forked from ggml-org/llama.cpp
-
Notifications
You must be signed in to change notification settings - Fork 0
Pull requests: AvalancheHQ/llama.cpp
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
ggml-cpu: write matmul results straight into dst in the nrows==1 fast path
ggml
#50
opened Jul 31, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: run the token-generation matmul without a global barrier
ggml
#49
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: fuse element-wise epilogues (ADD, MUL, SILU) into
ggml_compute_forward_mul_mat
ggml
#48
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: add 2x2 register-blocked AVX2 kernel for ggml_vec_dot_q8_0_q8_0
ggml
#47
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
benchmarks: build ggml with the tinyBLAS GEMM path (GGML_LLAMAFILE) llama.cpp enables by default
#46
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: add 2x2 register-blocked AVX2 kernel for ggml_vec_dot_q8_0_q8_0
ggml
#45
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: unroll the AVX2 q4_K x q8_K sub-block loop
ggml
#44
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-quants: vectorize the per-block extremes scan in the reference quantizers
ggml
#43
opened Jul 30, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: 2x2 register-blocked kernel for
ggml_vec_dot_f16
ggml
#42
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: dequantize each weight row once per matmul tile (4-column Q4_0/Q4_K AVX2 kernels)
ggml
#41
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: keep the generic SIMD vec kernel accumulators in registers
ggml
#40
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: 2x2 register-blocked AVX2 kernel for
ggml_vec_dot_q4_0_q8_0
ggml
#39
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: remove the per-matmul global barrier in ggml_compute_forward_mul_mat
ggml
#38
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: cache-aware tiling in ggml_compute_forward_mul_mat
ggml
#37
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: 2x2 register-blocked AVX2 kernel for ggml_vec_dot_q4_K_q8_K
ggml
#36
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: unpack the Q4_K scales/mins with SIMD in the AVX2 ggml_vec_dot_q4_K_q8_K
ggml
#35
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: 2x2 register-blocked kernel for
ggml_vec_dot_f32 (F32 matmul)
ggml
#34
opened Jul 29, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: vectorize sum-of-squares reduction in RMS-norm F32 kernel
ggml
#33
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: vectorize the fused RMS-norm kernel (ggml_compute_forward_rms_norm_f32)
ggml
#32
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
Optimize AVX2 ggml_vec_dot_q4_0_q8_0: two FMA accumulator chains
ggml
#31
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
Optimize AVX2 ggml_vec_dot_q8_0_q8_0: hardware F16C scale conversion + 2-accumulator unroll
ggml
#30
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: vectorize the fused RMS-norm × weight kernel
ggml
#29
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: unroll ggml_vec_dot_f16 into four accumulator groups to hide FMA latency
ggml
#28
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
ggml-quants: evaluate the k-quant scale search error in O(1) in
make_qkx2_quants
ggml
#27
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
ggml-cpu: vectorize
ggml_vec_max_f32 across SIMD targets (soft-max row-max scan)
ggml
#26
opened Jul 28, 2026 by
codspeed-hq
Bot
Loading…
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.