|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
#include <assert.h> |
|
|
|
#include <arm_neon.h> |
|
|
|
#include <xnnpack/igemm.h> |
|
#include <xnnpack/intrinsics-polyfill.h> |
|
|
|
|
|
void xnn_f16_igemm_minmax_ukernel_4x16__neonfp16arith_ld64( |
|
size_t mr, |
|
size_t nc, |
|
size_t kc, |
|
size_t ks, |
|
const void** restrict a, |
|
const void* restrict w, |
|
void* restrict c, |
|
size_t cm_stride, |
|
size_t cn_stride, |
|
size_t a_offset, |
|
const void* zero, |
|
const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)]) |
|
{ |
|
assert(mr != 0); |
|
assert(mr <= 4); |
|
assert(nc != 0); |
|
assert(kc != 0); |
|
assert(kc % sizeof(uint16_t) == 0); |
|
assert(ks != 0); |
|
assert(ks % (4 * sizeof(void*)) == 0); |
|
assert(a_offset % sizeof(uint16_t) == 0); |
|
assert(a != NULL); |
|
assert(w != NULL); |
|
assert(c != NULL); |
|
|
|
uint16_t* c0 = (uint16_t*) c; |
|
uint16_t* c1 = (uint16_t*) ((uintptr_t) c0 + cm_stride); |
|
if XNN_UNPREDICTABLE(mr < 2) { |
|
c1 = c0; |
|
} |
|
uint16_t* c2 = (uint16_t*) ((uintptr_t) c1 + cm_stride); |
|
if XNN_UNPREDICTABLE(mr <= 2) { |
|
c2 = c1; |
|
} |
|
uint16_t* c3 = (uint16_t*) ((uintptr_t) c2 + cm_stride); |
|
if XNN_UNPREDICTABLE(mr != 4) { |
|
c3 = c2; |
|
} |
|
|
|
do { |
|
float16x8_t vacc0x01234567 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
float16x8_t vacc0x89ABCDEF = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
float16x8_t vacc1x01234567 = vacc0x01234567; |
|
float16x8_t vacc1x89ABCDEF = vacc0x89ABCDEF; |
|
float16x8_t vacc2x01234567 = vacc0x01234567; |
|
float16x8_t vacc2x89ABCDEF = vacc0x89ABCDEF; |
|
float16x8_t vacc3x01234567 = vacc0x01234567; |
|
float16x8_t vacc3x89ABCDEF = vacc0x89ABCDEF; |
|
|
|
size_t p = ks; |
|
do { |
|
const uint16_t* restrict a0 = (const uint16_t*) a[0]; |
|
assert(a0 != NULL); |
|
if XNN_UNPREDICTABLE(a0 != zero) { |
|
a0 = (const uint16_t*) ((uintptr_t) a0 + a_offset); |
|
} |
|
const uint16_t* restrict a1 = (const uint16_t*) a[1]; |
|
assert(a1 != NULL); |
|
if XNN_UNPREDICTABLE(a1 != zero) { |
|
a1 = (const uint16_t*) ((uintptr_t) a1 + a_offset); |
|
} |
|
const uint16_t* restrict a2 = (const uint16_t*) a[2]; |
|
assert(a2 != NULL); |
|
if XNN_UNPREDICTABLE(a2 != zero) { |
|
a2 = (const uint16_t*) ((uintptr_t) a2 + a_offset); |
|
} |
|
const uint16_t* restrict a3 = (const uint16_t*) a[3]; |
|
assert(a3 != NULL); |
|
if XNN_UNPREDICTABLE(a3 != zero) { |
|
a3 = (const uint16_t*) ((uintptr_t) a3 + a_offset); |
|
} |
|
a += 4; |
|
|
|
size_t k = kc; |
|
for (; k >= 4 * sizeof(uint16_t); k -= 4 * sizeof(uint16_t)) { |
|
const float16x4_t va0 = vreinterpret_f16_u16(vld1_u16(a0)); a0 += 4; |
|
const float16x4_t va1 = vreinterpret_f16_u16(vld1_u16(a1)); a1 += 4; |
|
const float16x4_t va2 = vreinterpret_f16_u16(vld1_u16(a2)); a2 += 4; |
|
