[llvm] [DAGCombine] Remove oneuse restrictions for RISCV in folding (shl (add_nsw x, c1)), c2) and folding (shl(sext(add x, c1)), c2) in some scenarios (PR #101294)

via llvm-commits llvm-commits at lists.llvm.org
Sat Nov 23 19:42:19 PST 2024


https://github.com/LiqinWeng updated https://github.com/llvm/llvm-project/pull/101294

>From 328e22351c0c570389932aa4d387e1fb1f0141ab Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Sat, 3 Aug 2024 19:13:35 +0800
Subject: [PATCH 1/7] [Test] Pre-submit tests for #101294

---
 .../CodeGen/RISCV/add_sext_shl_constant.ll    | 195 ++++++++++++++++++
 llvm/test/CodeGen/RISCV/add_shl_constant.ll   | 132 ++++++++++++
 .../CodeGen/RISCV/riscv-shifted-extend.ll     | 124 -----------
 3 files changed, 327 insertions(+), 124 deletions(-)
 create mode 100644 llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
 create mode 100644 llvm/test/CodeGen/RISCV/add_shl_constant.ll
 delete mode 100644 llvm/test/CodeGen/RISCV/riscv-shifted-extend.ll

diff --git a/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
new file mode 100644
index 00000000000000..35f3656e868681
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
@@ -0,0 +1,195 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
+; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV64 %s
+
+define void @add_sext_shl_moreOneUse_add(ptr %array1, i32 %a, i32 %b) {
+; RV64-LABEL: add_sext_shl_moreOneUse_add:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addiw a3, a1, 5
+; RV64-NEXT:    slli a4, a3, 2
+; RV64-NEXT:    add a4, a0, a4
+; RV64-NEXT:    sw a2, 0(a4)
+; RV64-NEXT:    sext.w a1, a1
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sw a2, 24(a0)
+; RV64-NEXT:    sw a3, 140(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 5
+  %idxprom = sext i32 %add to i64
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
+  store i32 %b, ptr %arrayidx
+  %add3 = add nsw i32 %a, 6
+  %idxprom4 = sext i32 %add3 to i64
+  %arrayidx5 = getelementptr inbounds i32, ptr %array1, i64 %idxprom4
+  store i32 %b, ptr %arrayidx5
+  %add6 = add nsw i32 %a, 35
+  %idxprom7 = sext i32 %add6 to i64
+  %arrayidx8 = getelementptr inbounds i32, ptr %array1, i64 %idxprom7
+  store i32 %add, ptr %arrayidx8
+  ret void
+}
+
+define void @add_sext_shl_moreOneUse_addexceedsign12(ptr %array1, i32 %a, i32 %b) {
+; RV64-LABEL: add_sext_shl_moreOneUse_addexceedsign12:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addi a3, a1, 2047
+; RV64-NEXT:    addiw a3, a3, 1
+; RV64-NEXT:    slli a4, a3, 2
+; RV64-NEXT:    add a4, a0, a4
+; RV64-NEXT:    sw a2, 0(a4)
+; RV64-NEXT:    sext.w a1, a1
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    lui a4, 2
+; RV64-NEXT:    add a0, a0, a4
+; RV64-NEXT:    add a0, a0, a1
+; RV64-NEXT:    sw a3, 4(a0)
+; RV64-NEXT:    sw a2, 120(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 2048
+  %idxprom = sext i32 %add to i64
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
+  store i32 %b, ptr %arrayidx
+  %0 = sext i32 %a to i64
+  %1 = getelementptr i32, ptr %array1, i64 %0
+  %arrayidx3 = getelementptr i8, ptr %1, i64 8196
+  store i32 %add, ptr %arrayidx3
+  %arrayidx6 = getelementptr i8, ptr %1, i64 8312
+  store i32 %b, ptr %arrayidx6
+  ret void
+}
+
+define void @add_sext_shl_moreOneUse_sext(ptr %array1, i32 %a, i32 %b) {
+; RV64-LABEL: add_sext_shl_moreOneUse_sext:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addiw a3, a1, 5
+; RV64-NEXT:    slli a4, a3, 2
+; RV64-NEXT:    add a4, a0, a4
+; RV64-NEXT:    sw a2, 0(a4)
+; RV64-NEXT:    sext.w a1, a1
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sw a2, 24(a0)
+; RV64-NEXT:    sd a3, 140(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 5
+  %idxprom = sext i32 %add to i64
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
+  store i32 %b, ptr %arrayidx
+  %add3 = add nsw i32 %a, 6
+  %idxprom4 = sext i32 %add3 to i64
+  %arrayidx5 = getelementptr inbounds i32, ptr %array1, i64 %idxprom4
+  store i32 %b, ptr %arrayidx5
+  %add6 = add nsw i32 %a, 35
+  %idxprom7 = sext i32 %add6 to i64
+  %arrayidx8 = getelementptr inbounds i32, ptr %array1, i64 %idxprom7
+  store i64 %idxprom, ptr %arrayidx8
+  ret void
+}
+
+; test of jumpping, find add's operand has one more use can simplified
+define void @add_sext_shl_moreOneUse_add_inSelect(ptr %array1, i32 signext  %a, i32 %b, i32 signext %x) {
+; RV64-LABEL: add_sext_shl_moreOneUse_add_inSelect:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addiw a4, a1, 5
+; RV64-NEXT:    slli a5, a4, 2
+; RV64-NEXT:    add a5, a0, a5
+; RV64-NEXT:    mv a6, a4
+; RV64-NEXT:    bgtz a3, .LBB3_2
+; RV64-NEXT:  # %bb.1: # %entry
+; RV64-NEXT:    mv a6, a2
+; RV64-NEXT:  .LBB3_2: # %entry
+; RV64-NEXT:    sw a6, 0(a5)
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sw a6, 24(a0)
+; RV64-NEXT:    sw a4, 140(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 5
+  %cmp = icmp sgt i32 %x, 0
+  %idxprom = sext i32 %add to i64
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
+  %add.b = select i1 %cmp, i32 %add, i32 %b
+  store i32 %add.b, ptr %arrayidx
+  %add5 = add nsw i32 %a, 6
+  %idxprom6 = sext i32 %add5 to i64
+  %arrayidx7 = getelementptr inbounds i32, ptr %array1, i64 %idxprom6
+  store i32 %add.b, ptr %arrayidx7
+  %add8 = add nsw i32 %a, 35
+  %idxprom9 = sext i32 %add8 to i64
+  %arrayidx10 = getelementptr inbounds i32, ptr %array1, i64 %idxprom9
+  store i32 %add, ptr %arrayidx10
+  ret void
+}
+
+define void @add_sext_shl_moreOneUse_add_inSelect_addexceedsign12(ptr %array1, i32 signext  %a, i32 %b, i32 signext %x) {
+; RV64-LABEL: add_sext_shl_moreOneUse_add_inSelect_addexceedsign12:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addi a4, a1, 2047
+; RV64-NEXT:    addiw a4, a4, 1
+; RV64-NEXT:    slli a6, a4, 2
+; RV64-NEXT:    add a6, a0, a6
+; RV64-NEXT:    mv a5, a4
+; RV64-NEXT:    bgtz a3, .LBB4_2
+; RV64-NEXT:  # %bb.1: # %entry
+; RV64-NEXT:    mv a5, a2
+; RV64-NEXT:  .LBB4_2: # %entry
+; RV64-NEXT:    sw a5, 0(a6)
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    add a0, a0, a1
+; RV64-NEXT:    lui a1, 2
+; RV64-NEXT:    add a0, a0, a1
+; RV64-NEXT:    sw a5, 4(a0)
+; RV64-NEXT:    sw a4, 120(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 2048
+  %cmp = icmp sgt i32 %x, 0
+  %idxprom = sext i32 %add to i64
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
+  %add.b = select i1 %cmp, i32 %add, i32 %b
+  store i32 %add.b, ptr %arrayidx
+  %0 = sext i32 %a to i64
+  %1 = getelementptr i32, ptr %array1, i64 %0
+  %arrayidx7 = getelementptr i8, ptr %1, i64 8196
+  store i32 %add.b, ptr %arrayidx7
+  %arrayidx10 = getelementptr i8, ptr %1, i64 8312
+  store i32 %add, ptr %arrayidx10
+  ret void
+}
+
+define void @add_shl_moreOneUse_inSelect(ptr %array1, i64 %a, i64 %b, i64 %x) {
+; RV64-LABEL: add_shl_moreOneUse_inSelect:
+; RV64:       # %bb.0: # %entry
+; RV64-NEXT:    addi a4, a1, 5
+; RV64-NEXT:    mv a5, a4
+; RV64-NEXT:    bgtz a3, .LBB5_2
+; RV64-NEXT:  # %bb.1: # %entry
+; RV64-NEXT:    mv a5, a2
+; RV64-NEXT:  .LBB5_2: # %entry
+; RV64-NEXT:    slli a2, a4, 3
+; RV64-NEXT:    add a2, a0, a2
+; RV64-NEXT:    sd a5, 0(a2)
+; RV64-NEXT:    slli a1, a1, 3
+; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sd a5, 48(a0)
+; RV64-NEXT:    sd a4, 280(a0)
+; RV64-NEXT:    ret
+entry:
+  %add = add nsw i64 %a, 5
+  %cmp = icmp sgt i64 %x, 0
+  %spec.select = select i1 %cmp, i64 %add, i64 %b
+  %0 = getelementptr inbounds i64, ptr %array1, i64 %add
+  store i64 %spec.select, ptr %0
+  %add3 = add nsw i64 %a, 6
+  %arrayidx4 = getelementptr inbounds i64, ptr %array1, i64 %add3
+  store i64 %spec.select, ptr %arrayidx4
+  %add5 = add nsw i64 %a, 35
+  %arrayidx6 = getelementptr inbounds i64, ptr %array1, i64 %add5
+  store i64 %add, ptr %arrayidx6
+  ret void
+}
diff --git a/llvm/test/CodeGen/RISCV/add_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
new file mode 100644
index 00000000000000..5c71a3c5449940
--- /dev/null
+++ b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
@@ -0,0 +1,132 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
+; RUN:   | FileCheck -check-prefix=RV32 %s
+
+define i32 @add_shl_oneUse(i32 %x, i32 %y) nounwind {
+; RV32-LABEL: add_shl_oneUse:
+; RV32:       # %bb.0:
+; RV32-NEXT:    slli a0, a0, 3
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    addi a0, a0, 984
+; RV32-NEXT:    ret
+  %add.0 = add i32 %x, 123
+  %shl = shl i32 %add.0, 3
+  %add.1 = add i32 %shl, %y
+  ret i32 %add.1
+}
+
+define void @add_shl_moreOneUse_inStore(ptr %array1, i32 %a, i32 %b)  {
