Skip to main content

Browse the archive

https://github.com/halide/Halide
21 March 2024, 02:33:35 UTC
Revision 474b9e193c460b7b620b610f58f9c7a075128548 authored by Andrew Adams on 13 October 2020, 23:01:42 UTC, committed by Andrew Adams on 13 October 2020, 23:01:42 UTC 
Use a binary reduction tree for outer iterations
 
Also it's better to vectorize at the native width, for both the new and
baseline schedules. New inner loop (which does the same number of
multiply-adds, but in a 16x4 tile instead of an 8x8 tile):

```
	vmovdqu64	(%rcx,%rdx,8), %zmm5
	vmovq	(%r14,%rdx,8), %xmm6    # xmm6 = mem[0],zero
	vpermw	%zmm5, %zmm0, %zmm7
	vpermw	%zmm5, %zmm1, %zmm5
	vpermd	%zmm6, %zmm2, %zmm8
	vpermd	%zmm8, %zmm3, %zmm8
	vpmaddwd	%zmm8, %zmm7, %zmm7
	vpbroadcastd	%xmm6, %zmm6
	vpmaddwd	%zmm6, %zmm5, %zmm5
	vpaddd	%zmm5, %zmm4, %zmm4
	vpaddd	%zmm7, %zmm4, %zmm4
	incq	%rdx
	cmpq	$32, %rdx

```
1 parent 21b7492
History
Tip revision: 474b9e193c460b7b620b610f58f9c7a075128548 authored by Andrew Adams on 13 October 2020, 23:01:42 UTC
Use a binary reduction tree for outer iterations
Tip revision: 474b9e1
File Mode Size
.github
apps
cmake
dependencies
doc
packaging
python_bindings
src
test
tools
tutorial
util
.clang-format -rw-r--r-- 1.5 KB
.clang-format-ignore -rw-r--r-- 265 bytes
.clang-tidy -rw-r--r-- 506 bytes
.gitattributes -rw-r--r-- 342 bytes
.gitignore -rw-r--r-- 1.1 KB
.gitmodules -rw-r--r-- 0 bytes
CMakeLists.txt -rw-r--r-- 4.5 KB
CODE_OF_CONDUCT.md -rw-r--r-- 3.5 KB
LICENSE.txt -rw-r--r-- 3.2 KB
Makefile -rw-r--r-- 95.7 KB
README.md -rw-r--r-- 19.6 KB
README_cmake.md -rw-r--r-- 66.5 KB
README_rungen.md -rw-r--r-- 12.1 KB
README_webassembly.md -rw-r--r-- 7.5 KB

README.md

back to top