[llvm] [DAG] Reassociate (add (add X, Y), X) --> add(add(X, X), Y) (PR #162242)
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Tue Oct 7 01:34:39 PDT 2025
https://github.com/RKSimon created https://github.com/llvm/llvm-project/pull/162242
Attempt to bring together self-additions, to help with folding to shift/mul/address patterns
>From 280a47128023ff56475a0b8db616044976d46461 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev at redking.me.uk>
Date: Tue, 7 Oct 2025 09:33:17 +0100
Subject: [PATCH] [DAG] Reassociate (add (add X, Y), X) --> add(add(X, X), Y)
Attempt to bring together self-additions, to help with folding to shift/mul/address patterns
---
llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp | 8 ++
.../AArch64/dag-ReplaceAllUsesOfValuesWith.ll | 5 +-
llvm/test/CodeGen/AMDGPU/idot2.ll | 122 ++++++++++--------
.../CodeGen/Hexagon/isel-fold-shl-zext.ll | 3 +-
llvm/test/CodeGen/X86/avx-vinsertf128.ll | 12 +-
5 files changed, 87 insertions(+), 63 deletions(-)
diff --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index 309f1bea8b77c..ea0aeffe7f02e 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -2926,6 +2926,14 @@ SDValue DAGCombiner::visitADDLike(SDNode *N) {
if (SDValue RADD = reassociateOps(ISD::ADD, DL, N0, N1, N->getFlags()))
return RADD;
+ // (X + Y) + X --> Y + (X + X)
+ SDValue X, Y;
+ if (sd_match(N, m_AddLike(m_OneUse(m_AddLike(m_Value(X), m_Value(Y))),
+ m_Deferred(X))))
+ if (X != Y)
+ return DAG.getNode(ISD::ADD, DL, VT, Y,
+ DAG.getNode(ISD::ADD, DL, VT, X, X));
+
// Reassociate (add (or x, c), y) -> (add add(x, y), c)) if (or x, c) is
// equivalent to (add x, c).
// Reassociate (add (xor x, c), y) -> (add add(x, y), c)) if (xor x, c) is
diff --git a/llvm/test/CodeGen/AArch64/dag-ReplaceAllUsesOfValuesWith.ll b/llvm/test/CodeGen/AArch64/dag-ReplaceAllUsesOfValuesWith.ll
index d76e817e62a49..ce657aa1f0b5b 100644
--- a/llvm/test/CodeGen/AArch64/dag-ReplaceAllUsesOfValuesWith.ll
+++ b/llvm/test/CodeGen/AArch64/dag-ReplaceAllUsesOfValuesWith.ll
@@ -27,10 +27,7 @@
define i64 @g(ptr %p) {
; CHECK-LABEL: g:
; CHECK: // %bb.0:
-; CHECK-NEXT: ldr x8, [x0, #8]
-; CHECK-NEXT: add x9, x8, x8
-; CHECK-NEXT: add x8, x9, x8
-; CHECK-NEXT: sub x0, x8, x8
+; CHECK-NEXT: mov x0, xzr
; CHECK-NEXT: ret
%vec = load <2 x i64>, ptr %p, align 1
%elt = extractelement <2 x i64> %vec, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/idot2.ll b/llvm/test/CodeGen/AMDGPU/idot2.ll
index b443e654350c5..32e26f4f387d1 100644
--- a/llvm/test/CodeGen/AMDGPU/idot2.ll
+++ b/llvm/test/CodeGen/AMDGPU/idot2.ll
@@ -1717,11 +1717,11 @@ define amdgpu_kernel void @udot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX7-NEXT: v_and_b32_e32 v0, 0xffff, v0
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mad_u32_u24 v1, v3, v1, s0
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX7-NEXT: v_add_i32_e32 v1, vcc, v1, v1
; GFX7-NEXT: v_mad_u32_u24 v0, v0, v2, v1
-; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v1
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GFX7-NEXT: s_endpgm
