mirrors/cosmopolitan
1
0
Fork 0
mirror of https://github.com/jart/cosmopolitan.git synced 2025-04-01 05:18:45 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 1

Fix the AARCH64 build

Browse source
This commit is contained in:
Justine Tunney 2023-05-13 08:19:44 -07:00
parent 5a4cf9560f
commit 410c8785c9
No known key found for this signature in database
GPG key ID: BE714B4575D6E328
4 changed files with 22 additions and 19 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

6
third_party/ggml/fp16.c vendored
View file

@ -45,8 +45,10 @@ asm(".include \"libc/disclaimer.inc\"");
#define B6(c,s,n) B5(c,s,n ## c), B5(c,s,n ## s) #define B6(c,s,n) B5(c,s,n ## c), B5(c,s,n ## s)
#define B7(c,s,n) B6(c,s,n ## c), B6(c,s,n ## s) #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) #define B8(c,s ) B7(c,s, c), B7(c,s, s)
// precomputed tables for expanding 8bits to 8 bytes (shl 4)
const uint64_t table_b2b_u[1 << 8] = { B8(00, 10) }; // precomputed tables for expanding 8bits to 8 bytes:
const uint64_t table_b2b_0[1 << 8] = { B8(00, 10) }; // ( b) << 4
const uint64_t table_b2b_1[1 << 8] = { B8(10, 00) }; // (!b) << 4
#endif #endif
// //

3
third_party/ggml/fp16.internal.h vendored
View file

@ -13,7 +13,8 @@ extern ggml_fp16_t table_silu_f16[1 << 16];
extern ggml_fp16_t table_exp_f16[1 << 16]; extern ggml_fp16_t table_exp_f16[1 << 16];
extern float table_f32_f16[1 << 16]; extern float table_f32_f16[1 << 16];
#if defined(__ARM_NEON) || defined(__wasm_simd128__) #if defined(__ARM_NEON) || defined(__wasm_simd128__)
extern const uint64_t table_b2b_u[1 << 8]; extern const uint64_t table_b2b_0[1 << 8];
extern const uint64_t table_b2b_1[1 << 8];
#endif #endif
inline static void ggml_vec_gelu_f16(const int n, ggml_fp16_t* y, inline static void ggml_vec_gelu_f16(const int n, ggml_fp16_t* y,

16
third_party/ggml/ggjt.v1.q5_0.c vendored
View file

@ -179,10 +179,10 @@ void ggml_vec_dot_v1_q5_0_q8_0(const int n, float * restrict s, const void * res
uint32_t qh; uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh)); memcpy(&qh, x0->qh, sizeof(qh));
tmp[0] = table_b2b_u[(qh >> 0) & 0xFF]; tmp[0] = table_b2b_0[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF]; tmp[1] = table_b2b_0[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF]; tmp[2] = table_b2b_0[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ]; tmp[3] = table_b2b_0[(qh >> 24) ];
const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0)); const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0));
const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2)); const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2));
@ -241,10 +241,10 @@ void ggml_vec_dot_v1_q5_0_q8_0(const int n, float * restrict s, const void * res
uint32_t qh; uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh)); memcpy(&qh, x0->qh, sizeof(qh));
tmp[0] = table_b2b_u[(qh >> 0) & 0xFF]; tmp[0] = table_b2b_0[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF]; tmp[1] = table_b2b_0[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF]; tmp[2] = table_b2b_0[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ]; tmp[3] = table_b2b_0[(qh >> 24) ];
const v128_t qhl = wasm_v128_load(tmp + 0); const v128_t qhl = wasm_v128_load(tmp + 0);
const v128_t qhh = wasm_v128_load(tmp + 2); const v128_t qhh = wasm_v128_load(tmp + 2);

16
third_party/ggml/ggjt.v1.q5_1.c vendored
View file

@ -172,10 +172,10 @@ void ggml_vec_dot_v1_q5_1_q8_1(const int n, float * restrict s, const void * res
uint32_t qh; uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh)); memcpy(&qh, x0->qh, sizeof(qh));
tmp[0] = table_b2b_u[(qh >> 0) & 0xFF]; tmp[0] = table_b2b_0[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF]; tmp[1] = table_b2b_0[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF]; tmp[2] = table_b2b_0[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ]; tmp[3] = table_b2b_0[(qh >> 24) ];
const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0)); const int8x16_t qhl = vld1q_s8((const int8_t *)(tmp + 0));
const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2)); const int8x16_t qhh = vld1q_s8((const int8_t *)(tmp + 2));
@ -237,10 +237,10 @@ void ggml_vec_dot_v1_q5_1_q8_1(const int n, float * restrict s, const void * res
uint32_t qh; uint32_t qh;
memcpy(&qh, x0->qh, sizeof(qh)); memcpy(&qh, x0->qh, sizeof(qh));
tmp[0] = table_b2b_u[(qh >> 0) & 0xFF]; tmp[0] = table_b2b_0[(qh >> 0) & 0xFF];
tmp[1] = table_b2b_u[(qh >> 8) & 0xFF]; tmp[1] = table_b2b_0[(qh >> 8) & 0xFF];
tmp[2] = table_b2b_u[(qh >> 16) & 0xFF]; tmp[2] = table_b2b_0[(qh >> 16) & 0xFF];
tmp[3] = table_b2b_u[(qh >> 24) ]; tmp[3] = table_b2b_0[(qh >> 24) ];
const v128_t qhl = wasm_v128_load(tmp + 0); const v128_t qhl = wasm_v128_load(tmp + 0);
const v128_t qhh = wasm_v128_load(tmp + 2); const v128_t qhh = wasm_v128_load(tmp + 2);