+; RV32-LABEL: add_shl_moreOneUse_inStore:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    addi a3, a1, 5
+; RV32-NEXT:    slli a4, a3, 2
+; RV32-NEXT:    add a4, a0, a4
+; RV32-NEXT:    sw a2, 0(a4)
+; RV32-NEXT:    slli a1, a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a2, 24(a0)
+; RV32-NEXT:    sw a3, 140(a0)
+; RV32-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 5
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i32 %add
+  store i32 %b, ptr %arrayidx
+  %0 = getelementptr i32, ptr %array1, i32 %a
+  %arrayidx3 = getelementptr i8, ptr %0, i32 24
+  store i32 %b, ptr %arrayidx3
+  %arrayidx5 = getelementptr i8, ptr %0, i32 140
+  store i32 %add, ptr %arrayidx5
+  ret void
+}
+
+define void @add_shl_moreOneUse_inStore_addexceedsign12(ptr %array1, i32 %a, i32 %b)  {
+; RV32-LABEL: add_shl_moreOneUse_inStore_addexceedsign12:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    addi a3, a1, 2047
+; RV32-NEXT:    addi a3, a3, 1
+; RV32-NEXT:    slli a4, a3, 2
+; RV32-NEXT:    add a4, a0, a4
+; RV32-NEXT:    sw a2, 0(a4)
+; RV32-NEXT:    slli a1, a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    lui a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a3, 4(a0)
+; RV32-NEXT:    sw a2, 120(a0)
+; RV32-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 2048
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i32 %add
+  store i32 %b, ptr %arrayidx
+  %0 = getelementptr i32, ptr %array1, i32 %a
+  %arrayidx2 = getelementptr i8, ptr %0, i32 8196
+  store i32 %add, ptr %arrayidx2
+  %arrayidx4 = getelementptr i8, ptr %0, i32 8312
+  store i32 %b, ptr %arrayidx4
+  ret void
+}
+
+define void @add_shl_moreOneUse_inSelect(ptr %array1, i32 %a, i32 %b, i32 %x) {
+; RV32-LABEL: add_shl_moreOneUse_inSelect:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    addi a4, a1, 5
+; RV32-NEXT:    mv a5, a4
+; RV32-NEXT:    bgtz a3, .LBB3_2
+; RV32-NEXT:  # %bb.1: # %entry
+; RV32-NEXT:    mv a5, a2
+; RV32-NEXT:  .LBB3_2: # %entry
+; RV32-NEXT:    slli a2, a4, 2
+; RV32-NEXT:    add a2, a0, a2
+; RV32-NEXT:    sw a5, 0(a2)
+; RV32-NEXT:    slli a1, a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a5, 24(a0)
+; RV32-NEXT:    sw a4, 140(a0)
+; RV32-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 5
+  %cmp = icmp sgt i32 %x, 0
+  %cond = select i1 %cmp, i32 %add, i32 %b
+  %arrayidx = getelementptr inbounds i32, ptr %array1, i32 %add
+  store i32 %cond, ptr %arrayidx
+  %0 = getelementptr i32, ptr %array1, i32 %a
+  %arrayidx2 = getelementptr i32, ptr %0, i32 6
+  store i32 %cond, ptr %arrayidx2
+  %arrayidx4 = getelementptr i32, ptr %0, i32 35
+  store i32 %add, ptr %arrayidx4
+  ret void
+}
+
+define void @add_shl_moreOneUse_inSelect_addexceedsign12(ptr %array1, i32 %a, i32 %b, i32 %x) {
+; RV32-LABEL: add_shl_moreOneUse_inSelect_addexceedsign12:
+; RV32:       # %bb.0: # %entry
+; RV32-NEXT:    addi a4, a1, 2047
+; RV32-NEXT:    addi a4, a4, 1
+; RV32-NEXT:    mv a5, a4
+; RV32-NEXT:    bgtz a3, .LBB4_2
+; RV32-NEXT:  # %bb.1: # %entry
+; RV32-NEXT:    mv a5, a2
+; RV32-NEXT:  .LBB4_2: # %entry
+; RV32-NEXT:    slli a2, a4, 2
+; RV32-NEXT:    add a2, a0, a2
+; RV32-NEXT:    sw a5, 0(a2)
+; RV32-NEXT:    slli a1, a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    lui a1, 2
+; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a5, 4(a0)
+; RV32-NEXT:    sw a4, 120(a0)
+; RV32-NEXT:    ret
+entry:
+  %add = add nsw i32 %a, 2048
+  %cmp = icmp sgt i32 %x, 0
+  %spec.select = select i1 %cmp, i32 %add, i32 %b
+  %0 = getelementptr inbounds i32, ptr %array1, i32 %add
+  store i32 %spec.select, ptr %0, align 4
+  %1 = getelementptr i32, ptr %array1, i32 %a
+  %arrayidx4 = getelementptr i8, ptr %1, i32 8196
+  store i32 %spec.select, ptr %arrayidx4
+  %arrayidx6 = getelementptr i8, ptr %1, i32 8312
+  store i32 %add, ptr %arrayidx6
+  ret void
+}
diff --git a/llvm/test/CodeGen/RISCV/riscv-shifted-extend.ll b/llvm/test/CodeGen/RISCV/riscv-shifted-extend.ll
deleted file mode 100644
index 4901e268ec11a0..00000000000000
--- a/llvm/test/CodeGen/RISCV/riscv-shifted-extend.ll
+++ /dev/null
@@ -1,124 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
-; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
-; RUN:   | FileCheck -check-prefix=RV64 %s
-
-define void @test(ptr nocapture noundef writeonly %array1, i32 noundef signext %a, i32 noundef signext %b) {
-; RV64-LABEL: test:
-; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addiw a3, a1, 5
-; RV64-NEXT:    slli a4, a3, 2
-; RV64-NEXT:    add a4, a0, a4
-; RV64-NEXT:    slli a1, a1, 2
-; RV64-NEXT:    add a0, a1, a0
-; RV64-NEXT:    sw a2, 0(a4)
-; RV64-NEXT:    sw a2, 24(a0)
-; RV64-NEXT:    sw a3, 140(a0)
-; RV64-NEXT:    ret
-entry:
-  %add = add nsw i32 %a, 5
-  %idxprom = sext i32 %add to i64
-  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
-  store i32 %b, ptr %arrayidx, align 4
-  %add3 = add nsw i32 %a, 6
-  %idxprom4 = sext i32 %add3 to i64
-  %arrayidx5 = getelementptr inbounds i32, ptr %array1, i64 %idxprom4
-  store i32 %b, ptr %arrayidx5, align 4
-  %add6 = add nsw i32 %a, 35
-  %idxprom7 = sext i32 %add6 to i64
-  %arrayidx8 = getelementptr inbounds i32, ptr %array1, i64 %idxprom7
-  store i32 %add, ptr %arrayidx8, align 4
-  ret void
-}
-
-; test of jumpping, find add's operand has one more use can simplified
-define void @test1(ptr nocapture noundef %array1, i32 noundef signext %a, i32 noundef signext %b, i32 noundef signext %x) {
-; RV64-LABEL: test1:
-; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addiw a4, a1, 5
-; RV64-NEXT:    slli a5, a4, 2
-; RV64-NEXT:    add a5, a0, a5
-; RV64-NEXT:    mv a6, a4
-; RV64-NEXT:    bgtz a3, .LBB1_2
-; RV64-NEXT:  # %bb.1: # %entry
-; RV64-NEXT:    mv a6, a2
-; RV64-NEXT:  .LBB1_2: # %entry
-; RV64-NEXT:    slli a1, a1, 2
-; RV64-NEXT:    add a0, a1, a0
-; RV64-NEXT:    sw a6, 0(a5)
-; RV64-NEXT:    sw a6, 24(a0)
-; RV64-NEXT:    sw a4, 140(a0)
-; RV64-NEXT:    ret
-entry:
-  %add = add nsw i32 %a, 5
-  %cmp = icmp sgt i32 %x, 0
-  %idxprom = sext i32 %add to i64
-  %arrayidx = getelementptr inbounds i32, ptr %array1, i64 %idxprom
-  %add.b = select i1 %cmp, i32 %add, i32 %b
-  store i32 %add.b, ptr %arrayidx, align 4
-  %add5 = add nsw i32 %a, 6
-  %idxprom6 = sext i32 %add5 to i64
-  %arrayidx7 = getelementptr inbounds i32, ptr %array1, i64 %idxprom6
-  store i32 %add.b, ptr %arrayidx7, align 4
-  %add8 = add nsw i32 %a, 35
-  %idxprom9 = sext i32 %add8 to i64
-  %arrayidx10 = getelementptr inbounds i32, ptr %array1, i64 %idxprom9
-  store i32 %add, ptr %arrayidx10, align 4
-  ret void
-}
-
-define void @test2(ptr nocapture noundef writeonly %array1, i64 noundef %a, i64 noundef %b) local_unnamed_addr #0 {
-; RV64-LABEL: test2:
-; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addi a3, a1, 5
-; RV64-NEXT:    slli a4, a3, 3
-; RV64-NEXT:    add a4, a0, a4
-; RV64-NEXT:    slli a1, a1, 3
-; RV64-NEXT:    add a0, a1, a0
-; RV64-NEXT:    sd a2, 0(a4)
-; RV64-NEXT:    sd a2, 48(a0)
-; RV64-NEXT:    sd a3, 280(a0)
-; RV64-NEXT:    ret
-entry:
-  %add = add nsw i64 %a, 5
-  %arrayidx = getelementptr inbounds i64, ptr %array1, i64 %add
-  store i64 %b, ptr %arrayidx, align 8
-  %add2 = add nsw i64 %a, 6
-  %arrayidx3 = getelementptr inbounds i64, ptr %array1, i64 %add2
-  store i64 %b, ptr %arrayidx3, align 8
-  %add4 = add nsw i64 %a, 35
-  %arrayidx5 = getelementptr inbounds i64, ptr %array1, i64 %add4
-  store i64 %add, ptr %arrayidx5, align 8
-  ret void
-}
-
-define void @test3(ptr nocapture noundef %array1, i64 noundef %a, i64 noundef %b, i64 noundef %x) {
-; RV64-LABEL: test3:
-; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addi a4, a1, 5
-; RV64-NEXT:    mv a5, a4
-; RV64-NEXT:    bgtz a3, .LBB3_2
-; RV64-NEXT:  # %bb.1: # %entry
-; RV64-NEXT:    mv a5, a2
-; RV64-NEXT:  .LBB3_2: # %entry
-; RV64-NEXT:    slli a2, a4, 3
-; RV64-NEXT:    add a2, a0, a2
-; RV64-NEXT:    slli a1, a1, 3
-; RV64-NEXT:    add a0, a1, a0
-; RV64-NEXT:    sd a5, 0(a2)
-; RV64-NEXT:    sd a5, 48(a0)
-; RV64-NEXT:    sd a4, 280(a0)
-; RV64-NEXT:    ret
-entry:
-  %add = add nsw i64 %a, 5
-  %cmp = icmp sgt i64 %x, 0
-  %spec.select = select i1 %cmp, i64 %add, i64 %b
-  %0 = getelementptr inbounds i64, ptr %array1, i64 %add
-  store i64 %spec.select, ptr %0, align 8
-  %add3 = add nsw i64 %a, 6
-  %arrayidx4 = getelementptr inbounds i64, ptr %array1, i64 %add3
-  store i64 %spec.select, ptr %arrayidx4, align 8
-  %add5 = add nsw i64 %a, 35
-  %arrayidx6 = getelementptr inbounds i64, ptr %array1, i64 %add5
-  store i64 %add, ptr %arrayidx6, align 8
-  ret void
-}

