Skip to content

[CodeGenPrepare] Consider target memory intrinics as memory use #159638

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Sep 19, 2025
Merged

[CodeGenPrepare] Consider target memory intrinics as memory use #159638

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4c52b42
[CodeGenPrepare] Consider target memory intrinics as memory use
jrbyrnes Sep 18, 2025
035f5c7
Review comments
jrbyrnes Sep 19, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions llvm/lib/CodeGen/CodeGenPrepare.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5596,6 +5596,19 @@ static bool FindAllMemoryUses(
continue;
}

if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(UserI)) {
SmallVector<Value *, 2> PtrOps;
Type *AccessTy;
if (!TLI.getAddrModeArguments(II, PtrOps, AccessTy))
return true;

if (!find(PtrOps, U.get()))
return true;

MemoryUses.push_back({&U, AccessTy});
continue;
}

if (CallInst *CI = dyn_cast<CallInst>(UserI)) {
if (CI->hasFnAttr(Attribute::Cold)) {
// If this is a cold call, we can sink the addressing calculation into
Expand Down
20 changes: 9 additions & 11 deletions llvm/test/CodeGen/AMDGPU/sink-addr-memory-intrinsics.ll
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,25 +6,23 @@ define amdgpu_kernel void @memoryIntrinstic(ptr addrspace(3) %inptr, i1 %cond, p
; CHECK: ; %bb.0:
; CHECK-NEXT: s_load_dwordx4 s[0:3], s[4:5], 0x0
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: s_and_b32 s1, s1, 1
; CHECK-NEXT: s_add_i32 s3, s0, 0x2000
; CHECK-NEXT: s_cmp_eq_u32 s1, 0
; CHECK-NEXT: s_bitcmp0_b32 s1, 0
; CHECK-NEXT: s_cbranch_scc0 .LBB0_2
; CHECK-NEXT: ; %bb.1: ; %else
; CHECK-NEXT: v_mov_b32_e32 v0, s3
; CHECK-NEXT: ds_read_b64_tr_b16 v[2:3], v0
; CHECK-NEXT: s_mov_b32 s0, 0x7060302
; CHECK-NEXT: s_mov_b32 s1, 0x5040100
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: ds_read_b64_tr_b16 v[2:3], v0 offset:8192
; CHECK-NEXT: s_mov_b32 s1, 0x7060302
; CHECK-NEXT: s_mov_b32 s3, 0x5040100
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
; CHECK-NEXT: v_perm_b32 v0, v3, v2, s0
; CHECK-NEXT: v_perm_b32 v1, v3, v2, s1
; CHECK-NEXT: v_perm_b32 v0, v3, v2, s1
; CHECK-NEXT: v_perm_b32 v1, v3, v2, s3
; CHECK-NEXT: s_cbranch_execz .LBB0_3
; CHECK-NEXT: s_branch .LBB0_4
; CHECK-NEXT: .LBB0_2:
; CHECK-NEXT: ; implicit-def: $vgpr1
; CHECK-NEXT: .LBB0_3: ; %then
; CHECK-NEXT: v_mov_b32_e32 v0, s3
; CHECK-NEXT: ds_read_b64_tr_b16 v[2:3], v0
; CHECK-NEXT: v_mov_b32_e32 v0, s0
; CHECK-NEXT: ds_read_b64_tr_b16 v[2:3], v0 offset:8192
; CHECK-NEXT: s_mov_b32 s0, 0x5040100
; CHECK-NEXT: s_mov_b32 s1, 0x7060302
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
Expand Down