From 1fb178385c2a343d3cf7148505c27bd88f7990c3 Mon Sep 17 00:00:00 2001
From: Stefan Krastanov <stefan@krastanov.org>
Date: Tue, 23 Jul 2024 13:07:30 -0400
Subject: [PATCH] fix for wrong SIMD mul_left! #320 (#322)

---
 CHANGELOG.md               |  3 +++
 Project.toml               |  2 +-
 src/mul_leftright.jl       |  2 +-
 test/test_mul_leftright.jl | 17 ++++++++++++++++-
 test/test_stabcanon.jl     | 13 +++++++++++++
 5 files changed, 34 insertions(+), 3 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index ec4e15c24..1c8952376 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -5,6 +5,9 @@
 
 # News
 
+## v0.9.7 - 2024-07-23
+
+- **(fix `#320`)** Fix a serious correctness bug in the SIMD implementation of Pauli string multiplication (affects the correctness of canonicalization and traceout for tableaux bigger than ~500 qubits; does not affect symbolic gates or Pauli frame simulations of any scale)
 ## v0.9.6 - 2024-07-12
 
 - `inv` implementation for single-qubit "symbolic" Clifford operators (subtypes of `AbstractSingleQubitOperator`).
diff --git a/Project.toml b/Project.toml
index a2ca448b8..196ddf832 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "QuantumClifford"
 uuid = "0525e862-1e90-11e9-3e4d-1b39d7109de1"
 authors = ["Stefan Krastanov <stefan@krastanov.org> and QuantumSavory community members"]
-version = "0.9.6"
+version = "0.9.7"
 
 [deps]
 Combinatorics = "861a8166-3701-5b0c-9a16-15d98fcdc6aa"
diff --git a/src/mul_leftright.jl b/src/mul_leftright.jl
index 43a3731cd..122fb1e61 100644
--- a/src/mul_leftright.jl
+++ b/src/mul_leftright.jl
@@ -89,7 +89,7 @@ function mul_ordered!(r::AbstractVector{T}, l::AbstractVector{T}; phases::Val{B}
             r[i+len+lane] = newz1 = z1 ⊻ z2
             x1z2 = x1 & z2
             anti_comm = (x2 & z1) ⊻ x1z2
-            cnt2 ⊻= (newx1 ⊻ newz1 ⊻ x1z2) & anti_comm
+            cnt2 ⊻= (cnt1 ⊻ newx1 ⊻ newz1 ⊻ x1z2) & anti_comm
             cnt1 ⊻= anti_comm
         end
         for i in 1:length(cnt1)
diff --git a/test/test_mul_leftright.jl b/test/test_mul_leftright.jl
index ecf2b36eb..1d7c14e45 100644
--- a/test/test_mul_leftright.jl
+++ b/test/test_mul_leftright.jl
@@ -5,7 +5,7 @@ using Test
 
 test_sizes = [1,2,10,63,64,65,127,128,129] # Including sizes that would test off-by-one errors in the bit encoding.
 
-@testset "Inner product between stabilizer states" begin
+@testset "Pauli string multiplication" begin
     for n in test_sizes
         for _ in 1:20
             p1 = random_pauli(n)
@@ -22,3 +22,18 @@ test_sizes = [1,2,10,63,64,65,127,128,129] # Including sizes that would test off
         end
     end
 end
+
+# test for #320
+@testset "verify SIMD implementation" begin
+    for i in 1:10,
+        n in 1:30,
+        T in [UInt8, UInt16, UInt32, UInt64]
+        a = rand(T, n)
+        b = rand(T, n)
+        c1,c2 = QuantumClifford.mul_ordered!(copy(a),copy(b))
+        n1,n2 = QuantumClifford._mul_ordered_nonvec!(copy(a),copy(b))
+        np = ((n1 ⊻ (n2<<1))&0x3)
+        cp = ((c1 ⊻ (c2<<1))&0x3)
+        @test np==cp
+    end
+end
diff --git a/test/test_stabcanon.jl b/test/test_stabcanon.jl
index 13442e4eb..5c113c69b 100644
--- a/test/test_stabcanon.jl
+++ b/test/test_stabcanon.jl
@@ -74,3 +74,16 @@ test_sizes = [1,2,10,63,64,65,127,128,129] # Including sizes that would test off
         end
     end
 end
+
+@testset "canonicalization invariants" begin
+    s = random_stabilizer(40,100)
+    ss = tensor_pow(s,20)
+    sa1 = canonicalize!(canonicalize_rref!(copy(ss))[1])
+    sa2 = canonicalize!(copy(ss))
+    @test sa1 == sa2
+    ms = MixedDestabilizer(s)
+    mss = tensor_pow(ms, 20)
+    msa1 = canonicalize!(canonicalize_rref!(copy(mss))[1])
+    msa2 = canonicalize!(copy(mss))
+    @test stabilizerview(msa1) == stabilizerview(msa2) == sa1
+end