[llvm] [LoongArch] Select {DIV,MOD}.{W,WU} instruction to eliminate explicit sign extension (PR #92205)
via llvm-commits
llvm-commits at lists.llvm.org
Tue May 14 21:03:29 PDT 2024
https://github.com/heiher updated https://github.com/llvm/llvm-project/pull/92205
>From 2c1a5671989bf735407822be0972100e3d15d3d3 Mon Sep 17 00:00:00 2001
From: WANG Rui <wangrui at loongson.cn>
Date: Wed, 15 May 2024 10:06:13 +0800
Subject: [PATCH] [LoongArch] Select {DIV,MOD}.{W,WU} instruction to eliminate
explicit sign extension
---
.../LoongArch/LoongArchISelLowering.cpp | 13 +++
.../Target/LoongArch/LoongArchISelLowering.h | 4 +
.../Target/LoongArch/LoongArchInstrInfo.td | 6 ++
.../ir-instruction/sdiv-udiv-srem-urem.ll | 96 +++++++------------
4 files changed, 57 insertions(+), 62 deletions(-)
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index 21d520656091c..abce5fe9efa05 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -139,6 +139,7 @@ LoongArchTargetLowering::LoongArchTargetLowering(const TargetMachine &TM,
setOperationAction(ISD::BITREVERSE, MVT::i32, Custom);
setOperationAction(ISD::BSWAP, MVT::i32, Custom);
+ setOperationAction({ISD::UDIV, ISD::UREM}, MVT::i32, Custom);
}
// Set operations for LA32 only.
@@ -1665,6 +1666,10 @@ static LoongArchISD::NodeType getLoongArchWOpcode(unsigned Opcode) {
switch (Opcode) {
default:
llvm_unreachable("Unexpected opcode");
+ case ISD::UDIV:
+ return LoongArchISD::DIV_WU;
+ case ISD::UREM:
+ return LoongArchISD::MOD_WU;
case ISD::SHL:
return LoongArchISD::SLL_W;
case ISD::SRA:
@@ -1841,6 +1846,12 @@ void LoongArchTargetLowering::ReplaceNodeResults(
switch (N->getOpcode()) {
default:
llvm_unreachable("Don't know how to legalize this operation");
+ case ISD::UDIV:
+ case ISD::UREM:
+ assert(VT == MVT::i32 && Subtarget.is64Bit() &&
+ "Unexpected custom legalisation");
+ Results.push_back(customLegalizeToWOp(N, DAG, 2, ISD::SIGN_EXTEND));
+ break;
case ISD::SHL:
case ISD::SRA:
case ISD::SRL:
@@ -3445,6 +3456,8 @@ const char *LoongArchTargetLowering::getTargetNodeName(unsigned Opcode) const {
NODE_NAME_CASE(BITREV_W)
NODE_NAME_CASE(ROTR_W)
NODE_NAME_CASE(ROTL_W)
+ NODE_NAME_CASE(DIV_WU)
+ NODE_NAME_CASE(MOD_WU)
NODE_NAME_CASE(CLZ_W)
NODE_NAME_CASE(CTZ_W)
NODE_NAME_CASE(DBAR)
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
index 31b4d65195634..8a2473cdfa68c 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.h
@@ -43,6 +43,10 @@ enum NodeType : unsigned {
ROTL_W,
ROTR_W,
+ // unsigned 32-bit integer division
+ DIV_WU,
+ MOD_WU,
+
// FPR<->GPR transfer operations
MOVGR2FR_W_LA64,
MOVFR2GR_S_LA64,
diff --git a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
index f56f8f7e1179c..35ea9f07866d5 100644
--- a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
@@ -85,6 +85,8 @@ def loongarch_sll_w : SDNode<"LoongArchISD::SLL_W", SDT_LoongArchIntBinOpW>;
def loongarch_sra_w : SDNode<"LoongArchISD::SRA_W", SDT_LoongArchIntBinOpW>;
def loongarch_srl_w : SDNode<"LoongArchISD::SRL_W", SDT_LoongArchIntBinOpW>;
def loongarch_rotr_w : SDNode<"LoongArchISD::ROTR_W", SDT_LoongArchIntBinOpW>;
+def loongarch_div_wu : SDNode<"LoongArchISD::DIV_WU", SDT_LoongArchIntBinOpW>;
+def loongarch_mod_wu : SDNode<"LoongArchISD::MOD_WU", SDT_LoongArchIntBinOpW>;
def loongarch_crc_w_b_w
: SDNode<"LoongArchISD::CRC_W_B_W", SDT_LoongArchIntBinOpW, [SDNPHasChain]>;
def loongarch_crc_w_h_w
@@ -1110,9 +1112,13 @@ def : PatGprImm_32<add, ADDI_W, simm12>;
def : PatGprGpr<sub, SUB_D>;
def : PatGprGpr_32<sub, SUB_W>;
def : PatGprGpr<sdiv, DIV_D>;
+def : PatGprGpr_32<sdiv, DIV_W>;
def : PatGprGpr<udiv, DIV_DU>;
+def : PatGprGpr<loongarch_div_wu, DIV_WU>;
def : PatGprGpr<srem, MOD_D>;
+def : PatGprGpr_32<srem, MOD_W>;
def : PatGprGpr<urem, MOD_DU>;
+def : PatGprGpr<loongarch_mod_wu, MOD_WU>;
def : PatGprGpr<rotr, ROTR_D>;
def : PatGprGpr<loongarch_rotr_w, ROTR_W>;
def : PatGprGpr_32<rotr, ROTR_W>;
diff --git a/llvm/test/CodeGen/LoongArch/ir-instruction/sdiv-udiv-srem-urem.ll b/llvm/test/CodeGen/LoongArch/ir-instruction/sdiv-udiv-srem-urem.ll
