whispercpp 1.3.6 → 1.3.7

data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.c

@@ -50,6 +50,10 @@
 #include "llamafile/sgemm.h"
 #endif
 
+#ifdef GGML_USE_CPU_RISCV64_SPACEMIT
+#    include "spacemit/ime.h"
+#endif
+
 // Note: once we move threading into a separate C++ file
 // will use std::hardware_destructive_interference_size instead of hardcoding it here
 // and we'll use C++ attribute syntax.
@@ -217,6 +221,12 @@ static const struct ggml_type_traits_cpu type_traits_cpu[GGML_TYPE_COUNT] = {
         .vec_dot_type             = GGML_TYPE_F16,
         .nrows                    = 1,
     },
+    [GGML_TYPE_Q1_0] = {
+        .from_float               = quantize_row_q1_0,
+        .vec_dot                  = ggml_vec_dot_q1_0_q8_0,
+        .vec_dot_type             = GGML_TYPE_Q8_0,
+        .nrows                    = 1,
+    },
     [GGML_TYPE_Q4_0] = {
         .from_float               = quantize_row_q4_0,
         .vec_dot                  = ggml_vec_dot_q4_0_q8_0,
@@ -1239,6 +1249,12 @@ void ggml_compute_forward_mul_mat(
     const struct ggml_tensor * src0 = dst->src[0];
     const struct ggml_tensor * src1 = dst->src[1];
 
+    const int32_t hint = ggml_get_op_params_i32(dst, 1);
+    if (hint == GGML_HINT_SRC0_IS_HADAMARD && !params->use_ref) {
+        ggml_compute_forward_fwht(params, dst);
+        return;
+    }
+
     GGML_TENSOR_BINARY_OP_LOCALS
 
     const int ith = params->ith;
@@ -1896,6 +1912,10 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_im2col_3d(params, tensor);
             } break;
+        case GGML_OP_COL2IM_1D:
+            {
+                ggml_compute_forward_col2im_1d(params, tensor);
+            } break;
         case GGML_OP_CONV_2D:
             {
                 ggml_compute_forward_conv_2d(params, tensor);
@@ -2327,6 +2347,7 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
         case GGML_OP_CONV_2D:
         case GGML_OP_CONV_3D:
         case GGML_OP_CONV_2D_DW:
+        case GGML_OP_COL2IM_1D:
         case GGML_OP_CONV_TRANSPOSE_1D:
         case GGML_OP_CONV_TRANSPOSE_2D:
             {
@@ -2350,11 +2371,15 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
         case GGML_OP_FLASH_ATTN_BACK:
         case GGML_OP_SSM_CONV:
         case GGML_OP_SSM_SCAN:
+            {
+                n_tasks = n_threads;
+            } break;
         case GGML_OP_RWKV_WKV6:
         case GGML_OP_GATED_LINEAR_ATTN:
         case GGML_OP_RWKV_WKV7:
             {
-                n_tasks = n_threads;
+                const int64_t n_heads = node->src[1]->ne[1];
+                n_tasks = MIN(n_threads, n_heads);
             } break;
         case GGML_OP_WIN_PART:
         case GGML_OP_WIN_UNPART:
@@ -2871,8 +2896,12 @@ struct ggml_cplan ggml_graph_plan(
                         const int64_t ne11 = node->src[1]->ne[1]; // H
                         const int64_t ne12 = node->src[1]->ne[2]; // Channels In
 