;
@@ -1748,8 +1748,8 @@ define amdgpu_kernel void @udot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: v_mad_u32_u24 v0, v0, v3, s0
-; GFX8-NEXT: v_mad_u32_u24 v1, v2, v1, v0
-; GFX8-NEXT: v_add_u32_e32 v2, vcc, v1, v0
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v0
+; GFX8-NEXT: v_mad_u32_u24 v2, v2, v1, v0
; GFX8-NEXT: v_mov_b32_e32 v0, s4
; GFX8-NEXT: v_mov_b32_e32 v1, s5
; GFX8-NEXT: flat_store_dword v[0:1], v2
@@ -1765,13 +1765,16 @@ define amdgpu_kernel void @udot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX9-NODL-NEXT: global_load_dword v2, v0, s[2:3]
; GFX9-NODL-NEXT: s_load_dword s0, s[6:7], 0x0
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NODL-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NODL-NEXT: v_and_b32_e32 v3, 0xffff, v1
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NODL-NEXT: v_mul_u32_u24_sdwa v3, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
+; GFX9-NODL-NEXT: v_and_b32_e32 v4, 0xffff, v2
; GFX9-NODL-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX9-NODL-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NODL-NEXT: v_mad_u32_u24 v1, v2, v1, s0
-; GFX9-NODL-NEXT: v_add3_u32 v1, v1, v3, v1
+; GFX9-NODL-NEXT: v_add_u32_e32 v1, v1, v1
+; GFX9-NODL-NEXT: v_mad_u32_u24 v1, v4, v3, v1
; GFX9-NODL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-NODL-NEXT: s_endpgm
;
@@ -1785,13 +1788,16 @@ define amdgpu_kernel void @udot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX9-DL-NEXT: global_load_dword v2, v0, s[2:3]
; GFX9-DL-NEXT: s_load_dword s0, s[6:7], 0x0
; GFX9-DL-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-DL-NEXT: s_waitcnt vmcnt(1)
+; GFX9-DL-NEXT: v_and_b32_e32 v3, 0xffff, v1
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
-; GFX9-DL-NEXT: v_mul_u32_u24_sdwa v3, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
+; GFX9-DL-NEXT: v_and_b32_e32 v4, 0xffff, v2
; GFX9-DL-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX9-DL-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-DL-NEXT: v_mad_u32_u24 v1, v2, v1, s0
-; GFX9-DL-NEXT: v_add3_u32 v1, v1, v3, v1
+; GFX9-DL-NEXT: v_add_u32_e32 v1, v1, v1
+; GFX9-DL-NEXT: v_mad_u32_u24 v1, v4, v3, v1
; GFX9-DL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-DL-NEXT: s_endpgm
;
@@ -1811,12 +1817,14 @@ define amdgpu_kernel void @udot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v0, 16, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 16, v2
-; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
-; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-DL-NEXT: v_and_b32_e32 v1, 0xffff, v1
+; GFX10-DL-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_mad_u32_u24 v0, v3, v0, s0
-; GFX10-DL-NEXT: v_add3_u32 v0, v0, v1, v0
-; GFX10-DL-NEXT: global_store_dword v2, v0, s[6:7]
+; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0
+; GFX10-DL-NEXT: v_add_nc_u32_e32 v0, v0, v0
+; GFX10-DL-NEXT: v_mad_u32_u24 v0, v2, v1, v0
+; GFX10-DL-NEXT: global_store_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: s_endpgm
