diff --git a/src/modules/meson.build b/src/modules/meson.build
index f90878c82..c267e681c 100644
--- a/src/modules/meson.build
+++ b/src/modules/meson.build
@@ -61,33 +61,36 @@ simd_cargs = []
 simd_dependencies = []
 
 if have_sse
-  pffft_sse = static_library('pffft_sse',
-    ['module-filter-chain/pffft.c' ],
+  filter_chain_sse = static_library('filter_chain_sse',
+    ['module-filter-chain/pffft.c',
+     'module-filter-chain/dsp-ops-sse.c' ],
     c_args : [sse_args, '-O3', '-DHAVE_SSE'],
     dependencies : [ spa_dep ],
     install : false
     )
   simd_cargs += ['-DHAVE_SSE']
-  simd_dependencies += pffft_sse
+  simd_dependencies += filter_chain_sse
 endif
 if have_neon
-  pffft_neon = static_library('pffft_neon',
+  filter_chain_neon = static_library('filter_chain_neon',
     ['module-filter-chain/pffft.c' ],
     c_args : [neon_args, '-O3', '-DHAVE_NEON'],
     dependencies : [ spa_dep ],
     install : false
     )
   simd_cargs += ['-DHAVE_NEON']
-  simd_dependencies += pffft_neon
+  simd_dependencies += filter_chain_neon
 endif
 
-pffft_c = static_library('pffft_c',
-  ['module-filter-chain/pffft.c' ],
+filter_chain_c = static_library('filter_chain_c',
+  ['module-filter-chain/pffft.c',
+   'module-filter-chain/dsp-ops.c',
+   'module-filter-chain/dsp-ops-c.c' ],
   c_args : [simd_cargs, '-O3', '-DPFFFT_SIMD_DISABLE'],
   dependencies : [ spa_dep ],
   install : false
 )
-simd_dependencies += pffft_c
+simd_dependencies += filter_chain_c
 
 filter_chain_sources = [
   'module-filter-chain.c',
diff --git a/src/modules/module-filter-chain/builtin_plugin.c b/src/modules/module-filter-chain/builtin_plugin.c
index 3966422c3..6918ad91d 100644
--- a/src/modules/module-filter-chain/builtin_plugin.c
+++ b/src/modules/module-filter-chain/builtin_plugin.c
@@ -40,6 +40,9 @@
 #include "biquad.h"
 #include "pffft.h"
 #include "convolver.h"
+#include "dsp-ops.h"
+
+static struct dsp_ops dsp_ops;
 
 struct builtin {
 	unsigned long rate;
@@ -82,7 +85,7 @@ static void copy_run(void * Instance, unsigned long SampleCount)
 {
 	struct builtin *impl = Instance;
 	float *in = impl->port[1], *out = impl->port[0];
-	memcpy(out, in, SampleCount * sizeof(float));
+	dsp_ops_copy(&dsp_ops, out, in, SampleCount);
 }
 
 static struct fc_port copy_ports[] = {
@@ -112,10 +115,10 @@ static const struct fc_descriptor copy_desc = {
 static void mixer_run(void * Instance, unsigned long SampleCount)
 {
 	struct builtin *impl = Instance;
-	int i;
-	unsigned long j;
+	int i, n_src = 0;
 	float *out = impl->port[0];
-	bool first = true;
+	const void *src[8];
+	float gains[8];
 
 	if (out == NULL)
 		return;
@@ -127,24 +130,10 @@ static void mixer_run(void * Instance, unsigned long SampleCount)
 		if (in == NULL || gain == 0.0f)
 			continue;
 
