SetFmatrix
功能说明
用于调用Load3Dv1/Load3Dv2时设置FeatureMap的属性描述。Load3Dv1/Load3Dv2的模板参数isSetFMatrix设置为false时,表示Load3Dv1/Load3Dv2传入的FeatureMap的属性(包括l1H、l1W、padList,参数介绍参考表4 LoadData3DParamsV1结构体内参数说明、表5 LoadData3DParamsV2结构体内参数说明)描述不生效,开发者需要通过该接口进行设置。
函数原型
__aicore__ inline void SetFmatrix(uint16_t l1H, uint16_t l1W, const uint8_t padList[4], const FmatrixMode& fmatrixMode)
参数说明
参数名称 |
输入/输出 |
含义 |
|---|---|---|
l1H |
输入 |
源操作数height,取值范围:l1H∈[1, 32767]。 |
l1W |
输入 |
源操作数width,取值范围:l1W∈[1, 32767] 。 |
padList |
输入 |
padding列表 [padding_left, padding_right, padding_top, padding_bottom],每个元素取值范围:[0,255]。默认为{0, 0, 0, 0}。 |
fmatrixMode |
输入 |
FmatrixMode类型,定义如下: enum class FmatrixMode : uint8_t {
FMATRIX_LEFT = 0,
FMATRIX_RIGHT = 1,
};
当前只支持FMATRIX_LEFT,左右矩阵均使用该配置;
|
注意事项
- 该接口需要配合load3Dv1/load3Dv2接口一起使用,需要在load3Dv1/load3Dv2接口之前调用。
- 操作数地址偏移对齐要求请参见通用约束。
支持的型号
Atlas推理系列产品AI Core
调用示例
#include "kernel_operator.h"
using namespace AscendC;
namespace AscendC {
template <typename dst_T, typename fmap_T, typename weight_T, typename dstCO1_T> class KernelLoad3d {
public:
__aicore__ inline KernelLoad3d()
{
// ceiling of 16
C0 = 32 / sizeof(fmap_T);
C1 = channelSize / C0;
coutBlocks = (Cout + 16 - 1) / 16;
ho = H - dilationH * (Kh - 1);
wo = W - dilationW * (Kw - 1);
howo = ho * wo;
howoRound = ((howo + 16 - 1) / 16) * 16;
featureMapA1Size = C1 * H * W * C0; // shape: [C1, H, W, C0]
weightA1Size = C1 * Kh * Kw * Cout * C0; // shape: [C1, Kh, Kw, Cout, C0]
featureMapA2Size = howoRound * (C1 * Kh * Kw * C0);
weightB2Size = (C1 * Kh * Kw * C0) * coutBlocks * 16;
m = howo;
k = C1 * Kh * Kw * C0;
n = Cout;
biasSize = Cout; // shape: [Cout]
dstSize = coutBlocks * howo * 16; // shape: [coutBlocks, howo, 16]
dstCO1Size = coutBlocks * howoRound * 16;
fmRepeat = featureMapA2Size / (16 * C0);
weRepeat = weightB2Size / (16 * C0);
}
__aicore__ inline void Init(__gm__ uint8_t* fmGm, __gm__ uint8_t* weGm, __gm__ uint8_t* biasGm,
__gm__ uint8_t* dstGm)
{
fmGlobal.SetGlobalBuffer((__gm__ fmap_T*)fmGm);
weGlobal.SetGlobalBuffer((__gm__ weight_T*)weGm);
biasGlobal.SetGlobalBuffer((__gm__ dstCO1_T*)biasGm);
dstGlobal.SetGlobalBuffer((__gm__ dst_T*)dstGm);
pipe.InitBuffer(inQueueFmA1, 1, featureMapA1Size * sizeof(fmap_T));
pipe.InitBuffer(inQueueFmA2, 1, featureMapA2Size * sizeof(fmap_T));
pipe.InitBuffer(inQueueWeB1, 1, weightA1Size * sizeof(weight_T));
pipe.InitBuffer(inQueueWeB2, 1, weightB2Size * sizeof(weight_T));
pipe.InitBuffer(outQueueCO1, 1, dstCO1Size * sizeof(dstCO1_T));
}
__aicore__ inline void Process()
{
CopyIn();
Split();
Compute();
CopyOut();
}
private:
__aicore__ inline void CopyIn()
{
LocalTensor<fmap_T> featureMapA1 = inQueueFmA1.AllocTensor<fmap_T>();
LocalTensor<weight_T> weightB1 = inQueueWeB1.AllocTensor<weight_T>();
DataCopy(featureMapA1, fmGlobal, { 1, static_cast<uint16_t>(featureMapA1Size * sizeof(fmap_T) / 32), 0, 0 });
DataCopy(weightB1, weGlobal, { 1, static_cast<uint16_t>(weightA1Size * sizeof(weight_T) / 32), 0, 0 });
