Skip to content

@FIR-999 - Create SOFT_MAX for tsavorite-backend for GGML #63

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 17, 2025
Merged

@FIR-999 - Create SOFT_MAX for tsavorite-backend for GGML #63

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
85b71fd
@FIR-999 - Create SOFT_MAX for tsavorite-backend for GGML
Oct 15, 2025
41137ce
Added memory Alignment for 128 Bytes
Oct 17, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion ggml-tsi-kernel
Open in desktop
Submodule ggml-tsi-kernel updated from 7dd322 to ab6aec
7 changes: 5 additions & 2 deletions ggml/include/ggml-tsavorite.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -140,6 +140,8 @@ enum ggml_tsavorite_kernel_type {
GGML_TSAVORITE_KERNEL_TYPE_GEGLU_ERF,
GGML_TSAVORITE_KERNEL_TYPE_GEGLU_QUICK,

GGML_TSAVORITE_KERNEL_TYPE_SOFT_MAX,

GGML_TSAVORITE_KERNEL_TYPE_COUNT
};

Expand All @@ -156,7 +158,7 @@ typedef struct tensor_log_ {
uint32_t leaf2_len;
uint32_t node_len;
enum ggml_tsavorite_tensor_data_type data_type;
enum ggml_tsavorite_kernel_type kernel_type;
enum ggml_op kernel_type;
uint64_t num_of_op;
FILE *log_file;
const ggml_tensor *tensor;
Expand Down Expand Up @@ -185,6 +187,7 @@ extern void _mlir_ciface_txe_sin_host(void *a, void *res);
extern void _mlir_ciface_txe_sigmoid_host(void *a, void *res);
extern void _mlir_ciface_txe_silu_host(void *a, void *res);
extern void _mlir_ciface_txe_swiglu_host(void *a, void *b, void *res);
extern void _mlir_ciface_txe_soft_max_host(void *a, void *b, void *res, void *buf);
extern void _mlir_ciface_txe_rms_norm_host(void *a, void *res, void *buf);

/*
Expand All @@ -210,7 +213,7 @@ extern void ggml_tsi_log_tensor_data(tensor_log log_data);

// GGML supports tensors with a maximum rank of 4
#define MEM_REF_DESCRIPTOR_RANK 4
#define TSI_TVU_LOAD_SIZE 32
#define TSI_TVU_MEM_ALIGN 128

//
// backend API
Expand Down
Loading