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
@@ -1873,8 +1881,8 @@ define amdgpu_kernel void @idot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX7-NEXT: v_ashrrev_i32_e32 v0, 16, v0
; GFX7-NEXT: s_waitcnt lgkmcnt(0)
; GFX7-NEXT: v_mad_i32_i24 v0, v0, v2, s0
-; GFX7-NEXT: v_mad_i32_i24 v1, v3, v1, v0
-; GFX7-NEXT: v_add_i32_e32 v0, vcc, v1, v0
+; GFX7-NEXT: v_add_i32_e32 v0, vcc, v0, v0
+; GFX7-NEXT: v_mad_i32_i24 v0, v3, v1, v0
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GFX7-NEXT: s_endpgm
;
@@ -1901,8 +1909,8 @@ define amdgpu_kernel void @idot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX8-NEXT: v_ashrrev_i32_e32 v0, 16, v0
; GFX8-NEXT: s_waitcnt lgkmcnt(0)
; GFX8-NEXT: v_mad_i32_i24 v0, v0, v3, s0
-; GFX8-NEXT: v_mad_i32_i24 v1, v2, v1, v0
-; GFX8-NEXT: v_add_u32_e32 v2, vcc, v1, v0
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, v0, v0
+; GFX8-NEXT: v_mad_i32_i24 v2, v2, v1, v0
; GFX8-NEXT: v_mov_b32_e32 v0, s4
; GFX8-NEXT: v_mov_b32_e32 v1, s5
; GFX8-NEXT: flat_store_dword v[0:1], v2
@@ -1918,13 +1926,16 @@ define amdgpu_kernel void @idot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX9-NODL-NEXT: global_load_dword v2, v0, s[2:3]
; GFX9-NODL-NEXT: s_load_dword s0, s[6:7], 0x0
; GFX9-NODL-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-NODL-NEXT: s_waitcnt vmcnt(1)
+; GFX9-NODL-NEXT: v_bfe_i32 v3, v1, 0, 16
; GFX9-NODL-NEXT: s_waitcnt vmcnt(0)
-; GFX9-NODL-NEXT: v_mul_i32_i24_sdwa v3, sext(v2), sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
+; GFX9-NODL-NEXT: v_bfe_i32 v4, v2, 0, 16
; GFX9-NODL-NEXT: v_ashrrev_i32_e32 v1, 16, v1
; GFX9-NODL-NEXT: v_ashrrev_i32_e32 v2, 16, v2
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-NODL-NEXT: v_mad_i32_i24 v1, v2, v1, s0
-; GFX9-NODL-NEXT: v_add3_u32 v1, v1, v3, v1
+; GFX9-NODL-NEXT: v_add_u32_e32 v1, v1, v1
+; GFX9-NODL-NEXT: v_mad_i32_i24 v1, v4, v3, v1
; GFX9-NODL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-NODL-NEXT: s_endpgm
;
@@ -1938,13 +1949,16 @@ define amdgpu_kernel void @idot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX9-DL-NEXT: global_load_dword v2, v0, s[2:3]
; GFX9-DL-NEXT: s_load_dword s0, s[6:7], 0x0
; GFX9-DL-NEXT: v_mov_b32_e32 v0, 0
+; GFX9-DL-NEXT: s_waitcnt vmcnt(1)
+; GFX9-DL-NEXT: v_bfe_i32 v3, v1, 0, 16
; GFX9-DL-NEXT: s_waitcnt vmcnt(0)
-; GFX9-DL-NEXT: v_mul_i32_i24_sdwa v3, sext(v2), sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
+; GFX9-DL-NEXT: v_bfe_i32 v4, v2, 0, 16
; GFX9-DL-NEXT: v_ashrrev_i32_e32 v1, 16, v1
; GFX9-DL-NEXT: v_ashrrev_i32_e32 v2, 16, v2
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX9-DL-NEXT: v_mad_i32_i24 v1, v2, v1, s0
-; GFX9-DL-NEXT: v_add3_u32 v1, v1, v3, v1
+; GFX9-DL-NEXT: v_add_u32_e32 v1, v1, v1
+; GFX9-DL-NEXT: v_mad_i32_i24 v1, v4, v3, v1
; GFX9-DL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-DL-NEXT: s_endpgm
;
@@ -1964,12 +1978,14 @@ define amdgpu_kernel void @idot2_MultipleUses_add1(ptr addrspace(1) %src1,
; GFX10-DL-NEXT: v_ashrrev_i32_e32 v0, 16, v1