const float16x4_t va3 = vreinterpret_f16_u16(vld1_u16(a3)); a3 += 4; |
|
|
|
const float16x8_t vb01234567c0 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
const float16x8_t vb89ABCDEFc0 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
|
|
#if XNN_ARCH_ARM64 |
|
vacc0x01234567 = vfmaq_lane_f16(vacc0x01234567, vb01234567c0, va0, 0); |
|
vacc1x01234567 = vfmaq_lane_f16(vacc1x01234567, vb01234567c0, va1, 0); |
|
vacc2x01234567 = vfmaq_lane_f16(vacc2x01234567, vb01234567c0, va2, 0); |
|
vacc3x01234567 = vfmaq_lane_f16(vacc3x01234567, vb01234567c0, va3, 0); |
|
vacc0x89ABCDEF = vfmaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc0, va0, 0); |
|
vacc1x89ABCDEF = vfmaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc0, va1, 0); |
|
vacc2x89ABCDEF = vfmaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc0, va2, 0); |
|
vacc3x89ABCDEF = vfmaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc0, va3, 0); |
|
#else |
|
vacc0x01234567 = vmlaq_lane_f16(vacc0x01234567, vb01234567c0, va0, 0); |
|
vacc1x01234567 = vmlaq_lane_f16(vacc1x01234567, vb01234567c0, va1, 0); |
|
vacc2x01234567 = vmlaq_lane_f16(vacc2x01234567, vb01234567c0, va2, 0); |
|
vacc3x01234567 = vmlaq_lane_f16(vacc3x01234567, vb01234567c0, va3, 0); |
|
vacc0x89ABCDEF = vmlaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc0, va0, 0); |
|
vacc1x89ABCDEF = vmlaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc0, va1, 0); |
|
vacc2x89ABCDEF = vmlaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc0, va2, 0); |
|
vacc3x89ABCDEF = vmlaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc0, va3, 0); |
|
#endif |
|
const float16x8_t vb01234567c1 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
const float16x8_t vb89ABCDEFc1 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
|
|
#if XNN_ARCH_ARM64 |
|
vacc0x01234567 = vfmaq_lane_f16(vacc0x01234567, vb01234567c1, va0, 1); |
|
vacc1x01234567 = vfmaq_lane_f16(vacc1x01234567, vb01234567c1, va1, 1); |
|
vacc2x01234567 = vfmaq_lane_f16(vacc2x01234567, vb01234567c1, va2, 1); |
|
vacc3x01234567 = vfmaq_lane_f16(vacc3x01234567, vb01234567c1, va3, 1); |
|
vacc0x89ABCDEF = vfmaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc1, va0, 1); |
|
vacc1x89ABCDEF = vfmaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc1, va1, 1); |
|
vacc2x89ABCDEF = vfmaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc1, va2, 1); |
|
vacc3x89ABCDEF = vfmaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc1, va3, 1); |
|
#else |
|
vacc0x01234567 = vmlaq_lane_f16(vacc0x01234567, vb01234567c1, va0, 1); |
|
vacc1x01234567 = vmlaq_lane_f16(vacc1x01234567, vb01234567c1, va1, 1); |
|
vacc2x01234567 = vmlaq_lane_f16(vacc2x01234567, vb01234567c1, va2, 1); |
|
vacc3x01234567 = vmlaq_lane_f16(vacc3x01234567, vb01234567c1, va3, 1); |
|