index 2064c398948fe..ab3eec240db3c 100644
--- a/llvm/test/CodeGen/LoongArch/ir-instruction/sdiv-udiv-srem-urem.ll
+++ b/llvm/test/CodeGen/LoongArch/ir-instruction/sdiv-udiv-srem-urem.ll
@@ -191,8 +191,7 @@ define signext i32 @sdiv_si32_ui32_ui32(i32 %a, i32 %b) {
; LA64: # %bb.0: # %entry
; LA64-NEXT: addi.w $a1, $a1, 0
; LA64-NEXT: addi.w $a0, $a0, 0
-; LA64-NEXT: div.d $a0, $a0, $a1
-; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: div.w $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: sdiv_si32_ui32_ui32:
@@ -208,12 +207,11 @@ define signext i32 @sdiv_si32_ui32_ui32(i32 %a, i32 %b) {
; LA64-TRAP: # %bb.0: # %entry
; LA64-TRAP-NEXT: addi.w $a1, $a1, 0
; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
-; LA64-TRAP-NEXT: div.d $a0, $a0, $a1
+; LA64-TRAP-NEXT: div.w $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB5_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB5_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = sdiv i32 %a, %b
@@ -228,8 +226,7 @@ define signext i32 @sdiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-LABEL: sdiv_si32_si32_si32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: div.d $a0, $a0, $a1
-; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: div.w $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: sdiv_si32_si32_si32:
@@ -243,12 +240,11 @@ define signext i32 @sdiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-TRAP-LABEL: sdiv_si32_si32_si32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: div.d $a0, $a0, $a1
+; LA64-TRAP-NEXT: div.w $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB6_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB6_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = sdiv i32 %a, %b
@@ -407,9 +403,9 @@ define i32 @udiv_i32(i32 %a, i32 %b) {
;
; LA64-LABEL: udiv_i32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: div.du $a0, $a0, $a1
+; LA64-NEXT: addi.w $a1, $a1, 0
+; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: div.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: udiv_i32:
@@ -423,9 +419,9 @@ define i32 @udiv_i32(i32 %a, i32 %b) {
;
; LA64-TRAP-LABEL: udiv_i32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: div.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: addi.w $a1, $a1, 0
+; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
+; LA64-TRAP-NEXT: div.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB11_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
@@ -444,9 +440,7 @@ define i32 @udiv_ui32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-LABEL: udiv_ui32_si32_si32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: div.du $a0, $a0, $a1
+; LA64-NEXT: div.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: udiv_ui32_si32_si32:
@@ -460,9 +454,7 @@ define i32 @udiv_ui32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-TRAP-LABEL: udiv_ui32_si32_si32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: div.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: div.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB12_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
@@ -481,10 +473,9 @@ define signext i32 @udiv_si32_ui32_ui32(i32 %a, i32 %b) {
;
; LA64-LABEL: udiv_si32_ui32_ui32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: div.du $a0, $a0, $a1
+; LA64-NEXT: addi.w $a1, $a1, 0
; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: div.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: udiv_si32_ui32_ui32:
@@ -498,14 +489,13 @@ define signext i32 @udiv_si32_ui32_ui32(i32 %a, i32 %b) {
;
; LA64-TRAP-LABEL: udiv_si32_ui32_ui32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: div.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: addi.w $a1, $a1, 0
+; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
+; LA64-TRAP-NEXT: div.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB13_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB13_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = udiv i32 %a, %b
@@ -520,10 +510,7 @@ define signext i32 @udiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-LABEL: udiv_si32_si32_si32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: div.du $a0, $a0, $a1
-; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: div.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: udiv_si32_si32_si32:
@@ -537,14 +524,11 @@ define signext i32 @udiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-TRAP-LABEL: udiv_si32_si32_si32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: div.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: div.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB14_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB14_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = udiv i32 %a, %b
@@ -995,9 +979,9 @@ define i32 @urem_i32(i32 %a, i32 %b) {
;
; LA64-LABEL: urem_i32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: mod.du $a0, $a0, $a1
+; LA64-NEXT: addi.w $a1, $a1, 0
+; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: mod.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: urem_i32:
@@ -1011,9 +995,9 @@ define i32 @urem_i32(i32 %a, i32 %b) {
;
; LA64-TRAP-LABEL: urem_i32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: mod.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: addi.w $a1, $a1, 0
+; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
+; LA64-TRAP-NEXT: mod.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB27_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
@@ -1032,9 +1016,7 @@ define i32 @urem_ui32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-LABEL: urem_ui32_si32_si32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: mod.du $a0, $a0, $a1
+; LA64-NEXT: mod.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: urem_ui32_si32_si32:
@@ -1048,9 +1030,7 @@ define i32 @urem_ui32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-TRAP-LABEL: urem_ui32_si32_si32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: mod.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: mod.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB28_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
@@ -1069,10 +1049,9 @@ define signext i32 @urem_si32_ui32_ui32(i32 %a, i32 %b) {
;
; LA64-LABEL: urem_si32_ui32_ui32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: mod.du $a0, $a0, $a1
+; LA64-NEXT: addi.w $a1, $a1, 0
; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: mod.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: urem_si32_ui32_ui32:
@@ -1086,14 +1065,13 @@ define signext i32 @urem_si32_ui32_ui32(i32 %a, i32 %b) {
;
; LA64-TRAP-LABEL: urem_si32_ui32_ui32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: mod.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: addi.w $a1, $a1, 0
+; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
+; LA64-TRAP-NEXT: mod.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB29_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB29_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = urem i32 %a, %b
@@ -1108,10 +1086,7 @@ define signext i32 @urem_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-LABEL: urem_si32_si32_si32:
; LA64: # %bb.0: # %entry
-; LA64-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-NEXT: mod.du $a0, $a0, $a1
-; LA64-NEXT: addi.w $a0, $a0, 0
+; LA64-NEXT: mod.wu $a0, $a0, $a1
; LA64-NEXT: ret
;
; LA32-TRAP-LABEL: urem_si32_si32_si32:
@@ -1125,14 +1100,11 @@ define signext i32 @urem_si32_si32_si32(i32 signext %a, i32 signext %b) {
;
; LA64-TRAP-LABEL: urem_si32_si32_si32:
; LA64-TRAP: # %bb.0: # %entry
-; LA64-TRAP-NEXT: bstrpick.d $a1, $a1, 31, 0
-; LA64-TRAP-NEXT: bstrpick.d $a0, $a0, 31, 0
-; LA64-TRAP-NEXT: mod.du $a0, $a0, $a1
+; LA64-TRAP-NEXT: mod.wu $a0, $a0, $a1
; LA64-TRAP-NEXT: bnez $a1, .LBB30_2
; LA64-TRAP-NEXT: # %bb.1: # %entry
; LA64-TRAP-NEXT: break 7
; LA64-TRAP-NEXT: .LBB30_2: # %entry
-; LA64-TRAP-NEXT: addi.w $a0, $a0, 0
; LA64-TRAP-NEXT: ret
entry:
%r = urem i32 %a, %b
More information about the llvm-commits
mailing list