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_ashrrev_i32_e32 v3, 16, v2
-; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v2), sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
-; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-DL-NEXT: v_bfe_i32 v1, v1, 0, 16
+; GFX10-DL-NEXT: v_bfe_i32 v2, v2, 0, 16
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
; GFX10-DL-NEXT: v_mad_i32_i24 v0, v3, v0, s0
-; GFX10-DL-NEXT: v_add3_u32 v0, v0, v1, v0
-; GFX10-DL-NEXT: global_store_dword v2, v0, s[6:7]
+; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0
+; GFX10-DL-NEXT: v_add_nc_u32_e32 v0, v0, v0
+; GFX10-DL-NEXT: v_mad_i32_i24 v0, v2, v1, v0
+; GFX10-DL-NEXT: global_store_dword v3, v0, s[6:7]
; GFX10-DL-NEXT: s_endpgm
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
@@ -2341,10 +2357,11 @@ define amdgpu_kernel void @udot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX7-NEXT: v_and_b32_e32 v2, 0xffff, v2
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_lshrrev_b32_e32 v3, 16, v0
-; GFX7-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7-NEXT: v_mad_u32_u24 v4, v3, v1, s0
-; GFX7-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX7-NEXT: v_mul_u32_u24_e32 v4, v3, v1
; GFX7-NEXT: v_mad_u32_u24 v1, v3, v1, v4
+; GFX7-NEXT: v_and_b32_e32 v0, 0xffff, v0
+; GFX7-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7-NEXT: v_add_i32_e32 v1, vcc, s0, v1
; GFX7-NEXT: v_mad_u32_u24 v0, v0, v2, v1
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GFX7-NEXT: s_endpgm
@@ -2370,9 +2387,10 @@ define amdgpu_kernel void @udot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_and_b32_e32 v2, 0xffff, v0
; GFX8-NEXT: v_lshrrev_b32_e32 v0, 16, v0
-; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mad_u32_u24 v4, v0, v3, s0
+; GFX8-NEXT: v_mul_u32_u24_e32 v4, v0, v3
; GFX8-NEXT: v_mad_u32_u24 v0, v0, v3, v4
+; GFX8-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, s0, v0
; GFX8-NEXT: v_mad_u32_u24 v2, v2, v1, v0
; GFX8-NEXT: v_mov_b32_e32 v0, s4
; GFX8-NEXT: v_mov_b32_e32 v1, s5
@@ -2394,9 +2412,9 @@ define amdgpu_kernel void @udot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX9-NODL-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX9-NODL-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; GFX9-NODL-NEXT: v_mul_u32_u24_e32 v4, v2, v1
+; GFX9-NODL-NEXT: v_mad_u32_u24 v1, v2, v1, v4
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NODL-NEXT: v_mad_u32_u24 v1, v2, v1, s0
-; GFX9-NODL-NEXT: v_add3_u32 v1, v4, v1, v3
+; GFX9-NODL-NEXT: v_add3_u32 v1, s0, v1, v3
; GFX9-NODL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-NODL-NEXT: s_endpgm
;
@@ -2415,9 +2433,9 @@ define amdgpu_kernel void @udot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX9-DL-NEXT: v_lshrrev_b32_e32 v1, 16, v1
; GFX9-DL-NEXT: v_lshrrev_b32_e32 v2, 16, v2
; GFX9-DL-NEXT: v_mul_u32_u24_e32 v4, v2, v1
+; GFX9-DL-NEXT: v_mad_u32_u24 v1, v2, v1, v4
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-DL-NEXT: v_mad_u32_u24 v1, v2, v1, s0
-; GFX9-DL-NEXT: v_add3_u32 v1, v4, v1, v3
+; GFX9-DL-NEXT: v_add3_u32 v1, s0, v1, v3
; GFX9-DL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-DL-NEXT: s_endpgm
;
@@ -2438,12 +2456,12 @@ define amdgpu_kernel void @udot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_lshrrev_b32_e32 v3, 16, v2
; GFX10-DL-NEXT: v_mul_u32_u24_sdwa v1, v2, v1 dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
-; GFX10-DL-NEXT: v_mul_u32_u24_e32 v2, v3, v0
+; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-DL-NEXT: v_mul_u32_u24_e32 v4, v3, v0
+; GFX10-DL-NEXT: v_mad_u32_u24 v0, v3, v0, v4
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX10-DL-NEXT: v_mad_u32_u24 v0, v3, v0, s0
-; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0
-; GFX10-DL-NEXT: v_add3_u32 v0, v2, v0, v1
-; GFX10-DL-NEXT: global_store_dword v3, v0, s[6:7]
+; GFX10-DL-NEXT: v_add3_u32 v0, s0, v0, v1
+; GFX10-DL-NEXT: global_store_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_endpgm
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
@@ -2499,9 +2517,10 @@ define amdgpu_kernel void @idot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX7-NEXT: s_waitcnt vmcnt(0)
; GFX7-NEXT: v_bfe_i32 v3, v0, 0, 16
; GFX7-NEXT: v_ashrrev_i32_e32 v0, 16, v0
-; GFX7-NEXT: s_waitcnt lgkmcnt(0)
-; GFX7-NEXT: v_mad_i32_i24 v4, v0, v2, s0
+; GFX7-NEXT: v_mul_i32_i24_e32 v4, v0, v2
; GFX7-NEXT: v_mad_i32_i24 v0, v0, v2, v4
+; GFX7-NEXT: s_waitcnt lgkmcnt(0)
+; GFX7-NEXT: v_add_i32_e32 v0, vcc, s0, v0
; GFX7-NEXT: v_mad_i32_i24 v0, v3, v1, v0
; GFX7-NEXT: buffer_store_dword v0, off, s[4:7], 0
; GFX7-NEXT: s_endpgm
@@ -2527,9 +2546,10 @@ define amdgpu_kernel void @idot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX8-NEXT: s_waitcnt vmcnt(0)
; GFX8-NEXT: v_bfe_i32 v2, v0, 0, 16
; GFX8-NEXT: v_ashrrev_i32_e32 v0, 16, v0
-; GFX8-NEXT: s_waitcnt lgkmcnt(0)
-; GFX8-NEXT: v_mad_i32_i24 v4, v0, v3, s0
+; GFX8-NEXT: v_mul_i32_i24_e32 v4, v0, v3
; GFX8-NEXT: v_mad_i32_i24 v0, v0, v3, v4
+; GFX8-NEXT: s_waitcnt lgkmcnt(0)
+; GFX8-NEXT: v_add_u32_e32 v0, vcc, s0, v0
; GFX8-NEXT: v_mad_i32_i24 v2, v2, v1, v0
; GFX8-NEXT: v_mov_b32_e32 v0, s4
; GFX8-NEXT: v_mov_b32_e32 v1, s5
@@ -2551,9 +2571,9 @@ define amdgpu_kernel void @idot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX9-NODL-NEXT: v_ashrrev_i32_e32 v1, 16, v1
; GFX9-NODL-NEXT: v_ashrrev_i32_e32 v2, 16, v2
; GFX9-NODL-NEXT: v_mul_i32_i24_e32 v4, v2, v1
+; GFX9-NODL-NEXT: v_mad_i32_i24 v1, v2, v1, v4
; GFX9-NODL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-NODL-NEXT: v_mad_i32_i24 v1, v2, v1, s0
-; GFX9-NODL-NEXT: v_add3_u32 v1, v4, v1, v3
+; GFX9-NODL-NEXT: v_add3_u32 v1, s0, v1, v3
; GFX9-NODL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-NODL-NEXT: s_endpgm
;
@@ -2572,9 +2592,9 @@ define amdgpu_kernel void @idot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX9-DL-NEXT: v_ashrrev_i32_e32 v1, 16, v1
; GFX9-DL-NEXT: v_ashrrev_i32_e32 v2, 16, v2
; GFX9-DL-NEXT: v_mul_i32_i24_e32 v4, v2, v1
+; GFX9-DL-NEXT: v_mad_i32_i24 v1, v2, v1, v4
