Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

ggml : remove bit shuffling #1305

Closed
wants to merge 20 commits into from
Closed

ggml : remove bit shuffling #1305

Changes from 1 commit
Commits
Show all changes
20 commits
Select commit Hold shift + click to select a range
a546dc6
ggml : remove Q4_0 bit shufling (ARM NEON)
ggerganov May 3, 2023
edb6c8b
ggml : remove Q4_1 bit shuffling (ARM NEON + reference)
ggerganov May 4, 2023
086cfea
ggml : nibbles_from_floats() + bytes_from_nibbles() (ARM NEON)
ggerganov May 4, 2023
a6a1d96
ggml : remove Q4_2 bit shuffling (WIP, BROKEN)
ggerganov May 4, 2023
796f8ae
ggml : remove Q5_0 bit shuffling (ARM NEON)
ggerganov May 4, 2023
39bb8e7
ggml : 2x faster scalar implementations
ggerganov May 4, 2023
c7af904
ggml : remove Q5_1 bit shuffling (ARM NEON + scalar)
ggerganov May 5, 2023
ba953d6
ggml : simplify scalar dot
ggerganov May 5, 2023
4991499
ggml : remove WASM SIMD bit shuffling + remove vzip for ARM 32-bit
ggerganov May 5, 2023
c216656
ggml : fix Q4_1 quantization
ggerganov May 7, 2023
b47bd28
ggml : update cuBLAS + normalize variable names
ggerganov May 7, 2023
7cdc08a
ggml : remove Q4_2 mode
ggerganov May 7, 2023
60f62bb
ggml : minor formatting
ggerganov May 7, 2023
8fbf777
ggml : fix Q5_0 quantization
ggerganov May 7, 2023
d155f0f
scripts : add script for measuring the time per token
ggerganov May 8, 2023
948d124
AVX implementations (#1370)
sw May 8, 2023
0e48eb6
ggml : uniform 5th bit extraction
ggerganov May 8, 2023
4201fa5
llama : produce error upon loading old model files
ggerganov May 9, 2023
ffd76e1
llama : fix model magic/version write
ggerganov May 9, 2023
e116eb6
ggml : speed-up Q5_0 + Q5_1 at 4 threads
ggerganov May 10, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Prev Previous commit
ggml : speed-up Q5_0 + Q5_1 at 4 threads
  • Loading branch information
@ggerganov
ggerganov committed May 11, 2023
commit e116eb638c9a426d679a9f756d27f13eb6ce626a
223 changes: 147 additions & 76 deletions ggml.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -339,8 +339,9 @@ static float table_f32_f16[1 << 16];
#define B7(c,s,n) B6(c,s,n ## c), B6(c,s,n ## s)
#define B8(c,s ) B7(c,s, c), B7(c,s, s)

// precomputed tables for expanding 8bits to 8 bytes (shl 4)
static const uint64_t table_b2b_u[1 << 8] = { B8(00, 10) };
// precomputed tables for expanding 8bits to 8 bytes:
static const uint64_t table_b2b_0[1 << 8] = { B8(00, 10) }; // ( b) << 4
static const uint64_t table_b2b_1[1 << 8] = { B8(10, 00) }; // (!b) << 4
#endif

// On ARM NEON, it's quicker to directly convert x -> x instead of calling into ggml_lookup_fp16_to_fp32,
Expand Down Expand Up @@ -2307,68 +2308,102 @@ static void ggml_vec_dot_q5_0_q8_0(const int n, float * restrict s, const void *
const block_q8_0 * restrict y = vy;

#if defined(__ARM_NEON)
float32x4_t sumv = vdupq_n_f32(0.0f);
float32x4_t sumv0 = vdupq_n_f32(0.0f);
float32x4_t sumv1 = vdupq_n_f32(0.0f);

uint64_t tmp[4];
uint32_t qh0;
uint32_t qh1;

for (int i = 0; i < nb; ++i) {
uint64_t tmp0[4];
uint64_t tmp1[4];

for (int i = 0; i < nb; i += 2) {
const block_q5_0 * restrict x0 = &x[i];
const block_q5_0 * restrict x1 = &x[i + 1];
const block_q8_0 * restrict y0 = &y[i];
const block_q8_0 * restrict y1 = &y[i + 1];

const uint8x16_t m4b = vdupq_n_u8(0x0F);
const int8x16_t s16b = vdupq_n_s8(0x10);
const uint8x16_t m4b = vdupq_n_u8(0x0F);

