vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

k-quants : fix build on armv7 (android only) (#2920)

Browse source 
* k-quants : fix build on armv7

* ggml : cleanup unused arm32 specific impl

* k-quants : avoid some unused vzero / mzero define

* ggml-alloc : use 4g for MEASURE_MAX_SIZE in 32-bit arm
This commit is contained in:
Jhen-Jie Hong 2023-09-02 20:23:45 +08:00 • committed by GitHub
parent 571083f508
commit 21f3d1be86
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
3 changed files with 42 additions and 51 deletions
Download patch file Download diff file Expand all files Collapse all files

46
ggml.c
View file

@ -817,46 +817,6 @@ static inline float hsum_float_4x4(const __m128 a, const __m128 b, const __m128
#if !defined(__aarch64__)
inline static uint16_t vaddvq_u8(uint8x16_t v) {
return
(uint16_t)vgetq_lane_u8(v, 0) + (uint16_t)vgetq_lane_u8(v, 1) +
(uint16_t)vgetq_lane_u8(v, 2) + (uint16_t)vgetq_lane_u8(v, 3) +
(uint16_t)vgetq_lane_u8(v, 4) + (uint16_t)vgetq_lane_u8(v, 5) +
(uint16_t)vgetq_lane_u8(v, 6) + (uint16_t)vgetq_lane_u8(v, 7) +
(uint16_t)vgetq_lane_u8(v, 8) + (uint16_t)vgetq_lane_u8(v, 9) +
(uint16_t)vgetq_lane_u8(v, 10) + (uint16_t)vgetq_lane_u8(v, 11) +
(uint16_t)vgetq_lane_u8(v, 12) + (uint16_t)vgetq_lane_u8(v, 13) +
(uint16_t)vgetq_lane_u8(v, 14) + (uint16_t)vgetq_lane_u8(v, 15);
}
inline static int16_t vaddvq_s8(int8x16_t v) {
return
(int16_t)vgetq_lane_s8(v, 0) + (int16_t)vgetq_lane_s8(v, 1) +
(int16_t)vgetq_lane_s8(v, 2) + (int16_t)vgetq_lane_s8(v, 3) +
(int16_t)vgetq_lane_s8(v, 4) + (int16_t)vgetq_lane_s8(v, 5) +
(int16_t)vgetq_lane_s8(v, 6) + (int16_t)vgetq_lane_s8(v, 7) +
(int16_t)vgetq_lane_s8(v, 8) + (int16_t)vgetq_lane_s8(v, 9) +
(int16_t)vgetq_lane_s8(v, 10) + (int16_t)vgetq_lane_s8(v, 11) +
(int16_t)vgetq_lane_s8(v, 12) + (int16_t)vgetq_lane_s8(v, 13) +
(int16_t)vgetq_lane_s8(v, 14) + (int16_t)vgetq_lane_s8(v, 15);
}
inline static int32_t vaddvq_s16(int16x8_t v) {
return
(int32_t)vgetq_lane_s16(v, 0) + (int32_t)vgetq_lane_s16(v, 1) +
(int32_t)vgetq_lane_s16(v, 2) + (int32_t)vgetq_lane_s16(v, 3) +
(int32_t)vgetq_lane_s16(v, 4) + (int32_t)vgetq_lane_s16(v, 5) +
(int32_t)vgetq_lane_s16(v, 6) + (int32_t)vgetq_lane_s16(v, 7);
}
inline static uint32_t vaddvq_u16(uint16x8_t v) {
return
(uint32_t)vgetq_lane_u16(v, 0) + (uint32_t)vgetq_lane_u16(v, 1) +
(uint32_t)vgetq_lane_u16(v, 2) + (uint32_t)vgetq_lane_u16(v, 3) +
(uint32_t)vgetq_lane_u16(v, 4) + (uint32_t)vgetq_lane_u16(v, 5) +
(uint32_t)vgetq_lane_u16(v, 6) + (uint32_t)vgetq_lane_u16(v, 7);
}
inline static int32_t vaddvq_s32(int32x4_t v) {
return vgetq_lane_s32(v, 0) + vgetq_lane_s32(v, 1) + vgetq_lane_s32(v, 2) + vgetq_lane_s32(v, 3);
}
@ -865,12 +825,6 @@ inline static float vaddvq_f32(float32x4_t v) {
return vgetq_lane_f32(v, 0) + vgetq_lane_f32(v, 1) + vgetq_lane_f32(v, 2) + vgetq_lane_f32(v, 3);
}
inline static float vminvq_f32(float32x4_t v) {
return
MIN(MIN(vgetq_lane_f32(v, 0), vgetq_lane_f32(v, 1)),
MIN(vgetq_lane_f32(v, 2), vgetq_lane_f32(v, 3)));
}
inline static float vmaxvq_f32(float32x4_t v) {
return
MAX(MAX(vgetq_lane_f32(v, 0), vgetq_lane_f32(v, 1)),