Attachment #241124 for bug #270056




PORTNAME=	ardour
DISTVERSION=	7.3.0
PORTREVISION=	1
CATEGORIES=	audio
# Official source download is tailored to browsers, downloads from github are
# deliberately made empty. Put the badly named distfile into a subdirectory.
MASTER_SITES=	https://community.ardour.org/download/release/
DISTNAME=	Ardour-${DISTVERSION}
DISTFILES=	522
DIST_SUBDIR=	${DISTNAME}

MAINTAINER=	dev@submerge.ch


BUILD_DEPENDS=	lv2>=1.18.0:audio/lv2 \
		itstool>2.0.0:textproc/itstool \
		openssl>0:security/openssl \
		${BUILD_DEPENDS_${ARCH}}
BUILD_DEPENDS_aarch64=	as:devel/binutils
LIB_DEPENDS=	libserd-0.so:devel/serd \

Lines 1-3 Link Here

(-)b/audio/ardour/distinfo (-3 / +3 lines)
1	TIMESTAMP = 1671141631	1	TIMESTAMP = 1676815269
2	SHA256 (Ardour-7.2.0/516) = c92d5b1dd09e7b8a580121478934b30cd9322fe052f45229dcf49363ca6e377f	2	SHA256 (Ardour-7.3.0/522) = 1f45263b4fbfd69e605125bd9094b43fac85d164f1839d29681b8d4ee5695f29
3	SIZE (Ardour-7.2.0/516) = 12882609	3	SIZE (Ardour-7.3.0/522) = 12897398




--- libs/ardour/ardour/cycles.h.orig	2023-01-03 16:39:59 UTC
+++ libs/ardour/ardour/cycles.h
@@ -72,12 +72,22 @@ static inline cycles_t get_cycles (void)
 
 #elif defined(__powerpc64__)
 
+#ifdef __linux__
 #include <sys/platform/ppc.h>
 typedef uint64_t cycles_t;
 static inline cycles_t get_cycles(void)
 {
 	return __ppc_get_timebase();
 }
+#elif defined(__FreeBSD__)
+typedef uint64_t cycles_t;
+static inline cycles_t get_cycles(void)
+{
+	cycles_t tbr;
+	asm volatile("mfspr %0, 268" : "=r"(tbr));
+	return tbr;
+}
+#endif
 
 #elif defined(__powerpc__)
 

Added Link Here

(-)b/audio/ardour/files/patch-libs_ardour_sse__functions__avx.cc (+11 lines)
1	--- libs/ardour/sse_functions_avx.cc.orig 2023-02-21 01:18:30 UTC
2	+++ libs/ardour/sse_functions_avx.cc
3	@@ -47,7 +47,7 @@ x86_sse_avx_find_peaks(const float* buf, uint32_t nfra
4	// load each 64 bytes into cash before processing
5	while (nframes >= 16) {
6	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
7	- _mm_prefetch(((char*)buf+64), _mm_hint(0) );
8	+ _mm_prefetch(((char*)buf+64), _MM_HINT_NTA );
9	#else
10	__builtin_prefetch(buf+64,0,0);
11	#endif




--- libs/ardour/sse_functions_avx_linux.cc.orig	2023-02-21 01:17:58 UTC
+++ libs/ardour/sse_functions_avx_linux.cc
@@ -106,7 +106,7 @@ x86_sse_avx_compute_peak(const float *src, uint32_t nf
 	while (nframes >= 32)
 	{
 #ifdef _WIN32
-		_mm_prefetch(reinterpret_cast<char const *>(src + 32), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<char const *>(src + 32), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
 #endif
@@ -196,7 +196,7 @@ x86_sse_avx_find_peaks(const float *src, uint32_t nfra
 	while (nframes >= 32)
 	{
 #ifdef _WIN32
-		_mm_prefetch(reinterpret_cast<char const *>(src + 32), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<char const *>(src + 32), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
 #endif
@@ -278,7 +278,7 @@ x86_sse_avx_apply_gain_to_buffer(float *dst, uint32_t 
 	while (frames >= 16)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
 #endif
@@ -410,8 +410,8 @@ x86_sse_avx_mix_buffers_with_gain_unaligned(float *dst
 	while (nframes >= 16)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
-		_mm_prefetch(((char *)src + (16 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
+		_mm_prefetch(((char *)src + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
@@ -503,8 +503,8 @@ x86_sse_avx_mix_buffers_with_gain_aligned(float *dst, 
 	while (nframes >= 16)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
-		_mm_prefetch(((char *)src + (16 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
+		_mm_prefetch(((char *)src + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
@@ -592,8 +592,8 @@ x86_sse_avx_mix_buffers_no_gain_unaligned(float *dst, 
 	while (nframes >= 16)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
-		_mm_prefetch(((char *)src + (16 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
+		_mm_prefetch(((char *)src + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
@@ -674,8 +674,8 @@ x86_sse_avx_mix_buffers_no_gain_aligned(float *dst, co
 	while (nframes >= 32)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (32 * sizeof(float))), _mm_hint(0));
-		_mm_prefetch(((char *)src + (32 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (32 * sizeof(float))), _MM_HINT_NTA);
+		_mm_prefetch(((char *)src + (32 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 32), 0, 0);
@@ -717,8 +717,8 @@ x86_sse_avx_mix_buffers_no_gain_aligned(float *dst, co
 	while (nframes >= 16)
 	{
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
-		_mm_prefetch(((char *)src + (16 * sizeof(float))), _mm_hint(0));
+		_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
+		_mm_prefetch(((char *)src + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 		__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
 		__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);




--- libs/ardour/x86_functions_avx512f.cc.orig	2023-02-21 01:05:04 UTC
+++ libs/ardour/x86_functions_avx512f.cc
@@ -83,7 +83,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
 	}
 
 	while (frames >= 256) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -142,7 +142,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
 	}
 
 	while (frames >= 128) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -176,7 +176,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
 	}
 
 	while (frames >= 64) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -310,7 +310,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
 	}
 
 	while (frames >= 256) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -371,7 +371,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
 	}
 
 	while (frames >= 128) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -405,7 +405,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
 	}
 
 	while (frames >= 64) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -532,7 +532,7 @@ x86_avx512f_apply_gain_to_buffer(float *dst, uint32_t 
 
 	// Process the remaining samples 128 at a time
 	while (frames >= 128) {
-		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(dst + 0);
 		__m512 x1 = _mm512_load_ps(dst + 16);
@@ -679,8 +679,8 @@ x86_avx512f_mix_buffers_with_gain(float *dst, const fl
 
 	// Process the remaining samples 128 at a time
 	while (frames >= 128) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
-		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
+		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -836,8 +836,8 @@ x86_avx512f_mix_buffers_no_gain(float *dst, const floa
 
 	// Process the remaining samples 128 at a time
 	while (frames >= 128) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
-		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
+		_mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -987,8 +987,8 @@ x86_avx512f_copy_vector(float *dst, const float *src, 
 
 	// Process 256 samples at a time
 	while (frames >= 256) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
-		_mm_prefetch(reinterpret_cast<void const *>(dst + 256), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
+		_mm_prefetch(reinterpret_cast<void const *>(dst + 256), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);
@@ -1033,8 +1033,8 @@ x86_avx512f_copy_vector(float *dst, const float *src, 
 
 	// Process remaining samples 64 at a time
 	while (frames >= 64) {
-		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
-		_mm_prefetch(reinterpret_cast<void const *>(dst + 64), _mm_hint(0));
+		_mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
+		_mm_prefetch(reinterpret_cast<void const *>(dst + 64), _MM_HINT_NTA);
 
 		__m512 x0 = _mm512_load_ps(src + 0);
 		__m512 x1 = _mm512_load_ps(src + 16);




--- libs/ardour/x86_functions_fma.cc.orig	2023-02-21 01:16:53 UTC
+++ libs/ardour/x86_functions_fma.cc
@@ -67,8 +67,8 @@ x86_fma_mix_buffers_with_gain(
 
 		while (nframes >= 16) {
 #if defined(COMPILER_MSVC) || defined(COMPILER_MINGW)
-			_mm_prefetch(((char *)dst + (16 * sizeof(float))), _mm_hint(0));
-			_mm_prefetch(((char *)src + (16 * sizeof(float))), _mm_hint(0));
+			_mm_prefetch(((char *)dst + (16 * sizeof(float))), _MM_HINT_NTA);
+			_mm_prefetch(((char *)src + (16 * sizeof(float))), _MM_HINT_NTA);
 #else
 			__builtin_prefetch(src + (16 * sizeof(float)), 0, 0);
 			__builtin_prefetch(dst + (16 * sizeof(float)), 0, 0);

