[llvm] [RegAlloc] Add coverage leading to revert of pr160765 (PR #161614)
Philip Reames via llvm-commits
llvm-commits at lists.llvm.org
Wed Oct 1 17:53:06 PDT 2025
https://github.com/preames created https://github.com/llvm/llvm-project/pull/161614
Essentially what happened is the following series of events: 1) We rematerialized the vmv.v.x into the loop.
2) As this was the last use of the instruction, we deleted the
instruction, and removed it from the original live range.
3) We split the live range for the remat.
4) We tried to rematerialize the uses of that split interval, and
crashed because the assert about the def being available in
the original live interval does not hold.
>From 5104be6561572d43ac9e2605c9db82549efab57c Mon Sep 17 00:00:00 2001
From: Philip Reames <preames at rivosinc.com>
Date: Wed, 1 Oct 2025 17:04:53 -0700
Subject: [PATCH] [RegAlloc] Add coverage leading to revert of pr160765
Essentially what happened is the following series of events:
1) We rematerialized the vmv.v.x into the loop.
2) As this was the last use of the instruction, we deleted the
instruction, and removed it from the original live range.
3) We split the live range for the remat.
4) We tried to rematerialize the uses of that split interval, and
crashed because the assert about the def being available in
the original live interval does not hold.
---
llvm/test/CodeGen/RISCV/rvv/remat.ll | 132 +++++++++++++++++++++++++++
1 file changed, 132 insertions(+)
diff --git a/llvm/test/CodeGen/RISCV/rvv/remat.ll b/llvm/test/CodeGen/RISCV/rvv/remat.ll
index 06d54fadaeffd..95bff27fe8ca6 100644
--- a/llvm/test/CodeGen/RISCV/rvv/remat.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/remat.ll
@@ -301,3 +301,135 @@ define void @vfmv.s.f(ptr %p, double %x) {
store volatile double %x, ptr %p
ret void
}
+
+; This test is fairly fragile, but it's trying to cover the case which
+; caused the revert of bba9172 due to interaction with how rematerialize
+; instructions are pruned from the original live interval. In the result
+; below, we remat the vmv.v.x into the loop, but fail to remat the vmv.v.x
+; a second time after further splitting it's live range. We shouldn't need
+; to spill it to the stack at all.
+define i64 @dual_remat(i64 %0, <vscale x 16 x i64> %1, <vscale x 16 x i64> %2, ptr %p) #0 {
+; CHECK-LABEL: dual_remat:
+; CHECK: # %bb.0: # %entry
+; CHECK-NEXT: addi sp, sp, -16
+; CHECK-NEXT: .cfi_def_cfa_offset 16
+; CHECK-NEXT: csrr a1, vlenb
+; CHECK-NEXT: slli a2, a1, 5
+; CHECK-NEXT: add a1, a2, a1
+; CHECK-NEXT: sub sp, sp, a1
+; CHECK-NEXT: .cfi_escape 0x0f, 0x0d, 0x72, 0x00, 0x11, 0x10, 0x22, 0x11, 0x21, 0x92, 0xa2, 0x38, 0x00, 0x1e, 0x22 # sp + 16 + 33 * vlenb
+; CHECK-NEXT: csrr a1, vlenb
+; CHECK-NEXT: slli a1, a1, 3
+; CHECK-NEXT: add a1, sp, a1
+; CHECK-NEXT: addi a1, a1, 16
+; CHECK-NEXT: vs8r.v v16, (a1) # vscale x 64-byte Folded Spill
+; CHECK-NEXT: addi a1, sp, 16
+; CHECK-NEXT: vs8r.v v8, (a1) # vscale x 64-byte Folded Spill
+; CHECK-NEXT: vsetvli a1, zero, e64, m8, ta, ma
+; CHECK-NEXT: vmv.v.i v16, 0
+; CHECK-NEXT: csrr a2, vlenb
+; CHECK-NEXT: srli a1, a2, 3
+; CHECK-NEXT: slli a2, a2, 3
+; CHECK-NEXT: add a2, a3, a2
+; CHECK-NEXT: vmv.v.i v0, 0
+; CHECK-NEXT: .LBB8_1: # %vector.body
+; CHECK-NEXT: # =>This Inner Loop Header: Depth=1
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: mv a5, a4
+; CHECK-NEXT: slli a4, a4, 3
+; CHECK-NEXT: add a5, a5, a4
+; CHECK-NEXT: slli a4, a4, 1
+; CHECK-NEXT: add a4, a4, a5
