happyz
/
llama.cpp
mirror of https://github.com/ggerganov/llama.cpp.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge 0e3a8a0c48 into d612901116

This commit is contained in:
En Yao 2026-02-16 17:45:28 +01:00 committed by GitHub
parent d612901116 0e3a8a0c48
commit 14a06f49cc
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 71 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

74
ggml/src/ggml-cpu/arch/wasm/quants.c
View File

@ -269,6 +269,10 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
v128_t y0_l = wasm_v128_load(y0->qs);
v128_t y0_h = wasm_v128_load(y0->qs + 16);
# if defined(__wasm_relaxed_simd__)
v128_t dp0 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0_0ls, y0_l, wasm_i32x4_splat(0));
dp0 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0_0hs, y0_h, dp0);
# else
// Extend to i16x8 and compute dot products
v128_t dx0l = wasm_i16x8_extend_low_i8x16(v0_0ls);
v128_t dx0h = wasm_i16x8_extend_high_i8x16(v0_0ls);
@ -290,6 +294,7 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i32x4_dot_i16x8(dx0hh, dy0hh)
)
);
# endif
// Load and process x1
v128_t v0_1 = wasm_v128_load(x1->qs);
@ -302,6 +307,10 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
v128_t y1_l = wasm_v128_load(y1->qs);
v128_t y1_h = wasm_v128_load(y1->qs + 16);
# if defined(__wasm_relaxed_simd__)
v128_t dp1 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0_1ls, y1_l, wasm_i32x4_splat(0));
dp1 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0_1hs, y1_h, dp1);
# else
// Extend to i16x8 and compute dot products
v128_t dx1l = wasm_i16x8_extend_low_i8x16(v0_1ls);
v128_t dx1h = wasm_i16x8_extend_high_i8x16(v0_1ls);
@ -323,6 +332,7 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i32x4_dot_i16x8(dx1hh, dy1hh)
)
);
# endif
// Accumulate results with scaling
float scale0 = GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d);
@ -411,6 +421,13 @@ void ggml_vec_dot_q5_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
const v128_t v1l = wasm_v128_load(y0->qs);
const v128_t v1h = wasm_v128_load(y0->qs + 16);
# if defined(__wasm_relaxed_simd__)
v128_t dp = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0lf, v1l, wasm_i32x4_splat(0));
dp = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0hf, v1h, dp);
sumv = wasm_f32x4_add(sumv, wasm_f32x4_mul(wasm_f32x4_convert_i32x4(dp),
wasm_f32x4_splat(GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d))));
# else
// int8x16 -> int16x8
const v128_t v0lfl = wasm_i16x8_extend_low_i8x16 (v0lf);
const v128_t v0lfh = wasm_i16x8_extend_high_i8x16(v0lf);
@ -430,6 +447,7 @@ void ggml_vec_dot_q5_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i32x4_add(wasm_i32x4_dot_i16x8(v0hfl, v1hl),
wasm_i32x4_dot_i16x8(v0hfh, v1hh)))),
wasm_f32x4_splat(GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d))));
# endif
}
sumf = wasm_f32x4_extract_lane(sumv, 0) + wasm_f32x4_extract_lane(sumv, 1) +
@ -506,6 +524,13 @@ void ggml_vec_dot_q5_1_q8_1(int n, float * GGML_RESTRICT s, size_t bs, const voi
const v128_t v1l = wasm_v128_load(y0->qs);
const v128_t v1h = wasm_v128_load(y0->qs + 16);
# if defined(__wasm_relaxed_simd__)
v128_t dp = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0lf, v1l, wasm_i32x4_splat(0));
dp = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(v0hf, v1h, dp);
sumv = wasm_f32x4_add(sumv, wasm_f32x4_mul(wasm_f32x4_convert_i32x4(dp),
wasm_f32x4_splat(GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d))));
# else
// int8x16 -> int16x8
const v128_t v0lfl = wasm_i16x8_extend_low_i8x16 (v0lf);
const v128_t v0lfh = wasm_i16x8_extend_high_i8x16(v0lf);
@ -525,6 +550,7 @@ void ggml_vec_dot_q5_1_q8_1(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i32x4_add(wasm_i32x4_dot_i16x8(v0hfl, v1hl),
wasm_i32x4_dot_i16x8(v0hfh, v1hh)))),
wasm_f32x4_splat(GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d))));
# endif
}
sumf = wasm_f32x4_extract_lane(sumv, 0) + wasm_f32x4_extract_lane(sumv, 1) +
@ -570,6 +596,10 @@ void ggml_vec_dot_q8_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
const v128_t y0_0 = wasm_v128_load(y0->qs);
const v128_t y0_1 = wasm_v128_load(y0->qs + 16);
# if defined(__wasm_relaxed_simd__)
v128_t sum_dots = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(x0_0, y0_0, wasm_i32x4_splat(0));
sum_dots = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(x0_1, y0_1, sum_dots);
