ayo
7c5d3c9157
Add native K/M/RR inference engines, HIP kernels, replay and validation harnesses, and Windows DLSS setup and runtime support. Include GPU profiling tools, build and packaging scripts, platform integration patches, and implementation documentation.
9 lines
401 B
Plaintext
9 lines
401 B
Plaintext
// fast numerics: intermediate activations f16 (no FP8 re-quantisation), f32 accumulation through GEMM chains
|
|
#define SWIN_NO_Q8
|
|
#define SWIN_F32ACC
|
|
#define SWIN_MAX_VGPR 192
|
|
#define SWIN_MLP_R(NW) 1
|
|
#define SWIN_SCALAR_BLOAD // scalar-base weight loads: faster here (no spills), slower in the 8-wave layers
|
|
#include "swin_block.h"
|
|
SWIN_MODULE_W(rrlite_enc1_4x4, 64, 2, 4, 96, false, CommonParams, 0)
|