http://llvm.org/bugs/show_bug.cgi?id=17170
Bug ID: 17170
Summary: missed opportunity to use wider (AVX2) vectors
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Classification: Unclassified
The following code sequence is correctly identified as an opportunity to use a
parallel (AVX) addition operation, but it could have been optimized further by
using 32-byte (8-integer) vector instructions rather than 16-byte (4-integer)
vector instructions:
$ ./clang -v
clang version 3.4 (trunk 189776)
Target: x86_64-apple-darwin11.4.2
Thread model: posix
$ cat paddd.c
void foo(int *restrict x, int *restrict y, int *restrict z) {
int i;
for (i=0; i<8; i++) {
z[i] = x[i] + y[i];
}
}
$ ./clang -S -std=c99 -O3 -funroll-loops -fomit-frame-pointer -march=core-avx2
-o /dev/stdout paddd.c
.section __TEXT,__text,regular,pure_instructions
.globl _foo
.align 4, 0x90
_foo: ## @foo
.cfi_startproc
## BB#0: ## %entry
vmovdqu (%rdi), %xmm0
vpaddd (%rsi), %xmm0, %xmm0
vmovdqu %xmm0, (%rdx)
vmovdqu 16(%rdi), %xmm0
vpaddd 16(%rsi), %xmm0, %xmm0
vmovdqu %xmm0, 16(%rdx)
ret
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