vacc0x89ABCDEF = vmlaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc1, va0, 1); |
|
vacc1x89ABCDEF = vmlaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc1, va1, 1); |
|
vacc2x89ABCDEF = vmlaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc1, va2, 1); |
|
vacc3x89ABCDEF = vmlaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc1, va3, 1); |
|
#endif |
|
const float16x8_t vb01234567c2 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
const float16x8_t vb89ABCDEFc2 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
|
|
#if XNN_ARCH_ARM64 |
|
vacc0x01234567 = vfmaq_lane_f16(vacc0x01234567, vb01234567c2, va0, 2); |
|
vacc1x01234567 = vfmaq_lane_f16(vacc1x01234567, vb01234567c2, va1, 2); |
|
vacc2x01234567 = vfmaq_lane_f16(vacc2x01234567, vb01234567c2, va2, 2); |
|
vacc3x01234567 = vfmaq_lane_f16(vacc3x01234567, vb01234567c2, va3, 2); |
|
vacc0x89ABCDEF = vfmaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc2, va0, 2); |
|
vacc1x89ABCDEF = vfmaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc2, va1, 2); |
|
vacc2x89ABCDEF = vfmaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc2, va2, 2); |
|
vacc3x89ABCDEF = vfmaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc2, va3, 2); |
|
#else |
|
vacc0x01234567 = vmlaq_lane_f16(vacc0x01234567, vb01234567c2, va0, 2); |
|
vacc1x01234567 = vmlaq_lane_f16(vacc1x01234567, vb01234567c2, va1, 2); |
|
vacc2x01234567 = vmlaq_lane_f16(vacc2x01234567, vb01234567c2, va2, 2); |
|
vacc3x01234567 = vmlaq_lane_f16(vacc3x01234567, vb01234567c2, va3, 2); |
|
vacc0x89ABCDEF = vmlaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc2, va0, 2); |
|
vacc1x89ABCDEF = vmlaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc2, va1, 2); |
|
vacc2x89ABCDEF = vmlaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc2, va2, 2); |
|
vacc3x89ABCDEF = vmlaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc2, va3, 2); |
|
#endif |
|
const float16x8_t vb01234567c3 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
const float16x8_t vb89ABCDEFc3 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
|
|
#if XNN_ARCH_ARM64 |
|
vacc0x01234567 = vfmaq_lane_f16(vacc0x01234567, vb01234567c3, va0, 3); |
|
vacc1x01234567 = vfmaq_lane_f16(vacc1x01234567, vb01234567c3, va1, 3); |
|
vacc2x01234567 = vfmaq_lane_f16(vacc2x01234567, vb01234567c3, va2, 3); |
|
vacc3x01234567 = vfmaq_lane_f16(vacc3x01234567, vb01234567c3, va3, 3); |
|
vacc0x89ABCDEF = vfmaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc3, va0, 3); |
|
vacc1x89ABCDEF = vfmaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc3, va1, 3); |
|
vacc2x89ABCDEF = vfmaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc3, va2, 3); |
|
vacc3x89ABCDEF = vfmaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc3, va3, 3); |
|
#else |
|
vacc0x01234567 = vmlaq_lane_f16(vacc0x01234567, vb01234567c3, va0, 3); |