-		if (first) {
-			if (gain == 1.0f)
-				memcpy(out, in, SampleCount * sizeof(float));
-			else
-				for (j = 0; j < SampleCount; j++)
-					out[j] = in[j] * gain;
-			first = false;
-		} else {
-			if (gain == 1.0f)
-				for (j = 0; j < SampleCount; j++)
-					out[j] += in[j];
-			else
-				for (j = 0; j < SampleCount; j++)
-					out[j] += in[j] * gain;
-		}
+		src[n_src] = in;
+		gains[n_src++] = gain;
 	}
-	if (first)
-		memset(out, 0, SampleCount * sizeof(float));
+	dsp_ops_mix_gain(&dsp_ops, out, src, gains, n_src, SampleCount);
 }
 
 static struct fc_port mixer_ports[] = {
@@ -941,7 +930,11 @@ struct fc_plugin *load_builtin_plugin(const struct spa_support *support, uint32_
 		const char *plugin, const char *config)
 {
 	struct spa_cpu *cpu_iface;
+	uint32_t cpu_flags;
 	cpu_iface = spa_support_find(support, n_support, SPA_TYPE_INTERFACE_CPU);
-	pffft_select_cpu(cpu_iface ? spa_cpu_get_flags(cpu_iface) : 0);
+	cpu_flags = cpu_iface ? spa_cpu_get_flags(cpu_iface) : 0;
+	dsp_ops.cpu_flags = cpu_flags;
+	dsp_ops_init(&dsp_ops);
+	pffft_select_cpu(cpu_flags);
 	return &builtin_plugin;
 }
diff --git a/src/modules/module-filter-chain/dsp-ops-c.c b/src/modules/module-filter-chain/dsp-ops-c.c
new file mode 100644
index 000000000..885c8cd38
--- /dev/null
+++ b/src/modules/module-filter-chain/dsp-ops-c.c
@@ -0,0 +1,100 @@
+/* Spa
+ *
+ * Copyright © 2022 Wim Taymans
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+
+#include <spa/utils/defs.h>
+
+#include "dsp-ops.h"
+
+static inline void dsp_clear_c(struct dsp_ops *ops, void * SPA_RESTRICT dst, uint32_t n_samples)
+{
+	memset(dst, 0, sizeof(float) * n_samples);
+}
+
+static inline void dsp_add_c(struct dsp_ops *ops, void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, uint32_t n_samples)
+{
+	uint32_t i;
+	const float *s = src;
+	float *d = dst;
+	for (i = 0; i < n_samples; i++)
+		d[i] += s[i];
+}
+
+static inline void dsp_gain_c(struct dsp_ops *ops, void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, float gain, uint32_t n_samples)
+{
+	uint32_t i;
+	const float *s = src;
+	float *d = dst;
+	for (i = 0; i < n_samples; i++)
+		d[i] = s[i] * gain;
+}
+
+static inline void dsp_gain_add_c(struct dsp_ops *ops, void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, float gain, uint32_t n_samples)
+{
+	uint32_t i;
+	const float *s = src;
+	float *d = dst;
+	for (i = 0; i < n_samples; i++)
+		d[i] += s[i] * gain;
+}
+
+
+void dsp_copy_c(struct dsp_ops *ops, void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, uint32_t n_samples)
+{
+	if (dst != src)
+		spa_memcpy(dst, src, sizeof(float) * n_samples);
+}
+
+void dsp_mix_gain_c(struct dsp_ops *ops,
+		void * SPA_RESTRICT dst,
+		const void * SPA_RESTRICT src[],
+		float gain[], uint32_t n_src, uint32_t n_samples)
+{
+	uint32_t i;
+	if (n_src == 0) {
+		dsp_clear_c(ops, dst, n_samples);
+	} else if (n_src == 1) {
+		if (dst != src[0])
