mirror of
https://github.com/hanzoai/traces.git
synced 2026-10-30 18:20:41 +00:00
Martin DisibioandGitHub
9aaf851965
* Upgrade parquet-go to latest for StructTag option * Upgrade parquet-go to get latest alloc fix
106 lines
2.1 KiB
ArmAsm
106 lines
2.1 KiB
ArmAsm
//go:build !purego
|
|
|
|
#include "funcdata.h"
|
|
#include "textflag.h"
|
|
|
|
// unpackInt64x8bitAVX2 implements optimized unpacking for bitWidth=8 using AVX2
|
|
// Each byte is already a complete value, processes 8 values at a time
|
|
//
|
|
// func unpackInt64x8bitAVX2(dst []int64, src []byte, bitWidth uint)
|
|
TEXT ·unpackInt64x8bitAVX2(SB), NOSPLIT, $0-56
|
|
MOVQ dst_base+0(FP), AX // AX = dst pointer
|
|
MOVQ dst_len+8(FP), DX // DX = dst length
|
|
MOVQ src_base+24(FP), BX // BX = src pointer
|
|
MOVQ bitWidth+48(FP), CX // CX = bitWidth (should be 8)
|
|
|
|
// Check if we have at least 8 values to process
|
|
CMPQ DX, $8
|
|
JB avx2_8bit_tail
|
|
|
|
// Round down to multiple of 8 for AVX2 processing
|
|
MOVQ DX, DI
|
|
SHRQ $3, DI // DI = len / 8
|
|
SHLQ $3, DI // DI = aligned length (multiple of 8)
|
|
XORQ SI, SI // SI = index
|
|
|
|
avx2_8bit_loop:
|
|
// Load 8 bytes (8 x 8-bit values)
|
|
MOVQ (BX), R8
|
|
|
|
// Extract each byte and store as int64
|
|
// Value 0 (byte 0)
|
|
MOVQ R8, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, (AX)
|
|
|
|
// Value 1 (byte 1)
|
|
MOVQ R8, R9
|
|
SHRQ $8, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 8(AX)
|
|
|
|
// Value 2 (byte 2)
|
|
MOVQ R8, R9
|
|
SHRQ $16, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 16(AX)
|
|
|
|
// Value 3 (byte 3)
|
|
MOVQ R8, R9
|
|
SHRQ $24, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 24(AX)
|
|
|
|
// Value 4 (byte 4)
|
|
MOVQ R8, R9
|
|
SHRQ $32, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 32(AX)
|
|
|
|
// Value 5 (byte 5)
|
|
MOVQ R8, R9
|
|
SHRQ $40, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 40(AX)
|
|
|
|
// Value 6 (byte 6)
|
|
MOVQ R8, R9
|
|
SHRQ $48, R9
|
|
ANDQ $0xFF, R9
|
|
MOVQ R9, 48(AX)
|
|
|
|
// Value 7 (byte 7)
|
|
MOVQ R8, R9
|
|
SHRQ $56, R9
|
|
MOVQ R9, 56(AX)
|
|
|
|
// Advance pointers
|
|
ADDQ $8, BX // src += 8 bytes
|
|
ADDQ $64, AX // dst += 8 int64 (64 bytes)
|
|
ADDQ $8, SI // index += 8
|
|
|
|
CMPQ SI, DI
|
|
JNE avx2_8bit_loop
|
|
|
|
avx2_8bit_tail:
|
|
// Handle remaining elements with scalar fallback
|
|
CMPQ SI, DX
|
|
JE avx2_8bit_done
|
|
|
|
// Compute remaining elements
|
|
SUBQ SI, DX
|
|
|
|
avx2_8bit_tail_loop:
|
|
MOVBQZX (BX), R8 // Load byte
|
|
MOVQ R8, (AX) // Store as int64 (zero-extended)
|
|
|
|
ADDQ $1, BX // src++
|
|
ADDQ $8, AX // dst++
|
|
DECQ DX // remaining--
|
|
|
|
CMPQ DX, $0
|
|
JNE avx2_8bit_tail_loop
|
|
|
|
avx2_8bit_done:
|
|
RET
|