From 1e519b9d407fd35538b8d4dfdc723448355e9fe1 Mon Sep 17 00:00:00 2001
From: Michael Niedermayer <michaelni@gmx.at>
Date: Thu, 7 Aug 2014 18:34:35 +0200
Subject: [PATCH] avutil: turn arm setend into a cpuflag
this allows disabling and enabling it
it also prevents crashes if vfpv3 and neon are disabled which previously
would have enabled the flag
And last but not least one can enable setend on cpus like cortex-a8 where
its fast but disabled by default
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
---
configure | 3 +++
libavutil/arm/cpu.c | 9 ++++++++-
libavutil/arm/cpu.h | 7 +------
libavutil/cpu.c | 2 ++
libavutil/cpu.h | 1 +
5 files changed, 15 insertions(+), 7 deletions(-)
diff --git a/configure b/configure
index dfbadd0933e..1224362e77c 100755
--- a/configure
+++ b/configure
@@ -1525,6 +1525,7 @@ ARCH_EXT_LIST_ARM="
neon
vfp
vfpv3
+ setend
"
ARCH_EXT_LIST_MIPS="
@@ -1951,6 +1952,7 @@ neon_deps_any="aarch64 arm"
intrinsics_neon_deps="neon"
vfp_deps_any="aarch64 arm"
vfpv3_deps="vfp"
+setend_deps="arm"
map 'eval ${v}_inline_deps=inline_asm' $ARCH_EXT_LIST_ARM
@@ -4447,6 +4449,7 @@ EOF
enabled neon && check_insn neon 'vadd.i16 q0, q0, q0'
enabled vfp && check_insn vfp 'fadds s0, s0, s0'
enabled vfpv3 && check_insn vfpv3 'vmov.f32 s0, #1.0'
+ enabled setend && check_insn setend 'setend be'
[ $target_os = linux ] || [ $target_os = android ] ||
map 'enabled_any ${v}_external ${v}_inline || disable $v' \
diff --git a/libavutil/arm/cpu.c b/libavutil/arm/cpu.c
index f6cf3385914..f1683e8d767 100644
--- a/libavutil/arm/cpu.c
+++ b/libavutil/arm/cpu.c
@@ -128,6 +128,12 @@ int ff_get_cpu_flags_arm(void)
trickle down. */
if (flags & (AV_CPU_FLAG_VFPV3 | AV_CPU_FLAG_NEON))
flags |= AV_CPU_FLAG_ARMV6T2;
+ else
+ /* Some functions use the 'setend' instruction which is deprecated on ARMv8
+ * and serializing on some ARMv7 cores. This ensures such functions
+ * are only enabled on ARMv6. */
+ flags |= AV_CPU_FLAG_SETEND;
+
if (flags & AV_CPU_FLAG_ARMV6T2)
flags |= AV_CPU_FLAG_ARMV6;
@@ -143,7 +149,8 @@ int ff_get_cpu_flags_arm(void)
AV_CPU_FLAG_ARMV6T2 * HAVE_ARMV6T2 |
AV_CPU_FLAG_VFP * HAVE_VFP |
AV_CPU_FLAG_VFPV3 * HAVE_VFPV3 |
- AV_CPU_FLAG_NEON * HAVE_NEON;
+ AV_CPU_FLAG_NEON * HAVE_NEON |
+ AV_CPU_FLAG_SETEND * !(HAVE_NEON | HAVE_VFPV3);
}
#endif
diff --git a/libavutil/arm/cpu.h b/libavutil/arm/cpu.h
index f056a91ca56..9b3b6ff58b7 100644
--- a/libavutil/arm/cpu.h
+++ b/libavutil/arm/cpu.h
@@ -29,11 +29,6 @@
#define have_vfp(flags) CPUEXT(flags, VFP)
#define have_vfpv3(flags) CPUEXT(flags, VFPV3)
#define have_neon(flags) CPUEXT(flags, NEON)
-
-/* Some functions use the 'setend' instruction which is deprecated on ARMv8
- * and serializing on some ARMv7 cores. This macro ensures such functions
- * are only enabled on ARMv6. */
-#define have_setend(flags) \
- (have_armv6(flags) && !(have_vfpv3(flags) || have_neon(flags)))
+#define have_setend(flags) CPUEXT(flags, SETEND)
#endif /* AVUTIL_ARM_CPU_H */
diff --git a/libavutil/cpu.c b/libavutil/cpu.c
index 1cd0522dc6f..53c12273ef5 100644
--- a/libavutil/cpu.c
+++ b/libavutil/cpu.c
@@ -225,6 +225,7 @@ int av_parse_cpu_caps(unsigned *flags, const char *s)
{ "vfp", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_VFP }, .unit = "flags" },
{ "vfpv3", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_VFPV3 }, .unit = "flags" },
{ "neon", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_NEON }, .unit = "flags" },
+ { "setend", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_SETEND }, .unit = "flags" },
#elif ARCH_AARCH64
{ "armv8", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_ARMV8 }, .unit = "flags" },
{ "neon", NULL, 0, AV_OPT_TYPE_CONST, { .i64 = AV_CPU_FLAG_NEON }, .unit = "flags" },
@@ -303,6 +304,7 @@ static const struct {
{ AV_CPU_FLAG_VFP, "vfp" },
{ AV_CPU_FLAG_VFPV3, "vfpv3" },
{ AV_CPU_FLAG_NEON, "neon" },
+ { AV_CPU_FLAG_SETEND, "setend" },
#elif ARCH_PPC
{ AV_CPU_FLAG_ALTIVEC, "altivec" },
#elif ARCH_X86
diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index 8ff14d3adaa..277e489788c 100644
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -64,6 +64,7 @@
#define AV_CPU_FLAG_VFPV3 (1 << 4)
#define AV_CPU_FLAG_NEON (1 << 5)
#define AV_CPU_FLAG_ARMV8 (1 << 6)
+#define AV_CPU_FLAG_SETEND (1 <<16)
/**
* Return the flags which specify extensions supported by the CPU.
--
GitLab