Lines 836-841 share/applications/%%ARDOUR_MAJOR%%.desktop Link Here

(-)b/audio/ardour/pkg-plist (+1 lines)
836	%%DATADIR%%/scripts/reset_mixer.lua	836	%%DATADIR%%/scripts/reset_mixer.lua
837	%%DATADIR%%/scripts/rubberband_swing.lua	837	%%DATADIR%%/scripts/rubberband_swing.lua
838	%%DATADIR%%/scripts/s_chanmap.lua	838	%%DATADIR%%/scripts/s_chanmap.lua
		839	%%DATADIR%%/scripts/s_cut_copy_section.lua
839	%%DATADIR%%/scripts/s_ducks.lua	840	%%DATADIR%%/scripts/s_ducks.lua
840	%%DATADIR%%/scripts/s_fader_automation.lua	841	%%DATADIR%%/scripts/s_fader_automation.lua
841	%%DATADIR%%/scripts/s_foreach_track.lua	842	%%DATADIR%%/scripts/s_foreach_track.lua

Return to bug 270056

Lines 1-12 Link Here

(-)b/audio/ardour/Makefile (-3 / +3 lines)
1	PORTNAME= ardour	1	PORTNAME= ardour
2	DISTVERSION= 7.2.0	2	DISTVERSION= 7.3.0
3	PORTREVISION= 1
4	CATEGORIES= audio	3	CATEGORIES= audio
5	# Official source download is tailored to browsers, downloads from github are	4	# Official source download is tailored to browsers, downloads from github are
6	# deliberately made empty. Put the badly named distfile into a subdirectory.	5	# deliberately made empty. Put the badly named distfile into a subdirectory.
7	MASTER_SITES= https://community.ardour.org/download/release/	6	MASTER_SITES= https://community.ardour.org/download/release/
8	DISTNAME= Ardour-${DISTVERSION}	7	DISTNAME= Ardour-${DISTVERSION}
9	DISTFILES= 516	8	DISTFILES= 522
10	DIST_SUBDIR= ${DISTNAME}	9	DIST_SUBDIR= ${DISTNAME}
11		10
12	MAINTAINER= dev@submerge.ch	11	MAINTAINER= dev@submerge.ch
Lines 18-23 LICENSE_FILE= ${WRKSRC}/COPYING Link Here
18		17
19	BUILD_DEPENDS= lv2>=1.18.0:audio/lv2 \	18	BUILD_DEPENDS= lv2>=1.18.0:audio/lv2 \
20	itstool>2.0.0:textproc/itstool \	19	itstool>2.0.0:textproc/itstool \
		20	openssl>0:security/openssl \
21	${BUILD_DEPENDS_${ARCH}}	21	${BUILD_DEPENDS_${ARCH}}
22	BUILD_DEPENDS_aarch64= as:devel/binutils	22	BUILD_DEPENDS_aarch64= as:devel/binutils
23	LIB_DEPENDS= libserd-0.so:devel/serd \	23	LIB_DEPENDS= libserd-0.so:devel/serd \

Removed Link Here

(-)a/audio/ardour/files/patch-libs_ardour_ardour_cycles.h (-25 lines)
1	--- libs/ardour/ardour/cycles.h.orig 2023-01-03 16:39:59 UTC
2	+++ libs/ardour/ardour/cycles.h
3	@@ -72,12 +72,22 @@ static inline cycles_t get_cycles (void)
4
5	#elif defined(__powerpc64__)
6
7	+#ifdef __linux__
8	#include <sys/platform/ppc.h>
9	typedef uint64_t cycles_t;
10	static inline cycles_t get_cycles(void)
11	{
12	return __ppc_get_timebase();
13	}
14	+#elif defined(__FreeBSD__)
15	+typedef uint64_t cycles_t;
16	+static inline cycles_t get_cycles(void)
17	+{
18	+ cycles_t tbr;
19	+ asm volatile("mfspr %0, 268" : "=r"(tbr));
20	+ return tbr;
21	+}
22	+#endif
23
24	#elif defined(__powerpc__)
25

