maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

Elaborate on the entry on integer vector multiplication by constants. 

llvm-svn: 51491
This commit is contained in:
Dan Gohman 2008-05-23 18:05:39 +00:00
parent 01b7fffb29
commit 66eea1b9b3
1 changed files with 6 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

7
llvm/lib/Target/X86/README-SSE.txt
View File

@ -814,7 +814,7 @@ define <4 x i32> @f(<4 x i32> %i) nounwind {
ret <4 x i32> %A ret <4 x i32> %A
} }
Compiles into: On targets without SSE4.1, this compiles into:
LCPI1_0: ## <4 x i32> LCPI1_0: ## <4 x i32>
.long 10 .long 10
@ -846,6 +846,11 @@ _f:
punpckldq %xmm2, %xmm0 punpckldq %xmm2, %xmm0
ret ret
It would be better to synthesize integer vector multiplication by constants
using shifts and adds, pslld and paddd here. And even on targets with SSE4.1,
simple cases such as multiplication by powers of two would be better as
vector shifts than as multiplications.
//===---------------------------------------------------------------------===// //===---------------------------------------------------------------------===//
We compile this: We compile this: