[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 19:22:54 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-loongarch
Author: hev (heiher)
<details>
<summary>Changes</summary>
---
Full diff: https://github.com/llvm/llvm-project/pull/92205.diff
2 Files Affected:
- (modified) llvm/lib/Target/LoongArch/LoongArchInstrInfo.td (+4)
- (modified) llvm/test/CodeGen/LoongArch/ir-instruction/sdiv-udiv-srem-urem.ll (+12-24)
``````````diff
diff --git a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
index f56f8f7e1179c..d3b577c48cb5f 100644
--- a/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
+++ b/llvm/lib/Target/LoongArch/LoongArchInstrInfo.td
@@ -1110,9 +1110,13 @@ 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_32<udiv, DIV_WU>;
def : PatGprGpr<srem, MOD_D>;
+def : PatGprGpr_32<srem, MOD_W>;
def : PatGprGpr<urem, MOD_DU>;
+def : PatGprGpr_32<urem, 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..aa4492b19e3ff 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
@@ -483,8 +479,7 @@ define signext i32 @udiv_si32_ui32_ui32(i32 %a, i32 %b) {
; 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_ui32_ui32:
@@ -500,12 +495,11 @@ define signext i32 @udiv_si32_ui32_ui32(i32 %a, i32 %b) {
; 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, .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
@@ -522,8 +516,7 @@ define signext i32 @udiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
; 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:
@@ -539,12 +532,11 @@ define signext i32 @udiv_si32_si32_si32(i32 signext %a, i32 signext %b) {
; 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
@@ -1071,8 +1063,7 @@ define signext i32 @urem_si32_ui32_ui32(i32 %a, i32 %b) {
; 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_ui32_ui32:
@@ -1088,12 +1079,11 @@ define signext i32 @urem_si32_ui32_ui32(i32 %a, i32 %b) {
; 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, .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
@@ -1110,8 +1100,7 @@ define signext i32 @urem_si32_si32_si32(i32 signext %a, i32 signext %b) {
; 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:
@@ -1127,12 +1116,11 @@ define signext i32 @urem_si32_si32_si32(i32 signext %a, i32 signext %b) {
; 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
``````````
</details>
https://github.com/llvm/llvm-project/pull/92205
More information about the llvm-commits
mailing list