#include "common.h"
#include "dequantize.h"
#include "fa_common.metal"
// TODO: this is quite ugly. in the future these types will be hardcoded in the kernel, but for now keep them as
// template to be able to explore different combinations
#define FA_TYPES \
half, half4, simdgroup_half8x8, \
half, half4x4, simdgroup_half8x8, \
half, half4x4, simdgroup_half8x8, \
float, simdgroup_float8x8, \
float, float2, simdgroup_float8x8, \
float, float4, simdgroup_float8x8
//half, half4, simdgroup_half8x8
#define FA_TYPES_BF \
bfloat, bfloat4, simdgroup_bfloat8x8, \
bfloat, bfloat4x4, simdgroup_bfloat8x8, \
bfloat, bfloat4x4, simdgroup_bfloat8x8, \
float, simdgroup_float8x8, \
float, float2, simdgroup_float8x8, \
half, half4, simdgroup_half8x8
//float, float4, simdgroup_float8x8
#define FA_TYPES_F32 \
half, half4, simdgroup_half8x8, \
float, float4x4, simdgroup_float8x8, \
float, float4x4, simdgroup_float8x8, \
float, simdgroup_float8x8, \
float, float2, simdgroup_float8x8, \
float, float4, simdgroup_float8x8
//half, half4, simdgroup_half8x8
typedef decltype(kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 64, 64>) flash_attn_ext_t;
template [[host_name("kernel_flash_attn_ext_f16_dk32_dv32" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 32, 32>;
template [[host_name("kernel_flash_attn_ext_f16_dk40_dv40" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 40, 40>;
template [[host_name("kernel_flash_attn_ext_f16_dk48_dv48" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 48, 48>;
template [[host_name("kernel_flash_attn_ext_f16_dk64_dv64" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 64, 64>;
template [[host_name("kernel_flash_attn_ext_f16_dk72_dv72" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 72, 72>;
template [[host_name("kernel_flash_attn_ext_f16_dk80_dv80" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 80, 80>;
template [[host_name("kernel_flash_attn_ext_f16_dk96_dv96" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 96, 96>;
template [[host_name("kernel_flash_attn_ext_f16_dk96_dv64" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 96, 64>;
template [[host_name("kernel_flash_attn_ext_f16_dk112_dv112")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 112, 112>;
template [[host_name("kernel_flash_attn_ext_f16_dk128_dv128")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 128, 128>;
template [[host_name("kernel_flash_attn_ext_f16_dk128_dv96" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 128, 96>;
template [[host_name("kernel_flash_attn_ext_f16_dk192_dv192")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 192, 192>;
template [[host_name("kernel_flash_attn_ext_f16_dk192_dv128")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 192, 128>;
template [[host_name("kernel_flash_attn_ext_f16_dk256_dv256")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 256, 256>;
template [[host_name("kernel_flash_attn_ext_f16_dk320_dv256")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 320, 256>;
template [[host_name("kernel_flash_attn_ext_f16_dk512_dv512")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 512, 512>;
template [[host_name("kernel_flash_attn_ext_f16_dk576_dv512")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES, half4x4, 1, dequantize_f16, half4x4, 1, dequantize_f16, 576, 512>;
#if defined(GGML_METAL_HAS_BF16)
template [[host_name("kernel_flash_attn_ext_bf16_dk32_dv32" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 32, 32>;
template [[host_name("kernel_flash_attn_ext_bf16_dk40_dv40" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 40, 40>;
template [[host_name("kernel_flash_attn_ext_bf16_dk48_dv48" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 48, 48>;
template [[host_name("kernel_flash_attn_ext_bf16_dk64_dv64" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 64, 64>;
template [[host_name("kernel_flash_attn_ext_bf16_dk72_dv72" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 72, 72>;
template [[host_name("kernel_flash_attn_ext_bf16_dk80_dv80" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 80, 80>;
template [[host_name("kernel_flash_attn_ext_bf16_dk96_dv96" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 96, 96>;
template [[host_name("kernel_flash_attn_ext_bf16_dk96_dv64" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 96, 64>;
template [[host_name("kernel_flash_attn_ext_bf16_dk112_dv112")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 112, 112>;
template [[host_name("kernel_flash_attn_ext_bf16_dk128_dv128")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 128, 128>;
template [[host_name("kernel_flash_attn_ext_bf16_dk128_dv96" )]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 128, 96>;
template [[host_name("kernel_flash_attn_ext_bf16_dk192_dv192")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 192, 192>;
template [[host_name("kernel_flash_attn_ext_bf16_dk192_dv128")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 192, 128>;
template [[host_name("kernel_flash_attn_ext_bf16_dk256_dv256")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 256, 256>;
template [[host_name("kernel_flash_attn_ext_bf16_dk320_dv256")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 320, 256>;
template [[host_name("kernel_flash_attn_ext_bf16_dk512_dv512")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 512, 512>;
template [[host_name("kernel_flash_attn_ext_bf16_dk576_dv512")]] kernel flash_attn_ext_t kernel_flash_attn_ext<FA_TYPES_BF, bfloat4x4, 1, dequantize_bf16, bfloat4x4, 1, dequantize_bf16, 576, 512>;
#endif
#undef FA_TYPES
#undef FA_TYPES_BF
#undef FA_TYPES_F32