-                        cur += sizeof(ggml_fp16_t)*ne00*ne01*ne02*ne03;
-                        cur += sizeof(ggml_fp16_t)*ne10*ne11*ne12;
+                        GGML_ASSERT(node->src[0]->type == GGML_TYPE_F16 || node->src[0]->type == GGML_TYPE_F32);
+                        GGML_ASSERT(node->src[1]->type == GGML_TYPE_F32);
+
+                        cur += ggml_type_size(node->src[0]->type) * ne00 * ne01 * ne02 * ne03;
+                        cur += ggml_type_size(node->src[0]->type) * ne10 * ne11 * ne12;
+
                     } break;
                 case GGML_OP_TOP_K:
                     {
@@ -2919,7 +2948,9 @@ struct ggml_cplan ggml_graph_plan(
                 case GGML_OP_GATED_DELTA_NET:
                     {
                         const int64_t S_v = node->src[2]->ne[0];
-                        cur = S_v * sizeof(float) * n_tasks;
+                        const int64_t K   = ggml_get_op_params_i32(node, 0);
+                        const int64_t per_thread = S_v + (K > 1 ? S_v * S_v : 0);
+                        cur = per_thread * sizeof(float) * n_tasks;
                     } break;
                 case GGML_OP_COUNT:
                     {
@@ -2945,6 +2976,45 @@ struct ggml_cplan ggml_graph_plan(
     return cplan;
 }
 
+
+// Try to fuse the current node with subsequent nodes for better performance.
+// Returns the number of nodes skipped by fusion (>=1), or 0 if no fusion was applied.
+static bool ggml_cpu_disable_fusion = false;  // initialized once in ggml_cpu_init(), read-only afterwards
+
+static int ggml_cpu_try_fuse_ops(
+        const struct ggml_cgraph * cgraph,
+        const int node_n,
+        const struct ggml_compute_params * params,
+        const struct ggml_cplan * cplan) {
+
+    if (ggml_cpu_disable_fusion || cplan->use_ref) {
+        return 0;
+    }
+
+    struct ggml_tensor * node = cgraph->nodes[node_n];
+
+    if (node->op == GGML_OP_RMS_NORM) {
+        // RMS_NORM + MUL fusion
+        const enum ggml_op fuse_ops[] = { GGML_OP_RMS_NORM, GGML_OP_MUL };
+        if (ggml_can_fuse(cgraph, node_n, fuse_ops, 2)) {
+            struct ggml_tensor * mul_node = cgraph->nodes[node_n + 1];
+            const struct ggml_tensor * mul_w = (mul_node->src[0] == node)
+                ? mul_node->src[1] : mul_node->src[0];
+            if (node->src[0]->type  == GGML_TYPE_F32 &&
+                mul_node->type      == GGML_TYPE_F32 &&
+                mul_w->type         == GGML_TYPE_F32 &&
+                mul_w->ne[0]        == node->ne[0]   &&
+                mul_w->nb[0]        == sizeof(float)) {
+
+                ggml_compute_forward_rms_norm_mul_fused(params, node, mul_node);
+                return 1;
+            }
+        }
+    }
+
+    return 0;
+}
+
 static thread_ret_t ggml_graph_compute_thread(void * data) {
     struct ggml_compute_state * state = (struct ggml_compute_state *) data;
     struct ggml_threadpool    * tp    = state->threadpool;
@@ -2952,7 +3022,11 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
     const struct ggml_cgraph * cgraph = tp->cgraph;
     const struct ggml_cplan  * cplan  = tp->cplan;
 
+#ifdef GGML_USE_CPU_RISCV64_SPACEMIT
+    ggml_backend_cpu_riscv64_spacemit_set_numa_thread_affinity(state->ith);
+#else
     set_numa_thread_affinity(state->ith);
+#endif
 
     struct ggml_compute_params params = {
         /*.ith        =*/ state->ith,
@@ -2981,7 +3055,14 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
             continue;
         }
 
-        ggml_compute_forward(&params, node);
+        // TODO: move fused-op detection into ggml_graph_plan so fusion decisions are made once at planning time
+        // Try fused ops, fall back to normal compute
+        const int n_fused = ggml_cpu_try_fuse_ops(cgraph, node_n, &params, cplan);
+        if (n_fused > 0) {
+            node_n += n_fused;
+        } else {
+            ggml_compute_forward(&params, node);
+        }
 
         if (state->ith == 0 && cplan->abort_callback &&
                 cplan->abort_callback(cplan->abort_callback_data)) {
@@ -3002,6 +3083,10 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
 
     ggml_barrier(state->threadpool);
 
