diff --git a/BUILD b/BUILD
index c2b02ce44..fe07940f3 100644
--- a/BUILD
+++ b/BUILD
@@ -93,7 +93,7 @@ cc_binary(
         "-fvisibility=hidden",
         "-march=native",
         "-DSTIM_PYBIND11_MODULE_NAME=stim",
-        "-DVERSION_INFO=0.0.dev"
+        "-DVERSION_INFO=0.0.dev",
     ],
     includes = ["src/"],
     linkopts = ["-lpthread"],
diff --git a/setup.py b/setup.py
index 0b0a79df2..14ad2847e 100644
--- a/setup.py
+++ b/setup.py
@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import sys
 
 from setuptools import setup, Extension
 import glob
@@ -26,13 +27,27 @@
 
 version = '1.11.dev0'
 
-common_compile_args = [
-    '-std=c++11',
-    '-fno-strict-aliasing',
-    '-O3',
-    '-g0',
-    f'-DVERSION_INFO={version}',
-]
+if sys.platform.startswith('win'):
+    common_compile_args = [
+        '/std:c++11',
+        '/O2',
+        f'/DVERSION_INFO={version}',
+    ]
+    arch_avx = ['/arch:AVX2']
+    arch_sse = ['/arch:SSE2']
+    arch_basic = []
+else:
+    common_compile_args = [
+        '-std=c++11',
+        '-fno-strict-aliasing',
+        '-O3',
+        '-g0',
+        f'-DVERSION_INFO={version}',
+    ]
+    arch_avx = ['-mavx2']
+    arch_sse = ['-msse2', '-mno-avx2']
+    arch_basic = []
+
 stim_detect_machine_architecture = Extension(
     'stim._detect_machine_architecture',
     sources=MUX_SOURCE_FILES,
@@ -40,8 +55,7 @@
     language='c++',
     extra_compile_args=[
         *common_compile_args,
-        '-mno-sse2',
-        '-mno-avx2',
+        *arch_basic,
     ],
 )
 stim_polyfill = Extension(
@@ -51,8 +65,7 @@
     language='c++',
     extra_compile_args=[
         *common_compile_args,
-        # I would specify -mno-sse2 but that causes build failures in non-stim code...?
-        '-mno-avx2',
+        *arch_basic,
         '-DSTIM_PYBIND11_MODULE_NAME=_stim_polyfill',
     ],
 )
@@ -63,8 +76,7 @@
     language='c++',
     extra_compile_args=[
         *common_compile_args,
-        '-msse2',
-        '-mno-avx2',
+        *arch_sse,
         '-DSTIM_PYBIND11_MODULE_NAME=_stim_sse2',
     ],
 )
@@ -77,8 +89,7 @@
 #     language='c++',
 #     extra_compile_args=[
 #         *common_compile_args,
-#         '-msse2',
-#         '-mavx2',
+#         *arch_avx,
 #         '-DSTIM_PYBIND11_MODULE_NAME=_stim_avx2',
 #     ],
 # )
diff --git a/src/stim/benchmark_main.perf.cc b/src/stim/benchmark_main.perf.cc
index 0c9661f0f..ee55968c8 100644
--- a/src/stim/benchmark_main.perf.cc
+++ b/src/stim/benchmark_main.perf.cc
@@ -25,40 +25,40 @@ std::vector<RegisteredBenchmark> all_registered_benchmarks{};
 
