ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-06-06 15:54:31 +00:00
Code Issues Packages Projects Releases Wiki Activity

Pack conversion in a loop

Browse Source
This commit is contained in:
Rostyslav Geyyer
2025-02-19 20:54:42 +00:00
parent bb953dad7e
commit e971702a8c
1 changed files with 7 additions and 63 deletions
Download Patch File Download Diff File Expand all files Collapse all files

70
include/ck/utility/scaled_type_convert.hpp
View File

@@ -405,70 +405,14 @@ inline __host__ __device__ float32_t scaled_type_convert<float32_t, f4x32_t>(e8m
value.f4x32_array = x;
float2_t op;
float32_t ret;
// TODO: pack in a loop
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[0], type_convert<float>(scale), 0);
ret[0] = op[1];
ret[1] = op[0];
float f_scale = type_convert<float>(scale);
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[1], type_convert<float>(scale), 0);
ret[2] = op[1];
ret[3] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[2], type_convert<float>(scale), 0);
ret[4] = op[1];
ret[5] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[3], type_convert<float>(scale), 0);
ret[6] = op[1];
ret[7] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[4], type_convert<float>(scale), 0);
ret[8] = op[1];
ret[9] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[5], type_convert<float>(scale), 0);
ret[10] = op[1];
ret[11] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[6], type_convert<float>(scale), 0);
ret[12] = op[1];
ret[13] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[7], type_convert<float>(scale), 0);
ret[14] = op[1];
ret[15] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[8], type_convert<float>(scale), 0);
ret[16] = op[1];
ret[17] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[9], type_convert<float>(scale), 0);
ret[18] = op[1];
ret[19] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[10], type_convert<float>(scale), 0);
ret[20] = op[1];
ret[21] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[11], type_convert<float>(scale), 0);
ret[22] = op[1];
ret[23] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[12], type_convert<float>(scale), 0);
ret[24] = op[1];
ret[25] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[13], type_convert<float>(scale), 0);
ret[26] = op[1];
ret[27] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[14], type_convert<float>(scale), 0);
ret[28] = op[1];
ret[29] = op[0];
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[15], type_convert<float>(scale), 0);
ret[30] = op[1];
ret[31] = op[0];
ck::static_for<0, 32 / 2, 1>{}([&](auto idx) {
op = __builtin_amdgcn_cvt_scalef32_pk_f32_fp4(value.fp4x2[idx], f_scale, 0);
// permute high bits and low bits to match the order of the original vector
ret[2 * idx] = op[1];
ret[2 * idx + 1] = op[0];
});
return ret;
#else