+#ifdef GGML_USE_CPU_RISCV64_SPACEMIT
+    ggml_backend_cpu_riscv64_spacemit_clear_numa_thread_affinity_threaded(state->ith);
+#endif
+
     return 0;
 }
 
@@ -3743,6 +3828,11 @@ void ggml_cpu_init(void) {
         ggml_init_riscv_arch_features();
 #endif
 
+        {
+            const char * env = getenv("GGML_CPU_DISABLE_FUSION");
+            ggml_cpu_disable_fusion = (env != NULL && atoi(env) == 1);
+        }
+
         is_first_call = false;
     }
 

data/ext/sources/ggml/src/ggml-cpu/ggml-cpu.cpp

@@ -195,6 +195,8 @@ static const struct ggml_backend_i ggml_backend_cpu_i = {
     /* .free                    = */ ggml_backend_cpu_free,
     /* .set_tensor_async        = */ NULL,
     /* .get_tensor_async        = */ NULL,
+    /* .set_tensor_2d_async     = */ NULL,
+    /* .get_tensor_2d_async     = */ NULL,
     /* .cpy_tensor_async        = */ NULL,
     /* .synchronize             = */ NULL,
     /* .graph_plan_create       = */ ggml_backend_cpu_graph_plan_create,

data/ext/sources/ggml/src/ggml-cpu/kleidiai/kleidiai.cpp

@@ -38,6 +38,7 @@
 #include "kleidiai.h"
 
 #include "ggml-cpu.h"
+#include "ggml-cpu-impl.h"
 #include "ggml-impl.h"
 #include "ggml-backend-impl.h"
 #include "ggml-threading.h"
@@ -61,7 +62,8 @@ struct ggml_kleidiai_context {
     ggml_kleidiai_kernels * kernels_q8;
     int sme_thread_cap; // <= 0 means “SME disabled/unknown”;
     int thread_hint;    // <= 0 means “no hint”
-} static ctx = { CPU_FEATURE_NONE, nullptr, nullptr, 0, -1 };
+    int chunk_multiplier;
+} static ctx = { CPU_FEATURE_NONE, nullptr, nullptr, 0, -1, 4 };
 
 static const char* cpu_feature_to_string(cpu_feature f) {
     if (f == CPU_FEATURE_NONE) {
@@ -186,8 +188,9 @@ static void init_kleidiai_context(void) {
     if (!initialized) {
         initialized = true;
 
-        const char *env_sme     = getenv("GGML_KLEIDIAI_SME");
-        const char *env_threads = getenv("GGML_TOTAL_THREADS");
+        const char *env_sme         = getenv("GGML_KLEIDIAI_SME");
+        const char *env_threads     = getenv("GGML_TOTAL_THREADS");
+        const char *env_chunk_mult  = getenv("GGML_KLEIDIAI_CHUNK_MULTIPLIER");
 
         const bool cpu_has_sme = ggml_cpu_has_sme();
         size_t detected_smcus = 0;
@@ -204,6 +207,14 @@ static void init_kleidiai_context(void) {
             }
         }
 
+        if (env_chunk_mult) {
+            bool ok = false;
+            int multiplier = parse_uint_env(env_chunk_mult, "GGML_KLEIDIAI_CHUNK_MULTIPLIER", &ok);
+            if (ok && multiplier > 0) {
+                ctx.chunk_multiplier = multiplier;
+            }
+        }
+
         // SME policy:
         // - If CPU doesn't support SME: SME always off.
         // - Else:
@@ -296,6 +307,50 @@ static inline size_t align_up(size_t value, size_t alignment) {
     return remainder == 0 ? value : value + (alignment - remainder);
 }
 