 /// Describe quantity as an SI-prefixed value with two significant figures.
 std::string si2(double val) {
-    std::string unit = "";
+    char unit = ' ';
     if (val < 1) {
         if (val < 1) {
             val *= 1000;
-            unit = "m";
+            unit = 'm';
         }
         if (val < 1) {
             val *= 1000;
-            unit = "u";
+            unit = 'u';
         }
         if (val < 1) {
             val *= 1000;
-            unit = "n";
+            unit = 'n';
         }
         if (val < 1) {
             val *= 1000;
-            unit = "p";
+            unit = 'p';
         }
     } else {
         if (val > 1000) {
             val /= 1000;
-            unit = "k";
+            unit = 'k';
         }
         if (val > 1000) {
             val /= 1000;
-            unit = "M";
+            unit = 'M';
         }
         if (val > 1000) {
             val /= 1000;
-            unit = "G";
+            unit = 'G';
         }
         if (val > 1000) {
             val /= 1000;
-            unit = "T";
+            unit = 'T';
         }
     }
     std::stringstream ss;
diff --git a/src/stim/diagram/timeline/timeline_ascii_drawer.cc b/src/stim/diagram/timeline/timeline_ascii_drawer.cc
index 207abfa60..b65f83f95 100644
--- a/src/stim/diagram/timeline/timeline_ascii_drawer.cc
+++ b/src/stim/diagram/timeline/timeline_ascii_drawer.cc
@@ -482,9 +482,11 @@ AsciiDiagram DiagramTimelineAsciiDrawer::make_diagram(const Circuit &circuit) {
             {0, obj.q2y(q), 1.0, 0.5},
             {obj.m2x(obj.cur_moment), obj.q2y(q), 0.0, 0.5},
         };
+        std::stringstream qubit;
+        qubit << 'q' << q << ": ";
         obj.diagram.add_entry(AsciiDiagramEntry{
             {0, obj.q2y(q), 1.0, 0.5},
-            "q" + std::to_string(q) + ": ",
+            qubit.str(),
         });
     }
 
diff --git a/src/stim/gen/circuit_gen_params.cc b/src/stim/gen/circuit_gen_params.cc
index 7dc2e559f..12b2ddf29 100644
--- a/src/stim/gen/circuit_gen_params.cc
+++ b/src/stim/gen/circuit_gen_params.cc
@@ -57,19 +57,25 @@ void CircuitGenParameters::append_unitary_2(
 }
 
 void CircuitGenParameters::append_reset(Circuit &circuit, const std::vector<uint32_t> targets, char basis) const {
-    circuit.safe_append_u(std::string("R") + basis, targets);
+    std::string gate("R");
+    gate.push_back(basis);
+    circuit.safe_append_u(gate, targets);
     append_anti_basis_error(circuit, targets, after_reset_flip_probability, basis);
 }
 
 void CircuitGenParameters::append_measure(Circuit &circuit, const std::vector<uint32_t> targets, char basis) const {
+    std::string gate("M");
+    gate.push_back(basis);
     append_anti_basis_error(circuit, targets, before_measure_flip_probability, basis);
-    circuit.safe_append_u(std::string("M") + basis, targets);
+    circuit.safe_append_u(gate, targets);
 }
 
 void CircuitGenParameters::append_measure_reset(
     Circuit &circuit, const std::vector<uint32_t> targets, char basis) const {
+    std::string gate("MR");
+    gate.push_back(basis);
     append_anti_basis_error(circuit, targets, before_measure_flip_probability, basis);
-    circuit.safe_append_u(std::string("MR") + basis, targets);
+    circuit.safe_append_u(gate, targets);
     append_anti_basis_error(circuit, targets, after_reset_flip_probability, basis);
 }
 
diff --git a/src/stim/gen/circuit_gen_params.h b/src/stim/gen/circuit_gen_params.h
index b223058f1..3920d2396 100644
--- a/src/stim/gen/circuit_gen_params.h
+++ b/src/stim/gen/circuit_gen_params.h
@@ -30,7 +30,7 @@ struct CircuitGenParameters {
 
 struct GeneratedCircuit {
     Circuit circuit;
-    std::map<std::pair<uint32_t, uint32_t>, std::pair<std::string, uint32_t>> layout;
+    std::map<std::pair<uint32_t, uint32_t>, std::pair<char, uint32_t>> layout;
     std::string hint_str;
     std::string layout_str() const;
 };
diff --git a/src/stim/gen/gen_color_code.cc b/src/stim/gen/gen_color_code.cc
index 5f0f000f4..56cc4733c 100644
--- a/src/stim/gen/gen_color_code.cc
+++ b/src/stim/gen/gen_color_code.cc
@@ -191,11 +191,11 @@ GeneratedCircuit stim::generate_color_code_circuit(const CircuitGenParameters &p
     auto full_circuit = head + body * (params.rounds - 2) + tail;
 