|
vacc1x01234567 = vmlaq_lane_f16(vacc1x01234567, vb01234567c3, va1, 3); |
|
vacc2x01234567 = vmlaq_lane_f16(vacc2x01234567, vb01234567c3, va2, 3); |
|
vacc3x01234567 = vmlaq_lane_f16(vacc3x01234567, vb01234567c3, va3, 3); |
|
vacc0x89ABCDEF = vmlaq_lane_f16(vacc0x89ABCDEF, vb89ABCDEFc3, va0, 3); |
|
vacc1x89ABCDEF = vmlaq_lane_f16(vacc1x89ABCDEF, vb89ABCDEFc3, va1, 3); |
|
vacc2x89ABCDEF = vmlaq_lane_f16(vacc2x89ABCDEF, vb89ABCDEFc3, va2, 3); |
|
vacc3x89ABCDEF = vmlaq_lane_f16(vacc3x89ABCDEF, vb89ABCDEFc3, va3, 3); |
|
#endif |
|
} |
|
if XNN_UNLIKELY(k != 0) { |
|
do { |
|
const float16x8_t va0 = vreinterpretq_f16_u16(vld1q_dup_u16(a0)); a0 += 1; |
|
const float16x8_t va1 = vreinterpretq_f16_u16(vld1q_dup_u16(a1)); a1 += 1; |
|
const float16x8_t va2 = vreinterpretq_f16_u16(vld1q_dup_u16(a2)); a2 += 1; |
|
const float16x8_t va3 = vreinterpretq_f16_u16(vld1q_dup_u16(a3)); a3 += 1; |
|
|
|
const float16x8_t vb01234567 = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
const float16x8_t vb89ABCDEF = vreinterpretq_f16_u16(vld1q_u16(w)); w = (const void*) ((uintptr_t) w + sizeof(float16x8_t)); |
|
|
|
vacc0x01234567 = vfmaq_f16(vacc0x01234567, va0, vb01234567); |
|
vacc1x01234567 = vfmaq_f16(vacc1x01234567, va1, vb01234567); |
|
vacc2x01234567 = vfmaq_f16(vacc2x01234567, va2, vb01234567); |
|
vacc3x01234567 = vfmaq_f16(vacc3x01234567, va3, vb01234567); |
|
vacc0x89ABCDEF = vfmaq_f16(vacc0x89ABCDEF, va0, vb89ABCDEF); |
|
vacc1x89ABCDEF = vfmaq_f16(vacc1x89ABCDEF, va1, vb89ABCDEF); |
|
vacc2x89ABCDEF = vfmaq_f16(vacc2x89ABCDEF, va2, vb89ABCDEF); |
|
vacc3x89ABCDEF = vfmaq_f16(vacc3x89ABCDEF, va3, vb89ABCDEF); |
|
|
|
k -= sizeof(uint16_t); |
|
} while (k != 0); |
|
} |
|
p -= 4 * sizeof(void*); |
|
} while (p != 0); |
|
|
|
|
|
const float16x8_t vmin = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->fp16arith.min)); |
|
vacc0x01234567 = vmaxq_f16(vacc0x01234567, vmin); |
|
vacc1x01234567 = vmaxq_f16(vacc1x01234567, vmin); |
|
vacc2x01234567 = vmaxq_f16(vacc2x01234567, vmin); |
|
vacc3x01234567 = vmaxq_f16(vacc3x01234567, vmin); |
|
vacc0x89ABCDEF = vmaxq_f16(vacc0x89ABCDEF, vmin); |
|
vacc1x89ABCDEF = vmaxq_f16(vacc1x89ABCDEF, vmin); |
|
vacc2x89ABCDEF = vmaxq_f16(vacc2x89ABCDEF, vmin); |
|
vacc3x89ABCDEF = vmaxq_f16(vacc3x89ABCDEF, vmin); |
|
|
|
const float16x8_t vmax = vreinterpretq_f16_u16(vld1q_dup_u16(¶ms->fp16arith.max)); |
|
vacc0x01234567 = vminq_f16(vacc0x01234567, vmax); |
|
vacc1x01234567 = vminq_f16(vacc1x01234567, vmax); |
|
vacc2x01234567 = vminq_f16(vacc2x01234567, vmax); |
|
vacc3x01234567 = vminq_f16(vacc3x01234567, vmax); |
|
vacc0x89ABCDEF = vminq_f16(vacc0x89ABCDEF, vmax); |
|
vacc1x89ABCDEF = vminq_f16(vacc1x89ABCDEF, vmax); |
|
vacc2x89ABCDEF = vminq_f16(vacc2x89ABCDEF, vmax); |
|
vacc3x89ABCDEF = vminq_f16(vacc3x89ABCDEF, vmax); |
