vulkan : support sum, sum_rows and mean with non-contiguous tensors

Author: Acly

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -1014,6 +1014,40 @@ struct vk_op_upscale_push_constants {
     float sf0; float sf1; float sf2; float sf3;
 };
 
+struct vk_op_sum_rows_push_constants
+{
+    uint32_t n_cols;
+    uint32_t ne01, ne02;
+    uint32_t nb00, nb01, nb02, nb03;
+    uint32_t nb11, nb12, nb13;
+    float weight;
+    uint32_t misalign_offsets;
+    uint32_t ne0_12mp, ne0_12L;
+    uint32_t ne0_1mp, ne0_1L;
+};
+
+vk_op_sum_rows_push_constants vk_op_sum_rows_push_constants_init(const ggml_tensor * src, const ggml_tensor * dst, int64_t n_cols) {
+    uint32_t type_size = (uint32_t)ggml_type_size(src->type);
+    vk_op_sum_rows_push_constants p = {};
+    p.n_cols = (uint32_t)n_cols;
+    p.ne01 = (uint32_t)src->ne[1];
+    p.ne02 = (uint32_t)src->ne[2];
+    p.nb00 = (uint32_t)src->nb[0] / type_size;
+    p.nb01 = (uint32_t)src->nb[1] / type_size;
+    p.nb02 = (uint32_t)src->nb[2] / type_size;
+    p.nb03 = (uint32_t)src->nb[3] / type_size;
+    p.nb11 = (uint32_t)dst->nb[1] / type_size;
+    p.nb12 = (uint32_t)dst->nb[2] / type_size;
+    p.nb13 = (uint32_t)dst->nb[3] / type_size;
+    p.weight = 1.0f;
+    return p;
+}
+
+template <> void init_pushconst_fastdiv(vk_op_sum_rows_push_constants &p) {
+    init_fastdiv_values(p.ne01*p.ne02, p.ne0_12mp, p.ne0_12L);
+    init_fastdiv_values(p.ne01,        p.ne0_1mp,  p.ne0_1L);
+}
+
 // Allow pre-recording command buffers
 struct vk_staging_memcpy {
     vk_staging_memcpy(void * _dst, const void * _src, size_t _n) : dst(_dst), src(_src), n(_n) {}
@@ -3122,7 +3156,7 @@ static void ggml_vk_load_shaders(vk_device& device) {
 
     ggml_vk_create_pipeline(device, device->pipeline_argmax_f32, "argmax_f32", argmax_f32_len, argmax_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
 
-    ggml_vk_create_pipeline(device, device->pipeline_sum_rows_f32, "sum_rows_f32", sum_rows_f32_len, sum_rows_f32_data, "main", 2, sizeof(vk_op_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
+    ggml_vk_create_pipeline(device, device->pipeline_sum_rows_f32, "sum_rows_f32", sum_rows_f32_len, sum_rows_f32_data, "main", 2, sizeof(vk_op_sum_rows_push_constants), {1, 1, 1}, { device->subgroup_size }, 1);
 
     ggml_vk_create_pipeline(device, device->pipeline_count_equal_i32, "count_equal_i32", count_equal_i32_len, count_equal_i32_data, "main", 3, sizeof(vk_op_push_constants), {512, 1, 1}, { device->subgroup_size }, 1);
 
@@ -7340,6 +7374,9 @@ static bool ggml_vk_op_supports_incontiguous(ggml_op op) {
     case GGML_OP_CONV_2D_DW:
     case GGML_OP_IM2COL:
     case GGML_OP_SET_ROWS:
+    case GGML_OP_SUM:
+    case GGML_OP_SUM_ROWS:
+    case GGML_OP_MEAN:
         return true;
     default:
         return false;
@@ -7374,6 +7411,16 @@ template <> void init_pushconst_tensor_offsets(ggml_backend_vk_context * ctx, vk
     GGML_UNUSED(src2);
 }
 
+template <> void init_pushconst_tensor_offsets(ggml_backend_vk_context * ctx, vk_op_sum_rows_push_constants &p, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst) {
+    const uint32_t a_offset = get_misalign_bytes(ctx, src0) / ggml_type_size(src0->type);
+    const uint32_t d_offset = get_misalign_bytes(ctx, dst) / ggml_type_size(dst->type);
+
+    p.misalign_offsets = (a_offset << 16) | d_offset;
+
+    GGML_UNUSED(src1);
+    GGML_UNUSED(src2);
+}
+
 template <> void init_pushconst_tensor_offsets(ggml_backend_vk_context * ctx, vk_op_binary_push_constants &p, const ggml_tensor * src0, const ggml_tensor * src1, const ggml_tensor * src2, ggml_tensor * dst) {
     const uint32_t a_offset = get_misalign_bytes(ctx, src0) / ggml_type_size(src0->type);
     const uint32_t b_offset = get_misalign_bytes(ctx, src1) / ggml_type_size(src1->type);
@@ -8542,15 +8589,20 @@ static void ggml_vk_argsort(ggml_backend_vk_context * ctx, vk_context& subctx, c
 }
 