>From d8ee1b0c7fcad6ca44cfd25f31d7c7a52a4e55e1 Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Fri, 9 Aug 2024 13:08:36 +0800
Subject: [PATCH 2/7] [DAGCombine] Remove OneUse restriction when folding (shl
 (add x, c1), c2) and (shl (sext (add x, c1)), c2)

---
 llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp |  5 +-
 .../Target/AArch64/AArch64ISelLowering.cpp    |  5 ++
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp |  7 +++
 llvm/lib/Target/ARM/ARMISelLowering.cpp       |  6 ++
 .../Target/Hexagon/HexagonISelLowering.cpp    | 15 +++++
 llvm/lib/Target/Hexagon/HexagonISelLowering.h |  3 +
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp   | 15 +++++
 llvm/lib/Target/PowerPC/PPCISelLowering.h     |  3 +
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp   | 48 ++++++++++++++++
 llvm/lib/Target/X86/X86ISelLowering.cpp       | 16 ++++++
 llvm/lib/Target/X86/X86ISelLowering.h         |  3 +
 .../CodeGen/RISCV/add_sext_shl_constant.ll    | 56 ++++++++-----------
 llvm/test/CodeGen/RISCV/add_shl_constant.ll   | 24 +++-----
 13 files changed, 153 insertions(+), 53 deletions(-)

diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index b800204d917503..2377e1ccf379ab 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -10192,7 +10192,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
   // Variant of version done on multiply, except mul by a power of 2 is turned
   // into a shift.
   if ((N0.getOpcode() == ISD::ADD || N0.getOpcode() == ISD::OR) &&
-      N0->hasOneUse() && TLI.isDesirableToCommuteWithShift(N, Level)) {
+      TLI.isDesirableToCommuteWithShift(N, Level)) {
     SDValue N01 = N0.getOperand(1);
     if (SDValue Shl1 =
             DAG.FoldConstantArithmetic(ISD::SHL, SDLoc(N1), VT, {N01, N1})) {
@@ -10211,8 +10211,7 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
   // TODO: Should we limit this with isLegalAddImmediate?
   if (N0.getOpcode() == ISD::SIGN_EXTEND &&
       N0.getOperand(0).getOpcode() == ISD::ADD &&
-      N0.getOperand(0)->getFlags().hasNoSignedWrap() && N0->hasOneUse() &&
-      N0.getOperand(0)->hasOneUse() &&
+      N0.getOperand(0)->getFlags().hasNoSignedWrap() &&
       TLI.isDesirableToCommuteWithShift(N, Level)) {
     SDValue Add = N0.getOperand(0);
     SDLoc DL(N0);
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 31a720ed7b5c77..5b1d8209e766c6 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -17912,6 +17912,11 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
   SDValue ShiftLHS = N->getOperand(0);
   EVT VT = N->getValueType(0);
 
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
+      !ShiftLHS->hasOneUse())
+    return false;
+
   // If ShiftLHS is unsigned bit extraction: ((x >> C) & mask), then do not
   // combine it with shift 'N' to let it be lowered to UBFX except:
   // ((x >> C) & mask) << C.
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index e4b54c7d72b083..fc439e0be578a2 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -1072,6 +1072,13 @@ bool AMDGPUTargetLowering::isDesirableToCommuteWithShift(
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
+
+  SDValue ShiftLHS = N->getOperand(0);
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
+      !ShiftLHS->hasOneUse())
+    return false;
+
   // Always commute pre-type legalization and right shifts.
   // We're looking for shl(or(x,y),z) patterns.
   if (Level < CombineLevel::AfterLegalizeTypes ||
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index a98b7a8420927e..13a4008bf32f23 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -13883,6 +13883,12 @@ ARMTargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
+  SDValue ShiftLHS = N->getOperand(0);
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
+      !ShiftLHS->hasOneUse())
+    return false;
+
   if (Level == BeforeLegalizeTypes)
     return true;
 
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index ab9bc559367787..c1f7253353b2a6 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -2156,6 +2156,21 @@ bool HexagonTargetLowering::hasBitTest(SDValue X, SDValue Y) const {
   return X.getValueType().isScalarInteger(); // 'tstbit'
 }
 
+bool HexagonTargetLowering::isDesirableToCommuteWithShift(
+    const SDNode *N, CombineLevel Level) const {
+  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
+          N->getOpcode() == ISD::SRL) &&
+         "Expected shift op");
+
+  SDValue ShiftLHS = N->getOperand(0);
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
+      !ShiftLHS->hasOneUse())
+    return false;
+
+  return true;
+}
+
 bool HexagonTargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const {
   return isTruncateFree(EVT::getEVT(Ty1), EVT::getEVT(Ty2));
 }
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.h b/llvm/lib/Target/Hexagon/HexagonISelLowering.h
index 3fd961f5a74623..a6bd57630031c4 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.h
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.h
@@ -155,6 +155,9 @@ class HexagonTargetLowering : public TargetLowering {
 
   bool hasBitTest(SDValue X, SDValue Y) const override;
 
+  bool isDesirableToCommuteWithShift(const SDNode *N,
+                                     CombineLevel Level) const override;
+
   bool allowTruncateForTailCall(Type *Ty1, Type *Ty2) const override;
 