// extract the 5th bit
uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh));
// extract the 5th bit via lookup table ((!b) << 4)
memcpy(&qh0, x0->qh, sizeof(qh0));
memcpy(&qh1, x1->qh, sizeof(qh1));

tmp[0] = table_b2b_u[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ];
tmp0[0] = table_b2b_1[(qh0 >> 0) & 0xFF];
tmp0[1] = table_b2b_1[(qh0 >> 8) & 0xFF];
tmp0[2] = table_b2b_1[(qh0 >> 16) & 0xFF];
tmp0[3] = table_b2b_1[(qh0 >> 24) ];

const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0));
const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2));
tmp1[0] = table_b2b_1[(qh1 >> 0) & 0xFF];
tmp1[1] = table_b2b_1[(qh1 >> 8) & 0xFF];
tmp1[2] = table_b2b_1[(qh1 >> 16) & 0xFF];
tmp1[3] = table_b2b_1[(qh1 >> 24) ];

const uint8x16_t v0 = vld1q_u8(x0->qs);
const int8x16_t qhl0 = vld1q_s8((const int8_t *)(tmp0 + 0));
const int8x16_t qhh0 = vld1q_s8((const int8_t *)(tmp0 + 2));
const int8x16_t qhl1 = vld1q_s8((const int8_t *)(tmp1 + 0));
const int8x16_t qhh1 = vld1q_s8((const int8_t *)(tmp1 + 2));

const uint8x16_t v0_0 = vld1q_u8(x0->qs);
const uint8x16_t v0_1 = vld1q_u8(x1->qs);

// 4-bit -> 8-bit
const int8x16_t v0l = vreinterpretq_s8_u8(vandq_u8 (v0, m4b));
const int8x16_t v0h = vreinterpretq_s8_u8(vshrq_n_u8(v0, 4));
int8x16_t v0_0l = vreinterpretq_s8_u8(vandq_u8 (v0_0, m4b));
int8x16_t v0_0h = vreinterpretq_s8_u8(vshrq_n_u8(v0_0, 4));
int8x16_t v0_1l = vreinterpretq_s8_u8(vandq_u8 (v0_1, m4b));
int8x16_t v0_1h = vreinterpretq_s8_u8(vshrq_n_u8(v0_1, 4));

// add high bit and sub 16
const int8x16_t v0lf = vsubq_s8(vorrq_s8(v0l, qhl), s16b);
const int8x16_t v0hf = vsubq_s8(vorrq_s8(v0h, qhh), s16b);
// add high bit and sub 16 (equivalent to sub 0x10 when bit is zero)
const int8x16_t v0_0lf = vsubq_s8(v0_0l, qhl0);
const int8x16_t v0_0hf = vsubq_s8(v0_0h, qhh0);
const int8x16_t v0_1lf = vsubq_s8(v0_1l, qhl1);
const int8x16_t v0_1hf = vsubq_s8(v0_1h, qhh1);

// load y
const int8x16_t v1l = vld1q_s8(y0->qs);
const int8x16_t v1h = vld1q_s8(y0->qs + 16);
const int8x16_t v1_0l = vld1q_s8(y0->qs);
const int8x16_t v1_0h = vld1q_s8(y0->qs + 16);
const int8x16_t v1_1l = vld1q_s8(y1->qs);
const int8x16_t v1_1h = vld1q_s8(y1->qs + 16);

const float x0d = GGML_FP16_TO_FP32(x0->d);
const float x1d = GGML_FP16_TO_FP32(x1->d);

