3outeille

• Developped under Python 3.7.12
• Standalone script which parse hackmd markdown (including imgur link) and create directory (from tags) with downloaded imgur link
• To use, python3 script.py --zip_file <file.zip>
• To export your hackmd markdown as zip file

3outeille

NAME = libasm.a
SRCS =	dot_product.asm
OBJS = $(SRCS:.asm=.o) 

all:
	nasm -g -f elf64 ${SRCS}
	gcc -g -c main.c -o main.o
	ar rc $(NAME) $(OBJS)
	gcc -g main.o $(NAME) -o main
	

3outeille

NAME = libasm.a
SRCS =	fill_mat.asm
OBJS = $(SRCS:.asm=.o) 

all:
	nasm -g -f elf64 ${SRCS}
	gcc -g -c main.c -o main.o
	ar rc $(NAME) $(OBJS)
	gcc -g main.o $(NAME) -o main

3outeille

NAME = libasm.a
SRCS =	block_avg.asm
OBJS = $(SRCS:.asm=.o) 

all:
	nasm -g -f elf64 ${SRCS}
	gcc -g -c main.c -o main.o
	ar rc $(NAME) $(OBJS)
	gcc -g main.o $(NAME) -o main

3outeille

all:
	nasm -f elf64 main.asm
	gcc -g -nostartfiles -no-pie -o main main.o

debug:
	echo "dashboard -output /dev/pts/0\nb _start\nrun >> gdb-log.txt 2>&1\nlayout asm\ndashboard assembly\ndashboard source\ndashboard threads\nset logging file gdb-log.txt\nset logging on\nset trace-commands on\nrecord\n" > gdb-commands.txt
	gdb -x gdb-commands.txt main

clean:
	rm *.o main gdb-commands.txt gdb-log.txt .gdb_history

3outeille

%define ARCH_X86_64 1

; sad_16x16, 4, 7, 5, src, src_stride, dst, dst_stride, src_stride3, dst_stride3, cnt   
; function_name, #args (%1), #regs (%2), #xmm_regs (%3), [stack_size,] (%4) arg_names... (%5-*)

%macro PUSH_IF_USED 1-*
    %rep %0
        %if %1 < regs_used
            push r%1
        %endif

3outeille

struct BenchWrapper
{
    size_t nb_fct_call;
    size_t iter_per_fct;
    // Define function arguments here
    int (*func)(const OVSEI *sei, OVFrame **frame);
};

void bench_decorator(struct BenchWrapper bw, char* name, const OVSEI *sei, OVFrame **frame)
{

3outeille

void fg_simulate_grain_blk8x8(int32_t *grainStripe, uint32_t grainStripeOffsetBlk8,
  uint32_t width, uint8_t log2ScaleFactor, int16_t scaleFactor, uint32_t kOffset, uint32_t lOffset, uint8_t h, uint8_t v, uint32_t xSize)
{
  uint32_t k, l;
  uint32_t idx, idx_offset, idx_offset_l, grainStripeOffsetBlk8_l;

  idx_offset = ( h*NUM_CUT_OFF_FREQ +  v ) * DATA_BASE_SIZE  * DATA_BASE_SIZE;
  for (l = 0; l < 8; l++) /* y direction */
  {
    idx_offset_l            = idx_offset + (l + lOffset) * DATA_BASE_SIZE;

3outeille

void fg_simulate_grain_blk8x8_sse4(int32_t *grainStripe, uint32_t grainStripeOffsetBlk8,
  uint32_t width, uint8_t log2ScaleFactor, int16_t scaleFactor, uint32_t kOffset, uint32_t lOffset, uint8_t h, uint8_t v, uint32_t xSize)
{
    uint32_t idx_offset_l1, idx_offset_l2, idx_offset_l3, idx_offset_l4;
    uint32_t grainStripeOffsetBlk8_l1, grainStripeOffsetBlk8_l2, grainStripeOffsetBlk8_l3, grainStripeOffsetBlk8_l4;

    uint32_t idx_offset = ( h*NUM_CUT_OFF_FREQ +  v ) * DATA_BASE_SIZE  * DATA_BASE_SIZE;

    __m128i scale = _mm_set_epi32(scaleFactor, scaleFactor, scaleFactor, scaleFactor);

3outeille

static inline uint32_t ov_clip_uintp2(int32_t val, uint32_t a)
{
    if (val > 0) {
        int32_t mask  = (1 << a) - 1;
        int32_t overflow = !!(val & (~mask));
        return ((-overflow) & mask) | (val & mask);
    } else {
        return 0;
    }
    #if 0