   /// Return true if an FMA operation is faster than a pair of mul and add
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index d8f3095ed7fb68..a529701201ccdf 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -19137,3 +19137,18 @@ Value *PPCTargetLowering::emitMaskedAtomicCmpXchgIntrinsic(
   return Builder.CreateOr(
       Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 64)), "val64");
 }
+
+bool PPCTargetLowering::isDesirableToCommuteWithShift(
+    const SDNode *N, CombineLevel Level) const {
+  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
+          N->getOpcode() == ISD::SRL) &&
+         "Expected shift op");
+
+  SDValue ShiftLHS = N->getOperand(0);
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
+      !ShiftLHS->hasOneUse())
+    return false;
+
+  return true;
+}
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.h b/llvm/lib/Target/PowerPC/PPCISelLowering.h
index dde45e4cf6f4ae..1562c0e0a70eeb 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.h
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.h
@@ -1492,6 +1492,9 @@ namespace llvm {
     /// through to determine the optimal load/store instruction format.
     unsigned computeMOFlags(const SDNode *Parent, SDValue N,
                             SelectionDAG &DAG) const;
+
+    bool isDesirableToCommuteWithShift(const SDNode *N,
+                                       CombineLevel Level) const override;
   }; // end class PPCTargetLowering
 
   namespace PPC {
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index af7a39b2580a37..cbd2d33f00c9eb 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -17983,8 +17983,46 @@ bool RISCVTargetLowering::isDesirableToCommuteWithShift(
   //   (shl (or x, c1), c2) -> (or (shl x, c2), c1 << c2)
   SDValue N0 = N->getOperand(0);
   EVT Ty = N0.getValueType();
+
+  // LD/ST will optimize constant Offset extraction, so when AddNode is used by
+  // LD/ST, it can still complete the folding optimization operation performed
+  // above.
+  auto isLDST = [&]() {
+    bool canOptAwlays = false;
+    if (N0->getOpcode() == ISD::ADD && !N0->hasOneUse()) {
+      for (SDNode *Use : N0->uses()) {
+        // This use is the one we're on right now. Skip it
+        if (Use == N || Use->getOpcode() == ISD::SELECT)
+          continue;
+        if (!isa<StoreSDNode>(Use) && !isa<LoadSDNode>(Use)) {
+          canOptAwlays = false;
+          break;
+        }
+        canOptAwlays = true;
+      }
+    }
+
+    if (N0->getOpcode() == ISD::SIGN_EXTEND &&
+        !N0->getOperand(0)->hasOneUse()) {
+      for (SDNode *Use : N0->getOperand(0)->uses()) {
+        // This use is the one we're on right now. Skip it
+        if (Use == N0.getNode() || Use->getOpcode() == ISD::SELECT)
+          continue;
+        if (!isa<StoreSDNode>(Use) && !isa<LoadSDNode>(Use)) {
+          canOptAwlays = false;
+          break;
+        }
+        canOptAwlays = true;
+      }
+    }
+    return canOptAwlays;
+  };
+
   if (Ty.isScalarInteger() &&
       (N0.getOpcode() == ISD::ADD || N0.getOpcode() == ISD::OR)) {
+    if (N0.getOpcode() == ISD::ADD && !N0->hasOneUse()) {
+      return isLDST();
+    }
     auto *C1 = dyn_cast<ConstantSDNode>(N0->getOperand(1));
     auto *C2 = dyn_cast<ConstantSDNode>(N->getOperand(1));
     if (C1 && C2) {
@@ -18019,6 +18057,16 @@ bool RISCVTargetLowering::isDesirableToCommuteWithShift(
         return false;
     }
   }
+
+  if ((N0->getOpcode() == ISD::ADD || N0->getOpcode() == ISD::OR) &&
+      !N0->hasOneUse())
+    return false;
+
+  if (N0->getOpcode() == ISD::SIGN_EXTEND &&
+      N0->getOperand(0)->getOpcode() == ISD::ADD &&
+      !(N0->hasOneUse() && N0->getOperand(0)->hasOneUse()))
+    return isLDST();
+
   return true;
 }
 
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 1c790f3813b7a4..d5e82a453803f7 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -60225,3 +60225,19 @@ Align X86TargetLowering::getPrefLoopAlignment(MachineLoop *ML) const {
     return Align(1ULL << ExperimentalPrefInnermostLoopAlignment);
   return TargetLowering::getPrefLoopAlignment();
 }
+
+bool X86TargetLowering::isDesirableToCommuteWithShift(
+    const SDNode *N, CombineLevel Level) const {
+  using namespace llvm::SDPatternMatch;
+  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
+          N->getOpcode() == ISD::SRL) &&
+         "Expected shift op");
+
+  SDValue ShiftLHS = N->getOperand(0);
+  SDValue Add;
+  if (sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))) ||
+      ShiftLHS->hasOneUse())
+    return true;
+
+  return false;
+}
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index 14ada1721fd40e..ae1a3a10a74b00 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -1904,6 +1904,9 @@ namespace llvm {
 
     SDValue getMOVL(SelectionDAG &DAG, const SDLoc &dl, MVT VT, SDValue V1,
                     SDValue V2) const;
+
+    bool isDesirableToCommuteWithShift(const SDNode *N,
+                                       CombineLevel Level) const override;
   };
 
   namespace X86 {
diff --git a/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
index 35f3656e868681..891230c10cc460 100644
--- a/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
+++ b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
@@ -5,13 +5,11 @@
 define void @add_sext_shl_moreOneUse_add(ptr %array1, i32 %a, i32 %b) {
 ; RV64-LABEL: add_sext_shl_moreOneUse_add:
 ; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addiw a3, a1, 5
-; RV64-NEXT:    slli a4, a3, 2
-; RV64-NEXT:    add a4, a0, a4
-; RV64-NEXT:    sw a2, 0(a4)
+; RV64-NEXT:    addi a3, a1, 5
 ; RV64-NEXT:    sext.w a1, a1
 ; RV64-NEXT:    slli a1, a1, 2
 ; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sw a2, 20(a0)
 ; RV64-NEXT:    sw a2, 24(a0)
 ; RV64-NEXT:    sw a3, 140(a0)
 ; RV64-NEXT:    ret
@@ -35,15 +33,13 @@ define void @add_sext_shl_moreOneUse_addexceedsign12(ptr %array1, i32 %a, i32 %b
 ; RV64-LABEL: add_sext_shl_moreOneUse_addexceedsign12:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi a3, a1, 2047
-; RV64-NEXT:    addiw a3, a3, 1
-; RV64-NEXT:    slli a4, a3, 2
-; RV64-NEXT:    add a4, a0, a4
-; RV64-NEXT:    sw a2, 0(a4)
+; RV64-NEXT:    addi a3, a3, 1
+; RV64-NEXT:    lui a4, 2
 ; RV64-NEXT:    sext.w a1, a1
 ; RV64-NEXT:    slli a1, a1, 2
-; RV64-NEXT:    lui a4, 2
 ; RV64-NEXT:    add a0, a0, a4
 ; RV64-NEXT:    add a0, a0, a1
+; RV64-NEXT:    sw a2, 0(a0)
 ; RV64-NEXT:    sw a3, 4(a0)
 ; RV64-NEXT:    sw a2, 120(a0)
 ; RV64-NEXT:    ret
@@ -64,13 +60,11 @@ entry:
 define void @add_sext_shl_moreOneUse_sext(ptr %array1, i32 %a, i32 %b) {
 ; RV64-LABEL: add_sext_shl_moreOneUse_sext:
 ; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addiw a3, a1, 5
-; RV64-NEXT:    slli a4, a3, 2
-; RV64-NEXT:    add a4, a0, a4
-; RV64-NEXT:    sw a2, 0(a4)
 ; RV64-NEXT:    sext.w a1, a1
+; RV64-NEXT:    addi a3, a1, 5
 ; RV64-NEXT:    slli a1, a1, 2
 ; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sw a2, 20(a0)
 ; RV64-NEXT:    sw a2, 24(a0)
 ; RV64-NEXT:    sd a3, 140(a0)
 ; RV64-NEXT:    ret
@@ -94,18 +88,16 @@ entry:
 define void @add_sext_shl_moreOneUse_add_inSelect(ptr %array1, i32 signext  %a, i32 %b, i32 signext %x) {
 ; RV64-LABEL: add_sext_shl_moreOneUse_add_inSelect:
 ; RV64:       # %bb.0: # %entry
-; RV64-NEXT:    addiw a4, a1, 5
-; RV64-NEXT:    slli a5, a4, 2
-; RV64-NEXT:    add a5, a0, a5
-; RV64-NEXT:    mv a6, a4
+; RV64-NEXT:    addi a4, a1, 5
+; RV64-NEXT:    mv a5, a4
 ; RV64-NEXT:    bgtz a3, .LBB3_2
 ; RV64-NEXT:  # %bb.1: # %entry
-; RV64-NEXT:    mv a6, a2
+; RV64-NEXT:    mv a5, a2
 ; RV64-NEXT:  .LBB3_2: # %entry
-; RV64-NEXT:    sw a6, 0(a5)
 ; RV64-NEXT:    slli a1, a1, 2
 ; RV64-NEXT:    add a0, a1, a0
-; RV64-NEXT:    sw a6, 24(a0)
+; RV64-NEXT:    sw a5, 20(a0)
+; RV64-NEXT:    sw a5, 24(a0)
 ; RV64-NEXT:    sw a4, 140(a0)
 ; RV64-NEXT:    ret
 entry:
@@ -130,20 +122,18 @@ define void @add_sext_shl_moreOneUse_add_inSelect_addexceedsign12(ptr %array1, i
 ; RV64-LABEL: add_sext_shl_moreOneUse_add_inSelect_addexceedsign12:
 ; RV64:       # %bb.0: # %entry
 ; RV64-NEXT:    addi a4, a1, 2047
-; RV64-NEXT:    addiw a4, a4, 1
-; RV64-NEXT:    slli a6, a4, 2
-; RV64-NEXT:    add a6, a0, a6
-; RV64-NEXT:    mv a5, a4
+; RV64-NEXT:    addi a4, a4, 1
+; RV64-NEXT:    lui a5, 2
+; RV64-NEXT:    slli a1, a1, 2
+; RV64-NEXT:    add a0, a0, a1
+; RV64-NEXT:    add a0, a0, a5
+; RV64-NEXT:    mv a1, a4
 ; RV64-NEXT:    bgtz a3, .LBB4_2
 ; RV64-NEXT:  # %bb.1: # %entry
-; RV64-NEXT:    mv a5, a2
+; RV64-NEXT:    mv a1, a2
 ; RV64-NEXT:  .LBB4_2: # %entry
-; RV64-NEXT:    sw a5, 0(a6)
-; RV64-NEXT:    slli a1, a1, 2
-; RV64-NEXT:    add a0, a0, a1
-; RV64-NEXT:    lui a1, 2
-; RV64-NEXT:    add a0, a0, a1
-; RV64-NEXT:    sw a5, 4(a0)
+; RV64-NEXT:    sw a1, 0(a0)
+; RV64-NEXT:    sw a1, 4(a0)
 ; RV64-NEXT:    sw a4, 120(a0)
 ; RV64-NEXT:    ret
 entry:
@@ -171,11 +161,9 @@ define void @add_shl_moreOneUse_inSelect(ptr %array1, i64 %a, i64 %b, i64 %x) {
 ; RV64-NEXT:  # %bb.1: # %entry
 ; RV64-NEXT:    mv a5, a2
 ; RV64-NEXT:  .LBB5_2: # %entry
-; RV64-NEXT:    slli a2, a4, 3
-; RV64-NEXT:    add a2, a0, a2
-; RV64-NEXT:    sd a5, 0(a2)
 ; RV64-NEXT:    slli a1, a1, 3
 ; RV64-NEXT:    add a0, a1, a0
+; RV64-NEXT:    sd a5, 40(a0)
 ; RV64-NEXT:    sd a5, 48(a0)
 ; RV64-NEXT:    sd a4, 280(a0)
 ; RV64-NEXT:    ret
diff --git a/llvm/test/CodeGen/RISCV/add_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
index 5c71a3c5449940..d60fe73fb20bff 100644
--- a/llvm/test/CodeGen/RISCV/add_shl_constant.ll
+++ b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
@@ -19,11 +19,9 @@ define void @add_shl_moreOneUse_inStore(ptr %array1, i32 %a, i32 %b)  {
 ; RV32-LABEL: add_shl_moreOneUse_inStore:
 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    addi a3, a1, 5
-; RV32-NEXT:    slli a4, a3, 2
-; RV32-NEXT:    add a4, a0, a4
-; RV32-NEXT:    sw a2, 0(a4)
 ; RV32-NEXT:    slli a1, a1, 2
 ; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a2, 20(a0)
 ; RV32-NEXT:    sw a2, 24(a0)
 ; RV32-NEXT:    sw a3, 140(a0)
 ; RV32-NEXT:    ret
@@ -44,13 +42,11 @@ define void @add_shl_moreOneUse_inStore_addexceedsign12(ptr %array1, i32 %a, i32
 ; RV32:       # %bb.0: # %entry
 ; RV32-NEXT:    addi a3, a1, 2047
 ; RV32-NEXT:    addi a3, a3, 1
-; RV32-NEXT:    slli a4, a3, 2
-; RV32-NEXT:    add a4, a0, a4
-; RV32-NEXT:    sw a2, 0(a4)
+; RV32-NEXT:    lui a4, 2
 ; RV32-NEXT:    slli a1, a1, 2
 ; RV32-NEXT:    add a0, a0, a1
-; RV32-NEXT:    lui a1, 2
-; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    add a0, a0, a4
+; RV32-NEXT:    sw a2, 0(a0)
 ; RV32-NEXT:    sw a3, 4(a0)
 ; RV32-NEXT:    sw a2, 120(a0)
 ; RV32-NEXT:    ret
@@ -75,11 +71,9 @@ define void @add_shl_moreOneUse_inSelect(ptr %array1, i32 %a, i32 %b, i32 %x) {
 ; RV32-NEXT:  # %bb.1: # %entry
 ; RV32-NEXT:    mv a5, a2
 ; RV32-NEXT:  .LBB3_2: # %entry
-; RV32-NEXT:    slli a2, a4, 2
-; RV32-NEXT:    add a2, a0, a2
-; RV32-NEXT:    sw a5, 0(a2)
 ; RV32-NEXT:    slli a1, a1, 2
 ; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    sw a5, 20(a0)
 ; RV32-NEXT:    sw a5, 24(a0)
 ; RV32-NEXT:    sw a4, 140(a0)
 ; RV32-NEXT:    ret
@@ -107,13 +101,11 @@ define void @add_shl_moreOneUse_inSelect_addexceedsign12(ptr %array1, i32 %a, i3
 ; RV32-NEXT:  # %bb.1: # %entry
 ; RV32-NEXT:    mv a5, a2
 ; RV32-NEXT:  .LBB4_2: # %entry
-; RV32-NEXT:    slli a2, a4, 2
-; RV32-NEXT:    add a2, a0, a2
-; RV32-NEXT:    sw a5, 0(a2)
+; RV32-NEXT:    lui a2, 2
 ; RV32-NEXT:    slli a1, a1, 2
 ; RV32-NEXT:    add a0, a0, a1
-; RV32-NEXT:    lui a1, 2
-; RV32-NEXT:    add a0, a0, a1
+; RV32-NEXT:    add a0, a0, a2
+; RV32-NEXT:    sw a5, 0(a0)
 ; RV32-NEXT:    sw a5, 4(a0)
 ; RV32-NEXT:    sw a4, 120(a0)
 ; RV32-NEXT:    ret

