GitHub-Proxy/eigen
1
0
Fork 0
mirror of https://gitlab.com/libeigen/eigen.git synced 2025-08-11 19:29:02 +08:00
Code Issues Packages Projects Releases Wiki Activity

bug #1240: Remove any assumption on NEON vector types.

Browse Source
This commit is contained in:
Gael Guennebaud 2016-06-09 23:08:11 +02:00
parent bd21243821
commit 0028049380
2 changed files with 4 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
Eigen/src/Core/arch/NEON/Complex.h
View File

@ -275,7 +275,8 @@ ptranspose(PacketBlock<Packet2cf,2>& kernel) {
//---------- double ----------
#if EIGEN_ARCH_ARM64 && !EIGEN_APPLE_DOUBLE_NEON_BUG
static uint64x2_t p2ul_CONJ_XOR = EIGEN_INIT_NEON_PACKET2(0x0, 0x8000000000000000);
const uint64_t p2ul_conj_XOR_DATA[] = { 0x0, 0x8000000000000000 };
static uint64x2_t p2ul_CONJ_XOR = vld1q_u64( p2ul_conj_XOR_DATA );
struct Packet1cd
{

11
Eigen/src/Core/arch/NEON/PacketMath.h
View File

@ -49,14 +49,6 @@ typedef uint32x4_t Packet4ui;
#define _EIGEN_DECLARE_CONST_Packet4i(NAME,X) \
const Packet4i p4i_##NAME = pset1<Packet4i>(X)
#if EIGEN_COMP_LLVM && !EIGEN_COMP_CLANG
//Special treatment for Apple's llvm-gcc, its NEON packet types are unions
#define EIGEN_INIT_NEON_PACKET2D(X, Y) {{X, Y}}
#else
//Default initializer for packets
#define EIGEN_INIT_NEON_PACKET2D(X, Y) {X, Y}
#endif
// arm64 does have the pld instruction. If available, let's trust the __builtin_prefetch built-in function
// which available on LLVM and GCC (at least)
#if EIGEN_HAS_BUILTIN(__builtin_prefetch) || EIGEN_COMP_GNUC
@ -584,7 +576,8 @@ template<> EIGEN_STRONG_INLINE Packet2d pset1<Packet2d>(const double& from) { r
template<> EIGEN_STRONG_INLINE Packet2d plset<Packet2d>(const double& a)
{
Packet2d countdown = EIGEN_INIT_NEON_PACKET2D(0, 1);
const double countdown_raw[] = {0.0,1.0};
const Packet2d countdown = vld1q_f64(countdown_raw);
return vaddq_f64(pset1<Packet2d>(a), countdown);
}
template<> EIGEN_STRONG_INLINE Packet2d padd<Packet2d>(const Packet2d& a, const Packet2d& b) { return vaddq_f64(a,b); }