+			dsp_copy_c(ops, dst, src[0], n_samples);
+	} else {
+		if (gain[0] == 1.0f)
+			dsp_copy_c(ops, dst, src[0], n_samples);
+		else
+			dsp_gain_c(ops, dst, src[0], gain[0], n_samples);
+
+		for (i = 1; i < n_src; i++) {
+			if (gain[i] == 1.0f)
+				dsp_add_c(ops, dst, src[i], n_samples);
+			else
+				dsp_gain_add_c(ops, dst, src[i], gain[i], n_samples);
+		}
+	}
+}
diff --git a/src/modules/module-filter-chain/dsp-ops-sse.c b/src/modules/module-filter-chain/dsp-ops-sse.c
new file mode 100644
index 000000000..cabcae340
--- /dev/null
+++ b/src/modules/module-filter-chain/dsp-ops-sse.c
@@ -0,0 +1,91 @@
+/* Spa
+ *
+ * Copyright © 2022 Wim Taymans
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+
+#include <spa/utils/defs.h>
+
+#include "dsp-ops.h"
+
+#include <xmmintrin.h>
+
+void dsp_mix_gain_sse(struct dsp_ops *ops,
+		void * SPA_RESTRICT dst,
+		const void * SPA_RESTRICT src[],
+		float gain[], uint32_t n_src, uint32_t n_samples)
+{
+	if (n_src == 0) {
+		memset(dst, 0, n_samples * sizeof(float));
+	} else if (n_src == 1) {
+		if (dst != src[0])
+			spa_memcpy(dst, src[0], n_samples * sizeof(float));
+	} else {
+		uint32_t n, i, unrolled;
+		__m128 in[4], g;
+		const float **s = (const float **)src;
+		float *d = dst;
+
+		if (SPA_LIKELY(SPA_IS_ALIGNED(dst, 16))) {
+			unrolled = n_samples & ~15;
+			for (i = 0; i < n_src; i++) {
+				if (SPA_UNLIKELY(!SPA_IS_ALIGNED(src[i], 16))) {
+					unrolled = 0;
+					break;
+				}
+			}
+		} else
+			unrolled = 0;
+
+		for (n = 0; n < unrolled; n += 16) {
+			g = _mm_set1_ps(gain[0]);
+			in[0] = _mm_mul_ps(g, _mm_load_ps(&s[0][n+ 0]));
+			in[1] = _mm_mul_ps(g, _mm_load_ps(&s[0][n+ 4]));
+			in[2] = _mm_mul_ps(g, _mm_load_ps(&s[0][n+ 8]));
+			in[3] = _mm_mul_ps(g, _mm_load_ps(&s[0][n+12]));
+
+			for (i = 1; i < n_src; i++) {
+				g = _mm_set1_ps(gain[i]);
+				in[0] = _mm_add_ps(in[0], _mm_mul_ps(g, _mm_load_ps(&s[i][n+ 0])));
+				in[1] = _mm_add_ps(in[1], _mm_mul_ps(g, _mm_load_ps(&s[i][n+ 4])));
+				in[2] = _mm_add_ps(in[2], _mm_mul_ps(g, _mm_load_ps(&s[i][n+ 8])));
+				in[3] = _mm_add_ps(in[3], _mm_mul_ps(g, _mm_load_ps(&s[i][n+12])));
+			}
+			_mm_store_ps(&d[n+ 0], in[0]);
+			_mm_store_ps(&d[n+ 4], in[1]);
+			_mm_store_ps(&d[n+ 8], in[2]);
+			_mm_store_ps(&d[n+12], in[3]);
+		}
+		for (; n < n_samples; n++) {
+			g = _mm_set_ss(gain[0]);
+			in[0] = _mm_mul_ss(g, _mm_load_ss(&s[0][n]));
+			for (i = 1; i < n_src; i++) {
+				g = _mm_set_ss(gain[i]);
+				in[0] = _mm_add_ss(in[0], _mm_mul_ss(g, _mm_load_ss(&s[i][n])));
+			}
+			_mm_store_ss(&d[n], in[0]);
+		}
+	}
+}
diff --git a/src/modules/module-filter-chain/dsp-ops.c b/src/modules/module-filter-chain/dsp-ops.c
new file mode 100644
index 000000000..b40278bf6
--- /dev/null
+++ b/src/modules/module-filter-chain/dsp-ops.c
@@ -0,0 +1,92 @@
+/* Spa
+ *
+ * Copyright © 2022 Wim Taymans
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <string.h>
