R600/SI: implement range reduction for sin/cos

These instructions can only take a limited input range, and return
the constant value 1 out of range. We should do range reduction to
be able to process arbitrary values. Use a FRACT instruction after
normalization to achieve this. Also add a test for constant folding
with the lowered code with unsafe-fp-math enabled.

v2: use DAG lowering instead of intrinsic, adapt test
v3: calculate constant, fold pattern into instruction definition
v4: misc style fixes, add sin-fold testcase, cosmetics

Patch by Grigori Goronzy

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@213458 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Matt Arsenault
2014-07-19 18:44:39 +00:00
parent 5e1c96a632
commit 18ecf3fff3
5 changed files with 53 additions and 14 deletions
+20 -2
View File
@@ -1,5 +1,6 @@
;RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck %s -check-prefix=EG -check-prefix=FUNC
;RUN: llc < %s -march=r600 -mcpu=SI | FileCheck %s -check-prefix=SI -check-prefix=FUNC
;RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
;RUN: llc -march=r600 -mcpu=SI < %s | FileCheck -check-prefix=SI -check-prefix=SI-SAFE -check-prefix=FUNC %s
;RUN: llc -march=r600 -mcpu=SI -enable-unsafe-fp-math < %s | FileCheck -check-prefix=SI -check-prefix=SI-UNSAFE -check-prefix=FUNC %s
;FUNC-LABEL: test
;EG: MULADD_IEEE *
@@ -8,6 +9,7 @@
;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
;EG-NOT: SIN
;SI: V_MUL_F32
;SI: V_FRACT_F32
;SI: V_SIN_F32
;SI-NOT: V_SIN_F32
@@ -17,6 +19,22 @@ define void @test(float addrspace(1)* %out, float %x) #1 {
ret void
}
;FUNC-LABEL: testf
;SI-UNSAFE: 4.774
;SI-UNSAFE: V_MUL_F32
;SI-SAFE: V_MUL_F32
;SI-SAFE: V_MUL_F32
;SI: V_FRACT_F32
;SI: V_SIN_F32
;SI-NOT: V_SIN_F32
define void @testf(float addrspace(1)* %out, float %x) #1 {
%y = fmul float 3.0, %x
%sin = call float @llvm.sin.f32(float %y)
store float %sin, float addrspace(1)* %out
ret void
}
;FUNC-LABEL: testv
;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}
;EG: SIN * T{{[0-9]+\.[XYZW], PV\.[XYZW]}}