Added Link Here

(-)b/audio/ardour/files/patch-libs_ardour_sse__functions__avx__linux.cc (+84 lines)
1	--- libs/ardour/sse_functions_avx_linux.cc.orig 2023-02-21 01:17:58 UTC
2	+++ libs/ardour/sse_functions_avx_linux.cc
3	@@ -106,7 +106,7 @@ x86_sse_avx_compute_peak(const float *src, uint32_t nf
4	while (nframes >= 32)
5	{
6	#ifdef _WIN32
7	- _mm_prefetch(reinterpret_cast<char const *>(src + 32), _mm_hint(0));
8	+ _mm_prefetch(reinterpret_cast<char const *>(src + 32), _MM_HINT_NTA);
9	#else
10	__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
11	#endif
12	@@ -196,7 +196,7 @@ x86_sse_avx_find_peaks(const float *src, uint32_t nfra
13	while (nframes >= 32)
14	{
15	#ifdef _WIN32
16	- _mm_prefetch(reinterpret_cast<char const *>(src + 32), _mm_hint(0));
17	+ _mm_prefetch(reinterpret_cast<char const *>(src + 32), _MM_HINT_NTA);
18	#else
19	__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
20	#endif
21	@@ -278,7 +278,7 @@ x86_sse_avx_apply_gain_to_buffer(float *dst, uint32_t
22	while (frames >= 16)
23	{
24	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
25	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
26	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
27	#else
28	__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
29	#endif
30	@@ -410,8 +410,8 @@ x86_sse_avx_mix_buffers_with_gain_unaligned(float *dst
31	while (nframes >= 16)
32	{
33	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
34	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
35	- _mm_prefetch(((char )src + (16 sizeof(float))), _mm_hint(0));
36	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
37	+ _mm_prefetch(((char )src + (16 sizeof(float))), _MM_HINT_NTA);
38	#else
39	__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
40	__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
41	@@ -503,8 +503,8 @@ x86_sse_avx_mix_buffers_with_gain_aligned(float *dst,
42	while (nframes >= 16)
43	{
44	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
45	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
46	- _mm_prefetch(((char )src + (16 sizeof(float))), _mm_hint(0));
47	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
48	+ _mm_prefetch(((char )src + (16 sizeof(float))), _MM_HINT_NTA);
49	#else
50	__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
51	__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
52	@@ -592,8 +592,8 @@ x86_sse_avx_mix_buffers_no_gain_unaligned(float *dst,
53	while (nframes >= 16)
54	{
55	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
56	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
57	- _mm_prefetch(((char )src + (16 sizeof(float))), _mm_hint(0));
58	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
59	+ _mm_prefetch(((char )src + (16 sizeof(float))), _MM_HINT_NTA);
60	#else
61	__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
62	__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);
63	@@ -674,8 +674,8 @@ x86_sse_avx_mix_buffers_no_gain_aligned(float *dst, co
64	while (nframes >= 32)
65	{
66	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
67	- _mm_prefetch(((char )dst + (32 sizeof(float))), _mm_hint(0));
68	- _mm_prefetch(((char )src + (32 sizeof(float))), _mm_hint(0));
69	+ _mm_prefetch(((char )dst + (32 sizeof(float))), _MM_HINT_NTA);
70	+ _mm_prefetch(((char )src + (32 sizeof(float))), _MM_HINT_NTA);
71	#else
72	__builtin_prefetch(reinterpret_cast<void const *>(src + 32), 0, 0);
73	__builtin_prefetch(reinterpret_cast<void const *>(dst + 32), 0, 0);
74	@@ -717,8 +717,8 @@ x86_sse_avx_mix_buffers_no_gain_aligned(float *dst, co
75	while (nframes >= 16)
76	{
77	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
78	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
79	- _mm_prefetch(((char )src + (16 sizeof(float))), _mm_hint(0));
80	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
81	+ _mm_prefetch(((char )src + (16 sizeof(float))), _MM_HINT_NTA);
82	#else
83	__builtin_prefetch(reinterpret_cast<void const *>(src + 16), 0, 0);
84	__builtin_prefetch(reinterpret_cast<void const *>(dst + 16), 0, 0);