+static inline size_t gcd_size(size_t a, size_t b) {
+    while (b != 0) {
+        const size_t t = a % b;
+        a = b;
+        b = t;
+    }
+    return a;
+}
+
+static inline bool lcm_size(size_t a, size_t b, size_t & result) {
+    if (a == 0 || b == 0) {
+        result = 0;
+        return false;
+    }
+    const size_t g = gcd_size(a, b);
+    const size_t q = a / g;
+    if (q > SIZE_MAX / b) {
+        return false;
+    }
+    result = q * b;
+    return true;
+}
+
+static inline size_t ceil_div_size(size_t a, size_t b) {
+    return b == 0 ? 0 : (a + b - 1) / b;
+}
+
+struct kleidiai_block_args {
+    size_t lhs_bl;
+    size_t rhs_bl;
+    size_t pack_bl;
+};
+
+static inline kleidiai_block_args kleidiai_get_block_args(ggml_type rhs_type) {
+    switch (rhs_type) {
+        case GGML_TYPE_Q4_0:
+            return { QK4_0, QK4_0, QK4_0 };
+        case GGML_TYPE_Q8_0:
+            return { 0, 0, QK8_0 };
+        default:
+            return { 0, 0, 0 };
+    }
+}
+
 static inline bool kleidiai_pack_fallback_allowed() {
     if (ctx.sme_thread_cap <= 0) {
         return false;
@@ -746,8 +801,10 @@ class tensor_traits : public ggml::cpu::tensor_traits {
             size_t n_step;
             size_t lhs_packed_size;
             size_t lhs_offset;
-            size_t n_offset;
-            size_t n_cols;
+            size_t lhs_bl;
+            size_t rhs_bl;
+            size_t pack_bl;
+            size_t lhs_packed_offset0;
             int assigned_threads;
             int thread_begin;
             int thread_end;
@@ -772,6 +829,8 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                 continue;
             }
 
+            const kleidiai_block_args block_args = kleidiai_get_block_args(kernels->rhs_type);
+
             runtime[runtime_count] = {
                 slot,
                 kernels,
@@ -784,7 +843,9 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                 kinfo->get_n_step(),
                 0,
                 0,
-                0,
+                block_args.lhs_bl,
+                block_args.rhs_bl,
+                block_args.pack_bl,
                 0,
                 0,
                 0,
@@ -795,45 +856,8 @@ class tensor_traits : public ggml::cpu::tensor_traits {
         }
 
         if (runtime_count == 0) {
-            ggml_kleidiai_kernels * fallback = ggml_kleidiai_select_kernels(ctx.features, dst);
-            if (!fallback) {
-                return false;
-            }
-            kernel_info * kinfo      = is_gemv ? &fallback->gemv : &fallback->gemm;
-            lhs_packing_info * linfo = is_gemv ? &fallback->gemv_lhs_info : &fallback->gemm_lhs_info;
-            rhs_packing_info * rinfo = &fallback->rhs_info;
-            if (!kinfo || !linfo || !linfo->packed_size_ex || !linfo->pack_func_ex ||
-                !kinfo->get_rhs_packed_offset_ex || !kinfo->run_kernel_ex || !kinfo->get_dst_offset ||
-                !rinfo || !rinfo->pack_func_ex || !rinfo->packed_size_ex) {
-                return false;
-            }
-            kernel_chain[0] = fallback;
-            runtime[0] = {
-                0,
-                fallback,
-                kinfo,
-                linfo,
-                kinfo->get_mr(),
-                kinfo->get_nr(),
-                kinfo->get_kr(),
-                kinfo->get_sr(),
-                kinfo->get_n_step(),
-                0,
-                0,
-                0,
-                0,
-                0,
-                0,
-                0,
-                nullptr
-            };
-            size_t rhs_size_fallback = 0;
-            const uint8_t * rhs_base = weight_for_slot(0, rhs_size_fallback);
-            if (!rhs_base) {
-                rhs_base = static_cast<const uint8_t *>(src0->data);
-            }
-            runtime[0].rhs_base = rhs_base;
-            runtime_count = 1;
+            GGML_LOG_WARN("kleidiai: no runtime kernel slot available for supported op %s\n", dst->name);
+            return false;
         }
 