#if defined(__ARM_FEATURE_DOTPROD)
sumv = vmlaq_n_f32(sumv, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0lf, v1l),
vdotq_s32(vdupq_n_s32(0), v0hf, v1h))), x0d*y0->d);
sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_0lf, v1_0l),
vdotq_s32(vdupq_n_s32(0), v0_0hf, v1_0h))), x0d*y0->d);
sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_1lf, v1_1l),
vdotq_s32(vdupq_n_s32(0), v0_1hf, v1_1h))), x1d*y1->d);
#else
const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0lf), vget_low_s8 (v1l));
const int16x8_t pl0h = vmull_s8(vget_high_s8(v0lf), vget_high_s8(v1l));
const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0hf), vget_low_s8 (v1h));
const int16x8_t ph0h = vmull_s8(vget_high_s8(v0hf), vget_high_s8(v1h));
const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0_0lf), vget_low_s8 (v1_0l));
const int16x8_t pl0h = vmull_s8(vget_high_s8(v0_0lf), vget_high_s8(v1_0l));
const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0_0hf), vget_low_s8 (v1_0h));
const int16x8_t ph0h = vmull_s8(vget_high_s8(v0_0hf), vget_high_s8(v1_0h));

const int16x8_t pl1l = vmull_s8(vget_low_s8 (v0_1lf), vget_low_s8 (v1_1l));
const int16x8_t pl1h = vmull_s8(vget_high_s8(v0_1lf), vget_high_s8(v1_1l));
const int16x8_t ph1l = vmull_s8(vget_low_s8 (v0_1hf), vget_low_s8 (v1_1h));
const int16x8_t ph1h = vmull_s8(vget_high_s8(v0_1hf), vget_high_s8(v1_1h));

const int32x4_t pl0 = vaddq_s32(vpaddlq_s16(pl0l), vpaddlq_s16(pl0h));
const int32x4_t ph0 = vaddq_s32(vpaddlq_s16(ph0l), vpaddlq_s16(ph0h));
const int32x4_t pl1 = vaddq_s32(vpaddlq_s16(pl1l), vpaddlq_s16(pl1h));
const int32x4_t ph1 = vaddq_s32(vpaddlq_s16(ph1l), vpaddlq_s16(ph1h));

sumv = vmlaq_n_f32(sumv, vcvtq_f32_s32(vaddq_s32(pl0, ph0)), x0d*y0->d);
sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(vaddq_s32(pl0, ph0)), x0d*y0->d);
sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(vaddq_s32(pl1, ph1)), x1d*y1->d);
#endif
}

*s = vaddvq_f32(sumv);
*s = vaddvq_f32(sumv0) + vaddvq_f32(sumv1);
#elif defined(__wasm_simd128__)
v128_t sumv = wasm_f32x4_splat(0.0f);

uint32_t qh;
uint64_t tmp[4];

// TODO: check if unrolling this is better
for (int i = 0; i < nb; ++i) {
const block_q5_0 * restrict x0 = &x[i];
const block_q8_0 * restrict y0 = &y[i];
Expand All @@ -2377,13 +2412,12 @@ static void ggml_vec_dot_q5_0_q8_0(const int n, float * restrict s, const void *
const v128_t s16b = wasm_i8x16_splat(0x10);

// extract the 5th bit
uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh));

tmp[0] = table_b2b_u[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ];
tmp[0] = table_b2b_1[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_1[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_1[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_1[(qh >> 24) ];

const v128_t qhl = wasm_v128_load(tmp + 0);
const v128_t qhh = wasm_v128_load(tmp + 2);
Expand All @@ -2395,8 +2429,8 @@ static void ggml_vec_dot_q5_0_q8_0(const int n, float * restrict s, const void *
const v128_t v0h = wasm_u8x16_shr(v0, 4);

// add high bit and sub 16
const v128_t v0lf = wasm_i8x16_sub(wasm_v128_or(v0l, qhl), s16b);
const v128_t v0hf = wasm_i8x16_sub(wasm_v128_or(v0h, qhh), s16b);
const v128_t v0lf = wasm_i8x16_sub(v0l, qhl);
const v128_t v0hf = wasm_i8x16_sub(v0h, qhh);

// load y
const v128_t v1l = wasm_v128_load(y0->qs);
Expand Down Expand Up @@ -2488,69 +2522,107 @@ static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void *
const block_q8_1 * restrict y = vy;

