Optimize Metal Tensor API usage

ggml/src/ggml-metal/ggml-metal-device.cpp

@@ -677,7 +677,15 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_mul_mm(ggml_meta
     const ggml_type tsrc1 = op->src[1]->type;
 
     const bool bc_inp = op->src[0]->ne[0] % 32 != 0;
-    const bool bc_out = op->ne[0] % 64 != 0 || op->ne[1] % 32 != 0;
+
+    constexpr int NRA = SZ_SIMDGROUP * N_MM_BLOCK_Y * N_MM_SIMD_GROUP_Y;
+    constexpr int NRB = SZ_SIMDGROUP * N_MM_BLOCK_X * N_MM_SIMD_GROUP_X;
+
+    const bool has_tensor = ggml_metal_device_get_props(ggml_metal_library_get_device(lib))->has_tensor;
+
+    const bool bc_out = has_tensor
+        ? (op->ne[0] % NRA != 0 || op->ne[1] % NRB != 0)
+        : (op->ne[0] % 64  != 0 || op->ne[1] % 32  != 0);
 
     snprintf(base, 256, "kernel_mul_mm_%s_%s", ggml_type_name(tsrc0), ggml_type_name(tsrc1));
     snprintf(name, 256, "%s_bci=%d_bco=%d", base, bc_inp, bc_out);
@@ -694,8 +702,21 @@ ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline_mul_mm(ggml_meta
         ggml_metal_cv_free(cv);
     }
 
-    // when the output size is not multiple of 64x32, we need extra smem to prevent out-of-bounds writes
-    res.smem = bc_out ? 8192 : 4096 + 2048;
+    if (has_tensor) {
+        res.nr0 = SZ_SIMDGROUP * N_MM_BLOCK_Y * N_MM_SIMD_GROUP_Y;  // 64
+        res.nr1 = SZ_SIMDGROUP * N_MM_BLOCK_X * N_MM_SIMD_GROUP_X;  // 32
+
+        const size_t NRA    = res.nr0;
+        const size_t smem_a = NRA * N_MM_NK_TOTAL * sizeof(ggml_fp16_t);
+        res.smem = smem_a;
+    } else {
+        res.nr0 = 64;
+        res.nr1 = 32;
+
+        res.smem = bc_out ? 8192 : (4096 + 2048);
+    }
+
+    res.nsg = N_MM_SIMD_GROUP_X * N_MM_SIMD_GROUP_Y;
 
     return res;
 }

ggml/src/ggml-metal/ggml-metal-device.h

@@ -102,6 +102,8 @@ ggml_metal_library_t ggml_metal_library_init_from_source(ggml_metal_device_t dev
 
 void ggml_metal_library_free(ggml_metal_library_t lib);
 
+ggml_metal_device_t ggml_metal_library_get_device(ggml_metal_library_t lib);
+
 struct ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline    (ggml_metal_library_t lib, const char * name);
 struct ggml_metal_pipeline_with_params ggml_metal_library_compile_pipeline(ggml_metal_library_t lib, const char * base, const char * name, ggml_metal_cv_t cv);
 

ggml/src/ggml-metal/ggml-metal-device.m

@@ -95,8 +95,8 @@ int ggml_metal_pipeline_max_theads_per_threadgroup(struct ggml_metal_pipeline_wi
 
 struct ggml_metal_library {
     id<MTLLibrary> obj;
-    id<MTLDevice> device;
 
+    ggml_metal_device_t dev;
     ggml_metal_pipelines_t pipelines; // cache of compiled pipelines
 
     NSLock * lock;
@@ -251,7 +251,7 @@ ggml_metal_library_t ggml_metal_library_init(ggml_metal_device_t dev) {
     ggml_metal_library_t res = calloc(1, sizeof(struct ggml_metal_library));
 
     res->obj       = library;
-    res->device    = device;
+    res->dev       = dev;
     res->pipelines = ggml_metal_pipelines_init();
     res->lock      = [NSLock new];
 
