diff --git a/ggml-metal.m b/ggml-metal.m index bf3f68fe4..ff4f2b23a 100644 --- a/ggml-metal.m +++ b/ggml-metal.m @@ -36,6 +36,9 @@ struct ggml_metal_context { int n_buffers; struct ggml_metal_buffer buffers[GGML_METAL_MAX_BUFFERS]; + int concur_list[GGML_MAX_NODES]; + int concur_list_len; + // custom kernels #define GGML_METAL_DECL_KERNEL(name) \ id function_##name; \ @@ -98,6 +101,7 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) { ctx->device = MTLCreateSystemDefaultDevice(); ctx->queue = [ctx->device newCommandQueue]; ctx->n_buffers = 0; + ctx->concur_list_len = 0; // determine if we can use MPS if (MPSSupportsMTLDevice(ctx->device)) { @@ -355,11 +359,92 @@ void ggml_metal_get_tensor( memcpy(t->data, (void *) ((uint8_t *) id_src.contents + offs), ggml_nbytes(t)); } +void ggml_metal_graph_find_concurrency( + struct ggml_metal_context * ctx, + struct ggml_cgraph * gf) { + int search_depth = gf->n_nodes; //we only find concurrency in this range to avoid wasting too much time + int nodes_unused[GGML_MAX_NODES]; + + for (int i = 0; i < GGML_MAX_NODES; i++) {ctx->concur_list[i] = 0;} + for (int i = 0; i < gf->n_nodes; i++) {nodes_unused[i] = 1;} + ctx->concur_list_len = 0; + + int n_left = gf->n_nodes; + int n_start = 0; // all nodes before n_start at nodes_unused array have been sorted and store back to ctx->concur_list + int level_pos = 0; // at ctx->concur_list, the last layer (level) ends at level_pos + + while (n_left > 0) { + // number of nodes at a layer (that can be issued concurrently) + int concurrency = 0; + for (int i = n_start; i < ((n_start + search_depth > gf->n_nodes) ? gf->n_nodes : n_start + search_depth); i++) { + if (nodes_unused[i]) { + // if the requirements for gf->nodes[i] are satisfied + int exe_flag=1; + // scan all srcs + for (int src_ind = 0; src_ind < GGML_MAX_SRC; src_ind++) { + struct ggml_tensor * src_cur = gf->nodes[i]->src[src_ind]; + if (src_cur) { + // if is leaf nodes it's satisfied. + if (src_cur->op == GGML_OP_NONE && src_cur->grad == NULL) {continue;} + + // otherwise this src should be the output from previous nodes. + int is_found = 0; + // scan 2*search_depth back because we inserted barrier. + for (int j = ((level_pos - 2*search_depth) < 0 ? 0 : (level_pos - 2*search_depth)); j < level_pos; j++) { + if (gf->nodes[ctx->concur_list[j]] == src_cur) {is_found = 1; break;} + } + if (is_found == 0) {exe_flag = 0; break;} + } + } + if (exe_flag) { + // check if nodes[i]'s data will be overwritten by a node before nodes[i]. + // if node[5] and node[3] write to the same memory region, then we can't issue node[5] before node[3] + int64_t data_start = (int64_t) gf->nodes[i]->data; + int64_t length = (int64_t) ggml_nbytes(gf->nodes[i]); + for (int j = n_start; j < i; j++) { + if (nodes_unused[j] && gf->nodes[j]->op != GGML_OP_RESHAPE \ + && gf->nodes[j]->op != GGML_OP_VIEW \ + && gf->nodes[j]->op != GGML_OP_TRANSPOSE \ + && gf->nodes[j]->op != GGML_OP_PERMUTE) { + if (((int64_t)gf->nodes[j]->data) >= data_start + length || \ + ((int64_t)gf->nodes[j]->data) + (int64_t) ggml_nbytes(gf->nodes[j]) <= data_start) { + continue; + } else { + exe_flag = 0; + } + } + } + } + if (exe_flag) { + ctx->concur_list[level_pos + concurrency] = i; + nodes_unused[i] = 0; + concurrency++; + ctx->concur_list_len++; + } + } + } + n_left -= concurrency; + // adding a barrier different layer + ctx->concur_list[level_pos + concurrency] = -1; + ctx->concur_list_len++; + // jump all sorted nodes at nodes_bak + while (!nodes_unused[n_start]) {n_start++;} + level_pos += concurrency + 1; + } + + if (ctx->concur_list_len > GGML_MAX_NODES) { + fprintf(stderr, "%s: too many elements for metal ctx->concur_list!