>From bccad663018af34406e5ab5754bcb00efcdc42e9 Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Fri, 1 Nov 2024 10:41:19 +0800
Subject: [PATCH 3/7] fix the comments

---
 llvm/lib/Target/AArch64/AArch64ISelLowering.cpp |  8 ++++++--
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp   |  6 ++++--
 llvm/lib/Target/ARM/ARMISelLowering.cpp         |  6 ++++--
 llvm/lib/Target/Hexagon/HexagonISelLowering.cpp | 13 ++++++++-----
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp     | 13 ++++++++-----
 llvm/lib/Target/X86/X86ISelLowering.cpp         |  5 +++--
 6 files changed, 33 insertions(+), 18 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 5b1d8209e766c6..19e9349a97b85b 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -17911,10 +17911,13 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
 
   SDValue ShiftLHS = N->getOperand(0);
   EVT VT = N->getValueType(0);
+  SDValue Add;
+
+  if (!ShiftLHS->hasOneUse())
+    return false;
 
   if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
-      !ShiftLHS->hasOneUse())
+       !ShiftLHS.getOperand(0)->hasOneUse()))
     return false;
 
   // If ShiftLHS is unsigned bit extraction: ((x >> C) & mask), then do not
@@ -17935,6 +17938,7 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
       }
     }
   }
+
   return true;
 }
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index fc439e0be578a2..1cb344355d3229 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -1074,9 +1074,11 @@ bool AMDGPUTargetLowering::isDesirableToCommuteWithShift(
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
+  if (!ShiftLHS->hasOneUse())
+    return false;
+
   if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
-      !ShiftLHS->hasOneUse())
+       !ShiftLHS.getOperand(0)->hasOneUse()))
     return false;
 
   // Always commute pre-type legalization and right shifts.
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index 13a4008bf32f23..a4634002d5b733 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -13884,9 +13884,11 @@ ARMTargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
+  if (!ShiftLHS->hasOneUse())
+    return false;
+
   if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
-      !ShiftLHS->hasOneUse())
+       !ShiftLHS.getOperand(0)->hasOneUse()))
     return false;
 
   if (Level == BeforeLegalizeTypes)
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index c1f7253353b2a6..21b5d5348f392a 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -28,6 +28,7 @@
 #include "llvm/CodeGen/MachineMemOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RuntimeLibcallUtil.h"
+#include "llvm/CodeGen/SDPatternMatch.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/CodeGen/TargetCallingConv.h"
 #include "llvm/CodeGen/ValueTypes.h"
@@ -2158,17 +2159,19 @@ bool HexagonTargetLowering::hasBitTest(SDValue X, SDValue Y) const {
 
 bool HexagonTargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
+  using namespace llvm::SDPatternMatch;
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
-      !ShiftLHS->hasOneUse())
-    return false;
+  SDValue Add;
 
-  return true;
+  if (ShiftLHS->hasOneUse() ||
+      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
+    return true;
+
+  return false;
 }
 
 bool HexagonTargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const {
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index a529701201ccdf..553a7a7fbf43ec 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -49,6 +49,7 @@
 #include "llvm/CodeGen/MachineOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RuntimeLibcallUtil.h"
+#include "llvm/CodeGen/SDPatternMatch.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/CodeGen/SelectionDAGNodes.h"
 #include "llvm/CodeGen/TargetInstrInfo.h"
@@ -19140,15 +19141,17 @@ Value *PPCTargetLowering::emitMaskedAtomicCmpXchgIntrinsic(
 
 bool PPCTargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
+  using namespace llvm::SDPatternMatch;
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !(ShiftLHS->hasOneUse() && ShiftLHS.getOperand(0)->hasOneUse())) ||
-      !ShiftLHS->hasOneUse())
-    return false;
+  SDValue Add;
 
-  return true;
+  if (ShiftLHS->hasOneUse() ||
+      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
+    return true;
+
+  return false;
 }
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index d5e82a453803f7..3144c99bf4de17 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -60235,8 +60235,9 @@ bool X86TargetLowering::isDesirableToCommuteWithShift(
 
   SDValue ShiftLHS = N->getOperand(0);
   SDValue Add;
-  if (sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))) ||
-      ShiftLHS->hasOneUse())
+
+  if (ShiftLHS->hasOneUse() ||
+      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
     return true;
 
   return false;

>From 04a843e531b55d66953cd640c7ae71b7a8a1294e Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Tue, 5 Nov 2024 12:27:50 +0800
Subject: [PATCH 4/7] fix some comments