         const int nth_total = params->nth > 0 ? params->nth : 1;
@@ -846,6 +870,13 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                 break;
             }
         }
+        int non_sme_slot = -1;
+        for (int i = 0; i < runtime_count; ++i) {
+            if ((runtime[i].kernels->required_cpu & CPU_FEATURE_SME) != CPU_FEATURE_SME) {
+                non_sme_slot = i;
+                break;
+            }
+        }
 
         const int sme_cap_limit = ctx.sme_thread_cap;
         const bool use_hybrid = sme_cap_limit > 0 &&
@@ -864,12 +895,15 @@ class tensor_traits : public ggml::cpu::tensor_traits {
         if (!hybrid_enabled) {
             int chosen_slot = 0;
             if (too_small_for_hybrid && sme_slot != -1) {
-                chosen_slot = sme_slot;
+                chosen_slot = nth_total > sme_cap_limit && non_sme_slot != -1 ? non_sme_slot : sme_slot;
             } else if (runtime_count > 1 && ctx.sme_thread_cap > 0 && nth_total > ctx.sme_thread_cap) {
                 chosen_slot = 1;
             }
             if (chosen_slot != 0 && chosen_slot < runtime_count) {
                 runtime[0] = runtime[chosen_slot];
+                runtime[0].assigned_threads = 0;
+                runtime[0].thread_begin = 0;
+                runtime[0].thread_end = 0;
             }
             runtime_count = runtime_count > 0 ? 1 : 0;
 
@@ -896,6 +930,8 @@ class tensor_traits : public ggml::cpu::tensor_traits {
 
         int fallback_indices[GGML_KLEIDIAI_MAX_KERNEL_SLOTS];
         int fallback_count = 0;
+        // The current hybrid chain is bounded to SME + one non-SME fallback slot.
+        GGML_ASSERT(GGML_KLEIDIAI_MAX_KERNEL_SLOTS == 2);
         for (int i = 0; i < runtime_count; ++i) {
             if (i == sme_slot) {
                 continue;
@@ -952,73 +988,67 @@ class tensor_traits : public ggml::cpu::tensor_traits {
 
         size_t cursor = 0;
         for (int i = 0; i < runtime_count; ++i) {
-            const ggml_type slot_rhs_type = runtime[i].kernels->rhs_type;
-            const size_t slot_pack_size_arg = slot_rhs_type == GGML_TYPE_Q4_0 ? QK4_0 :
-                                              slot_rhs_type == GGML_TYPE_Q8_0 ? QK8_0 : 0;
-            runtime[i].lhs_packed_size = runtime[i].lhs_info->packed_size_ex(m, k, slot_pack_size_arg, runtime[i].mr, runtime[i].kr, runtime[i].sr);
+            runtime[i].lhs_packed_size = runtime[i].lhs_info->packed_size_ex(m, k, runtime[i].pack_bl, runtime[i].mr, runtime[i].kr, runtime[i].sr);
             cursor = align_up(cursor, GGML_KLEIDIAI_PACK_ALIGN);
             runtime[i].lhs_offset = cursor;
+            runtime[i].lhs_packed_offset0 = runtime[i].lhs_info->get_packed_offset_ex(0, k, runtime[i].lhs_bl, runtime[i].mr, runtime[i].kr, runtime[i].sr);
             cursor += runtime[i].lhs_packed_size;
         }
 
         GGML_ASSERT(cursor <= params->wsize);
         uint8_t * scratch = static_cast<uint8_t *>(params->wdata);
 
