[CK_TILE] Add Various Fusion Functions to RMSNorm (#1802)

* Add shortcut to RMSNorm

* Modify test for adding shortcut for RMSNorm

* Add fused parameter into tests

* 1. Add YDataType. 2. rmsnorm2d_fwd_traits_ from rmsnorm2d_fwd.hpp to rmsnorm2d_fwd_api.cpp and rmsnorm2d_fwd_instance_common.hpp

* 1. Supports various stride and percisions.

* Add support of Epilogue

* Add fuse and epilogue support to rmsnorm ref

* Modify rmsnorm example

* Refactor tests/examples

* Bug fix for newly added tests/examples

* Bug fix for new tests 2

* Modify smoke test scripts

remove dbg code

* Supports non-smooth dyanmic quant

* Update Rmsnorm2dFwd::GetName()

* rename xscale and prec_sx to smoothscale and prec_sm

Bug fix after rename

Remove files

* change example_rmsnorm2d_fwd.cpp

* update performance calculator

* Fix issue in two-pass when fuse add is enabled

* Remove comment of beta

---------

Co-authored-by: rocking <ChunYu.Lai@amd.com>
This commit is contained in:
ruanjm
2025-01-15 10:23:48 +08:00
committed by GitHub
parent c0b90f130f
commit 04dd314883
58 changed files with 1823 additions and 1045 deletions

View File

@@ -1,30 +1,34 @@
#!/bin/sh
EXE="$(find . -name tile_rmsnorm2d_fwd -type f | head -n 1)"
for fquant in "" "-fquant=1 -prec_o=int8" "-fquant=2 -prec_o=int8"; do
for pr_i in "fp16" "bf16" ; do
$EXE -prec=$pr_i -m=99 -n=13
$EXE -prec=$pr_i -m=17 -n=16
$EXE -prec=$pr_i -m=1 -n=100
$EXE -prec=$pr_i -m=4 -n=128
$EXE -prec=$pr_i -m=80 -n=127
$EXE -prec=$pr_i -m=22 -n=255 -stride=256
$EXE -prec=$pr_i -m=7 -n=599
$EXE -prec=$pr_i -m=19 -n=512
$EXE -prec=$pr_i -m=33 -n=313 -stride=1000
$EXE -prec=$pr_i -m=11 -n=510
$EXE -prec=$pr_i -m=171 -n=676 -stride=818
$EXE -prec=$pr_i -m=91 -n=636
$EXE -prec=$pr_i -m=12 -n=768 -stride=800
$EXE -prec=$pr_i -m=100 -n=766 -stride=812
$EXE -prec=$pr_i -m=31 -n=1024
$EXE -prec=$pr_i -m=64 -n=1000 -stride=1004
$EXE -prec=$pr_i -m=8 -n=1501
$EXE -prec=$pr_i -m=3 -n=1826
$EXE -prec=$pr_i -m=5 -n=2040
$EXE -prec=$pr_i -m=7 -n=2734
$EXE -prec=$pr_i -m=1 -n=3182
$EXE -prec=$pr_i -m=9 -n=4096
$EXE -prec=$pr_i -m=3 -n=8192
$EXE -prec=$pr_i -m=1 -n=10547
$EXE -prec=$pr_i -m=3 -n=17134
for fadd in "0" "1"; do
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=99 -n=13
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=17 -n=16
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=100
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=4 -n=128
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=80 -n=127
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=22 -n=255 -stride=256
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=7 -n=599
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=19 -n=512
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=33 -n=313 -stride=1000
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=11 -n=510
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=171 -n=676 -stride=818
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=91 -n=636
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=12 -n=768 -stride=800
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=100 -n=766 -stride=812
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=31 -n=1024
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=64 -n=1000 -stride=1004
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=8 -n=1501
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=1826
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=5 -n=2040
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=7 -n=2734
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=3182
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=9 -n=4096
$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=8192
#$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=1 -n=10547
#$EXE -prec_i=$pr_i -fadd=$fadd $fquant -m=3 -n=17134
done
done
done