---
 llvm/lib/Target/AArch64/AArch64ISelLowering.cpp  |  2 --
 llvm/lib/Target/Hexagon/HexagonISelLowering.cpp  | 14 +++++++-------
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp      | 13 ++++++-------
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp      |  5 ++---
 llvm/lib/Target/X86/X86ISelLowering.cpp          | 12 ++++++------
 llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll |  3 +--
 llvm/test/CodeGen/RISCV/add_shl_constant.ll      |  3 +--
 7 files changed, 23 insertions(+), 29 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index c22893702d5c25..e4e92300f86651 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -17904,7 +17904,6 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
 
   SDValue ShiftLHS = N->getOperand(0);
   EVT VT = N->getValueType(0);
-  SDValue Add;
 
   if (!ShiftLHS->hasOneUse())
     return false;
@@ -17931,7 +17930,6 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
       }
     }
   }
-
   return true;
 }
 
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index 21b5d5348f392a..87be876bb21fce 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -28,7 +28,6 @@
 #include "llvm/CodeGen/MachineMemOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RuntimeLibcallUtil.h"
-#include "llvm/CodeGen/SDPatternMatch.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/CodeGen/TargetCallingConv.h"
 #include "llvm/CodeGen/ValueTypes.h"
@@ -2159,19 +2158,20 @@ bool HexagonTargetLowering::hasBitTest(SDValue X, SDValue Y) const {
 
 bool HexagonTargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
-  using namespace llvm::SDPatternMatch;
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
-  SDValue Add;
 
-  if (ShiftLHS->hasOneUse() ||
-      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
-    return true;
+  if (!ShiftLHS->hasOneUse())
+    return false;
 
-  return false;
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !ShiftLHS.getOperand(0)->hasOneUse()))
+    return false;
+
+  return true;
 }
 
 bool HexagonTargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const {
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 553a7a7fbf43ec..740fa30425cea1 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -49,7 +49,6 @@
 #include "llvm/CodeGen/MachineOperand.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RuntimeLibcallUtil.h"
-#include "llvm/CodeGen/SDPatternMatch.h"
 #include "llvm/CodeGen/SelectionDAG.h"
 #include "llvm/CodeGen/SelectionDAGNodes.h"
 #include "llvm/CodeGen/TargetInstrInfo.h"
@@ -19141,17 +19140,17 @@ Value *PPCTargetLowering::emitMaskedAtomicCmpXchgIntrinsic(
 
 bool PPCTargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
-  using namespace llvm::SDPatternMatch;
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
-  SDValue Add;
+  if (!ShiftLHS->hasOneUse())
+    return false;
 
-  if (ShiftLHS->hasOneUse() ||
-      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
-    return true;
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !ShiftLHS.getOperand(0)->hasOneUse()))
+    return false;
 
-  return false;
+  return true;
 }
diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 24c94de396ad8f..e736743cc954aa 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -18099,13 +18099,12 @@ bool RISCVTargetLowering::isDesirableToCommuteWithShift(
     }
   }
 
-  if ((N0->getOpcode() == ISD::ADD || N0->getOpcode() == ISD::OR) &&
-      !N0->hasOneUse())
+  if (!N0->hasOneUse())
     return false;
 
   if (N0->getOpcode() == ISD::SIGN_EXTEND &&
       N0->getOperand(0)->getOpcode() == ISD::ADD &&
-      !(N0->hasOneUse() && N0->getOperand(0)->hasOneUse()))
+      !N0->getOperand(0)->hasOneUse())
     return isLDST();
 
   return true;
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index ced4ad052ca15d..8196141a405aea 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -60247,17 +60247,17 @@ Align X86TargetLowering::getPrefLoopAlignment(MachineLoop *ML) const {
 
 bool X86TargetLowering::isDesirableToCommuteWithShift(
     const SDNode *N, CombineLevel Level) const {
-  using namespace llvm::SDPatternMatch;
   assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
           N->getOpcode() == ISD::SRL) &&
          "Expected shift op");
 
   SDValue ShiftLHS = N->getOperand(0);
-  SDValue Add;
+  if (!ShiftLHS->hasOneUse())
+    return false;
 
-  if (ShiftLHS->hasOneUse() ||
-      sd_match(ShiftLHS, m_OneUse(m_SExt(m_OneUse(m_Value(Add))))))
-    return true;
+  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+       !ShiftLHS.getOperand(0)->hasOneUse()))
+    return false;
 
-  return false;
+  return true;
 }
diff --git a/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
index 891230c10cc460..2c933e68e8307b 100644
--- a/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
+++ b/llvm/test/CodeGen/RISCV/add_sext_shl_constant.ll
@@ -1,6 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 3
-; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s \
-; RUN:   | FileCheck -check-prefix=RV64 %s
+; RUN: llc -mtriple=riscv64 < %s | FileCheck -check-prefix=RV64 %s
 
 define void @add_sext_shl_moreOneUse_add(ptr %array1, i32 %a, i32 %b) {
 ; RV64-LABEL: add_sext_shl_moreOneUse_add:
diff --git a/llvm/test/CodeGen/RISCV/add_shl_constant.ll b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
index d60fe73fb20bff..6a10147ff54b91 100644
--- a/llvm/test/CodeGen/RISCV/add_shl_constant.ll
+++ b/llvm/test/CodeGen/RISCV/add_shl_constant.ll
@@ -1,6 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
-; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s \
-; RUN:   | FileCheck -check-prefix=RV32 %s
+; RUN: llc -mtriple=riscv32  < %s | FileCheck -check-prefix=RV32 %s
 
 define i32 @add_shl_oneUse(i32 %x, i32 %y) nounwind {
 ; RV32-LABEL: add_shl_oneUse:

>From f2cbb92bc83700fb8972b9a04f73e2ba90ba48b4 Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Mon, 11 Nov 2024 10:19:36 +0800
Subject: [PATCH 5/7] fix the comments

---
 llvm/lib/Target/AArch64/AArch64ISelLowering.cpp | 4 ++--
 llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp   | 4 ++--
 llvm/lib/Target/ARM/ARMISelLowering.cpp         | 4 ++--
 llvm/lib/Target/Hexagon/HexagonISelLowering.cpp | 4 ++--
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp     | 4 ++--
 llvm/lib/Target/X86/X86ISelLowering.cpp         | 4 ++--
 6 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index cd40ccbd966e37..c3282b1875af0a 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -17913,8 +17913,8 @@ AArch64TargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   // If ShiftLHS is unsigned bit extraction: ((x >> C) & mask), then do not
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 685f0c1be9b2c9..c5d473ac2b42e7 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -1078,8 +1078,8 @@ bool AMDGPUTargetLowering::isDesirableToCommuteWithShift(
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   // Always commute pre-type legalization and right shifts.
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index c8c4e6975b1d01..e7c53727923a4b 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -13889,8 +13889,8 @@ ARMTargetLowering::isDesirableToCommuteWithShift(const SDNode *N,
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   if (Level == BeforeLegalizeTypes)
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index 87be876bb21fce..91098511c18452 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -2167,8 +2167,8 @@ bool HexagonTargetLowering::isDesirableToCommuteWithShift(
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   return true;
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 2886a0a9a34e50..5c3db6f5a7fa65 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -19115,8 +19115,8 @@ bool PPCTargetLowering::isDesirableToCommuteWithShift(
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   return true;
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 8b84d3e1681910..5b279e4e831336 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -60479,8 +60479,8 @@ bool X86TargetLowering::isDesirableToCommuteWithShift(
   if (!ShiftLHS->hasOneUse())
     return false;
 
-  if ((ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-       !ShiftLHS.getOperand(0)->hasOneUse()))
+  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+      !ShiftLHS.getOperand(0)->hasOneUse())
     return false;
 
   return true;

>From 76a397f2efd5a8b149a6c27dbb6a4cc9fea371f5 Mon Sep 17 00:00:00 2001
From: "Liqin.Weng" <liqin.weng at spacemit.com>
Date: Thu, 14 Nov 2024 09:51:47 +0800
Subject: [PATCH 6/7] fix the comments

---
 llvm/lib/Target/RISCV/RISCVISelLowering.cpp | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)

diff --git a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
index 60f69508e4cb13..dbf2b1fcbe2c8b 100644
--- a/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVISelLowering.cpp
@@ -18046,17 +18046,17 @@ bool RISCVTargetLowering::isDesirableToCommuteWithShift(
   // LD/ST, it can still complete the folding optimization operation performed
   // above.
   auto isLDST = [&]() {
-    bool canOptAwlays = false;
+    bool CanOptAlways = false;
     if (N0->getOpcode() == ISD::ADD && !N0->hasOneUse()) {
       for (SDNode *Use : N0->uses()) {
         // This use is the one we're on right now. Skip it
         if (Use == N || Use->getOpcode() == ISD::SELECT)
           continue;
         if (!isa<StoreSDNode>(Use) && !isa<LoadSDNode>(Use)) {
-          canOptAwlays = false;
+          CanOptAlways = false;
           break;
         }
-        canOptAwlays = true;
+        CanOptAlways = true;
       }
     }
 