 static void ggml_vk_sum(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
-    ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SUM, { (uint32_t)ggml_nelements(src0), 0, 1.0f, 0.0f }, dryrun);
+    vk_op_sum_rows_push_constants p = vk_op_sum_rows_push_constants_init(src0, dst, ggml_nelements(src0));
+    p.nb00 = 1; // treat src0 as flattened 1D tensor
+    ggml_vk_op_f32(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SUM, p, dryrun);
 }
 
 static void ggml_vk_sum_rows(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
-    ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SUM_ROWS, { (uint32_t)src0->ne[0], 0, 1.0f, 0.0f }, dryrun);
+    vk_op_sum_rows_push_constants p = vk_op_sum_rows_push_constants_init(src0, dst, src0->ne[0]);
+    ggml_vk_op_f32(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_SUM_ROWS, p, dryrun);
 }
 
 static void ggml_vk_mean(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {
-    ggml_vk_op_f32<vk_op_push_constants>(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_MEAN, { (uint32_t)src0->ne[0], 0, 1.0f / (float)src0->ne[0], 0.0f }, dryrun);
+    vk_op_sum_rows_push_constants p = vk_op_sum_rows_push_constants_init(src0, dst, src0->ne[0]);
+    p.weight = 1.0f / (float)src0->ne[0];
+    ggml_vk_op_f32(ctx, subctx, src0, nullptr, nullptr, dst, GGML_OP_MEAN, p, dryrun);
 }
 
 static void ggml_vk_argmax(ggml_backend_vk_context * ctx, vk_context& subctx, const ggml_tensor * src0, ggml_tensor * dst, bool dryrun = false) {

ggml/src/ggml-vulkan/vulkan-shaders/sum_rows.comp

@@ -1,27 +1,59 @@
 #version 450
 
-#include "generic_head.comp"
 #include "types.comp"
 
 #extension GL_EXT_control_flow_attributes : enable
+
 layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
 
 layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
 layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
 
 layout (constant_id = 0) const uint BLOCK_SIZE = 32;
 
+layout (push_constant) uniform parameter
+{
+    uint n_cols;
+    uint ne01, ne02;
+    uint nb00, nb01, nb02, nb03;
+    uint nb11, nb12, nb13;
+    float weight;
+    uint misalign_offsets;
+    uint ne0_12mp, ne0_12L;
+    uint ne0_1mp, ne0_1L;
+} p;
+
+uint get_aoffset() { return p.misalign_offsets >> 16; }
+uint get_doffset() { return p.misalign_offsets & 0xFFFF; }
+
+// see init_fastdiv_values in ggml-vulkan.cpp
+uint fastdiv(uint n, uint mp, uint L) {
+    uint msbs, lsbs;
+    // msbs = mulhi(n, mp)
+    umulExtended(n, mp, msbs, lsbs);
+    return (msbs + n) >> L;
+}
+
+
 shared FLOAT_TYPE tmp[BLOCK_SIZE];
 
 void main() {
     const uint row = gl_WorkGroupID.z * 262144 + gl_WorkGroupID.y * 512 + gl_WorkGroupID.x;
     const uint col = gl_LocalInvocationID.x;
-    const float weight = p.param1;
+    const float weight = p.weight;
+
+    const uint i03 = fastdiv(row, p.ne0_12mp, p.ne0_12L);
+    const uint i03_offset = i03 * p.ne01*p.ne02;
+    const uint i02 = fastdiv(row - i03_offset, p.ne0_1mp, p.ne0_1L);
+    const uint i01 = row - i03_offset - i02*p.ne01;
+
+    const uint src_idx = get_aoffset() + i01 * p.nb01 + i02 * p.nb02 + i03 * p.nb03;
+    const uint dst_idx = get_doffset() + i01 * p.nb11 + i02 * p.nb12 + i03 * p.nb13;
 
-    tmp[col] = FLOAT_TYPE(0.0f);
+    tmp[col] = FLOAT_TYPE(0.0);
 
-    for (uint i = col; i < p.KX; i += BLOCK_SIZE) {
-        tmp[col] += FLOAT_TYPE(data_a[row*p.KX + i]);
+    for (uint i = col; i < p.n_cols; i += BLOCK_SIZE) {
+        tmp[col] += FLOAT_TYPE(data_a[src_idx + i * p.nb00]);
     }
 
     barrier();
@@ -33,6 +65,6 @@ void main() {
     }
 
     if (col == 0) {
-        data_d[row] = D_TYPE(tmp[0] * weight);
+        data_d[dst_idx] = D_TYPE(tmp[0] * weight);
     }
 }