@@ -318,7 +318,7 @@ ggml_metal_library_t ggml_metal_library_init_from_source(ggml_metal_device_t dev
     }
 
     res->obj       = library;
-    res->device    = device;
+    res->dev       = dev;
     res->pipelines = ggml_metal_pipelines_init();
     res->lock      = [NSLock new];
 
@@ -341,6 +341,10 @@ void ggml_metal_library_free(ggml_metal_library_t lib) {
     free(lib);
 }
 
+ggml_metal_device_t ggml_metal_library_get_device(ggml_metal_library_t lib) {
+    return lib->dev;
+}
+
 struct ggml_metal_pipeline_with_params ggml_metal_library_get_pipeline(ggml_metal_library_t lib, const char * name) {
     [lib->lock lock];
 
@@ -405,7 +409,8 @@ struct ggml_metal_pipeline_with_params ggml_metal_library_compile_pipeline(ggml_
             return res;
         }
 
-        id<MTLComputePipelineState> obj = [lib->device newComputePipelineStateWithFunction:mtl_function error:&error];
+        id<MTLDevice> device = ggml_metal_device_get_obj(lib->dev);
+        id<MTLComputePipelineState> obj = [device newComputePipelineStateWithFunction:mtl_function error:&error];
 
         [mtl_function release];
 
@@ -699,7 +704,7 @@ ggml_metal_device_t ggml_metal_device_init(int device) {
                     "    auto sB = tB.slice(0, 0); \n"
                     "    mm.run(sB, sA, cT); \n"
                     " \n"
-                    "    auto tC = tensor<device float, dextents<int32_t, 2>, tensor_inline>(C, dextents<int32_t, 2>(4, 4)); \n"
+                    "    auto tC = tensor<device float, dextents<int32_t, 2>, tensor_inline>(C, dextents<int32_t, 2>(16, 16)); \n"
                     " \n"
                     "    cT.store(tC); \n"
                     "}";
@@ -749,7 +754,7 @@ ggml_metal_device_t ggml_metal_device_init(int device) {
                     "    auto sB = tB.slice(0, 0); \n"
                     "    mm.run(sB, sA, cT); \n"
                     " \n"
-                    "    auto tC = tensor<device float, dextents<int32_t, 2>, tensor_inline>(C, dextents<int32_t, 2>(4, 4)); \n"
+                    "    auto tC = tensor<device float, dextents<int32_t, 2>, tensor_inline>(C, dextents<int32_t, 2>(16, 16)); \n"
                     " \n"
                     "    cT.store(tC); \n"
                     "}";

ggml/src/ggml-metal/ggml-metal-impl.h

@@ -8,6 +8,15 @@
 //
 // TODO: for optimal performance, become function of the device and work size
 
+#define SZ_SIMDGROUP 16
+#define N_MM_NK 2
+#define N_MM_NK_TOTAL (SZ_SIMDGROUP * N_MM_NK)
+
+#define N_MM_BLOCK_X 4
+#define N_MM_BLOCK_Y 2
+#define N_MM_SIMD_GROUP_X 2
+#define N_MM_SIMD_GROUP_Y 2
+
 #define N_R0_Q1_0 8
 #define N_SG_Q1_0 2
 

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -2195,7 +2195,12 @@ int ggml_metal_op_mul_mat(ggml_metal_op_t ctx, int idx) {
         const size_t smem = pipeline.smem;
 
         ggml_metal_encoder_set_threadgroup_memory_size(enc, smem, 0);
-        ggml_metal_encoder_dispatch_threadgroups(enc, ((ne11 + 31)/32), ((ne01 + 63)/64), ne12*ne13, 128, 1, 1);
+
+        const int nr0 = pipeline.nr0;
+        const int nr1 = pipeline.nr1;
+        const int nsg = pipeline.nsg;
+
+        ggml_metal_encoder_dispatch_threadgroups(enc, ((ne11 + nr1 - 1) / nr1), ((ne01 + nr0 - 1) / nr0), ne12 * ne13, 32, nsg, 1);
     } else {
         auto pipeline = ggml_metal_library_get_pipeline_mul_mv(lib, op);