On Thu, 10 Dec 2015, Janne Grunau wrote:
On 2015-12-10 15:46:17 +0200, Martin Storsjö wrote:
On Wed, 9 Dec 2015, Janne Grunau wrote:
+
+#define check_lfe_fir(decifactor, eps) \
+ do { \
+ LOCAL_ALIGNED_16(float, in, [256 / decifactor]); \
+ LOCAL_ALIGNED_16(float, out0, [decifactor * 2]); \
+ LOCAL_ALIGNED_16(float, out1, [decifactor * 2]); \
+ LOCAL_ALIGNED_16(float, coeffs, [256]); \
+ int i; \
+ const float * in_ptr = in + (256 / decifactor) - 1; \
+ declare_func(void, float *out, const float *in, const float *coeffs); \
+ /* repeat the test several times */ \
+ for (i = 0; i < 32; i++) { \
+ int j; \
+ memset(out0, 0, sizeof(*out0) * 2 * decifactor); \
+ memset(out1, 0xFF, sizeof(*out1) * 2 * decifactor); \
+ randomize_lfe_fir(256 / decifactor); \
+ call_ref(out0, in_ptr, coeffs); \
+ call_new(out1, in_ptr, coeffs); \
+ for (j = 0; j < 2 * decifactor; j++) { \
+ if (!float_near_abs_eps(out0[j], out1[j], eps)) { \
+ if (0) { \
+ union av_intfloat32 x, y; x.f = out0[j]; y.f =
out1[j]; \
+ fprintf(stderr, "%3d: %11g (0x%08x); %11g (0x%08x)\n",
\
+ j, x.f, x.i, y.f, y.i); \
+ } \
Is this a leftover, or is it intentionally kept in if (0)?
left semi-intentionally but I don't know how such debugging outputs
should be handled. I removed them in other checks.
In that case I guess it's ok with me. Compared to commented out code, this
at least gets compile tested so it won't rot.
// Martin
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