; GFX9-DL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX9-DL-NEXT: v_mad_i32_i24 v1, v2, v1, s0
-; GFX9-DL-NEXT: v_add3_u32 v1, v4, v1, v3
+; GFX9-DL-NEXT: v_add3_u32 v1, s0, v1, v3
; GFX9-DL-NEXT: global_store_dword v0, v1, s[6:7]
; GFX9-DL-NEXT: s_endpgm
;
@@ -2595,12 +2615,12 @@ define amdgpu_kernel void @idot2_MultipleUses_mul2(ptr addrspace(1) %src1,
; GFX10-DL-NEXT: s_waitcnt vmcnt(0)
; GFX10-DL-NEXT: v_ashrrev_i32_e32 v3, 16, v2
; GFX10-DL-NEXT: v_mul_i32_i24_sdwa v1, sext(v2), sext(v1) dst_sel:DWORD dst_unused:UNUSED_PAD src0_sel:WORD_0 src1_sel:WORD_0
-; GFX10-DL-NEXT: v_mul_i32_i24_e32 v2, v3, v0
+; GFX10-DL-NEXT: v_mov_b32_e32 v2, 0
+; GFX10-DL-NEXT: v_mul_i32_i24_e32 v4, v3, v0
+; GFX10-DL-NEXT: v_mad_i32_i24 v0, v3, v0, v4
; GFX10-DL-NEXT: s_waitcnt lgkmcnt(0)
-; GFX10-DL-NEXT: v_mad_i32_i24 v0, v3, v0, s0
-; GFX10-DL-NEXT: v_mov_b32_e32 v3, 0
-; GFX10-DL-NEXT: v_add3_u32 v0, v2, v0, v1
-; GFX10-DL-NEXT: global_store_dword v3, v0, s[6:7]
+; GFX10-DL-NEXT: v_add3_u32 v0, s0, v0, v1
+; GFX10-DL-NEXT: global_store_dword v2, v0, s[6:7]
; GFX10-DL-NEXT: s_endpgm
ptr addrspace(1) %src2,
ptr addrspace(1) nocapture %dst) {
diff --git a/llvm/test/CodeGen/Hexagon/isel-fold-shl-zext.ll b/llvm/test/CodeGen/Hexagon/isel-fold-shl-zext.ll
index fe0f7dd89babe..3697ff5ff79d5 100644
--- a/llvm/test/CodeGen/Hexagon/isel-fold-shl-zext.ll
+++ b/llvm/test/CodeGen/Hexagon/isel-fold-shl-zext.ll
@@ -15,11 +15,10 @@ define dso_local void @foo(i64* nocapture noundef %buf, i32 %a, i32 %b) local_un
; CHECK: .cfi_startproc
; CHECK-NEXT: // %bb.0: // %entry
; CHECK-NEXT: {
-; CHECK-NEXT: r2 = addasl(r2,r1,#1)
; CHECK-NEXT: r3 = asl(r1,#1)
; CHECK-NEXT: }
; CHECK-NEXT: {
-; CHECK-NEXT: r2 = addasl(r2,r1,#1)
+; CHECK-NEXT: r2 += add(r3,r3)
; CHECK-NEXT: }
; CHECK-NEXT: {
; CHECK-NEXT: jumpr r31
diff --git a/llvm/test/CodeGen/X86/avx-vinsertf128.ll b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
index 9a52ef3293d2e..2c997f77f280e 100644
--- a/llvm/test/CodeGen/X86/avx-vinsertf128.ll
+++ b/llvm/test/CodeGen/X86/avx-vinsertf128.ll
@@ -59,13 +59,13 @@ define <4 x i32> @DAGCombineA(<4 x i32> %v1) nounwind readonly {
define <8 x i32> @DAGCombineB(<8 x i32> %v1, <8 x i32> %v2) nounwind readonly {
; CHECK-LABEL: DAGCombineB:
; CHECK: # %bb.0:
-; CHECK-NEXT: vpaddd %xmm1, %xmm0, %xmm2
-; CHECK-NEXT: vextractf128 $1, %ymm1, %xmm1
+; CHECK-NEXT: vextractf128 $1, %ymm1, %xmm2
; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm3
-; CHECK-NEXT: vpaddd %xmm1, %xmm3, %xmm1
-; CHECK-NEXT: vpaddd %xmm3, %xmm1, %xmm1
-; CHECK-NEXT: vpaddd %xmm0, %xmm2, %xmm0
-; CHECK-NEXT: vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; CHECK-NEXT: vpaddd %xmm3, %xmm3, %xmm3
+; CHECK-NEXT: vpaddd %xmm3, %xmm2, %xmm2
+; CHECK-NEXT: vpaddd %xmm0, %xmm0, %xmm0
+; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0
+; CHECK-NEXT: vinsertf128 $1, %xmm2, %ymm0, %ymm0
; CHECK-NEXT: retq
%t1 = add <8 x i32> %v1, %v2
%t2 = add <8 x i32> %t1, %v1
More information about the llvm-commits
mailing list