cui-llama.rn 1.7.3 → 1.7.6

package/cpp/ggml-cpu/arch-fallback.h

@@ -0,0 +1,184 @@
+#pragma once
+
+// Rename `_generic` functions if no native implementation is available.
+// This effectively selects the generic implementation.
+
+#if defined(LM_GGML_CPU_GENERIC)
+// quants.c
+#define quantize_row_q8_0_generic quantize_row_q8_0
+#define quantize_row_q8_1_generic quantize_row_q8_1
+#define quantize_row_q8_K_generic quantize_row_q8_K
+#define lm_ggml_vec_dot_q4_0_q8_0_generic lm_ggml_vec_dot_q4_0_q8_0
+#define lm_ggml_vec_dot_q4_1_q8_1_generic lm_ggml_vec_dot_q4_1_q8_1
+#define lm_ggml_vec_dot_q5_0_q8_0_generic lm_ggml_vec_dot_q5_0_q8_0
+#define lm_ggml_vec_dot_q5_1_q8_1_generic lm_ggml_vec_dot_q5_1_q8_1
+#define lm_ggml_vec_dot_q8_0_q8_0_generic lm_ggml_vec_dot_q8_0_q8_0
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_q2_K_q8_K_generic lm_ggml_vec_dot_q2_K_q8_K
+#define lm_ggml_vec_dot_q3_K_q8_K_generic lm_ggml_vec_dot_q3_K_q8_K
+#define lm_ggml_vec_dot_q4_K_q8_K_generic lm_ggml_vec_dot_q4_K_q8_K
+#define lm_ggml_vec_dot_q5_K_q8_K_generic lm_ggml_vec_dot_q5_K_q8_K
+#define lm_ggml_vec_dot_q6_K_q8_K_generic lm_ggml_vec_dot_q6_K_q8_K
+#define lm_ggml_vec_dot_iq2_xxs_q8_K_generic lm_ggml_vec_dot_iq2_xxs_q8_K
+#define lm_ggml_vec_dot_iq2_xs_q8_K_generic lm_ggml_vec_dot_iq2_xs_q8_K
+#define lm_ggml_vec_dot_iq2_s_q8_K_generic lm_ggml_vec_dot_iq2_s_q8_K
+#define lm_ggml_vec_dot_iq3_xxs_q8_K_generic lm_ggml_vec_dot_iq3_xxs_q8_K
+#define lm_ggml_vec_dot_iq3_s_q8_K_generic lm_ggml_vec_dot_iq3_s_q8_K
+#define lm_ggml_vec_dot_iq1_s_q8_K_generic lm_ggml_vec_dot_iq1_s_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+#define lm_ggml_vec_dot_iq4_nl_q8_0_generic lm_ggml_vec_dot_iq4_nl_q8_0
+#define lm_ggml_vec_dot_iq4_xs_q8_K_generic lm_ggml_vec_dot_iq4_xs_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_0_8x8_q8_0_generic lm_ggml_gemv_q4_0_8x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_0_8x8_q8_0_generic lm_ggml_gemm_q4_0_8x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__aarch64__) || defined(__arm__) || defined(_M_ARM) || defined(_M_ARM64)
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#elif defined(__x86_64__) || defined(__i386__) || defined(_M_IX86) || defined(_M_X64)
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__POWERPC__) || defined(__powerpc__)
+// ref: https://github.com/ggml-org/llama.cpp/pull/14146#issuecomment-2972561679
+// quants.c
+#define quantize_row_q8_K_generic quantize_row_q8_K
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_0_8x8_q8_0_generic lm_ggml_gemv_q4_0_8x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_0_8x8_q8_0_generic lm_ggml_gemm_q4_0_8x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__loongarch64)
+// quants.c
+#define quantize_row_q8_K_generic quantize_row_q8_K
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_0_8x8_q8_0_generic lm_ggml_gemv_q4_0_8x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_0_8x8_q8_0_generic lm_ggml_gemm_q4_0_8x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__riscv)
+// quants.c
+#define quantize_row_q8_K_generic quantize_row_q8_K
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_iq2_xxs_q8_K_generic lm_ggml_vec_dot_iq2_xxs_q8_K
+#define lm_ggml_vec_dot_iq2_xs_q8_K_generic lm_ggml_vec_dot_iq2_xs_q8_K
+#define lm_ggml_vec_dot_iq2_s_q8_K_generic lm_ggml_vec_dot_iq2_s_q8_K
+#define lm_ggml_vec_dot_iq3_xxs_q8_K_generic lm_ggml_vec_dot_iq3_xxs_q8_K