|
|
|
if XNN_LIKELY(nc >= 16) { |
|
vst1q_u16(c3, vreinterpretq_u16_f16(vacc3x01234567)); |
|
vst1q_u16(c3 + 8, vreinterpretq_u16_f16(vacc3x89ABCDEF)); |
|
c3 = (uint16_t*) ((uintptr_t) c3 + cn_stride); |
|
vst1q_u16(c2, vreinterpretq_u16_f16(vacc2x01234567)); |
|
vst1q_u16(c2 + 8, vreinterpretq_u16_f16(vacc2x89ABCDEF)); |
|
c2 = (uint16_t*) ((uintptr_t) c2 + cn_stride); |
|
vst1q_u16(c1, vreinterpretq_u16_f16(vacc1x01234567)); |
|
vst1q_u16(c1 + 8, vreinterpretq_u16_f16(vacc1x89ABCDEF)); |
|
c1 = (uint16_t*) ((uintptr_t) c1 + cn_stride); |
|
vst1q_u16(c0, vreinterpretq_u16_f16(vacc0x01234567)); |
|
vst1q_u16(c0 + 8, vreinterpretq_u16_f16(vacc0x89ABCDEF)); |
|
c0 = (uint16_t*) ((uintptr_t) c0 + cn_stride); |
|
|
|
a = (const void**restrict) ((uintptr_t) a - ks); |
|
nc -= 16; |
|
} else { |
|
if (nc & 8) { |
|
vst1q_u16(c3, vreinterpretq_u16_f16(vacc3x01234567)); c3 += 8; |
|
vst1q_u16(c2, vreinterpretq_u16_f16(vacc2x01234567)); c2 += 8; |
|
vst1q_u16(c1, vreinterpretq_u16_f16(vacc1x01234567)); c1 += 8; |
|
vst1q_u16(c0, vreinterpretq_u16_f16(vacc0x01234567)); c0 += 8; |
|
|
|
vacc3x01234567 = vacc3x89ABCDEF; |
|
vacc2x01234567 = vacc2x89ABCDEF; |
|
vacc1x01234567 = vacc1x89ABCDEF; |
|
vacc0x01234567 = vacc0x89ABCDEF; |
|
} |
|
float16x4_t vacc3x0123 = vget_low_f16(vacc3x01234567); |
|
float16x4_t vacc2x0123 = vget_low_f16(vacc2x01234567); |
|
float16x4_t vacc1x0123 = vget_low_f16(vacc1x01234567); |
|
float16x4_t vacc0x0123 = vget_low_f16(vacc0x01234567); |
|
if (nc & 4) { |
|
vst1_u16(c3, vreinterpret_u16_f16(vacc3x0123)); c3 += 4; |
|
vst1_u16(c2, vreinterpret_u16_f16(vacc2x0123)); c2 += 4; |
|
vst1_u16(c1, vreinterpret_u16_f16(vacc1x0123)); c1 += 4; |
|
vst1_u16(c0, vreinterpret_u16_f16(vacc0x0123)); c0 += 4; |
|
|
|
vacc3x0123 = vget_high_f16(vacc3x01234567); |
|
vacc2x0123 = vget_high_f16(vacc2x01234567); |
|
vacc1x0123 = vget_high_f16(vacc1x01234567); |
|
vacc0x0123 = vget_high_f16(vacc0x01234567); |
|
} |
|
if (nc & 2) { |
|
vst1_lane_u32((void*) c3, vreinterpret_u32_f16(vacc3x0123), 0); c3 += 2; |
|
vst1_lane_u32((void*) c2, vreinterpret_u32_f16(vacc2x0123), 0); c2 += 2; |
|
vst1_lane_u32((void*) c1, vreinterpret_u32_f16(vacc1x0123), 0); c1 += 2; |
|
vst1_lane_u32((void*) c0, vreinterpret_u32_f16(vacc0x0123), 0); c0 += 2; |
|
|
|
vacc3x0123 = vext_f16(vacc3x0123, vacc3x0123, 2); |
|
vacc2x0123 = vext_f16(vacc2x0123, vacc2x0123, 2); |
|
vacc1x0123 = vext_f16(vacc1x0123, vacc1x0123, 2); |
|
vacc0x0123 = vext_f16(vacc0x0123, vacc0x0123, 2); |
|
} |
|
if (nc & 1) { |
|
vst1_lane_u16(c3, vreinterpret_u16_f16(vacc3x0123), 0); |
|
vst1_lane_u16(c2, vreinterpret_u16_f16(vacc2x0123), 0); |
|
vst1_lane_u16(c1, vreinterpret_u16_f16(vacc1x0123), 0); |
|
vst1_lane_u16(c0, vreinterpret_u16_f16(vacc0x0123), 0); |
|
} |
|
|
|
nc = 0; |
|
} |
|
} while (nc != 0); |
|
} |
|
|
