Re: [FFmpeg-devel] [PATCH] checkasm: add test for dcmul_add

[flow gg <hlefthleft@gmail.com>]

From 2785ce57f68dbb2373c951b9432afa73796f7cc1 Mon Sep 17 00:00:00 2001
From: sunyuechi <sunyuechi@iscas.ac.cn>
Date: Sat, 18 Nov 2023 10:58:17 +0800
Subject: [PATCH] checkasm: test for dcmul_add

---
 tests/checkasm/af_afir.c | 141 +++++++++++++++++++++++++++------------
 1 file changed, 98 insertions(+), 43 deletions(-)

diff --git a/tests/checkasm/af_afir.c b/tests/checkasm/af_afir.c
index 08c55dacfc..6cb59dbb6c 100644
--- a/tests/checkasm/af_afir.c
+++ b/tests/checkasm/af_afir.c
@@ -33,64 +33,119 @@ do {                                          \
     int i;                                    \
     double bmg[2], stddev = 10.0, mean = 0.0; \
                                               \
-    for (i = 0; i < LEN*2+8; i += 2) {        \
+    for (i = 0; i < BUF_SIZE; i += 2) {        \
         av_bmg_get(&checkasm_lfg, bmg);       \
         buf[i]     = bmg[0] * stddev + mean;  \
         buf[i + 1] = bmg[1] * stddev + mean;  \
     }                                         \
 } while(0);

-static void test_fcmul_add(const float *src0, const float *src1, const
float *src2)
+static void test_fcmul_add(AudioFIRDSPContext *fir)
 {
-    LOCAL_ALIGNED_32(float, cdst, [LEN*2+8]);
-    LOCAL_ALIGNED_32(float, odst, [LEN*2+8]);
-    int i;
-
-    declare_func(void, float *sum, const float *t, const float *c,
-                 ptrdiff_t len);
-
-    memcpy(cdst, src0, (LEN*2+8) * sizeof(float));
-    memcpy(odst, src0, (LEN*2+8) * sizeof(float));
-    call_ref(cdst, src1, src2, LEN);
-    call_new(odst, src1, src2, LEN);
-    for (i = 0; i <= LEN*2; i++) {
-        int idx = i & ~1;
-        float cre = src2[idx];
-        float cim = src2[idx + 1];
-        float tre = src1[idx];
-        float tim = src1[idx + 1];
-        double t = fabs(src0[i]) +
-                   fabs(tre) + fabs(tim) + fabs(cre) + fabs(cim) +
-                   fabs(tre * cre) + fabs(tim * cim) +
-                   fabs(tre * cim) + fabs(tim * cre) +
-                   fabs(tre * cre - tim * cim) +
-                   fabs(tre * cim + tim * cre) +
-                   fabs(cdst[i]) + 1.0;
-        if (!float_near_abs_eps(cdst[i], odst[i], t * 2 * FLT_EPSILON)) {
-            fprintf(stderr, "%d: %- .12f - %- .12f = % .12g\n",
-                    i, cdst[i], odst[i], cdst[i] - odst[i]);
-            fail();
-            break;
+#define BUF_SIZE LEN*2+8
+    LOCAL_ALIGNED_32(float, src0, [BUF_SIZE]);
+    LOCAL_ALIGNED_32(float, src1, [BUF_SIZE]);
+    LOCAL_ALIGNED_32(float, src2, [BUF_SIZE]);
+
+    randomize_buffer(src0);
+    randomize_buffer(src1);
+    randomize_buffer(src2);
+
+    if (check_func(fir->fcmul_add, "fcmul_add")) {
+        LOCAL_ALIGNED_32(float, cdst, [BUF_SIZE]);
+        LOCAL_ALIGNED_32(float, odst, [BUF_SIZE]);
+        int i;
+
+        declare_func(void, float *sum, const float *t, const float *c,
+                     ptrdiff_t len);
+
+        memcpy(cdst, src0, (BUF_SIZE) * sizeof(float));
+        memcpy(odst, src0, (BUF_SIZE) * sizeof(float));
+        call_ref(cdst, src1, src2, LEN);
+        call_new(odst, src1, src2, LEN);
+        for (i = 0; i <= LEN*2; i++) {
+            int idx = i & ~1;
+            float cre = src2[idx];
+            float cim = src2[idx + 1];
+            float tre = src1[idx];
+            float tim = src1[idx + 1];
+            double t = fabs(src0[i]) +
+                       fabs(tre) + fabs(tim) + fabs(cre) + fabs(cim) +
+                       fabs(tre * cre) + fabs(tim * cim) +
+                       fabs(tre * cim) + fabs(tim * cre) +
+                       fabs(tre * cre - tim * cim) +
+                       fabs(tre * cim + tim * cre) +
+                       fabs(cdst[i]) + 1.0;
+            if (!float_near_abs_eps(cdst[i], odst[i], t * 2 *
FLT_EPSILON)) {
+                fprintf(stderr, "%d: %- .12f - %- .12f = % .12g\n",
+                        i, cdst[i], odst[i], cdst[i] - odst[i]);
+                fail();
+                break;
+            }
         }
+        memcpy(odst, src0, (BUF_SIZE) * sizeof(float));
+        bench_new(odst, src1, src2, LEN);
     }
-    memcpy(odst, src0, (LEN*2+8) * sizeof(float));
-    bench_new(odst, src1, src2, LEN);
+
+    report("fcmul_add");
 }

