[PATCH] D67108: [ARM} Lower CTTZ on MVE
oliver cruickshank via Phabricator via llvm-commits
llvm-commits at lists.llvm.org
Tue Sep 3 09:51:50 PDT 2019
oliverlars created this revision.
oliverlars added reviewers: dmgreen, SjoerdMeijer, samparker, simon_tatham, t.p.northover.
Herald added subscribers: llvm-commits, hiraditya, kristof.beyls, javed.absar.
Herald added a project: LLVM.
Lower CTTZ on MVE using VBRSR and VCLS which will reverse the bits and count the leading zeros, equivalent to a count trailing zeros (CTTZ).
Repository:
rL LLVM
https://reviews.llvm.org/D67108
Files:
llvm/lib/Target/ARM/ARMISelLowering.cpp
llvm/test/CodeGen/Thumb2/mve-cttz.ll
Index: llvm/test/CodeGen/Thumb2/mve-cttz.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/Thumb2/mve-cttz.ll
@@ -0,0 +1,39 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=thumbv8.1m.main-arm-none-eabi -mattr=+mve %s -o - | FileCheck %s
+
+define arm_aapcs_vfpcc <4 x i32> @cttz_4i32_t(<4 x i32> %src){
+; CHECK-LABEL: cttz_4i32_t:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: vbrsr.32 q0, q0, #0
+; CHECK-NEXT: vclz.i32 q0, q0
+; CHECK-NEXT: bx lr
+entry:
+ %0 = call <4 x i32> @llvm.cttz.v4i32(<4 x i32> %src)
+ ret <4 x i32> %0
+}
+
+define arm_aapcs_vfpcc <8 x i16> @cttz_8i16_t(<8 x i16> %src){
+; CHECK-LABEL: cttz_8i16_t:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: vbrsr.16 q0, q0, #0
+; CHECK-NEXT: vclz.i16 q0, q0
+; CHECK-NEXT: bx lr
+entry:
+ %0 = call <8 x i16> @llvm.cttz.v8i16(<8 x i16> %src)
+ ret <8 x i16> %0
+}
+
+define arm_aapcs_vfpcc <16 x i8> @cttz_16i8_t(<16 x i8> %src){
+; CHECK-LABEL: cttz_16i8_t:
+; CHECK: @ %bb.0: @ %entry
+; CHECK-NEXT: vbrsr.8 q0, q0, #0
+; CHECK-NEXT: vclz.i8 q0, q0
+; CHECK-NEXT: bx lr
+entry:
+ %0 = call <16 x i8> @llvm.cttz.v16i8(<16 x i8> %src)
+ ret <16 x i8> %0
+}
+
+declare <4 x i32> @llvm.cttz.v4i32(<4 x i32>)
+declare <8 x i16> @llvm.cttz.v8i16(<8 x i16>)
+declare <16 x i8> @llvm.cttz.v16i8(<16 x i8>)
Index: llvm/lib/Target/ARM/ARMISelLowering.cpp
===================================================================
--- llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -260,6 +260,7 @@
setOperationAction(ISD::ABS, VT, Legal);
setOperationAction(ISD::SETCC, VT, Custom);
setOperationAction(ISD::CTLZ, VT, Legal);
+ setOperationAction(ISD::CTTZ, VT, Custom);
// No native support for these.
setOperationAction(ISD::UDIV, VT, Expand);
@@ -5738,9 +5739,7 @@
const ARMSubtarget *ST) {
SDLoc dl(N);
EVT VT = N->getValueType(0);
- if (VT.isVector()) {
- assert(ST->hasNEON());
-
+ if (VT.isVector() && ST->hasNEON()) {
// Compute the least significant set bit: LSB = X & -X
SDValue X = N->getOperand(0);
SDValue NX = DAG.getNode(ISD::SUB, dl, VT, getZeroVector(VT, DAG, dl), X);
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D67108.218472.patch
Type: text/x-patch
Size: 2318 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20190903/d667ee07/attachment.bin>
More information about the llvm-commits
mailing list