ardour/ardour
ardour
/
ardour
13
1
Fork 0
Code Pull Requests Releases Activity

Fix potential memory corruption in AVX512f DSP

This commit is contained in:
Robin Gareus 2024-04-10 03:09:05 +02:00
parent 8a7d92f196
commit d023ea0d77
Signed by: rgareus
GPG Key ID: A090BCE02CF57F04
1 changed files with 19 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
libs/ardour/x86_functions_avx512f.cc
View File

@ -1,5 +1,6 @@
/*
* Copyright (C) 2023 Ayan Shafqat <ayan.x.shafqat@gmail.com>
* Copyright (C) 2024 Robin Gareus <robin@gareus.org>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
@ -50,7 +51,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nframes, float current)
break;
}
if (IS_ALIGNED_TO(src, sizeof(__m256))) {
if (frames >=8 && IS_ALIGNED_TO(src, sizeof(__m256))) {
__m512 x = _mm512_castps256_ps512(_mm256_load_ps(src));
x = _mm512_abs_ps(x);
@ -61,7 +62,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nframes, float current)
continue;
}
if (IS_ALIGNED_TO(src, sizeof(__m128))) {
if (frames >= 4 && IS_ALIGNED_TO(src, sizeof(__m128))) {
__m512 x = _mm512_castps128_ps512(_mm_load_ps(src));
x = _mm512_abs_ps(x);
@ -286,7 +287,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nframes, float *minf, float *m
break;
}
if (IS_ALIGNED_TO(src, sizeof(__m256))) {
if (frames >= 8 && IS_ALIGNED_TO(src, sizeof(__m256))) {
__m512 x = _mm512_castps256_ps512(_mm256_load_ps(src));
zmin = _mm512_min_ps(zmin, x);
@ -297,7 +298,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nframes, float *minf, float *m
continue;
}
if (IS_ALIGNED_TO(src, sizeof(__m128))) {
if (frames >= 4 && IS_ALIGNED_TO(src, sizeof(__m128))) {
__m512 x = _mm512_castps128_ps512(_mm_load_ps(src));
zmin = _mm512_min_ps(zmin, x);
@ -522,7 +523,7 @@ x86_avx512f_apply_gain_to_buffer(float *dst, uint32_t nframes, float gain)
break;
}
if (IS_ALIGNED_TO(dst, sizeof(__m256))) {
if (frames >= 8 && IS_ALIGNED_TO(dst, sizeof(__m256))) {
__m256 x = _mm256_load_ps(dst);
__m256 y = _mm256_mul_ps(ygain, x);
_mm256_store_ps(dst, y);
@ -531,7 +532,7 @@ x86_avx512f_apply_gain_to_buffer(float *dst, uint32_t nframes, float gain)
continue;
}
if (IS_ALIGNED_TO(dst, sizeof(__m128))) {
if (frames >= 4 && IS_ALIGNED_TO(dst, sizeof(__m128))) {
__m128 x = _mm_load_ps(dst);
__m128 y = _mm_mul_ps(xgain, x);
_mm_store_ps(dst, y);
@ -658,7 +659,8 @@ x86_avx512f_mix_buffers_with_gain(float *dst, const float *src, uint32_t nframes
break;
}
if (IS_ALIGNED_TO(src, sizeof(__m256)) &&
if (frames >= 8 &&
IS_ALIGNED_TO(src, sizeof(__m256)) &&
IS_ALIGNED_TO(dst, sizeof(__m256))) {
__m256 x = _mm256_load_ps(src);
__m256 y = _mm256_load_ps(dst);
@ -672,7 +674,8 @@ x86_avx512f_mix_buffers_with_gain(float *dst, const float *src, uint32_t nframes
continue;
}
if (IS_ALIGNED_TO(src, sizeof(__m128)) &&
if (frames >= 4 &&
IS_ALIGNED_TO(src, sizeof(__m128)) &&
IS_ALIGNED_TO(dst, sizeof(__m128))) {
__m128 x = _mm_load_ps(src);
__m128 y = _mm_load_ps(dst);
@ -826,7 +829,8 @@ x86_avx512f_mix_buffers_no_gain(float *dst, const float *src, uint32_t nframes)
break;
}
if (IS_ALIGNED_TO(src, sizeof(__m256)) &&
if (frames >= 8 &&
IS_ALIGNED_TO(src, sizeof(__m256)) &&
IS_ALIGNED_TO(dst, sizeof(__m256))) {
__m256 x = _mm256_load_ps(src);
__m256 y = _mm256_load_ps(dst);
@ -838,7 +842,8 @@ x86_avx512f_mix_buffers_no_gain(float *dst, const float *src, uint32_t nframes)
continue;
}
if (IS_ALIGNED_TO(src, sizeof(__m128)) &&
if (frames >= 4 &&
IS_ALIGNED_TO(src, sizeof(__m128)) &&
IS_ALIGNED_TO(dst, sizeof(__m128))) {
__m128 x = _mm_load_ps(src);
__m128 y = _mm_load_ps(dst);
@ -989,7 +994,8 @@ x86_avx512f_copy_vector(float *dst, const float *src, uint32_t nframes)
break;
}
if (IS_ALIGNED_TO(src, sizeof(__m256)) &&
if (frames >= 8 &&
IS_ALIGNED_TO(src, sizeof(__m256)) &&
IS_ALIGNED_TO(dst, sizeof(__m256))) {
__m256 x = _mm256_load_ps(src);
_mm256_store_ps(dst, x);
@ -999,7 +1005,8 @@ x86_avx512f_copy_vector(float *dst, const float *src, uint32_t nframes)
continue;
}
if (IS_ALIGNED_TO(src, sizeof(__m128)) &&
if (frames >= 4 &&
IS_ALIGNED_TO(src, sizeof(__m128)) &&
IS_ALIGNED_TO(dst, sizeof(__m128))) {
__m128 x = _mm_load_ps(src);
_mm_store_ps(dst, x);