inQueueFmA1.EnQue(featureMapA1);
inQueueWeB1.EnQue(weightB1);
}
__aicore__ inline void Split()
{
LocalTensor<fmap_T> featureMapA1 = inQueueFmA1.DeQue<fmap_T>();
LocalTensor<weight_T> weightB1 = inQueueWeB1.DeQue<weight_T>();
LocalTensor<fmap_T> featureMapA2 = inQueueFmA2.AllocTensor<fmap_T>();
LocalTensor<weight_T> weightB2 = inQueueWeB2.AllocTensor<weight_T>();
uint8_t padList[PAD_SIZE] = {0, 0, 0, 0};
SetFmatrix(H, W, padList, FmatrixMode::FMATRIX_LEFT);
SetLoadDataPaddingValue(0);
SetLoadDataRepeat({0, 1, 0});
SetLoadDataBoundary((uint32_t)0);
static constexpr IsResetLoad3dConfig LOAD3D_CONFIG = {false,false};
LoadData<fmap_T, LOAD3D_CONFIG>(featureMapA2, featureMapA1,
{ padList, H, W, channelSize, k, howoRound, 0, 0, 1, 1, Kw, Kh, dilationW, dilationH, false, false, 0 });
LoadData(weightB2, weightB1, { 0, weRepeat, 1, 0, 0, false, 0 });
inQueueFmA2.EnQue<fmap_T>(featureMapA2);
inQueueWeB2.EnQue<weight_T>(weightB2);
inQueueFmA1.FreeTensor(featureMapA1);
inQueueWeB1.FreeTensor(weightB1);
}
__aicore__ inline void Compute()
{
LocalTensor<fmap_T> featureMapA2 = inQueueFmA2.DeQue<fmap_T>();
LocalTensor<weight_T> weightB2 = inQueueWeB2.DeQue<weight_T>();
LocalTensor<dstCO1_T> dstCO1 = outQueueCO1.AllocTensor<dstCO1_T>();
Mmad(dstCO1, featureMapA2, weightB2, { m, n, k, true, 0, false, false, false });
outQueueCO1.EnQue<dstCO1_T>(dstCO1);
inQueueFmA2.FreeTensor(featureMapA2);
inQueueWeB2.FreeTensor(weightB2);
}
__aicore__ inline void CopyOut()
{
LocalTensor<dstCO1_T> dstCO1 = outQueueCO1.DeQue<dstCO1_T>();
FixpipeParamsV220 fixpipeParams;
fixpipeParams.nSize = coutBlocks * 16;
fixpipeParams.mSize = howo;
fixpipeParams.srcStride = howo;
fixpipeParams.dstStride = howo * BLOCK_CUBE * sizeof(dst_T) / ONE_BLK_SIZE;
fixpipeParams.quantPre = deqMode;
Fixpipe<dst_T, dstCO1_T, CFG_NZ>(dstGlobal, dstCO1, fixpipeParams);
outQueueCO1.FreeTensor(dstCO1);
}
private:
TPipe pipe;
// feature map queue
TQue<QuePosition::A1, 1> inQueueFmA1;
TQue<QuePosition::A2, 1> inQueueFmA2;
// weight queue
TQue<QuePosition::B1, 1> inQueueWeB1;
TQue<QuePosition::B2, 1> inQueueWeB2;
// dst queue
TQue<QuePosition::CO1, 1> outQueueCO1;
GlobalTensor<fmap_T> fmGlobal;
GlobalTensor<weight_T> weGlobal;
GlobalTensor<dst_T> dstGlobal;
GlobalTensor<dstCO1_T> biasGlobal;
uint16_t channelSize = 32;
uint16_t H = 4, W = 4;
uint8_t Kh = 2, Kw = 2;
uint16_t Cout = 16;
uint16_t C0, C1;
uint8_t dilationH = 2, dilationW = 2;
uint16_t coutBlocks, ho, wo, howo, howoRound;
uint32_t featureMapA1Size, weightA1Size, featureMapA2Size, weightB2Size, biasSize, dstSize, dstCO1Size;
uint16_t m, k, n;
uint8_t fmRepeat, weRepeat;
QuantMode_t deqMode = QuantMode_t::F322F16;
};
} // namespace AscendC
extern "C" __global__ __aicore__ void load3d_simple_kernel(__gm__ uint8_t *fmGm, __gm__ uint8_t *weGm,
__gm__ uint8_t *biasGm, __gm__ uint8_t *dstGm)
{
AscendC::KernelLoad3d<dst_type, fmap_type, weight_type, dstCO1_type> op;
op.Init(fmGm, weGm, biasGm, dstGm);
op.Process();
}
父主题: 矩阵计算