From cf7d8169ec12cb3adf90d4ff78b3257e74985302 Mon Sep 17 00:00:00 2001 From: Onur Gündüz Date: Tue, 29 Sep 2026 13:05:44 +0200 Subject: Remove unused model families and inference paths --- libffwd/kernels.h | 147 ++---------------------------------------------------- 1 file changed, 4 insertions(+), 143 deletions(-) (limited to 'libffwd/kernels.h') diff --git a/libffwd/kernels.h b/libffwd/kernels.h index 11312db..1025602 100644 --- a/libffwd/kernels.h +++ b/libffwd/kernels.h @@ -1,78 +1,17 @@ -/* - * Transformer inference kernels (CPU, float32). - * - * Low-level math operations. All operate on float32 tensors in row-major order. - */ - #ifndef KERNELS_H #define KERNELS_H #include #include -/* Basic operations */ - void add_inplace(float *a, const float *b, int n); -/* Matrix operations */ - -/* C = A @ B^T: A[M,K], B[N,K], C[M,N] */ -void matmul_t(float *C, const float *A, const float *B, int M, int K, int N); - -/* y = x @ W^T + b: x[seq,in], W[out,in], b[out], y[seq,out] */ +/* y = x @ W^T + b: x[seq,in], W[out,in], b[out], y[seq,out]. */ void linear( float *y, const float *x, const float *W, const float *b, int seq_len, int in_dim, int out_dim); - void linear_nobias(float *y, const float *x, const float *W, int seq_len, int in_dim, int out_dim); -/* bf16 weight variants */ - -void bf16_to_f32_buf(float *dst, const uint16_t *src, size_t n); -void linear_nobias_bf16( - float *y, const float *x, const uint16_t *W_bf16, int seq_len, int in_dim, int out_dim); - -/* Small-sequence BF16 path: compute two projections with one threaded dispatch. */ -void linear_nobias_bf16_pair(float *a, - float *b, - const float *x, - const uint16_t *Wa_bf16, - const uint16_t *Wb_bf16, - int seq_len, - int in_dim, - int a_dim, - int b_dim); - -/* Small-sequence BF16 path: compute Q/K/V matvecs with one threaded dispatch. */ -void linear_nobias_bf16_qkv(float *q, - float *k, - float *v, - const float *x, - const uint16_t *Wq_bf16, - const uint16_t *Wk_bf16, - const uint16_t *Wv_bf16, - int seq_len, - int in_dim, - int q_dim, - int kv_dim); - -/* Dot product using the best available local SIMD implementation. */ -float dot_f32(const float *a, const float *b, int n); - -/* Normalization */ - -/* RMS Normalization: out = x / rms(x) * weight */ -void rms_norm(float *out, const float *x, const float *weight, int seq_len, int hidden, float eps); - -/* Per-head RMS Normalization for Q/K norms in decoder - * x: [seq, n_heads, head_dim], weight: [head_dim] - * Normalizes each head independently */ -void rms_norm_per_head( - float *x, const float *weight, int seq_len, int n_heads, int head_dim, float eps); - -/* Layer normalization for BERT-family models: - * out = gamma * (x - mean) / sqrt(var + eps) + beta. - * Mean and biased/population variance are computed over the hidden axis per row. - * Unlike RMSNorm, this subtracts the mean and adds a bias. */ +/* Mean and population variance are reduced over the hidden axis per row. */ void layer_norm(float *out, const float *x, const float *gamma, @@ -81,44 +20,11 @@ void layer_norm(float *out, int hidden, float eps); -/* Activation functions */ - void softmax(float *x, int rows, int cols); -/* gate = SiLU(gate) * up */ -void silu_mul_inplace(float *gate, const float *up, int n); - -/* Exact (erf) GeLU in place: x = 0.5 * x * (1 + erf(x / sqrt(2))). - * This is the "gelu" activation HF BERT/BGE/MiniLM use, not the tanh approximation. */ void gelu_inplace(float *x, int n); -/* Tanh-approximation GeLU in place (HF gelu_new / gelu_pytorch_tanh): - * x = 0.5 * x * (1 + tanh(sqrt(2/pi) * (x + 0.044715 * x^3))). */ -void gelu_tanh_inplace(float *x, int n); - -/* Attention operations */ - -/* - * Bidirectional GQA attention over a packed/ragged batch. - * - * Q: [total_seq, n_heads * head_dim] - * K,V: [total_seq, n_kv_heads * head_dim] - * offsets: [batch + 1], token spans for each sequence. - * - * Attention is block-diagonal: tokens only attend within their own - * offsets[b]..offsets[b+1] span. - */ -void bidirectional_gqa_attention_packed(float *out, - const float *Q, - const float *K, - const float *V, - const int *offsets, - int batch, - int n_heads, - int n_kv_heads, - int head_dim, - float scale); - -/* Optional reusable scratch for the BLAS-tiled packed attention path. */ +/* Attention is block-diagonal: each token attends only within its offsets span. + * Missing scratch uses the online-softmax path instead of tiled BLAS. */ size_t bidirectional_gqa_attention_packed_scratch_bytes(const int *offsets, int batch); void bidirectional_gqa_attention_packed_with_scratch(float *out, const float *Q, @@ -133,49 +39,4 @@ void bidirectional_gqa_attention_packed_with_scratch(float *out, float *scratch, size_t scratch_bytes); -/* - * Causal GQA attention over a packed/ragged batch. Query position i attends - * only to positions at or before i within its sequence. - */ -void causal_gqa_attention_packed(float *out, - const float *Q, - const float *K, - const float *V, - const int *offsets, - int batch, - int n_heads, - int n_kv_heads, - int head_dim, - float scale); -void causal_gqa_attention_packed_with_scratch(float *out, - const float *Q, - const float *K, - const float *V, - const int *offsets, - int batch, - int n_heads, - int n_kv_heads, - int head_dim, - float scale, - float *scratch, - size_t scratch_bytes); - -/* Position embeddings */ - -/* - * NeoX-style RoPE: compute cos/sin for positions. - * cos_out, sin_out: [seq, head_dim] - * cos[d] and cos[half+d] are the same (duplicated for full head_dim). - */ -void compute_rope_neox( - float *cos_out, float *sin_out, const int *positions, int seq, int head_dim, float theta); - -/* - * Apply NeoX-style RoPE to Q or K. - * x: [seq, n_heads * head_dim] (in-place) - * cos_vals, sin_vals: [seq, head_dim] - */ -void apply_rope_neox( - float *x, const float *cos_vals, const float *sin_vals, int seq, int n_heads, int head_dim); - #endif /* KERNELS_H */ -- cgit v1.3.1