Added Link Here

(-)b/audio/ardour/files/patch-libs_ardour_x86__functions__avx512f.cc (+109 lines)
1	--- libs/ardour/x86_functions_avx512f.cc.orig 2023-02-21 01:05:04 UTC
2	+++ libs/ardour/x86_functions_avx512f.cc
3	@@ -83,7 +83,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
4	}
5
6	while (frames >= 256) {
7	- _mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
8	+ _mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
9
10	__m512 x0 = _mm512_load_ps(src + 0);
11	__m512 x1 = _mm512_load_ps(src + 16);
12	@@ -142,7 +142,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
13	}
14
15	while (frames >= 128) {
16	- _mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
17	+ _mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
18
19	__m512 x0 = _mm512_load_ps(src + 0);
20	__m512 x1 = _mm512_load_ps(src + 16);
21	@@ -176,7 +176,7 @@ x86_avx512f_compute_peak(const float *src, uint32_t nf
22	}
23
24	while (frames >= 64) {
25	- _mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
26	+ _mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
27
28	__m512 x0 = _mm512_load_ps(src + 0);
29	__m512 x1 = _mm512_load_ps(src + 16);
30	@@ -310,7 +310,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
31	}
32
33	while (frames >= 256) {
34	- _mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
35	+ _mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
36
37	__m512 x0 = _mm512_load_ps(src + 0);
38	__m512 x1 = _mm512_load_ps(src + 16);
39	@@ -371,7 +371,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
40	}
41
42	while (frames >= 128) {
43	- _mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
44	+ _mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
45
46	__m512 x0 = _mm512_load_ps(src + 0);
47	__m512 x1 = _mm512_load_ps(src + 16);
48	@@ -405,7 +405,7 @@ x86_avx512f_find_peaks(const float *src, uint32_t nfra
49	}
50
51	while (frames >= 64) {
52	- _mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
53	+ _mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
54
55	__m512 x0 = _mm512_load_ps(src + 0);
56	__m512 x1 = _mm512_load_ps(src + 16);
57	@@ -532,7 +532,7 @@ x86_avx512f_apply_gain_to_buffer(float *dst, uint32_t
58
59	// Process the remaining samples 128 at a time
60	while (frames >= 128) {
61	- _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
62	+ _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
63
64	__m512 x0 = _mm512_load_ps(dst + 0);
65	__m512 x1 = _mm512_load_ps(dst + 16);
66	@@ -679,8 +679,8 @@ x86_avx512f_mix_buffers_with_gain(float *dst, const fl
67
68	// Process the remaining samples 128 at a time
69	while (frames >= 128) {
70	- _mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
71	- _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
72	+ _mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
73	+ _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
74
75	__m512 x0 = _mm512_load_ps(src + 0);
76	__m512 x1 = _mm512_load_ps(src + 16);
77	@@ -836,8 +836,8 @@ x86_avx512f_mix_buffers_no_gain(float *dst, const floa
78
79	// Process the remaining samples 128 at a time
80	while (frames >= 128) {
81	- _mm_prefetch(reinterpret_cast<void const *>(src + 128), _mm_hint(0));
82	- _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _mm_hint(0));
83	+ _mm_prefetch(reinterpret_cast<void const *>(src + 128), _MM_HINT_NTA);
84	+ _mm_prefetch(reinterpret_cast<void const *>(dst + 128), _MM_HINT_NTA);
85
86	__m512 x0 = _mm512_load_ps(src + 0);
87	__m512 x1 = _mm512_load_ps(src + 16);
88	@@ -987,8 +987,8 @@ x86_avx512f_copy_vector(float dst, const float src,
89
90	// Process 256 samples at a time
91	while (frames >= 256) {
92	- _mm_prefetch(reinterpret_cast<void const *>(src + 256), _mm_hint(0));
93	- _mm_prefetch(reinterpret_cast<void const *>(dst + 256), _mm_hint(0));
94	+ _mm_prefetch(reinterpret_cast<void const *>(src + 256), _MM_HINT_NTA);
95	+ _mm_prefetch(reinterpret_cast<void const *>(dst + 256), _MM_HINT_NTA);
96
97	__m512 x0 = _mm512_load_ps(src + 0);
98	__m512 x1 = _mm512_load_ps(src + 16);
99	@@ -1033,8 +1033,8 @@ x86_avx512f_copy_vector(float dst, const float src,
100
101	// Process remaining samples 64 at a time
102	while (frames >= 64) {
103	- _mm_prefetch(reinterpret_cast<void const *>(src + 64), _mm_hint(0));
104	- _mm_prefetch(reinterpret_cast<void const *>(dst + 64), _mm_hint(0));
105	+ _mm_prefetch(reinterpret_cast<void const *>(src + 64), _MM_HINT_NTA);
106	+ _mm_prefetch(reinterpret_cast<void const *>(dst + 64), _MM_HINT_NTA);
107
108	__m512 x0 = _mm512_load_ps(src + 0);
109	__m512 x1 = _mm512_load_ps(src + 16);

Added Link Here

(-)b/audio/ardour/files/patch-libs_ardour_x86__functions__fma.cc (+13 lines)
1	--- libs/ardour/x86_functions_fma.cc.orig 2023-02-21 01:16:53 UTC
2	+++ libs/ardour/x86_functions_fma.cc
3	@@ -67,8 +67,8 @@ x86_fma_mix_buffers_with_gain(
4
5	while (nframes >= 16) {
6	#if defined(COMPILER_MSVC) \|\| defined(COMPILER_MINGW)
7	- _mm_prefetch(((char )dst + (16 sizeof(float))), _mm_hint(0));
8	- _mm_prefetch(((char )src + (16 sizeof(float))), _mm_hint(0));
9	+ _mm_prefetch(((char )dst + (16 sizeof(float))), _MM_HINT_NTA);
10	+ _mm_prefetch(((char )src + (16 sizeof(float))), _MM_HINT_NTA);
11	#else
12	__builtin_prefetch(src + (16 * sizeof(float)), 0, 0);
13	__builtin_prefetch(dst + (16 * sizeof(float)), 0, 0);