#if defined(__ARM_NEON)
float32x4_t sumv = vdupq_n_f32(0.0f);
float32x4_t sumv0 = vdupq_n_f32(0.0f);
float32x4_t sumv1 = vdupq_n_f32(0.0f);

float summs = 0.0f;
float summs0 = 0.0f;
float summs1 = 0.0f;

uint64_t tmp[4];
uint32_t qh0;
uint32_t qh1;

for (int i = 0; i < nb; ++i) {
uint64_t tmp0[4];
uint64_t tmp1[4];

for (int i = 0; i < nb; i += 2) {
const block_q5_1 * restrict x0 = &x[i];
const block_q5_1 * restrict x1 = &x[i + 1];
const block_q8_1 * restrict y0 = &y[i];
const block_q8_1 * restrict y1 = &y[i + 1];

summs += GGML_FP16_TO_FP32(x0->m) * (y0->s0 + y0->s1);
const uint8x16_t m4b = vdupq_n_u8(0x0F);

// extract the 5th bit
uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh));
summs0 += GGML_FP16_TO_FP32(x0->m) * (y0->s0 + y0->s1);
summs1 += GGML_FP16_TO_FP32(x1->m) * (y1->s0 + y1->s1);

// extract the 5th bit via lookup table ((b) << 4)
memcpy(&qh0, x0->qh, sizeof(qh0));
memcpy(&qh1, x1->qh, sizeof(qh1));

tmp[0] = table_b2b_u[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ];
tmp0[0] = table_b2b_0[(qh0 >> 0) & 0xFF];
tmp0[1] = table_b2b_0[(qh0 >> 8) & 0xFF];
tmp0[2] = table_b2b_0[(qh0 >> 16) & 0xFF];
tmp0[3] = table_b2b_0[(qh0 >> 24) ];

const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0));
const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2));
tmp1[0] = table_b2b_0[(qh1 >> 0) & 0xFF];
tmp1[1] = table_b2b_0[(qh1 >> 8) & 0xFF];
tmp1[2] = table_b2b_0[(qh1 >> 16) & 0xFF];
tmp1[3] = table_b2b_0[(qh1 >> 24) ];

const uint8x16_t v0 = vld1q_u8(x0->qs);
const int8x16_t qhl0 = vld1q_s8((const int8_t *)(tmp0 + 0));
const int8x16_t qhh0 = vld1q_s8((const int8_t *)(tmp0 + 2));
const int8x16_t qhl1 = vld1q_s8((const int8_t *)(tmp1 + 0));
const int8x16_t qhh1 = vld1q_s8((const int8_t *)(tmp1 + 2));

const uint8x16_t v0_0 = vld1q_u8(x0->qs);
const uint8x16_t v0_1 = vld1q_u8(x1->qs);

// 4-bit -> 8-bit
const int8x16_t v0l = vreinterpretq_s8_u8(vandq_u8 (v0, vdupq_n_u8(0x0F)));
const int8x16_t v0h = vreinterpretq_s8_u8(vshrq_n_u8(v0, 4));
const int8x16_t v0_0l = vreinterpretq_s8_u8(vandq_u8 (v0_0, m4b));
const int8x16_t v0_0h = vreinterpretq_s8_u8(vshrq_n_u8(v0_0, 4));
const int8x16_t v0_1l = vreinterpretq_s8_u8(vandq_u8 (v0_1, m4b));
const int8x16_t v0_1h = vreinterpretq_s8_u8(vshrq_n_u8(v0_1, 4));

// add
const int8x16_t v0lf = vorrq_s8(v0l, qhl);
const int8x16_t v0hf = vorrq_s8(v0h, qhh);
// add 5th bit
const int8x16_t v0_0lf = vorrq_s8(v0_0l, qhl0);
const int8x16_t v0_0hf = vorrq_s8(v0_0h, qhh0);
const int8x16_t v0_1lf = vorrq_s8(v0_1l, qhl1);
const int8x16_t v0_1hf = vorrq_s8(v0_1h, qhh1);