+#define lm_ggml_vec_dot_iq3_s_q8_K_generic lm_ggml_vec_dot_iq3_s_q8_K
+#define lm_ggml_vec_dot_iq1_s_q8_K_generic lm_ggml_vec_dot_iq1_s_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+#define lm_ggml_vec_dot_iq4_nl_q8_0_generic lm_ggml_vec_dot_iq4_nl_q8_0
+#define lm_ggml_vec_dot_iq4_xs_q8_K_generic lm_ggml_vec_dot_iq4_xs_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__s390x__)
+// quants.c
+#define quantize_row_q8_K_generic quantize_row_q8_K
+#define lm_ggml_vec_dot_q5_0_q8_0_generic lm_ggml_vec_dot_q5_0_q8_0
+#define lm_ggml_vec_dot_q5_1_q8_1_generic lm_ggml_vec_dot_q5_1_q8_1
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_q2_K_q8_K_generic lm_ggml_vec_dot_q2_K_q8_K
+#define lm_ggml_vec_dot_iq2_xxs_q8_K_generic lm_ggml_vec_dot_iq2_xxs_q8_K
+#define lm_ggml_vec_dot_iq2_xs_q8_K_generic lm_ggml_vec_dot_iq2_xs_q8_K
+#define lm_ggml_vec_dot_iq2_s_q8_K_generic lm_ggml_vec_dot_iq2_s_q8_K
+#define lm_ggml_vec_dot_iq3_xxs_q8_K_generic lm_ggml_vec_dot_iq3_xxs_q8_K
+#define lm_ggml_vec_dot_iq3_s_q8_K_generic lm_ggml_vec_dot_iq3_s_q8_K
+#define lm_ggml_vec_dot_iq1_s_q8_K_generic lm_ggml_vec_dot_iq1_s_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_0_8x8_q8_0_generic lm_ggml_gemv_q4_0_8x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_0_8x8_q8_0_generic lm_ggml_gemm_q4_0_8x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#elif defined(__wasm__)
+// quants.c
+#define lm_ggml_vec_dot_q4_1_q8_1_generic lm_ggml_vec_dot_q4_1_q8_1
+#define lm_ggml_vec_dot_tq1_0_q8_K_generic lm_ggml_vec_dot_tq1_0_q8_K
+#define lm_ggml_vec_dot_tq2_0_q8_K_generic lm_ggml_vec_dot_tq2_0_q8_K
+#define lm_ggml_vec_dot_iq2_xxs_q8_K_generic lm_ggml_vec_dot_iq2_xxs_q8_K
+#define lm_ggml_vec_dot_iq2_xs_q8_K_generic lm_ggml_vec_dot_iq2_xs_q8_K
+#define lm_ggml_vec_dot_iq2_s_q8_K_generic lm_ggml_vec_dot_iq2_s_q8_K
+#define lm_ggml_vec_dot_iq3_xxs_q8_K_generic lm_ggml_vec_dot_iq3_xxs_q8_K
+#define lm_ggml_vec_dot_iq3_s_q8_K_generic lm_ggml_vec_dot_iq3_s_q8_K
+#define lm_ggml_vec_dot_iq1_s_q8_K_generic lm_ggml_vec_dot_iq1_s_q8_K
+#define lm_ggml_vec_dot_iq1_m_q8_K_generic lm_ggml_vec_dot_iq1_m_q8_K
+#define lm_ggml_vec_dot_iq4_nl_q8_0_generic lm_ggml_vec_dot_iq4_nl_q8_0
+#define lm_ggml_vec_dot_iq4_xs_q8_K_generic lm_ggml_vec_dot_iq4_xs_q8_K
+// repack.cpp
+#define lm_ggml_quantize_mat_q8_0_4x4_generic lm_ggml_quantize_mat_q8_0_4x4
+#define lm_ggml_quantize_mat_q8_0_4x8_generic lm_ggml_quantize_mat_q8_0_4x8
+#define lm_ggml_quantize_mat_q8_K_4x8_generic lm_ggml_quantize_mat_q8_K_4x8
+#define lm_ggml_gemv_q4_0_4x4_q8_0_generic lm_ggml_gemv_q4_0_4x4_q8_0
+#define lm_ggml_gemv_q4_0_4x8_q8_0_generic lm_ggml_gemv_q4_0_4x8_q8_0
+#define lm_ggml_gemv_q4_0_8x8_q8_0_generic lm_ggml_gemv_q4_0_8x8_q8_0
+#define lm_ggml_gemv_q4_K_8x8_q8_K_generic lm_ggml_gemv_q4_K_8x8_q8_K
+#define lm_ggml_gemv_iq4_nl_4x4_q8_0_generic lm_ggml_gemv_iq4_nl_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x4_q8_0_generic lm_ggml_gemm_q4_0_4x4_q8_0
+#define lm_ggml_gemm_q4_0_4x8_q8_0_generic lm_ggml_gemm_q4_0_4x8_q8_0
+#define lm_ggml_gemm_q4_0_8x8_q8_0_generic lm_ggml_gemm_q4_0_8x8_q8_0
+#define lm_ggml_gemm_q4_K_8x8_q8_K_generic lm_ggml_gemm_q4_K_8x8_q8_K
+#define lm_ggml_gemm_iq4_nl_4x4_q8_0_generic lm_ggml_gemm_iq4_nl_4x4_q8_0
+#endif

package/cpp/ggml-cpu/common.h

@@ -1,9 +1,10 @@
 #pragma once
 
 #include "ggml.h"
-#include "ggml-cpu-traits.h"
+#include "traits.h"
 #include "ggml-cpu-impl.h"
 #include "ggml-impl.h"
+#include "simd-mappings.h"
 