+#include <stdio.h>
+#include <math.h>
+
+#include <spa/support/cpu.h>
+#include <spa/utils/defs.h>
+#include <spa/param/audio/format-utils.h>
+
+#include "dsp-ops.h"
+
+struct dsp_info {
+	uint32_t cpu_flags;
+
+	void (*copy) (struct dsp_ops *ops,
+			void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, uint32_t n_samples);
+	void (*mix_gain) (struct dsp_ops *ops,
+			void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src[],
+			float gain[], uint32_t n_src, uint32_t n_samples);
+};
+
+static struct dsp_info dsp_table[] =
+{
+#if defined (HAVE_SSE)
+	{ SPA_CPU_FLAG_SSE,
+		.copy = dsp_copy_c,
+		.mix_gain = dsp_mix_gain_sse,
+	},
+#endif
+	{ 0,
+		.copy = dsp_copy_c,
+		.mix_gain = dsp_mix_gain_c,
+	},
+};
+
+#define MATCH_CPU_FLAGS(a,b)	((a) == 0 || ((a) & (b)) == a)
+
+static const struct dsp_info *find_dsp_info(uint32_t cpu_flags)
+{
+	SPA_FOR_EACH_ELEMENT_VAR(dsp_table, t) {
+		if (MATCH_CPU_FLAGS(t->cpu_flags, cpu_flags))
+			return t;
+	}
+	return NULL;
+}
+
+static void impl_dsp_ops_free(struct dsp_ops *ops)
+{
+	spa_zero(*ops);
+}
+
+int dsp_ops_init(struct dsp_ops *ops)
+{
+	const struct dsp_info *info;
+
+	info = find_dsp_info(ops->cpu_flags);
+	if (info == NULL)
+		return -ENOTSUP;
+
+	ops->priv = info;
+	ops->cpu_flags = info->cpu_flags;
+	ops->copy = info->copy;
+	ops->mix_gain = info->mix_gain;
+	ops->free = impl_dsp_ops_free;
+
+	return 0;
+}
diff --git a/src/modules/module-filter-chain/dsp-ops.h b/src/modules/module-filter-chain/dsp-ops.h
new file mode 100644
index 000000000..ffbca6ce7
--- /dev/null
+++ b/src/modules/module-filter-chain/dsp-ops.h
@@ -0,0 +1,62 @@
+/* Spa
+ *
+ * Copyright © 2022 Wim Taymans
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include <spa/utils/defs.h>
+
+struct dsp_ops {
+	uint32_t cpu_flags;
+
+	void (*clear) (struct dsp_ops *ops, void * SPA_RESTRICT dst, uint32_t n_samples);
+	void (*copy) (struct dsp_ops *ops,
+			void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src, uint32_t n_samples);
+	void (*mix_gain) (struct dsp_ops *ops,
+			void * SPA_RESTRICT dst,
+			const void * SPA_RESTRICT src[],
+			float gain[], uint32_t n_src, uint32_t n_samples);
+	void (*free) (struct dsp_ops *ops);
+
+	const void *priv;
+};
+
+int dsp_ops_init(struct dsp_ops *ops);
+
+#define dsp_ops_copy(ops,...)		(ops)->copy(ops, __VA_ARGS__)
+#define dsp_ops_mix_gain(ops,...)	(ops)->mix_gain(ops, __VA_ARGS__)
+#define dsp_ops_free(ops)		(ops)->free(ops)
+
+
+#define MAKE_COPY_FUNC(arch) \
+void dsp_copy_##arch(struct dsp_ops *ops, void * SPA_RESTRICT dst, \
+			const void * SPA_RESTRICT src, uint32_t n_samples)
+#define MAKE_MIX_GAIN_FUNC(arch) \
+void dsp_mix_gain_##arch(struct dsp_ops *ops, void * SPA_RESTRICT dst,	\
+	const void * SPA_RESTRICT src[], float gain[], uint32_t n_src, uint32_t n_samples)
+
+
+MAKE_COPY_FUNC(c);
+MAKE_MIX_GAIN_FUNC(c);
+#if defined (HAVE_SSE)
+MAKE_MIX_GAIN_FUNC(sse);
+#endif