Fix compile errors initializing packets on ARM DS-5 5.20

The ARM DS-5 5.20 compiler fails compiling with the following errors:

"src/Core/arch/NEON/PacketMath.h", line 113: Error:  #146: too many initializer values
    Packet4f countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
                         ^
"src/Core/arch/NEON/PacketMath.h", line 118: Error:  #146: too many initializer values
    Packet4i countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
                         ^
"src/Core/arch/NEON/Complex.h", line 30: Error:  #146: too many initializer values
  static uint32x4_t p4ui_CONJ_XOR = EIGEN_INIT_NEON_PACKET4(0x00000000, 0x80000000, 0x00000000, 0x80000000);
                                    ^
"src/Core/arch/NEON/Complex.h", line 31: Error:  #146: too many initializer values
  static uint32x2_t p2ui_CONJ_XOR = EIGEN_INIT_NEON_PACKET2(0x00000000, 0x80000000);
                                    ^

The vectors are implemented as two doubles, hence the too many initializer values error.
Changed the code to use intrinsic load functions which all compilers
implementing NEON should have.
This commit is contained in:
Sean Templeton 2016-06-03 10:51:35 -05:00
parent 1fc2746417
commit bd21243821
2 changed files with 10 additions and 10 deletions

View File

@ -14,8 +14,9 @@ namespace Eigen {
namespace internal {
static uint32x4_t p4ui_CONJ_XOR = EIGEN_INIT_NEON_PACKET4(0x00000000, 0x80000000, 0x00000000, 0x80000000);
static uint32x2_t p2ui_CONJ_XOR = EIGEN_INIT_NEON_PACKET2(0x00000000, 0x80000000);
const uint32_t conj_XOR_DATA[] = { 0x00000000, 0x80000000, 0x00000000, 0x80000000 };
static uint32x4_t p4ui_CONJ_XOR = vld1q_u32( conj_XOR_DATA );
static uint32x2_t p2ui_CONJ_XOR = vld1_u32( conj_XOR_DATA );
//---------- float ----------
struct Packet2cf

View File

@ -51,15 +51,12 @@ typedef uint32x4_t Packet4ui;
#if EIGEN_COMP_LLVM && !EIGEN_COMP_CLANG
//Special treatment for Apple's llvm-gcc, its NEON packet types are unions
#define EIGEN_INIT_NEON_PACKET2(X, Y) {{X, Y}}
#define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W) {{X, Y, Z, W}}
#define EIGEN_INIT_NEON_PACKET2D(X, Y) {{X, Y}}
#else
//Default initializer for packets
#define EIGEN_INIT_NEON_PACKET2(X, Y) {X, Y}
#define EIGEN_INIT_NEON_PACKET4(X, Y, Z, W) {X, Y, Z, W}
#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
@ -122,12 +119,14 @@ template<> EIGEN_STRONG_INLINE Packet4i pset1<Packet4i>(const int& from) {
template<> EIGEN_STRONG_INLINE Packet4f plset<Packet4f>(const float& a)
{
Packet4f countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
const float32_t f[] = {0, 1, 2, 3};
Packet4f countdown = vld1q_f32(f);
return vaddq_f32(pset1<Packet4f>(a), countdown);
}
template<> EIGEN_STRONG_INLINE Packet4i plset<Packet4i>(const int& a)
{
Packet4i countdown = EIGEN_INIT_NEON_PACKET4(0, 1, 2, 3);
const int32_t i[] = {0, 1, 2, 3};
Packet4i countdown = vld1q_s32(i);
return vaddq_s32(pset1<Packet4i>(a), countdown);
}
@ -585,7 +584,7 @@ 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_PACKET2(0, 1);
Packet2d countdown = EIGEN_INIT_NEON_PACKET2D(0, 1);
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); }