-void checkasm_check_afir(void)
+static void test_dcmul_add(AudioFIRDSPContext *fir)
 {
-    LOCAL_ALIGNED_32(float, src0, [LEN*2+8]);
-    LOCAL_ALIGNED_32(float, src1, [LEN*2+8]);
-    LOCAL_ALIGNED_32(float, src2, [LEN*2+8]);
-    AudioFIRDSPContext fir = { 0 };
-
-    ff_afir_init(&fir);
+#define BUF_SIZE LEN*2+8
+    LOCAL_ALIGNED_32(double, src0, [BUF_SIZE]);
+    LOCAL_ALIGNED_32(double, src1, [BUF_SIZE]);
+    LOCAL_ALIGNED_32(double, src2, [BUF_SIZE]);

     randomize_buffer(src0);
     randomize_buffer(src1);
     randomize_buffer(src2);

-    if (check_func(fir.fcmul_add, "fcmul_add"))
-        test_fcmul_add(src0, src1, src2);
-    report("fcmul_add");
+    if (check_func(fir->dcmul_add, "dcmul_add")) {
+        LOCAL_ALIGNED_32(double, cdst, [BUF_SIZE]);
+        LOCAL_ALIGNED_32(double, odst, [BUF_SIZE]);
+        int i;
+
+        declare_func(void, double *sum, const double *t, const double *c,
+                     ptrdiff_t len);
+
+        memcpy(cdst, src0, (BUF_SIZE) * sizeof(double));
+        memcpy(odst, src0, (BUF_SIZE) * sizeof(double));
+        call_ref(cdst, src1, src2, LEN);
+        call_new(odst, src1, src2, LEN);
+        for (i = 0; i <= LEN*2; i++) {
+            int idx = i & ~1;
+            double cre = src2[idx];
+            double cim = src2[idx + 1];
+            double tre = src1[idx];
+            double tim = src1[idx + 1];
+            double t = fabs(src0[i]) +
+                       fabs(tre) + fabs(tim) + fabs(cre) + fabs(cim) +
+                       fabs(tre * cre) + fabs(tim * cim) +
+                       fabs(tre * cim) + fabs(tim * cre) +
+                       fabs(tre * cre - tim * cim) +
+                       fabs(tre * cim + tim * cre) +
+                       fabs(cdst[i]) + 1.0;
+            if (!double_near_abs_eps(cdst[i], odst[i], t * 2 *
FLT_EPSILON)) {
+                fprintf(stderr, "%d: %- .12f - %- .12f = % .12g\n",
+                        i, cdst[i], odst[i], cdst[i] - odst[i]);
+                fail();
+                break;
+            }
+        }
+        memcpy(odst, src0, (BUF_SIZE) * sizeof(double));
+        bench_new(odst, src1, src2, LEN);
+    }
+
+    report("dcmul_add");
+}
+
+
+void checkasm_check_afir(void)
+{
+    AudioFIRDSPContext fir = { 0 };
+
+    ff_afir_init(&fir);
+    test_fcmul_add(&fir);
+    test_dcmul_add(&fir);
 }
-- 
2.42.1


flow gg <hlefthleft@gmail.com> 于2023年11月18日周六 11:21写道：

>
>
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".