     // Make 2d layout.
-    std::map<std::pair<uint32_t, uint32_t>, std::pair<std::string, uint32_t>> layout;
+    std::map<std::pair<uint32_t, uint32_t>, std::pair<char, uint32_t>> layout;
     for (auto q : data_coords) {
-        layout[{(uint32_t)(q.x * 2), (uint32_t)q.y}] = {q.y == 0 ? "L" : "d", p2q[q]};
+        layout[{(uint32_t)(q.x * 2), (uint32_t)q.y}] = {q.y == 0 ? 'L' : 'd', p2q[q]};
     }
-    std::array<const char *, 3> rgb{"R", "G", "B"};
+    std::array<char, 3> rgb{'R', 'G', 'B'};
     for (auto q : measure_coords) {
         auto x = (uint32_t)(q.x * 2);
         auto y = (uint32_t)q.y;
diff --git a/src/stim/gen/gen_rep_code.cc b/src/stim/gen/gen_rep_code.cc
index 1bb6a79f1..f66152df6 100644
--- a/src/stim/gen/gen_rep_code.cc
+++ b/src/stim/gen/gen_rep_code.cc
@@ -81,11 +81,11 @@ GeneratedCircuit stim::generate_rep_code_circuit(const CircuitGenParameters &par
     Circuit full_circuit = head + body * (params.rounds - 1) + tail;
 
     // Produce a 2d layout.
-    std::map<std::pair<uint32_t, uint32_t>, std::pair<std::string, uint32_t>> layout;
+    std::map<std::pair<uint32_t, uint32_t>, std::pair<char, uint32_t>> layout;
     for (uint32_t k = 0; k < n; k++) {
-        layout[{k, 0}] = {std::string() + "dZ"[k & 1], k};
+        layout[{k, 0}] = {"dZ"[k & 1], k};
     }
-    layout[{0, 0}].first = "L";
+    layout[{0, 0}].first = 'L';
 
     return {
         full_circuit,
diff --git a/src/stim/gen/gen_surface_code.cc b/src/stim/gen/gen_surface_code.cc
index 9ada5a3b5..4ff30577a 100644
--- a/src/stim/gen/gen_surface_code.cc
+++ b/src/stim/gen/gen_surface_code.cc
@@ -191,19 +191,19 @@ GeneratedCircuit _finish_surface_code_circuit(
     Circuit full_circuit = head + body * (params.rounds - 1) + tail;
 
     // Produce a 2d layout.
-    std::map<std::pair<uint32_t, uint32_t>, std::pair<std::string, uint32_t>> layout;
+    std::map<std::pair<uint32_t, uint32_t>, std::pair<char, uint32_t>> layout;
     float scale = x_order[0].x == 0.5 ? 2 : 1;
     for (auto q : data_coords) {
-        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {"d", p2q[q]};
+        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {'d', p2q[q]};
     }
     for (auto q : x_measure_coords) {
-        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {"X", p2q[q]};
+        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {'X', p2q[q]};
     }
     for (auto q : z_measure_coords) {
-        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {"Z", p2q[q]};
+        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}] = {'Z', p2q[q]};
     }
     for (auto q : chosen_basis_observable) {
-        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}].first = "L";
+        layout[{(uint32_t)(q.x * scale), (uint32_t)(q.y * scale)}].first = 'L';
     }
 
     return {
diff --git a/src/stim/mem/simd_bits.h b/src/stim/mem/simd_bits.h
index 1e00e2589..7b2f91bb5 100644
--- a/src/stim/mem/simd_bits.h
+++ b/src/stim/mem/simd_bits.h
@@ -71,8 +71,10 @@ struct simd_bits {
 
     // Equality.
     bool operator==(const simd_bits_range_ref<W> &other) const;
+    bool operator==(const simd_bits<W> &other) const;
     // Inequality.
     bool operator!=(const simd_bits_range_ref<W> &other) const;
+    bool operator!=(const simd_bits<W> &other) const;
     /// Determines whether or not any of the bits in the simd_bits are non-zero.
     bool not_zero() const;
 
diff --git a/src/stim/mem/simd_bits.inl b/src/stim/mem/simd_bits.inl
index af3fa1d27..fcfb3f71c 100644
--- a/src/stim/mem/simd_bits.inl
+++ b/src/stim/mem/simd_bits.inl
@@ -102,11 +102,21 @@ bool simd_bits<W>::operator==(const simd_bits_range_ref<W> &other) const {
     return simd_bits_range_ref<W>(*this) == other;
 }
 