-        size_t assigned_cols = 0;
-        uint64_t weighted_total = 0;
-        if (runtime_count > 1 && sme_slot != -1) {
-            for (int i = 0; i < runtime_count; ++i) {
-                const uint64_t weight = (i == sme_slot) ? (sme_cap << 1) : 1;
-                weighted_total += (uint64_t)runtime[i].assigned_threads * weight;
-            }
-        }
+        size_t common_step = 1;
         for (int i = 0; i < runtime_count; ++i) {
-            runtime[i].n_offset = assigned_cols;
             if (runtime[i].assigned_threads == 0) {
-                runtime[i].n_cols = 0;
-                continue;
-            }
-            const size_t remaining_cols = n - assigned_cols;
-            if (remaining_cols == 0) {
-                runtime[i].n_cols = 0;
                 continue;
             }
-            const size_t step = runtime[i].n_step ? runtime[i].n_step : 1;
-            size_t target      = 0;
-            if (weighted_total > 0) {
-                const uint64_t weight = (i == sme_slot) ? (sme_cap << 1) : 1;
-                target = (size_t)(((uint64_t)n * runtime[i].assigned_threads * weight) / weighted_total);
-            } else {
-                target = (size_t)(((uint64_t)n * runtime[i].assigned_threads) / nth_total);
-            }
-            target             = std::min(target, remaining_cols);
-            size_t aligned     = round_down(target, step);
-            if (aligned == 0 && remaining_cols >= step) {
-                aligned = step;
+            size_t next_step = 0;
+            if (!lcm_size(common_step, runtime[i].n_step ? runtime[i].n_step : 1, next_step)) {
+                return false;
             }
-            runtime[i].n_cols = aligned;
-            assigned_cols += aligned;
+            common_step = next_step;
         }
-
-        if (assigned_cols < n) {
-            for (int i = runtime_count - 1; i >= 0; --i) {
-                if (runtime[i].assigned_threads > 0) {
-                    runtime[i].n_cols += n - assigned_cols;
-                    break;
-                }
-            }
+        GGML_ASSERT(common_step > 0);
+
+        const bool disable_chunking = ggml_is_numa();
+        const size_t chunk_multiplier = std::max(1, ctx.chunk_multiplier);
+        const size_t chunk_divisor = (nth_total == 1 || disable_chunking) ? (size_t)nth_total : (size_t)nth_total * chunk_multiplier;
+        size_t chunk_cols = align_up(std::max<size_t>(1, ceil_div_size(n, chunk_divisor)), common_step);
+        if (chunk_cols == 0) {
+            chunk_cols = common_step;
         }
+        // If common_step is larger than n, the loop below runs one valid tail chunk
+        // with cols == n.
+        const size_t nchunk_size = std::max<size_t>(1, ceil_div_size(n, chunk_cols));
+        GGML_ASSERT(nchunk_size <= (size_t)INT_MAX);
+        const int nchunk = (int)nchunk_size;
         const size_t dst_stride = dst->nb[1];
 
+        auto run_chunk = [&](runtime_slot & slot, size_t global_start, size_t cols, uint8_t * dst_batch_base) {
+            const size_t rhs_packed_offset = slot.kernel->get_rhs_packed_offset_ex(global_start, k, slot.rhs_bl);
+            const size_t dst_offset        = slot.kernel->get_dst_offset(0, global_start, dst_stride);
+
+            const uint8_t * lhs_ptr = scratch + slot.lhs_offset + slot.lhs_packed_offset0;
+            const uint8_t * rhs_ptr = slot.rhs_base + rhs_packed_offset;
+            float * dst_ptr         = reinterpret_cast<float *>(dst_batch_base + dst_offset);
+
+            slot.kernel->run_kernel_ex(m, cols, k, slot.rhs_bl,
+                                       lhs_ptr,
+                                       rhs_ptr,
+                                       dst_ptr,
+                                       dst_stride,
+                                       sizeof(float),
+                                       -FLT_MAX,
+                                       FLT_MAX);
+        };
+
         for (int64_t batch_idx = 0; batch_idx < ne12; ++batch_idx) {
             const uint8_t * lhs_batch_base = static_cast<const uint8_t *>(src1->data) + batch_idx * src1->nb[2];
             uint8_t * dst_batch_base = static_cast<uint8_t *>(dst->data) + batch_idx * dst->nb[2];
 
