ltpf: Move analysis to fixed point

test/neon/ltpf_neon.c

@@ -0,0 +1,116 @@
+/******************************************************************************
+ *
+ *  Copyright 2022 Google LLC
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at:
+ *
+ *  http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  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.
+ *
+ ******************************************************************************/
+
+#include "neon.h"
+
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+
+/* -------------------------------------------------------------------------- */
+
+#define TEST_NEON
+#include <ltpf.c>
+
+void lc3_put_bits_generic(lc3_bits_t *a, unsigned b, int c)
+{ (void)a, (void)b, (void)c; }
+
+unsigned lc3_get_bits_generic(struct lc3_bits *a, int b)
+{ return (void)a, (void)b, 0; }
+
+/* -------------------------------------------------------------------------- */
+
+static int check_resampler()
+{
+    int16_t __x[60+480], *x = __x + 60;
+    for (int i = -60; i < 480; i++)
+          x[i] = rand() & 0xffff;
+
+    struct lc3_ltpf_hp50_state hp50 = { 0 }, hp50_neon = { 0 };
+    int16_t y[128], y_neon[128];
+
+    resample_16k_12k8(&hp50, x, y, 128);
+    neon_resample_16k_12k8(&hp50_neon, x, y_neon, 128);
+    if (memcmp(y, y_neon, 128 * sizeof(*y)) != 0)
+        return printf("Error\n"), -1;
+
+    resample_32k_12k8(&hp50, x, y, 128);
+    neon_resample_32k_12k8(&hp50_neon, x, y_neon, 128);
+    if (memcmp(y, y_neon, 128 * sizeof(*y)) != 0)
+        return printf("Error\n"), -1;
+
+    resample_48k_12k8(&hp50, x, y, 128);
+    neon_resample_48k_12k8(&hp50_neon, x, y_neon, 128);
+    if (memcmp(y, y_neon, 128 * sizeof(*y)) != 0)
+        return -1;
+
+    return 0;
+}
+
+static int check_dot()
+{
+    int16_t x[200];
+    for (int i = 0; i < 200; i++)
+        x[i] = rand() & 0xffff;
+
+    float y = dot(x, x+3, 128);
+    float y_neon = neon_dot(x, x+3, 128);
+    if (y != y_neon)
+        return -1;
+
+    return 0;
+}
+
+static int check_correlate()
+{
+    int16_t alignas(4) a[500], b[500];
+    float y[100], y_neon[100];
+
+    for (int i = 0; i < 500; i++) {
+        a[i] = rand() & 0xffff;
+        b[i] = rand() & 0xffff;
+    }
+
+    correlate(a, b+200, 128, y, 100);
+    neon_correlate(a, b+200, 128, y_neon, 100);
+    if (memcmp(y, y_neon, 100 * sizeof(*y)) != 0)
+        return -1;
+
+    correlate(a, b+199, 128, y, 99);
+    neon_correlate(a, b+199, 128, y_neon, 99);
+    if (memcmp(y, y_neon, 99 * sizeof(*y)) != 0)
+        return -1;
+
+    return 0;
+}
+
+int check_ltpf(void)
+{
+    int ret;
+
+    if ((ret = check_resampler()) < 0)
+        return ret;
+
+    if ((ret = check_dot()) < 0)
+        return ret;
+
+    if ((ret = check_correlate()) < 0)
+        return ret;
+
+    return 0;
+}

test/neon/makefile.mk

@@ -0,0 +1,31 @@
+#
+# Copyright 2022 Google LLC
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at:
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# 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.
+#
+
+test_neon_src += \
+    $(TEST_DIR)/neon/test_neon.c \
+    $(TEST_DIR)/neon/ltpf_neon.c \
+    $(SRC_DIR)/tables.c
+
+test_neon_include += $(SRC_DIR)
+test_neon_ldlibs += m
+
+$(eval $(call add-bin,test_neon))
+
+test_neon: $(test_neon_bin)
+	@echo "  RUN     $(notdir $<)"
+	$(V)$<
+
+test: test_neon

