woutermans
3490023668
The remaining two layers (96 channels, four heads) use the same swin_tok.h template: byte-identical on the same recorded launches and harness sequences, enc2 0.22 -> 0.14 ms and dec2 0.20 -> 0.14 ms at 1280x720 render, 0.75 -> 0.47 and 0.69 -> 0.47 ms at 2560x1440. Skipping windows that lie entirely outside the image is also exact but below the measurement noise (at most 2.4% of the windows of every other tube launch), so it is not kept; true16 code generation remains not tried.
22 lines
874 B
Plaintext
22 lines
874 B
Plaintext
// fast numerics: intermediate activations f16 (no FP8 re-quantisation), f32 accumulation through GEMM chains
|
|
// RX 7700 XT: packed FP8 input decode, packed norm1 sums, half-wave exchange via ds_bpermute, MLP loop unrolled 4x;
|
|
// byte-identical to the default build, gfx1101 only (not measured on other targets)
|
|
// d4r-build-flags-gfx1101: -DSWIN_PACKED_INPUT -DSWIN_PACKED_NORM1 -DSWIN_HALF_BPERMUTE -DSWIN_UNROLL_MLP=4
|
|
#define SWIN_NO_Q8
|
|
#define SWIN_F32ACC
|
|
#ifndef SWIN_MAX_VGPR
|
|
#define SWIN_MAX_VGPR 192
|
|
#endif
|
|
#ifndef SWIN_MLP_R
|
|
#define SWIN_MLP_R(NW) 2
|
|
#endif
|
|
// RDNA4 native FP8 builds use the token-lane block (swin_tok.h); -DSWIN_NO_TOKEN_LANES keeps swin_block.h
|
|
#ifndef SWIN_NO_TOKEN_LANES
|
|
#define SWIN_TOKEN_LANES
|
|
#endif
|
|
#include "swin_block.h"
|
|
#ifndef SWIN_NW
|
|
#define SWIN_NW 8
|
|
#endif
|
|
SWIN_MODULE_W(rrlite_enc2_4x4, 96, 4, SWIN_NW, 128, false, CommonParams, 0)
|