             if (runtime[local_slot].assigned_threads > 0) {
                 runtime_slot & slot = runtime[local_slot];
-                const ggml_type slot_rhs_type = slot.kernels->rhs_type;
-                const size_t slot_lhs_exec_arg = slot_rhs_type == GGML_TYPE_Q4_0 ? QK4_0 :
-                                                 slot_rhs_type == GGML_TYPE_Q8_0 ? 0 : 0;
                 const int64_t m_roundup_mr = kai_roundup((int64_t)m, (int64_t)slot.mr);
                 int64_t max_threads = slot.mr ? (m_roundup_mr / (int64_t)slot.mr) : slot.assigned_threads;
                 max_threads = std::max<int64_t>(1, max_threads);
@@ -1031,8 +1061,8 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                     const int64_t m_start = (int64_t)local_ith * num_m_per_thread0;
                     const int64_t m_count = (local_ith == use_threads - 1) ? num_m_per_threadN_1 : num_m_per_thread0;
 
-                    const size_t base_packed_off  = slot.lhs_info->get_packed_offset_ex(m_start, k, slot_lhs_exec_arg, slot.mr, slot.kr, slot.sr);
-                    const size_t next_block_off   = slot.lhs_info->get_packed_offset_ex(m_start + slot.mr, k, slot_lhs_exec_arg, slot.mr, slot.kr, slot.sr);
+                    const size_t base_packed_off  = slot.lhs_info->get_packed_offset_ex(m_start, k, slot.lhs_bl, slot.mr, slot.kr, slot.sr);
+                    const size_t next_block_off   = slot.lhs_info->get_packed_offset_ex(m_start + slot.mr, k, slot.lhs_bl, slot.mr, slot.kr, slot.sr);
                     const size_t row_stride_bytes = slot.mr ? (next_block_off - base_packed_off) / slot.mr : 0;
 
                     int64_t remaining = m_count;
@@ -1049,7 +1079,7 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                         const size_t dst_off = base_packed_off + (size_t)(cur - m_start) * row_stride_bytes;
                         void * dst_ptr       = lhs_packed + dst_off;
 
-                        slot.lhs_info->pack_func_ex(take, k, slot_lhs_exec_arg, slot.mr, slot.kr, slot.sr, 0, src_ptr, src1->nb[1], dst_ptr);
+                        slot.lhs_info->pack_func_ex(take, k, slot.lhs_bl, slot.mr, slot.kr, slot.sr, 0, src_ptr, src1->nb[1], dst_ptr);
 
                         cur       += take;
                         remaining -= take;
@@ -1057,49 +1087,29 @@ class tensor_traits : public ggml::cpu::tensor_traits {
                 }
             }
 
+            if (ith_total == 0) {
+                ggml_threadpool_chunk_set(params->threadpool, nth_total);
+            }
+
+            // Publishes both LHS packing and the initialized dynamic chunk queue.
             ggml_barrier(params->threadpool);
 