 #ifdef __cplusplus
 
@@ -12,11 +13,11 @@
 // convenience functions/macros for use in template calls
 // note: these won't be required after the 'traits' lookup table is used.
 static inline lm_ggml_fp16_t f32_to_f16(float x) {
-    return LM_GGML_FP32_TO_FP16(x);
+    return LM_GGML_CPU_FP32_TO_FP16(x);
 }
 
 static inline float f16_to_f32(lm_ggml_fp16_t x) {
-    return LM_GGML_FP16_TO_FP32(x);
+    return LM_GGML_CPU_FP16_TO_FP32(x);
 }
 
 static inline lm_ggml_bf16_t f32_to_bf16(float x) {

package/cpp/ggml-cpu/ggml-cpu-impl.h

@@ -62,11 +62,17 @@ struct lm_ggml_compute_params {
 #if defined(__s390x__) && defined(__VEC__)
 #ifndef __VXE__
 #define __VXE__
-#endif
+#endif  // __VXE__
 #ifndef __VXE2__
 #define __VXE2__
-#endif
-#endif
+#endif  // __VXE2__
+#endif  // __s390x__ && __VEC__
+
+#if defined(__s390x__) && defined(LM_GGML_NNPA)
+#ifndef __NNPA__
+#define __NNPA__
+#endif  // __NNPA__
+#endif  // __s390x__ && LM_GGML_NNPA
 
 #if defined(__ARM_FEATURE_SVE)
 #include <sys/prctl.h>
@@ -320,21 +326,17 @@ inline static int32x4_t lm_ggml_vdotq_s32(int32x4_t acc, int8x16_t a, int8x16_t
 
 #ifdef __wasm_simd128__
 #include <wasm_simd128.h>
-#else
+#endif
+
 #ifdef __POWER9_VECTOR__
 #include <altivec.h>
-#else
+#endif
+
 #if defined(_MSC_VER) || defined(__MINGW32__)
 #include <intrin.h>
-#else
-#if defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__) || defined(__SSSE3__) || defined(__SSE3__) || defined(__SSE__)
-#if !defined(__riscv)
+#elif defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__) || defined(__SSSE3__) || defined(__SSE3__) || defined(__SSE__)
 #include <immintrin.h>
 #endif
-#endif
-#endif
-#endif
-#endif
 
 #ifdef __riscv_v_intrinsic
 #include <riscv_vector.h>
@@ -375,7 +377,7 @@ inline static int32x4_t lm_ggml_vdotq_s32(int32x4_t acc, int8x16_t a, int8x16_t
 #define vec_xor(a, b) ((a) ^ (b)) // Vector XOR
 #endif
 
-typedef signed char char8x16_t __attribute__((vector_size(16)));
+typedef signed   char char8x16_t  __attribute__((vector_size(16)));
 typedef unsigned char uchar8x16_t __attribute__((vector_size(16)));
 
 typedef int8_t  int8x16_t __attribute__((vector_size(16)));
@@ -386,10 +388,10 @@ typedef uint8_t  uint8x16_t __attribute__((vector_size(16)));
 typedef uint16_t uint16x8_t __attribute__((vector_size(16)));
 typedef uint32_t uint32x4_t __attribute__((vector_size(16)));
 
-typedef float float32x4_t __attribute__((vector_size(16)));
-typedef double double64x2_t __attribute((vector_size(16)));
+typedef float  float32x4_t  __attribute__((vector_size(16)));
+typedef double double64x2_t __attribute__((vector_size(16)));
 
-typedef signed long long long64x2_t __attribute((vector_size(16)));
+typedef signed   long long long64x2_t  __attribute__((vector_size(16)));
 typedef unsigned long long ulong64x2_t __attribute__((vector_size(16)));
 
 typedef struct lm_ggml_uint8x16x2_t {
@@ -507,6 +509,9 @@ static __m256 __lasx_xvreplfr2vr_s(const float val) {
 // TODO: move to ggml-threading
 void lm_ggml_barrier(struct lm_ggml_threadpool * tp);
 
+void lm_ggml_threadpool_chunk_set(struct lm_ggml_threadpool * tp, int value);
+int  lm_ggml_threadpool_chunk_add(struct lm_ggml_threadpool * tp, int value);
+
 #ifdef __cplusplus
 }
 #endif