+template <size_t W>
+bool simd_bits<W>::operator==(const simd_bits<W> &other) const {
+    return simd_bits_range_ref<W>(*this) == simd_bits_range_ref<W>(other);
+}
+
 template <size_t W>
 bool simd_bits<W>::operator!=(const simd_bits_range_ref<W> &other) const {
     return !(*this == other);
 }
 
+template <size_t W>
+bool simd_bits<W>::operator!=(const simd_bits<W> &other) const {
+    return !(*this == other);
+}
+
 template <size_t W>
 simd_bits<W> simd_bits<W>::random(size_t min_bits, std::mt19937_64 &rng) {
     simd_bits<W> result(min_bits);
diff --git a/src/stim/mem/simd_bits_range_ref.h b/src/stim/mem/simd_bits_range_ref.h
index a24d7271e..721b086b9 100644
--- a/src/stim/mem/simd_bits_range_ref.h
+++ b/src/stim/mem/simd_bits_range_ref.h
@@ -70,9 +70,9 @@ struct simd_bits_range_ref {
     void swap_with(simd_bits_range_ref other);
 
     /// Equality.
-    bool operator==(const simd_bits_range_ref other) const;
+    bool operator==(const simd_bits_range_ref<W> &other) const;
     /// Inequality.
-    bool operator!=(const simd_bits_range_ref other) const;
+    bool operator!=(const simd_bits_range_ref<W> &other) const;
     /// Determines whether or not any of the bits in the referenced range are non-zero.
     bool not_zero() const;
 
diff --git a/src/stim/mem/simd_bits_range_ref.inl b/src/stim/mem/simd_bits_range_ref.inl
index 4eb9b0d4e..d795dbbf2 100644
--- a/src/stim/mem/simd_bits_range_ref.inl
+++ b/src/stim/mem/simd_bits_range_ref.inl
@@ -75,7 +75,7 @@ void simd_bits_range_ref<W>::clear() {
 }
 
 template <size_t W>
-bool simd_bits_range_ref<W>::operator==(const simd_bits_range_ref<W> other) const {
+bool simd_bits_range_ref<W>::operator==(const simd_bits_range_ref<W> &other) const {
     return num_simd_words == other.num_simd_words && memcmp(ptr_simd, other.ptr_simd, num_u8_padded()) == 0;
 }
 
@@ -89,7 +89,7 @@ bool simd_bits_range_ref<W>::not_zero() const {
 }
 
 template <size_t W>
-bool simd_bits_range_ref<W>::operator!=(const simd_bits_range_ref<W> other) const {
+bool simd_bits_range_ref<W>::operator!=(const simd_bits_range_ref<W> &other) const {
     return !(*this == other);
 }
 
diff --git a/src/stim/stabilizers/pauli_string.cc b/src/stim/stabilizers/pauli_string.cc
index 6b1ddcd88..9685ccbd4 100644
--- a/src/stim/stabilizers/pauli_string.cc
+++ b/src/stim/stabilizers/pauli_string.cc
@@ -115,10 +115,16 @@ PauliString PauliString::random(size_t num_qubits, std::mt19937_64 &rng) {
 bool PauliString::operator==(const PauliStringRef &other) const {
     return ref() == other;
 }
+bool PauliString::operator==(const PauliString &other) const {
+    return ref() == other.ref();
+}
 
 bool PauliString::operator!=(const PauliStringRef &other) const {
     return ref() != other;
 }
+bool PauliString::operator!=(const PauliString &other) const {
+    return ref() != other.ref();
+}
 
 void PauliString::ensure_num_qubits(size_t min_num_qubits, double resize_pad_factor) {
     assert(resize_pad_factor >= 1);
diff --git a/src/stim/stabilizers/pauli_string.h b/src/stim/stabilizers/pauli_string.h
index 3b5d5b197..fea16c13e 100644
--- a/src/stim/stabilizers/pauli_string.h
+++ b/src/stim/stabilizers/pauli_string.h
@@ -91,8 +91,10 @@ struct PauliString {
 
     /// Equality.
     bool operator==(const PauliStringRef &other) const;
+    bool operator==(const PauliString &other) const;
     /// Inequality.
     bool operator!=(const PauliStringRef &other) const;
+    bool operator!=(const PauliString &other) const;
 
     /// Implicit conversion to a reference.
     operator PauliStringRef();