\n", __func__); + } +} + void ggml_metal_graph_compute( struct ggml_metal_context * ctx, struct ggml_cgraph * gf) { metal_printf("%s: evaluating graph\n", __func__); + if (!ctx->concur_list_len) { + ggml_metal_graph_find_concurrency(ctx,gf); + } // create multiple command buffers and enqueue them // then, we encode the graph into the command buffers in parallel @@ -378,7 +463,7 @@ void ggml_metal_graph_compute( dispatch_queue_t queue = dispatch_queue_create("llama.cpp", DISPATCH_QUEUE_CONCURRENT); for (int cb_idx = 0; cb_idx < n_cb; ++cb_idx) { - const int n_nodes_per_cb = (gf->n_nodes + n_cb - 1) / n_cb; + const int n_nodes_per_cb = (ctx->concur_list_len + n_cb - 1) / n_cb; dispatch_async(queue, ^{ size_t offs_src0 = 0; @@ -390,9 +475,18 @@ void ggml_metal_graph_compute( id encoder = nil; const int node_start = (cb_idx + 0) * n_nodes_per_cb; - const int node_end = (cb_idx == n_cb - 1) ? gf->n_nodes : (cb_idx + 1) * n_nodes_per_cb; + const int node_end = (cb_idx == n_cb - 1) ? ctx->concur_list_len : (cb_idx + 1) * n_nodes_per_cb; - for (int i = node_start; i < node_end; ++i) { + for (int ind = node_start; ind < node_end; ++ind) { + int i = ctx->concur_list[ind]; + if (i == -1) { + if (encoder == nil) { + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; + continue; + } + [encoder memoryBarrierWithScope:MTLBarrierScopeBuffers]; + continue; + } metal_printf("%s: encoding node %3d, op = %8s\n", __func__, i, ggml_op_name(gf->nodes[i]->op)); struct ggml_tensor * src0 = gf->nodes[i]->src[0]; @@ -463,7 +557,7 @@ void ggml_metal_graph_compute( case GGML_OP_ADD: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } if (ggml_nelements(src1) == ne10) { @@ -484,7 +578,7 @@ void ggml_metal_graph_compute( case GGML_OP_MUL: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } if (ggml_nelements(src1) == ne10) { @@ -505,7 +599,7 @@ void ggml_metal_graph_compute( case GGML_OP_SCALE: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const float scale = *(const float *) src1->data; @@ -522,7 +616,7 @@ void ggml_metal_graph_compute( case GGML_OP_SILU: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } [encoder setComputePipelineState:ctx->pipeline_silu]; @@ -536,7 +630,7 @@ void ggml_metal_graph_compute( case GGML_OP_RELU: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } [encoder setComputePipelineState:ctx->pipeline_relu]; @@ -550,7 +644,7 @@ void ggml_metal_graph_compute( case GGML_OP_GELU: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } [encoder setComputePipelineState:ctx->pipeline_gelu]; @@ -564,7 +658,7 @@ void ggml_metal_graph_compute( case GGML_OP_SOFT_MAX: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const int nth = 32; @@ -582,7 +676,7 @@ void ggml_metal_graph_compute( case GGML_OP_DIAG_MASK_INF: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const int n_past = ((int32_t *)(dst->op_params))[0]; @@ -645,7 +739,7 @@ void ggml_metal_graph_compute( } } else { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } int nth0 = 32; @@ -772,7 +866,7 @@ void ggml_metal_graph_compute( case GGML_OP_GET_ROWS: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } switch (src0->type) { @@ -801,7 +895,7 @@ void ggml_metal_graph_compute( case GGML_OP_RMS_NORM: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const float eps = 1e-6f; @@ -823,7 +917,7 @@ void ggml_metal_graph_compute( case GGML_OP_NORM: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const float eps = 1e-5f; @@ -845,7 +939,7 @@ void ggml_metal_graph_compute( case GGML_OP_ALIBI: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } GGML_ASSERT((src0t == GGML_TYPE_F32)); @@ -888,7 +982,7 @@ void ggml_metal_graph_compute( case GGML_OP_ROPE: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const int n_past = ((int32_t *) dst->op_params)[0]; @@ -932,7 +1026,7 @@ void ggml_metal_graph_compute( case GGML_OP_CONT: { if (encoder == nil) { - encoder = [command_buffer computeCommandEncoder]; + encoder = [command_buffer computeCommandEncoderWithDispatchType: MTLDispatchTypeConcurrent]; } const int nth = 32;