz.ll (added)
+++ llvm/trunk/test/CodeGen/AArch64/vec_cttz.ll Tue Feb 12 10:55:53 2019
@@ -0,0 +1,124 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=aarch64-none-linux-gnu -mattr=+neon | FileCheck %s --check-prefix=CHECK
+
+declare <8 x i8> @llvm.cttz.v8i8(<8 x i8>, i1)
+declare <4 x i16> @llvm.cttz.v4i16(<4 x i16>, i1)
+declare <2 x i32> @llvm.cttz.v2i32(<2 x i32>, i1)
+declare <1 x i64> @llvm.cttz.v1i64(<1 x i64>, i1)
+
+declare <16 x i8> @llvm.cttz.v16i8(<16 x i8>, i1)
+declare <8 x i16> @llvm.cttz.v8i16(<8 x i16>, i1)
+declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>, i1)
+declare <2 x i64> @llvm.cttz.v2i64(<2 x i64>, i1)
+
+define <8 x i8> @cttz_v8i8(<8 x i8> %a) nounwind {
+; CHECK-LABEL: cttz_v8i8:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.8b, #1
+; CHECK-NEXT: sub v1.8b, v0.8b, v1.8b
+; CHECK-NEXT: bic v0.8b, v1.8b, v0.8b
+; CHECK-NEXT: cnt v0.8b, v0.8b
+; CHECK-NEXT: ret
+ %b = call <8 x i8> @llvm.cttz.v8i8(<8 x i8> %a, i1 true)
+ ret <8 x i8> %b
+}
+
+define <4 x i16> @cttz_v4i16(<4 x i16> %a) nounwind {
+; CHECK-LABEL: cttz_v4i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.4h, #1
+; CHECK-NEXT: sub v1.4h, v0.4h, v1.4h
+; CHECK-NEXT: bic v0.8b, v1.8b, v0.8b
+; CHECK-NEXT: clz v0.4h, v0.4h
+; CHECK-NEXT: movi v1.4h, #16
+; CHECK-NEXT: sub v0.4h, v1.4h, v0.4h
+; CHECK-NEXT: ret
+ %b = call <4 x i16> @llvm.cttz.v4i16(<4 x i16> %a, i1 true)
+ ret <4 x i16> %b
+}
+
+define <2 x i32> @cttz_v2i32(<2 x i32> %a) nounwind {
+; CHECK-LABEL: cttz_v2i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.2s, #1
+; CHECK-NEXT: sub v1.2s, v0.2s, v1.2s
+; CHECK-NEXT: bic v0.8b, v1.8b, v0.8b
+; CHECK-NEXT: clz v0.2s, v0.2s
+; CHECK-NEXT: movi v1.2s, #32
+; CHECK-NEXT: sub v0.2s, v1.2s, v0.2s
+; CHECK-NEXT: ret
+ %b = call <2 x i32> @llvm.cttz.v2i32(<2 x i32> %a, i1 true)
+ ret <2 x i32> %b
+}
+
+define <1 x i64> @cttz_v1i64(<1 x i64> %a) nounwind {
+; CHECK-LABEL: cttz_v1i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: orr w8, wzr, #0x1
+; CHECK-NEXT: fmov d1, x8
+; CHECK-NEXT: sub d1, d0, d1
+; CHECK-NEXT: bic v0.8b, v1.8b, v0.8b
+; CHECK-NEXT: cnt v0.8b, v0.8b
+; CHECK-NEXT: uaddlp v0.4h, v0.8b
+; CHECK-NEXT: uaddlp v0.2s, v0.4h
+; CHECK-NEXT: uaddlp v0.1d, v0.2s
+; CHECK-NEXT: ret
+ %b = call <1 x i64> @llvm.cttz.v1i64(<1 x i64> %a, i1 true)
+ ret <1 x i64> %b
+}
+
+define <16 x i8> @cttz_v16i8(<16 x i8> %a) nounwind {
+; CHECK-LABEL: cttz_v16i8:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.16b, #1
+; CHECK-NEXT: sub v1.16b, v0.16b, v1.16b
+; CHECK-NEXT: bic v0.16b, v1.16b, v0.16b
+; CHECK-NEXT: cnt v0.16b, v0.16b
+; CHECK-NEXT: ret
+ %b = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %a, i1 true)
+ ret <16 x i8> %b
+}
+
+define <8 x i16> @cttz_v8i16(<8 x i16> %a) nounwind {
+; CHECK-LABEL: cttz_v8i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.8h, #1
+; CHECK-NEXT: sub v1.8h, v0.8h, v1.8h
+; CHECK-NEXT: bic v0.16b, v1.16b, v0.16b
+; CHECK-NEXT: clz v0.8h, v0.8h
+; CHECK-NEXT: movi v1.8h, #16
+; CHECK-NEXT: sub v0.8h, v1.8h, v0.8h
+; CHECK-NEXT: ret
+ %b = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %a, i1 true)
+ ret <8 x i16> %b
+}
+
+define <4 x i32> @cttz_v4i32(<4 x i32> %a) nounwind {
+; CHECK-LABEL: cttz_v4i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: movi v1.4s, #1
+; CHECK-NEXT: sub v1.4s, v0.4s, v1.4s
+; CHECK-NEXT: bic v0.16b, v1.16b, v0.16b
+; CHECK-NEXT: clz v0.4s, v0.4s
+; CHECK-NEXT: movi v1.4s, #32
+; CHECK-NEXT: sub v0.4s, v1.4s, v0.4s
+; CHECK-NEXT: ret
+ %b = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %a, i1 true)
+ ret <4 x i32> %b
+}
+
+define <2 x i64> @cttz_v2i64(<2 x i64> %a) nounwind {
+; CHECK-LABEL: cttz_v2i64:
+; CHECK: // %bb.0:
+; CHECK-NEXT: orr w8, wzr, #0x1
+; CHECK-NEXT: dup v1.2d, x8
+; CHECK-NEXT: sub v1.2d, v0.2d, v1.2d
+; CHECK-NEXT: bic v0.16b, v1.16b, v0.16b
+; CHECK-NEXT: cnt v0.16b, v0.16b
+; CHECK-NEXT: uaddlp v0.8h, v0.16b
+; CHECK-NEXT: uaddlp v0.4s, v0.8h
+; CHECK-NEXT: uaddlp v0.2d, v0.4s
+; CHECK-NEXT: ret
+ %b = call <2 x i64> @llvm.cttz.v2i64(<2 x i64> %a, i1 true)
+ ret <2 x i64> %b
+}
More information about the llvm-commits
mailing list