Hi, all
I encountered a difficulty of applying "vrgather" instruction recently. The details are shown blow:
The date from source should be duplicated as pair in a upsample application.
Eg: src = [0, 1, 2, 3, 4, 5, 6, 7, 8]
dst = [0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8]
So, my relazation is:
--------------------------------------------------------------------------------------------------
int inex[64] = {0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5 ........，31，31}；// to be compatibled of all the VLEN(128 ~ 1024)
vfloat32m2_t data = vundefined_f32m2();
vfloat32m1_t zero = ( vfloat32m1_t )vmv_v_i_i32m1(0);
while(length >0)
{
int gvl = vsetvl_e32m1(length);
vuint32m2_t v_index = vle32_v_u32m2(index, gvl);
vfloat32m1_t src_data = vle32_v_f32m1(src, gvl);
data = vset_f32m2(data, src_data, 0);
data = vset_f32m2(data, zero, 1);
vfloat32m2_t res = vrrgather_vv_f32m2(data, v_index, gvl);
length -=gvl;
src += gvl;
vse32_v_f32m2(out, res, gvl);
dst +=gvl;
}
-----------------------------------------------------------------------------------------------
As shown before, the index data should be initialized as the max VLEN to make the code compatibled.
So do all the applications, that need a constant.
I think it is contrary to the idea of RISC-V, that one code can run on all the RISC-V hardware. Does anyone have a better method ?