# else
// Extend 8-bit to 16-bit
const v128_t x0_0l = wasm_i16x8_extend_low_i8x16(x0_0);
const v128_t x0_0h = wasm_i16x8_extend_high_i8x16(x0_0);
@ -589,7 +619,7 @@ void ggml_vec_dot_q8_0_q8_0(int n, float * GGML_RESTRICT s, size_t bs, const voi
// Sum all dot products
const v128_t sum_dots = wasm_i32x4_add(wasm_i32x4_add(dx0_0, dx0_1), wasm_i32x4_add(dx1_0, dx1_1));
# endif
// Convert to float and accumulate
const float scale = GGML_CPU_FP16_TO_FP32(x0->d) * GGML_CPU_FP16_TO_FP32(y0->d);
sumv = wasm_f32x4_add(sumv, wasm_f32x4_mul(wasm_f32x4_convert_i32x4(sum_dots), wasm_f32x4_splat(scale)));
@ -678,6 +708,10 @@ void ggml_vec_dot_q2_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
v128_t q2_shift_1 = wasm_u8x16_shr(q2_1, shift);
v128_t q2_bits_1 = wasm_v128_and(q2_shift_1, wasm_i8x16_splat(0x03));
# if defined(__wasm_relaxed_simd__)
v128_t p_sum0 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q2_bits_0, q8_0, wasm_i32x4_splat(0));
v128_t p_sum1 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q2_bits_1, q8_1, wasm_i32x4_splat(0));
# else
// Calculate dot products
v128_t p0 = wasm_i32x4_dot_i16x8(
wasm_i16x8_extend_low_i8x16(q8_0),
@ -695,11 +729,15 @@ void ggml_vec_dot_q2_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i16x8_extend_high_i8x16(q8_1),
wasm_i16x8_extend_high_i8x16(q2_bits_1)
);
v128_t p_sum0 = wasm_i32x4_add(p0, p1);
v128_t p_sum1 = wasm_i32x4_add(p2, p3);
# endif
// Accumulate scaled results
v128_t scaled = wasm_i32x4_add(
wasm_i32x4_mul(wasm_i32x4_add(p0, p1), wasm_i32x4_splat(d0)),
wasm_i32x4_mul(wasm_i32x4_add(p2, p3), wasm_i32x4_splat(d1))
wasm_i32x4_mul(p_sum0, wasm_i32x4_splat(d0)),
wasm_i32x4_mul(p_sum1, wasm_i32x4_splat(d1))
);
isum_vec = wasm_i32x4_add(isum_vec, scaled);
@ -914,6 +952,14 @@ void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
q8 += 64;
// Low nibble products
# if defined(__wasm_relaxed_simd__)
v128_t vacc1 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(
q4l0, q8x0, wasm_i32x4_splat(0)
);
vacc1 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(
q4l1, q8x1, vacc1
);
# else
v128_t vacc1 = wasm_i32x4_dot_i16x8(
wasm_i16x8_extend_low_i8x16(q4l0),
wasm_i16x8_extend_low_i8x16(q8x0)
@ -930,8 +976,17 @@ void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i16x8_extend_high_i8x16(q4l1),
wasm_i16x8_extend_high_i8x16(q8x1)
));
# endif
// High nibble products
# if defined(__wasm_relaxed_simd__)
v128_t vacc2 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(
q4h0, q8x2, wasm_i32x4_splat(0)
);
vacc2 = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(
q4h1, q8x3, vacc2
);
# else
v128_t vacc2 = wasm_i32x4_dot_i16x8(
wasm_i16x8_extend_low_i8x16(q4h0),
wasm_i16x8_extend_low_i8x16(q8x2)
@ -948,6 +1003,7 @@ void ggml_vec_dot_q4_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i16x8_extend_high_i8x16(q4h1),
wasm_i16x8_extend_high_i8x16(q8x3)
));
# endif
// Accumulate scaled results
int32_t vacc1_sum = wasm_i32x4_extract_lane(vacc1, 0) + wasm_i32x4_extract_lane(vacc1, 1) +
@ -1055,6 +1111,11 @@ void ggml_vec_dot_q5_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
v128_t q8_3 = wasm_v128_load(q8 + 48);
q8 += 64;
# if defined(__wasm_relaxed_simd__)
// Process low quants using relaxed SIMD
v128_t sum_low = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q5l_0, q8_0, wasm_i32x4_splat(0));
sum_low = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q5l_1, q8_1, sum_low);
# else
// Process low quants
v128_t pl0 = wasm_i32x4_dot_i16x8(
wasm_i16x8_extend_low_i8x16(q5l_0),
@ -1073,7 +1134,13 @@ void ggml_vec_dot_q5_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i16x8_extend_high_i8x16(q8_1)
));
v128_t sum_low = wasm_i32x4_add(pl0, pl1);
# endif
# if defined(__wasm_relaxed_simd__)
// Process high quants using relaxed SIMD
v128_t sum_high = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q5h_0, q8_2, wasm_i32x4_splat(0));
sum_high = wasm_i32x4_relaxed_dot_i8x16_i7x16_add(q5h_1, q8_3, sum_high);
# else
// Process high quants
v128_t ph0 = wasm_i32x4_dot_i16x8(
wasm_i16x8_extend_low_i8x16(q5h_0),
@ -1092,6 +1159,7 @@ void ggml_vec_dot_q5_K_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const voi
wasm_i16x8_extend_high_i8x16(q8_3)
));
v128_t sum_high = wasm_i32x4_add(ph0, ph1);
# endif
// Accumulate with scale factors
int32_t sl = wasm_i32x4_extract_lane(sum_low, 0) + wasm_i32x4_extract_lane(sum_low, 1) +