[LLVMbugs] [Bug 17236] New: missed opportunity to use vector FMA instruction
bugzilla-daemon at llvm.org
bugzilla-daemon at llvm.org
Fri Sep 13 11:34:22 PDT 2013
http://llvm.org/bugs/show_bug.cgi?id=17236
Bug ID: 17236
Summary: missed opportunity to use vector FMA instruction
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: Loop Optimizer
Assignee: unassignedbugs at nondot.org
Reporter: kkhoo at perfwizard.com
CC: llvmbugs at cs.uiuc.edu
Classification: Unclassified
I think this problem is similar to bug 17172. For a short vector, the SLP
vectorizer is not recognizing an FMA operation as a vector candidate?
$ ./clang -v
clang version 3.4 (trunk 189776)
Target: x86_64-apple-darwin11.4.2
Thread model: posix
$ cat fma.c
#pragma STDC FP_CONTRACT ON
void foo(double *x) {
int i;
for(i=0; i<8; i++) {
x[i] = x[i] * x[i] + x[i];
}
}
$ ./clang -S -O3 -fomit-frame-pointer -march=core-avx2 -o /dev/stdout fma.c
.section __TEXT,__text,regular,pure_instructions
.globl _foo
.align 4, 0x90
_foo: ## @foo
.cfi_startproc
## BB#0: ## %entry
vmovsd (%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, (%rdi)
vmovsd 8(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 8(%rdi)
vmovsd 16(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 16(%rdi)
vmovsd 24(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 24(%rdi)
vmovsd 32(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 32(%rdi)
vmovsd 40(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 40(%rdi)
vmovsd 48(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 48(%rdi)
vmovsd 56(%rdi), %xmm0
vfmadd213sd %xmm0, %xmm0, %xmm0
vmovsd %xmm0, 56(%rdi)
ret
--
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20130913/f782576a/attachment.html>
More information about the llvm-bugs
mailing list