2015-04-24 00:07:50 +08:00
|
|
|
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=avx -enable-unsafe-fp-math < %s | FileCheck %s
|
2012-08-31 07:35:16 +08:00
|
|
|
|
|
|
|
define float @test1(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test1:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2012-08-31 07:35:16 +08:00
|
|
|
%t1 = fadd float %a, %a
|
|
|
|
%r = fadd float %t1, %t1
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test2(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test2:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2012-08-31 07:35:16 +08:00
|
|
|
%t1 = fmul float 4.0, %a
|
|
|
|
%t2 = fadd float %a, %a
|
|
|
|
%r = fadd float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test3(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test3:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fmul float %a, 4.0
|
|
|
|
%t2 = fadd float %a, %a
|
|
|
|
%r = fadd float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test4(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test4:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fadd float %a, %a
|
|
|
|
%t2 = fmul float 4.0, %a
|
|
|
|
%r = fadd float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test5(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test5:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vmulss {{.*}}(%rip), %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fadd float %a, %a
|
|
|
|
%t2 = fmul float %a, 4.0
|
|
|
|
%r = fadd float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test6(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test6:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2012-08-31 07:35:16 +08:00
|
|
|
%t1 = fmul float 2.0, %a
|
|
|
|
%t2 = fadd float %a, %a
|
|
|
|
%r = fsub float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
2013-06-15 02:17:35 +08:00
|
|
|
define float @test7(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test7:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fmul float %a, 2.0
|
|
|
|
%t2 = fadd float %a, %a
|
|
|
|
%r = fsub float %t1, %t2
|
|
|
|
ret float %r
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test8(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test8:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: retq
|
2012-11-01 10:00:53 +08:00
|
|
|
%t1 = fmul float %a, 0.0
|
|
|
|
%t2 = fadd float %a, %t1
|
|
|
|
ret float %t2
|
|
|
|
}
|
|
|
|
|
2013-06-15 02:17:35 +08:00
|
|
|
define float @test9(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test9:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fmul float 0.0, %a
|
|
|
|
%t2 = fadd float %t1, %a
|
|
|
|
ret float %t2
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test10(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test10:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2013-06-15 02:17:35 +08:00
|
|
|
%t1 = fsub float -0.0, %a
|
|
|
|
%t2 = fadd float %a, %t1
|
|
|
|
ret float %t2
|
|
|
|
}
|
|
|
|
|
|
|
|
define float @test11(float %a) {
|
2015-04-24 00:07:50 +08:00
|
|
|
; CHECK-LABEL: test11:
|
|
|
|
; CHECK: # BB#0:
|
|
|
|
; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
|
|
|
|
; CHECK-NEXT: retq
|
2012-11-01 10:00:53 +08:00
|
|
|
%t1 = fsub float -0.0, %a
|
|
|
|
%t2 = fadd float %a, %t1
|
|
|
|
ret float %t2
|
|
|
|
}
|
transform fadd chains to increase parallelism
This is a compromise: with this simple patch, we should always handle a chain of exactly 3
operations optimally, but we're not generating the optimal balanced binary tree for a longer
sequence.
In general, this transform will reduce the dependency chain for a sequence of instructions
using N operands from a worst case N-1 dependent operations to N/2 dependent operations.
The optimal balanced binary tree would reduce the chain to log2(N).
The trade-off for not dealing with longer sequences is: (1) we have less complexity in the
compiler, (2) we avoid unknown compile-time blowup calculating a balanced tree, and (3) we
don't need to worry about the increased register pressure required to parallelize longer
sequences. It also seems unlikely that we would ever encounter really long strings of
dependent ops like that in the wild, but I'm not sure how to verify that speculation.
FWIW, I see no perf difference for test-suite running on btver2 (x86-64) with -ffast-math
and this patch.
We can extend this patch to cover other associative operations such as fmul, fmax, fmin,
integer add, integer mul.
This is a partial fix for:
https://llvm.org/bugs/show_bug.cgi?id=17305
and if extended:
https://llvm.org/bugs/show_bug.cgi?id=21768
https://llvm.org/bugs/show_bug.cgi?id=23116
The issue also came up in:
http://reviews.llvm.org/D8941
Differential Revision: http://reviews.llvm.org/D9232
llvm-svn: 236031
2015-04-29 05:03:22 +08:00
|
|
|
|