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vs8r.v v16, (a4) # vscale x 64-byte Folded Spill
+; CHECK-NEXT: vmv.v.x v8, a0
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: slli a5, a4, 4
+; CHECK-NEXT: add a4, a5, a4
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vs8r.v v8, (a4) # vscale x 64-byte Folded Spill
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: mv a5, a4
+; CHECK-NEXT: slli a4, a4, 3
+; CHECK-NEXT: add a5, a5, a4
+; CHECK-NEXT: slli a4, a4, 1
+; CHECK-NEXT: add a4, a4, a5
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vl8r.v v16, (a4) # vscale x 64-byte Folded Reload
+; CHECK-NEXT: vand.vv v16, v16, v8
+; CHECK-NEXT: vmsne.vi v24, v16, 0
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: slli a4, a4, 4
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vs1r.v v24, (a4) # vscale x 8-byte Folded Spill
+; CHECK-NEXT: vand.vv v16, v0, v8
+; CHECK-NEXT: vmsne.vi v8, v16, 0
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: mv a5, a4
+; CHECK-NEXT: slli a4, a4, 3
+; CHECK-NEXT: add a5, a5, a4
+; CHECK-NEXT: slli a4, a4, 1
+; CHECK-NEXT: add a4, a4, a5
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vl8r.v v16, (a4) # vscale x 64-byte Folded Reload
+; CHECK-NEXT: csrr a4, vlenb
+; CHECK-NEXT: slli a4, a4, 4
+; CHECK-NEXT: add a4, sp, a4
+; CHECK-NEXT: addi a4, a4, 16
+; CHECK-NEXT: vl1r.v v9, (a4) # vscale x 8-byte Folded Reload
+; CHECK-NEXT: vsetvli a4, zero, e8, mf4, ta, ma
+; CHECK-NEXT: vslideup.vx v9, v8, a1
+; CHECK-NEXT: vsetvli a4, zero, e8, m2, ta, ma
+; CHECK-NEXT: vcpop.m a4, v9
+; CHECK-NEXT: csrr a5, vlenb
+; CHECK-NEXT: slli a6, a5, 4
+; CHECK-NEXT: add a5, a6, a5
+; CHECK-NEXT: add a5, sp, a5
+; CHECK-NEXT: addi a5, a5, 16
+; CHECK-NEXT: vl8r.v v8, (a5) # vscale x 64-byte Folded Reload
+; CHECK-NEXT: vs8r.v v8, (a3)
+; CHECK-NEXT: vs8r.v v8, (a2)
+; CHECK-NEXT: addi a5, sp, 16
+; CHECK-NEXT: vl8r.v v8, (a5) # vscale x 64-byte Folded Reload
+; CHECK-NEXT: vsetvli a5, zero, e64, m8, ta, ma
+; CHECK-NEXT: vor.vv v16, v16, v8
+; CHECK-NEXT: csrr a5, vlenb
+; CHECK-NEXT: slli a5, a5, 3
+; CHECK-NEXT: add a5, sp, a5
+; CHECK-NEXT: addi a5, a5, 16
+; CHECK-NEXT: vl8r.v v8, (a5) # vscale x 64-byte Folded Reload
+; CHECK-NEXT: vor.vv v0, v0, v8
+; CHECK-NEXT: beqz a4, .LBB8_1
+; CHECK-NEXT: # %bb.2: # %middle.block
+; CHECK-NEXT: andi a0, a0, 1
+; CHECK-NEXT: csrr a1, vlenb
+; CHECK-NEXT: slli a2, a1, 5
+; CHECK-NEXT: add a1, a2, a1
+; CHECK-NEXT: add sp, sp, a1
+; CHECK-NEXT: .cfi_def_cfa sp, 16
+; CHECK-NEXT: addi sp, sp, 16
+; CHECK-NEXT: .cfi_def_cfa_offset 0
+; CHECK-NEXT: ret
+entry:
+ %broadcast.splatinsert = insertelement <vscale x 16 x i64> zeroinitializer, i64 %0, i64 0
+ %broadcast.splat = shufflevector <vscale x 16 x i64> %broadcast.splatinsert, <vscale x 16 x i64> zeroinitializer, <vscale x 16 x i32> zeroinitializer
+ br label %vector.body
+
+vector.body: ; preds = %vector.body, %entry
+ %vec.ind = phi <vscale x 16 x i64> [ zeroinitializer, %entry ], [ %vec.ind.next, %vector.body ]
+ %3 = and <vscale x 16 x i64> %vec.ind, %broadcast.splat
+ %4 = icmp ne <vscale x 16 x i64> %3, zeroinitializer
+ store <vscale x 16 x i64> %broadcast.splat, ptr %p
+ %5 = tail call i1 @llvm.vector.reduce.or.nxv16i1(<vscale x 16 x i1> %4)
+ %vec.ind.next = or <vscale x 16 x i64> %vec.ind, %1
+ br i1 %5, label %middle.block, label %vector.body
+
+middle.block: ; preds = %vector.body
+ %and.i = and i64 1, %0
+ ret i64 %and.i
+}
More information about the llvm-commits
mailing list