mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-05-02 04:31:25 +00:00
[Ck tile] support rmsnorm and related fusion (#1605)
* Add reduce2d new api
* Prevent user use cross warp reduction
* Fix bug of std caculation
* Add rmsnorm2d
* Add rmsnorm small example
* Remove static assert to prevent compile fail
* Add script to test performance and correctness
* Add missing cmake change
* refine naming
* refine example of rmsnorm
* Fix bug of rmsnorm
* Refine naming
* Fix cmake
* clang format
* Refine pipeline name
* Add add_rmsnorm2d_rdquant kernel
* Add reduce op
* host verification
* Fix bug of one pass pipeline
* Refine tile size
* Add two pass pipeline
* Rename two pass to three pass
* Fix bug of kSaveX == false
* Add instance library
* Add test script
* Fix bug of x verification
* Add save_x to trait
* Add README
* Move reduce2d into reduce folder
* Fix bug of welford when number of m warp > 1
* remove reduncant comment
* 1. move 06_rmsnorm2d to 10_rmsnorm2d
2. move 07_add_rmsnorm2d_rdquant to 11_add_rmsnorm2d_rdquant
* clang format and add missing header
* Add host validation of add + layernorm2d + rsquant
* Revert "Add host validation of add + layernorm2d + rsquant"
This reverts commit 936cb45797.
* Remove deprecated flag
This commit is contained in:
38
example/ck_tile/10_rmsnorm2d/script/perf_test.sh
Executable file
38
example/ck_tile/10_rmsnorm2d/script/perf_test.sh
Executable file
@@ -0,0 +1,38 @@
|
||||
|
||||
# run from top of ck folder
|
||||
EXE=build/bin/tile_rmsnorm2d_fwd
|
||||
|
||||
$EXE -m=1 -n=1 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=80 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=128 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=144 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=168 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=184 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=256 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=288 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=344 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=376 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=448 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=512 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=924 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1024 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1078 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=1996 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
$EXE -m=700 -n=4080 -e=1e-12 -v=1 -prec=bf16 -repeat=1000
|
||||
|
||||
$EXE -m=700 -n=80 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=128 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=144 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=168 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=184 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=256 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=288 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=344 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=376 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=448 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=512 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=924 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1024 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1078 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=1996 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
$EXE -m=700 -n=4080 -e=1e-12 -v=1 -prec=fp16 -repeat=1000
|
||||
31
example/ck_tile/10_rmsnorm2d/script/smoke_test.sh
Executable file
31
example/ck_tile/10_rmsnorm2d/script/smoke_test.sh
Executable file
@@ -0,0 +1,31 @@
|
||||
#!/bin/sh
|
||||
# call from top of CK folder
|
||||
EXE=./build/bin/tile_rmsnorm2d_fwd
|
||||
|
||||
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
|
||||
done
|
||||
Reference in New Issue
Block a user