// load y
const int8x16_t v1l = vld1q_s8(y0->qs);
const int8x16_t v1h = vld1q_s8(y0->qs + 16);
const int8x16_t v1_0l = vld1q_s8(y0->qs);
const int8x16_t v1_0h = vld1q_s8(y0->qs + 16);
const int8x16_t v1_1l = vld1q_s8(y1->qs);
const int8x16_t v1_1h = vld1q_s8(y1->qs + 16);

const float x0d = GGML_FP16_TO_FP32(x0->d);
const float x1d = GGML_FP16_TO_FP32(x1->d);

#if defined(__ARM_FEATURE_DOTPROD)
sumv = vmlaq_n_f32(sumv, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0lf, v1l),
vdotq_s32(vdupq_n_s32(0), v0hf, v1h))), x0d*y0->d);
sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_0lf, v1_0l),
vdotq_s32(vdupq_n_s32(0), v0_0hf, v1_0h))), x0d*y0->d);
sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(vaddq_s32(
vdotq_s32(vdupq_n_s32(0), v0_1lf, v1_1l),
vdotq_s32(vdupq_n_s32(0), v0_1hf, v1_1h))), x1d*y1->d);
#else
const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0lf), vget_low_s8 (v1l));
const int16x8_t pl0h = vmull_s8(vget_high_s8(v0lf), vget_high_s8(v1l));
const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0hf), vget_low_s8 (v1h));
const int16x8_t ph0h = vmull_s8(vget_high_s8(v0hf), vget_high_s8(v1h));
const int16x8_t pl0l = vmull_s8(vget_low_s8 (v0_0lf), vget_low_s8 (v1_0l));
const int16x8_t pl0h = vmull_s8(vget_high_s8(v0_0lf), vget_high_s8(v1_0l));
const int16x8_t ph0l = vmull_s8(vget_low_s8 (v0_0hf), vget_low_s8 (v1_0h));
const int16x8_t ph0h = vmull_s8(vget_high_s8(v0_0hf), vget_high_s8(v1_0h));

const int16x8_t pl1l = vmull_s8(vget_low_s8 (v0_1lf), vget_low_s8 (v1_1l));
const int16x8_t pl1h = vmull_s8(vget_high_s8(v0_1lf), vget_high_s8(v1_1l));
const int16x8_t ph1l = vmull_s8(vget_low_s8 (v0_1hf), vget_low_s8 (v1_1h));
const int16x8_t ph1h = vmull_s8(vget_high_s8(v0_1hf), vget_high_s8(v1_1h));

const int32x4_t pl0 = vaddq_s32(vpaddlq_s16(pl0l), vpaddlq_s16(pl0h));
const int32x4_t ph0 = vaddq_s32(vpaddlq_s16(ph0l), vpaddlq_s16(ph0h));
const int32x4_t pl1 = vaddq_s32(vpaddlq_s16(pl1l), vpaddlq_s16(pl1h));
const int32x4_t ph1 = vaddq_s32(vpaddlq_s16(ph1l), vpaddlq_s16(ph1h));

sumv = vmlaq_n_f32(sumv, vcvtq_f32_s32(vaddq_s32(pl0, ph0)), x0d*y0->d);
sumv0 = vmlaq_n_f32(sumv0, vcvtq_f32_s32(vaddq_s32(pl0, ph0)), x0d*y0->d);
sumv1 = vmlaq_n_f32(sumv1, vcvtq_f32_s32(vaddq_s32(pl1, ph1)), x1d*y1->d);
#endif
}

*s = vaddvq_f32(sumv) + summs;
*s = vaddvq_f32(sumv0) + vaddvq_f32(sumv1) + summs0 + summs1;
#elif defined(__wasm_simd128__)
v128_t sumv = wasm_f32x4_splat(0.0f);

float summs = 0.0f;

uint32_t qh;
uint64_t tmp[4];

for (int i = 0; i < nb; ++i) {
Expand All @@ -2562,13 +2634,12 @@ static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void *
const v128_t m4b = wasm_i8x16_splat(0x0F);

// extract the 5th bit
uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh));

tmp[0] = table_b2b_u[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ];
tmp[0] = table_b2b_0[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_0[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_0[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_0[(qh >> 24) ];

const v128_t qhl = wasm_v128_load(tmp + 0);
const v128_t qhh = wasm_v128_load(tmp + 2);
Expand Down