Commit 2a7f1602 authored by Clemens Backes's avatar Clemens Backes Committed by V8 LUCI CQ

[compiler] Fix SIMD register overlap check

We should only check the "SIMD sibling" register if we are handling a
SIMD register. This avoids unneeded spills, and in this particular case
ran into a DCHECK because there are only 29 registers, but we tried
checking #29.

R=thibaudm@chromium.org

Bug: v8:12330, v8:1285007
Change-Id: Ife8b295ac958990611ca8816bbfbfb5124a4297d
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/3372916Reviewed-by: 's avatarThibaud Michaud <thibaudm@chromium.org>
Commit-Queue: Clemens Backes <clemensb@chromium.org>
Cr-Commit-Position: refs/heads/main@{#78532}
parent b4870d28
......@@ -1763,9 +1763,10 @@ void SinglePassRegisterAllocator::MergeStateFrom(
if (processed_regs.Contains(reg, rep)) continue;
processed_regs.Add(reg, rep);
bool reg_in_use = register_state_->IsAllocated(reg) ||
(!kSimpleFPAliasing &&
register_state_->IsAllocated(reg.simdSibling()));
bool reg_in_use =
register_state_->IsAllocated(reg) ||
(!kSimpleFPAliasing && rep == MachineRepresentation::kSimd128 &&
register_state_->IsAllocated(reg.simdSibling()));
if (!reg_in_use) {
DCHECK(successor_registers->IsAllocated(reg));
......
// Copyright 2022 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
// Flags: --allow-natives-syntax --assert-types --interrupt-budget=1000
// Flags: --no-lazy-feedback-allocation --turboprop
var __v_12 = [];
var __v_30 = {};
function __f_0(x, len) {
var __v_1 = new Array();
var __v_2 = x + 0.5;
var __v_3 = x + 1.5;
var __v_4 = x + 2.5;
var __v_5 = x + 3.5;
var __v_7 = x + 5.5;
var __v_15 = x + 13.5;
var __v_16 = x + 14.5;
var __v_22 = x + 20.5;
var __v_25 = x + 23.5;
var __v_8 = x + 24.5;
__v_12[len] = 0;
__v_1[0] = __v_2;
__v_1[1] = __v_3;
__v_1[2] = __v_4;
__v_1[3] = __v_5;
__v_1 = __v_30;
__v_1[5] = __v_7;
__v_1[6] = __v_8;
__v_1[13] = __v_15;
__v_1[14] = __v_16;
__v_1[23] = __v_25;
for (var __v_27 = 0; __v_27 < __v_1.length; __v_27++) {
x + __v_22 + [__v_37];
}
}
%PrepareFunctionForOptimization(__f_0);
__f_0();
%OptimizeFunctionOnNextCall(__f_0);
__f_0();
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment