static inline void
i40e_rxq_rearm(struct i40e_rx_queue *rxq)
{
int i;
uint16_t rx_id;
volatile union i40e_rx_desc *rxdp;
struct i40e_rx_entry *rxep = &rxq->sw_ring[rxq->rxrearm_start];
struct rte_mbuf *mb0, *mb1;
uint64x2_t dma_addr0, dma_addr1;
uint64x2_t zero = vdupq_n_u64(0);
uint64_t paddr;
rxdp = rxq->rx_ring + rxq->rxrearm_start;
/* Pull 'n' more MBUFs into the software ring */
if (unlikely(rte_mempool_get_bulk(rxq->mp,
(void *)rxep,
RTE_I40E_RXQ_REARM_THRESH) < 0)) {
if (rxq->rxrearm_nb + RTE_I40E_RXQ_REARM_THRESH >=
rxq->nb_rx_desc) {
for (i = 0; i < RTE_I40E_DESCS_PER_LOOP; i++) {
rxep[i].mbuf = &rxq->fake_mbuf;
vst1q_u64((uint64_t *)&rxdp[i].read, zero);
}
}
rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed +=
RTE_I40E_RXQ_REARM_THRESH;
return;
}
/* Initialize the mbufs in vector, process 2 mbufs in one loop */
for (i = 0; i < RTE_I40E_RXQ_REARM_THRESH; i += 2, rxep += 2) {
mb0 = rxep[0].mbuf;
mb1 = rxep[1].mbuf;
paddr = mb0->buf_iova + RTE_PKTMBUF_HEADROOM;
dma_addr0 = vdupq_n_u64(paddr);
/* flush desc with pa dma_addr */
vst1q_u64((uint64_t *)&rxdp++->read, dma_addr0);
paddr = mb1->buf_iova + RTE_PKTMBUF_HEADROOM;
dma_addr1 = vdupq_n_u64(paddr);
vst1q_u64((uint64_t *)&rxdp++->read, dma_addr1);
}
rxq->rxrearm_start += RTE_I40E_RXQ_REARM_THRESH;
if (rxq->rxrearm_start >= rxq->nb_rx_desc)
rxq->rxrearm_start = 0;
rxq->rxrearm_nb -= RTE_I40E_RXQ_REARM_THRESH;
rx_id = (uint16_t)((rxq->rxrearm_start == 0) ?
(rxq->nb_rx_desc - 1) : (rxq->rxrearm_start - 1));
/* Update the tail pointer on the NIC */
I40E_PCI_REG_WRITE(rxq->qrx_tail, rx_id);
}
void test_vdupQ_nu64 (void)
{
uint64x2_t out_uint64x2_t;
uint64_t arg0_uint64_t;
out_uint64x2_t = vdupq_n_u64 (arg0_uint64_t);
}
static uint64_t SharpYUVUpdateY_NEON(const uint16_t* ref, const uint16_t* src,
uint16_t* dst, int len) {
int i;
const int16x8_t zero = vdupq_n_s16(0);
const int16x8_t max = vdupq_n_s16(MAX_Y);
uint64x2_t sum = vdupq_n_u64(0);
uint64_t diff;
for (i = 0; i + 8 <= len; i += 8) {
const int16x8_t A = vreinterpretq_s16_u16(vld1q_u16(ref + i));
const int16x8_t B = vreinterpretq_s16_u16(vld1q_u16(src + i));
const int16x8_t C = vreinterpretq_s16_u16(vld1q_u16(dst + i));
const int16x8_t D = vsubq_s16(A, B); // diff_y
const int16x8_t F = vaddq_s16(C, D); // new_y
const uint16x8_t H =
vreinterpretq_u16_s16(vmaxq_s16(vminq_s16(F, max), zero));
const int16x8_t I = vabsq_s16(D); // abs(diff_y)
vst1q_u16(dst + i, H);
sum = vpadalq_u32(sum, vpaddlq_u16(vreinterpretq_u16_s16(I)));
}
diff = vgetq_lane_u64(sum, 0) + vgetq_lane_u64(sum, 1);
for (; i < len; ++i) {
const int diff_y = ref[i] - src[i];
const int new_y = (int)(dst[i]) + diff_y;
dst[i] = clip_y(new_y);
diff += (uint64_t)(abs(diff_y));
}
return diff;
}
void
foo ()
{
int8x8_t val14;
int8x8_t val15;
uint8x8_t val16;
uint32x4_t val40;
val14 = vcreate_s8 (0xff0080f6807f807fUL);
val15 = vcreate_s8 (0x10807fff7f808080UL);
val16 = vcgt_s8 (val14, val15);
val40 = vreinterpretq_u32_u64 (
vdupq_n_u64 (
vget_lane_s64_1 (
vreinterpret_s64_u8 (val16), 0)
));
}
uint64x2_t test_vdupq_n_u64(uint64_t v1) {
// CHECK: test_vdupq_n_u64
return vdupq_n_u64(v1);
// CHECK: dup {{v[0-9]+}}.2d, {{x[0-9]+}}
}
bool CPU_ProbeNEON()
{
#if defined(CRYPTOPP_NO_CPU_FEATURE_PROBES)
return false;
#elif (CRYPTOPP_ARM_NEON_AVAILABLE)
# if defined(CRYPTOPP_MS_STYLE_INLINE_ASSEMBLY)
volatile bool result = true;
__try
{
uint32_t v1[4] = {1,1,1,1};
uint32x4_t x1 = vld1q_u32(v1);
uint64_t v2[2] = {1,1};
uint64x2_t x2 = vld1q_u64(v2);
uint32x4_t x3 = vdupq_n_u32(2);
x3 = vsetq_lane_u32(vgetq_lane_u32(x1,0),x3,0);
x3 = vsetq_lane_u32(vgetq_lane_u32(x1,3),x3,3);
uint64x2_t x4 = vdupq_n_u64(2);
x4 = vsetq_lane_u64(vgetq_lane_u64(x2,0),x4,0);
x4 = vsetq_lane_u64(vgetq_lane_u64(x2,1),x4,1);
result = !!(vgetq_lane_u32(x3,0) | vgetq_lane_u64(x4,1));
}
__except (EXCEPTION_EXECUTE_HANDLER)
{
return false;
}
return result;
# else
// longjmp and clobber warnings. Volatile is required.
// http://github.com/weidai11/cryptopp/issues/24 and http://stackoverflow.com/q/7721854
volatile bool result = true;
volatile SigHandler oldHandler = signal(SIGILL, SigIllHandler);
if (oldHandler == SIG_ERR)
return false;
volatile sigset_t oldMask;
if (sigprocmask(0, NULLPTR, (sigset_t*)&oldMask))
return false;
if (setjmp(s_jmpSIGILL))
result = false;
else
{
uint32_t v1[4] = {1,1,1,1};
uint32x4_t x1 = vld1q_u32(v1);
uint64_t v2[2] = {1,1};
uint64x2_t x2 = vld1q_u64(v2);
uint32x4_t x3 = {0,0,0,0};
x3 = vsetq_lane_u32(vgetq_lane_u32(x1,0),x3,0);
x3 = vsetq_lane_u32(vgetq_lane_u32(x1,3),x3,3);
uint64x2_t x4 = {0,0};
x4 = vsetq_lane_u64(vgetq_lane_u64(x2,0),x4,0);
x4 = vsetq_lane_u64(vgetq_lane_u64(x2,1),x4,1);
// Hack... GCC optimizes away the code and returns true
result = !!(vgetq_lane_u32(x3,0) | vgetq_lane_u64(x4,1));
}
sigprocmask(SIG_SETMASK, (sigset_t*)&oldMask, NULLPTR);
signal(SIGILL, oldHandler);
return result;
# endif
#else
return false;
#endif // CRYPTOPP_ARM_NEON_AVAILABLE
}
inline uint64x2_t vdupq_n(const u64 & val) { return vdupq_n_u64(val); }