@@ -18067,20 +18067,20 @@ bool RISCVTargetLowering::isDesirableToCommuteWithShift(
         if (Use == N0.getNode() || Use->getOpcode() == ISD::SELECT)
           continue;
         if (!isa<StoreSDNode>(Use) && !isa<LoadSDNode>(Use)) {
-          canOptAwlays = false;
+          CanOptAlways = false;
           break;
         }
-        canOptAwlays = true;
+        CanOptAlways = true;
       }
     }
-    return canOptAwlays;
+    return CanOptAlways;
   };
 
   if (Ty.isScalarInteger() &&
       (N0.getOpcode() == ISD::ADD || N0.getOpcode() == ISD::OR)) {
-    if (N0.getOpcode() == ISD::ADD && !N0->hasOneUse()) {
+    if (N0.getOpcode() == ISD::ADD && !N0->hasOneUse())
       return isLDST();
-    }
+
     auto *C1 = dyn_cast<ConstantSDNode>(N0->getOperand(1));
     auto *C2 = dyn_cast<ConstantSDNode>(N->getOperand(1));
     if (C1 && C2) {

>From eb29b6dd98486e233f9871dd09cffa54961a89d1 Mon Sep 17 00:00:00 2001
From: LiqinWeng <liqin.weng at spacemit.com>
Date: Sun, 24 Nov 2024 11:36:17 +0800
Subject: [PATCH 7/7] fix the comments

---
 llvm/include/llvm/CodeGen/TargetLowering.h    |   6 +
 .../Target/Hexagon/HexagonISelLowering.cpp    |  18 --
 llvm/lib/Target/Hexagon/HexagonISelLowering.h |   3 -
 llvm/lib/Target/PowerPC/PPCISelLowering.cpp   |  17 --
 llvm/lib/Target/PowerPC/PPCISelLowering.h     |   3 -
 llvm/test/CodeGen/AArch64/tbl-loops.ll        | 163 ++----------------
 6 files changed, 19 insertions(+), 191 deletions(-)

diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index e0b638201a0474..f52b4b53124b02 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -4297,6 +4297,12 @@ class TargetLowering : public TargetLoweringBase {
   /// @param Level the current DAGCombine legalization level.
   virtual bool isDesirableToCommuteWithShift(const SDNode *N,
                                              CombineLevel Level) const {
+    SDValue ShiftLHS = N->getOperand(0);
+    if (!ShiftLHS->hasOneUse())
+      return false;
+    if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
+        !ShiftLHS.getOperand(0)->hasOneUse())
+      return false;
     return true;
   }
 
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
index 91098511c18452..ab9bc559367787 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.cpp
@@ -2156,24 +2156,6 @@ bool HexagonTargetLowering::hasBitTest(SDValue X, SDValue Y) const {
   return X.getValueType().isScalarInteger(); // 'tstbit'
 }
 
-bool HexagonTargetLowering::isDesirableToCommuteWithShift(
-    const SDNode *N, CombineLevel Level) const {
-  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
-          N->getOpcode() == ISD::SRL) &&
-         "Expected shift op");
-
-  SDValue ShiftLHS = N->getOperand(0);
-
-  if (!ShiftLHS->hasOneUse())
-    return false;
-
-  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-      !ShiftLHS.getOperand(0)->hasOneUse())
-    return false;
-
-  return true;
-}
-
 bool HexagonTargetLowering::isTruncateFree(Type *Ty1, Type *Ty2) const {
   return isTruncateFree(EVT::getEVT(Ty1), EVT::getEVT(Ty2));
 }
diff --git a/llvm/lib/Target/Hexagon/HexagonISelLowering.h b/llvm/lib/Target/Hexagon/HexagonISelLowering.h
index a6bd57630031c4..3fd961f5a74623 100644
--- a/llvm/lib/Target/Hexagon/HexagonISelLowering.h
+++ b/llvm/lib/Target/Hexagon/HexagonISelLowering.h
@@ -155,9 +155,6 @@ class HexagonTargetLowering : public TargetLowering {
 
   bool hasBitTest(SDValue X, SDValue Y) const override;
 
-  bool isDesirableToCommuteWithShift(const SDNode *N,
-                                     CombineLevel Level) const override;
-
   bool allowTruncateForTailCall(Type *Ty1, Type *Ty2) const override;
 
   /// Return true if an FMA operation is faster than a pair of mul and add
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
index 5c3db6f5a7fa65..ec4f8f4be425ed 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -19104,20 +19104,3 @@ Value *PPCTargetLowering::emitMaskedAtomicCmpXchgIntrinsic(
   return Builder.CreateOr(
       Lo, Builder.CreateShl(Hi, ConstantInt::get(ValTy, 64)), "val64");
 }
-
-bool PPCTargetLowering::isDesirableToCommuteWithShift(
-    const SDNode *N, CombineLevel Level) const {
-  assert((N->getOpcode() == ISD::SHL || N->getOpcode() == ISD::SRA ||
-          N->getOpcode() == ISD::SRL) &&
-         "Expected shift op");
-
-  SDValue ShiftLHS = N->getOperand(0);
-  if (!ShiftLHS->hasOneUse())
-    return false;
-
-  if (ShiftLHS.getOpcode() == ISD::SIGN_EXTEND &&
-      !ShiftLHS.getOperand(0)->hasOneUse())
-    return false;
-
-  return true;
-}
diff --git a/llvm/lib/Target/PowerPC/PPCISelLowering.h b/llvm/lib/Target/PowerPC/PPCISelLowering.h
index 1d36a68960bca4..0adbad86845973 100644
--- a/llvm/lib/Target/PowerPC/PPCISelLowering.h
+++ b/llvm/lib/Target/PowerPC/PPCISelLowering.h
@@ -1491,9 +1491,6 @@ namespace llvm {
     /// through to determine the optimal load/store instruction format.
     unsigned computeMOFlags(const SDNode *Parent, SDValue N,
                             SelectionDAG &DAG) const;
-
-    bool isDesirableToCommuteWithShift(const SDNode *N,
-                                       CombineLevel Level) const override;
   }; // end class PPCTargetLowering
 
   namespace PPC {
diff --git a/llvm/test/CodeGen/AArch64/tbl-loops.ll b/llvm/test/CodeGen/AArch64/tbl-loops.ll
index 0ad99008655184..4bbadd0be3d543 100644
--- a/llvm/test/CodeGen/AArch64/tbl-loops.ll
+++ b/llvm/test/CodeGen/AArch64/tbl-loops.ll
@@ -1,145 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple=aarch64 < %s | FileCheck %s
 
-define void @loop1(ptr noalias nocapture noundef writeonly %dst, ptr nocapture noundef readonly %data, i32 noundef %width) {
-; CHECK-LABEL: loop1:
-; CHECK:       // %bb.0: // %entry
-; CHECK-NEXT:    subs w8, w2, #1
-; CHECK-NEXT:    b.lt .LBB0_8
-; CHECK-NEXT:  // %bb.1: // %for.body.preheader
-; CHECK-NEXT:    cmp w8, #6
-; CHECK-NEXT:    b.hi .LBB0_3
-; CHECK-NEXT:  // %bb.2:
-; CHECK-NEXT:    mov w10, wzr
-; CHECK-NEXT:    mov x8, x1
-; CHECK-NEXT:    mov x9, x0
-; CHECK-NEXT:    b .LBB0_6
-; CHECK-NEXT:  .LBB0_3: // %vector.ph
-; CHECK-NEXT:    add x11, x8, #1
-; CHECK-NEXT:    mov w8, #1132396544 // =0x437f0000
-; CHECK-NEXT:    add x12, x0, #4
-; CHECK-NEXT:    and x10, x11, #0x1fffffff8
-; CHECK-NEXT:    dup v0.4s, w8
-; CHECK-NEXT:    add x13, x1, #16
-; CHECK-NEXT:    add x8, x1, x10, lsl #2
-; CHECK-NEXT:    add x9, x0, x10
-; CHECK-NEXT:    mov x14, x10
-; CHECK-NEXT:  .LBB0_4: // %vector.body
-; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    ldp q1, q2, [x13, #-16]
-; CHECK-NEXT:    subs x14, x14, #8
-; CHECK-NEXT:    add x13, x13, #32
-; CHECK-NEXT:    fcmgt v3.4s, v1.4s, v0.4s
-; CHECK-NEXT:    fcmgt v4.4s, v2.4s, v0.4s
-; CHECK-NEXT:    fcmlt v5.4s, v1.4s, #0.0
-; CHECK-NEXT:    fcmlt v6.4s, v2.4s, #0.0
-; CHECK-NEXT:    bit v1.16b, v0.16b, v3.16b
-; CHECK-NEXT:    bit v2.16b, v0.16b, v4.16b
-; CHECK-NEXT:    bic v1.16b, v1.16b, v5.16b
-; CHECK-NEXT:    bic v2.16b, v2.16b, v6.16b
-; CHECK-NEXT:    fcvtzs v1.4s, v1.4s
-; CHECK-NEXT:    fcvtzs v2.4s, v2.4s
-; CHECK-NEXT:    xtn v1.4h, v1.4s
-; CHECK-NEXT:    xtn v2.4h, v2.4s
-; CHECK-NEXT:    uzp1 v1.8b, v1.8b, v0.8b
-; CHECK-NEXT:    uzp1 v2.8b, v2.8b, v0.8b
-; CHECK-NEXT:    mov v1.s[1], v2.s[0]
-; CHECK-NEXT:    stur d1, [x12, #-4]
-; CHECK-NEXT:    add x12, x12, #8
-; CHECK-NEXT:    b.ne .LBB0_4
-; CHECK-NEXT:  // %bb.5: // %middle.block
-; CHECK-NEXT:    cmp x11, x10
-; CHECK-NEXT:    b.eq .LBB0_8
-; CHECK-NEXT:  .LBB0_6: // %for.body.preheader1
-; CHECK-NEXT:    movi d0, #0000000000000000
-; CHECK-NEXT:    mov w11, #1132396544 // =0x437f0000
-; CHECK-NEXT:    sub w10, w2, w10
-; CHECK-NEXT:    fmov s1, w11
-; CHECK-NEXT:  .LBB0_7: // %for.body
-; CHECK-NEXT:    // =>This Inner Loop Header: Depth=1
-; CHECK-NEXT:    ldr s2, [x8], #4
-; CHECK-NEXT:    fcmp s2, s1
-; CHECK-NEXT:    fcsel s3, s1, s2, gt
-; CHECK-NEXT:    fcmp s2, #0.0
-; CHECK-NEXT:    fcsel s2, s0, s3, mi
-; CHECK-NEXT:    subs w10, w10, #1
-; CHECK-NEXT:    fcvtzs w11, s2
-; CHECK-NEXT:    strb w11, [x9], #1
-; CHECK-NEXT:    b.ne .LBB0_7
-; CHECK-NEXT:  .LBB0_8: // %for.cond.cleanup
-; CHECK-NEXT:    ret
-entry:
-  %cmp9 = icmp sgt i32 %width, 0
-  br i1 %cmp9, label %for.body.preheader, label %for.cond.cleanup
 