test/neon/neon.h

@@ -0,0 +1,141 @@
+/******************************************************************************
+ *
+ *  Copyright 2022 Google LLC
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at:
+ *
+ *  http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  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.
+ *
+ ******************************************************************************/
+
+#if __ARM_NEON
+
+#include <arm_neon.h>
+
+#else
+#define __ARM_NEON 1
+
+#include <stdint.h>
+
+typedef struct { int16_t e[4]; } int16x4_t;
+
+typedef struct { int16_t e[8]; } int16x8_t;
+typedef struct { int32_t e[4]; } int32x4_t;
+typedef struct { int64_t e[2]; } int64x2_t;
+
+
+/* ----------------------------------------------------------------------------
+ *  Load / Store
+ * -------------------------------------------------------------------------- */
+
+__attribute__((unused))
+static int16x4_t vld1_s16(const int16_t *p)
+{
+    int16x4_t r;
+
+    for (int i = 0; i < 4; i++)
+        r.e[i] = *(p++);
+
+    return r;
+}
+
+__attribute__((unused))
+static int64x2_t vmovq_n_s64(int64_t v)
+{
+    int64x2_t r;
+
+    r.e[0] = v;
+    r.e[1] = v;
+
+    return r;
+}
+
+
+/* ----------------------------------------------------------------------------
+ *  Move
+ * -------------------------------------------------------------------------- */
+
+__attribute__((unused))
+static int32x4_t vmovq_n_s32(uint32_t v)
+{
+    int32x4_t r;
+
+    for (int i = 0; i < 4; i++)
+        r.e[i] = v;
+
+    return r;
+}
+
+__attribute__((unused))
+static int16x4_t vext_s16(int16x4_t a, int16x4_t b, const int n)
+{
+    int16x4_t r;
+    int i = 0;
+
+    for (; i < n; i++) r.e[3-i] = b.e[(n-1)-i];
+    for (; i < 4; i++) r.e[3-i] = a.e[3-(i-n)];
+
+    return r;
+}
+
+/* ----------------------------------------------------------------------------
+ *  Arithmetic
+ * -------------------------------------------------------------------------- */
+
+__attribute__((unused))
+static int32x4_t vmull_s16(int16x4_t a, int16x4_t b)
+{
+    int32x4_t r;
+
+    for (int i = 0; i < 4; i++)
+        r.e[i] = (int32_t)a.e[i] * b.e[i];
+
+    return r;
+}
+
+__attribute__((unused))
+static int32x4_t vmlal_s16(int32x4_t r, int16x4_t a, int16x4_t b)
+{
+    for (int i = 0; i < 4; i++)
+        r.e[i] += (int32_t)a.e[i] * b.e[i];
+
+    return r;
+}
+
+__attribute__((unused))
+static int64x2_t vpadalq_s32(int64x2_t a, int32x4_t b)
+{
+    int64x2_t r;
+
+    r.e[0] = a.e[0] + ((int64_t)b.e[0] + b.e[1]);
+    r.e[1] = a.e[1] + ((int64_t)b.e[2] + b.e[3]);
+
+    return r;
+}
+
+
+/* ----------------------------------------------------------------------------
+ *  Reduce
+ * -------------------------------------------------------------------------- */
+
+__attribute__((unused))
+static int32_t vaddvq_s32(int32x4_t v)
+{
+    return v.e[0] + v.e[1] + v.e[2] + v.e[3];
+}
+
+__attribute__((unused))
+static int64_t vaddvq_s64(int64x2_t v)
+{
+    return v.e[0] + v.e[1];
+}
+
+#endif /* __ARM_NEON */

test/neon/test_neon.c

@@ -0,0 +1,32 @@
+/******************************************************************************
+ *
+ *  Copyright 2022 Google LLC
+ *
+ *  Licensed under the Apache License, Version 2.0 (the "License");
+ *  you may not use this file except in compliance with the License.
+ *  You may obtain a copy of the License at:
+ *
+ *  http://www.apache.org/licenses/LICENSE-2.0
+ *
+ *  Unless required by applicable law or agreed to in writing, software
+ *  distributed under the License is distributed on an "AS IS" BASIS,
+ *  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.
+ *
+ ******************************************************************************/
+
+#include <stdio.h>
+
+int check_ltpf(void);
+
+int main()
+{
+    int r, ret = 0;
+
+    printf("Checking LTPF Neon... "); fflush(stdout);
+    printf("%s\n", (r = check_ltpf()) == 0 ? "OK" : "Failed");
+    ret = ret || r;
+
+    return ret;
+}