             runtime_slot & slot = runtime[local_slot];
-            if (slot.n_cols > 0 && slot.assigned_threads > 0) {
-                int64_t active_threads = slot.assigned_threads;
-                const int64_t max_threads = slot.n_step ? (slot.n_cols / slot.n_step) : slot.assigned_threads;
-                if (max_threads > 0) {
-                    active_threads = std::min<int64_t>(active_threads, std::max<int64_t>(1, max_threads));
+            int current_chunk = ith_total;
+            while (current_chunk < nchunk) {
+                const size_t global_start = (size_t)current_chunk * chunk_cols;
+                if (global_start >= n) {
+                    break;
                 }
-                active_threads = std::max<int64_t>(1, active_threads);
-
-                if (local_ith < active_threads) {
-                    const size_t step = slot.n_step ? slot.n_step : 1;
-                    const size_t chunk0 = round_down((size_t)(slot.n_cols / active_threads), step);
-                    const size_t chunkN = slot.n_cols - (active_threads - 1) * chunk0;
-                    const size_t local_start = (size_t)local_ith * chunk0;
-                    const size_t cols = (local_ith == active_threads - 1) ? chunkN : chunk0;
-
-                    if (cols > 0) {
-                        const ggml_type slot_rhs_type = slot.kernels->rhs_type;
-                        const size_t slot_lhs_exec_arg = slot_rhs_type == GGML_TYPE_Q4_0 ? QK4_0 :
-                                                         slot_rhs_type == GGML_TYPE_Q8_0 ? 0 : 0;
-                        const size_t slot_rhs_block_arg = slot_rhs_type == GGML_TYPE_Q4_0 ? QK4_0 :
-                                                          slot_rhs_type == GGML_TYPE_Q8_0 ? 0 : 0;
-                        const size_t global_start = slot.n_offset + local_start;
-                        const size_t lhs_packed_offset = slot.lhs_info->get_packed_offset_ex(0, k, slot_lhs_exec_arg, slot.mr, slot.kr, slot.sr);
-                        const size_t rhs_packed_offset = slot.kernel->get_rhs_packed_offset_ex(global_start, k, slot_rhs_block_arg);
-                        const size_t dst_offset        = slot.kernel->get_dst_offset(0, global_start, dst_stride);
-
-                        const uint8_t * lhs_ptr = scratch + slot.lhs_offset + lhs_packed_offset;
-                        const uint8_t * rhs_ptr = slot.rhs_base + rhs_packed_offset;
-                        float * dst_ptr         = reinterpret_cast<float *>(dst_batch_base + dst_offset);
-
-                        slot.kernel->run_kernel_ex(m, cols, k, slot_rhs_block_arg,
-                                                   lhs_ptr,
-                                                   rhs_ptr,
-                                                   dst_ptr,
-                                                   dst_stride,
-                                                   sizeof(float),
-                                                   -FLT_MAX,
-                                                   FLT_MAX);
-                    }
+
+                const size_t cols = std::min(chunk_cols, n - global_start);
+                if (cols > 0) {
+                    // KleidiAI GEMM/GEMV kernels accept arbitrary final tail widths;
+                    // only non-tail chunks are guaranteed to be n_step-aligned.
+                    run_chunk(slot, global_start, cols, dst_batch_base);
                 }
+
+                current_chunk = ggml_threadpool_chunk_add(params->threadpool, 1);
             }
 
             if (batch_idx != ne12 - 1) {
@@ -1461,7 +1471,7 @@ class extra_buffer_type : ggml::cpu::extra_buffer_type {
                 return false;
             }
             if ((op->src[1]->type == GGML_TYPE_F32 || op->src[1]->type == GGML_TYPE_I32) &&
-                ggml_ne(op->src[1], 2) == 1 && ggml_ne(op->src[1], 3) == 1) {
+                ggml_ne(op->src[1], 3) == 1) {
                 return true;
             }
         }
@@ -1473,10 +1483,12 @@ class extra_buffer_type : ggml::cpu::extra_buffer_type {
             if (op->src[0]->buffer && op->src[0]->buffer->buft == ggml_backend_cpu_kleidiai_buffer_type()) {
                 return (ggml::cpu::tensor_traits *) op->src[0]->extra;
             } else {
+                if (op->src[0]->type != GGML_TYPE_F16) {
+                    return nullptr;
+                }
                 std::array<ggml_kleidiai_kernels *, GGML_KLEIDIAI_MAX_KERNEL_SLOTS> kernel_chain;
                 const int slot_total = kleidiai_collect_kernel_chain(op, kernel_chain);
-                const bool has_kernel = slot_total > 0;
-                if (has_kernel && op->src[1]->ne[1] > 1) {
+                if (slot_total > 0 && op->src[1]->ne[1] > 1) {
                     if ((op->src[0]->nb[1] * op->src[0]->ne[1] != op->src[0]->nb[2]) ||
                         (op->src[1]->nb[1] * op->src[1]->ne[1] != op->src[1]->nb[2])) {
                         return nullptr;