-for.body.preheader:                               ; preds = %entry
-  %0 = add i32 %width, -1
-  %1 = zext i32 %0 to i64
-  %2 = add nuw nsw i64 %1, 1
-  %min.iters.check = icmp ult i32 %0, 7
-  br i1 %min.iters.check, label %for.body.preheader21, label %vector.ph
-
-vector.ph:                                        ; preds = %for.body.preheader
-  %n.vec = and i64 %2, 8589934584
-  %ind.end = trunc i64 %n.vec to i32
-  %ind.end14 = getelementptr float, ptr %data, i64 %n.vec
-  %ind.end16 = getelementptr i8, ptr %dst, i64 %n.vec
-  br label %vector.body
-
-vector.body:                                      ; preds = %vector.body, %vector.ph
-  %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
-  %next.gep = getelementptr float, ptr %data, i64 %index
-  %next.gep18 = getelementptr i8, ptr %dst, i64 %index
-  %wide.load = load <4 x float>, ptr %next.gep, align 4
-  %3 = getelementptr float, ptr %next.gep, i64 4
-  %wide.load20 = load <4 x float>, ptr %3, align 4
-  %4 = fcmp olt <4 x float> %wide.load, zeroinitializer
-  %5 = fcmp olt <4 x float> %wide.load20, zeroinitializer
-  %6 = fcmp ogt <4 x float> %wide.load, <float 2.550000e+02, float 2.550000e+02, float 2.550000e+02, float 2.550000e+02>
-  %7 = fcmp ogt <4 x float> %wide.load20, <float 2.550000e+02, float 2.550000e+02, float 2.550000e+02, float 2.550000e+02>
-  %8 = select <4 x i1> %6, <4 x float> <float 2.550000e+02, float 2.550000e+02, float 2.550000e+02, float 2.550000e+02>, <4 x float> %wide.load
-  %9 = select <4 x i1> %7, <4 x float> <float 2.550000e+02, float 2.550000e+02, float 2.550000e+02, float 2.550000e+02>, <4 x float> %wide.load20
-  %10 = select <4 x i1> %4, <4 x float> zeroinitializer, <4 x float> %8
-  %11 = select <4 x i1> %5, <4 x float> zeroinitializer, <4 x float> %9
-  %12 = fptoui <4 x float> %10 to <4 x i8>
-  %13 = fptoui <4 x float> %11 to <4 x i8>
-  store <4 x i8> %12, ptr %next.gep18, align 1
-  %14 = getelementptr i8, ptr %next.gep18, i64 4
-  store <4 x i8> %13, ptr %14, align 1
-  %index.next = add nuw i64 %index, 8
-  %15 = icmp eq i64 %index.next, %n.vec
-  br i1 %15, label %middle.block, label %vector.body
-
-middle.block:                                     ; preds = %vector.body
-  %cmp.n = icmp eq i64 %2, %n.vec
-  br i1 %cmp.n, label %for.cond.cleanup, label %for.body.preheader21
-
-for.body.preheader21:                             ; preds = %for.body.preheader, %middle.block
-  %i.012.ph = phi i32 [ 0, %for.body.preheader ], [ %ind.end, %middle.block ]
-  %src.011.ph = phi ptr [ %data, %for.body.preheader ], [ %ind.end14, %middle.block ]
-  %dst.addr.010.ph = phi ptr [ %dst, %for.body.preheader ], [ %ind.end16, %middle.block ]
-  br label %for.body
-
-for.cond.cleanup:                                 ; preds = %for.body, %middle.block, %entry
-  ret void
-
-for.body:                                         ; preds = %for.body.preheader21, %for.body
-  %i.012 = phi i32 [ %inc, %for.body ], [ %i.012.ph, %for.body.preheader21 ]
-  %src.011 = phi ptr [ %add.ptr, %for.body ], [ %src.011.ph, %for.body.preheader21 ]
-  %dst.addr.010 = phi ptr [ %add.ptr2, %for.body ], [ %dst.addr.010.ph, %for.body.preheader21 ]
-  %16 = load float, ptr %src.011, align 4
-  %cmp.i = fcmp olt float %16, 0.000000e+00
-  %cmp1.i = fcmp ogt float %16, 2.550000e+02
-  %.x.i = select i1 %cmp1.i, float 2.550000e+02, float %16
-  %retval.0.i = select i1 %cmp.i, float 0.000000e+00, float %.x.i
-  %conv = fptoui float %retval.0.i to i8
-  store i8 %conv, ptr %dst.addr.010, align 1
-  %add.ptr = getelementptr inbounds float, ptr %src.011, i64 1
-  %add.ptr2 = getelementptr inbounds i8, ptr %dst.addr.010, i64 1
-  %inc = add nuw nsw i32 %i.012, 1
-  %exitcond.not = icmp eq i32 %inc, %width
-  br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
-}
 
 define void @loop2(ptr noalias nocapture noundef writeonly %dst, ptr nocapture noundef readonly %data, i32 noundef %width) {
 ; CHECK-LABEL: loop2:
@@ -150,13 +12,13 @@ define void @loop2(ptr noalias nocapture noundef writeonly %dst, ptr nocapture n
 ; CHECK-NEXT:    cmp w8, #2
 ; CHECK-NEXT:    b.ls .LBB1_4
 ; CHECK-NEXT:  // %bb.2: // %vector.memcheck
-; CHECK-NEXT:    ubfiz x9, x8, #1, #32
-; CHECK-NEXT:    add x9, x9, #2
-; CHECK-NEXT:    add x10, x1, x9, lsl #2
-; CHECK-NEXT:    cmp x10, x0
+; CHECK-NEXT:    add x9, x1, w8, uxtw #3
+; CHECK-NEXT:    add x9, x9, #8
+; CHECK-NEXT:    cmp x9, x0
 ; CHECK-NEXT:    b.ls .LBB1_8
 ; CHECK-NEXT:  // %bb.3: // %vector.memcheck
-; CHECK-NEXT:    add x9, x0, x9
+; CHECK-NEXT:    add x9, x0, w8, uxtw #1
+; CHECK-NEXT:    add x9, x9, #2
 ; CHECK-NEXT:    cmp x9, x1
 ; CHECK-NEXT:    b.ls .LBB1_8
 ; CHECK-NEXT:  .LBB1_4:
@@ -326,9 +188,10 @@ define void @loop3(ptr noalias nocapture noundef writeonly %dst, ptr nocapture n
 ; CHECK-NEXT:    b.ls .LBB2_6
 ; CHECK-NEXT:  // %bb.2: // %vector.memcheck
 ; CHECK-NEXT:    add x9, x8, w8, uxtw #1
-; CHECK-NEXT:    add x9, x9, #3
 ; CHECK-NEXT:    add x10, x1, x9, lsl #2
-; CHECK-NEXT:    add x9, x0, x9
+; CHECK-NEXT:    add x9, x9, x0
+; CHECK-NEXT:    add x9, x9, #3
+; CHECK-NEXT:    add x10, x10, #12
 ; CHECK-NEXT:    cmp x10, x0
 ; CHECK-NEXT:    ccmp x9, x1, #0, hi
 ; CHECK-NEXT:    b.hi .LBB2_6
@@ -534,13 +397,13 @@ define void @loop4(ptr noalias nocapture noundef writeonly %dst, ptr nocapture n
 ; CHECK-NEXT:    cmp w8, #2
 ; CHECK-NEXT:    b.ls .LBB3_4
 ; CHECK-NEXT:  // %bb.2: // %vector.memcheck
-; CHECK-NEXT:    ubfiz x9, x8, #2, #32
-; CHECK-NEXT:    add x9, x9, #4
-; CHECK-NEXT:    add x10, x1, x9, lsl #2
-; CHECK-NEXT:    cmp x10, x0
+; CHECK-NEXT:    add x9, x1, w8, uxtw #4
+; CHECK-NEXT:    add x9, x9, #16
+; CHECK-NEXT:    cmp x9, x0
 ; CHECK-NEXT:    b.ls .LBB3_8
 ; CHECK-NEXT:  // %bb.3: // %vector.memcheck
-; CHECK-NEXT:    add x9, x0, x9
+; CHECK-NEXT:    add x9, x0, w8, uxtw #2
+; CHECK-NEXT:    add x9, x9, #4
 ; CHECK-NEXT:    cmp x9, x1
 ; CHECK-NEXT:    b.ls .LBB3_8
 ; CHECK-NEXT:  .LBB3_4:



More information about the llvm-commits mailing list