mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-25 04:02:41 +01:00
3127a19140
instructions can be used to match combinations of multiply/divide and VCVT (between floating-point and integer, Advanced SIMD). Basically the VCVT immediate operand that specifies the number of fraction bits corresponds to a floating-point multiply or divide by the corresponding power of 2. For example, VCVT (floating-point to fixed-point, Advanced SIMD) can replace a combination of VMUL and VCVT (floating-point to integer) as follows: Example (assume d17 = <float 8.000000e+00, float 8.000000e+00>): vmul.f32 d16, d17, d16 vcvt.s32.f32 d16, d16 becomes: vcvt.s32.f32 d16, d16, #3 Similarly, VCVT (fixed-point to floating-point, Advanced SIMD) can replace a combinations of VCVT (integer to floating-point) and VDIV as follows: Example (assume d17 = <float 8.000000e+00, float 8.000000e+00>): vcvt.f32.s32 d16, d16 vdiv.f32 d16, d17, d16 becomes: vcvt.f32.s32 d16, d16, #3 llvm-svn: 133813
100 lines
3.5 KiB
LLVM
100 lines
3.5 KiB
LLVM
; RUN: llc < %s -mtriple=armv7-apple-ios | FileCheck %s
|
|
|
|
@in = global float 0x400921FA00000000, align 4
|
|
|
|
; Test signed conversion.
|
|
; CHECK: t0
|
|
; CHECK-NOT: vmul
|
|
define void @t0() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <2 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <2 x float> %vecinit.i, float %tmp, i32 1
|
|
%mul.i = fmul <2 x float> %vecinit2.i, <float 8.000000e+00, float 8.000000e+00>
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
|
tail call void @foo_int32x2_t(<2 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
declare void @foo_int32x2_t(<2 x i32>)
|
|
|
|
; Test unsigned conversion.
|
|
; CHECK: t1
|
|
; CHECK-NOT: vmul
|
|
define void @t1() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <2 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <2 x float> %vecinit.i, float %tmp, i32 1
|
|
%mul.i = fmul <2 x float> %vecinit2.i, <float 8.000000e+00, float 8.000000e+00>
|
|
%vcvt.i = fptoui <2 x float> %mul.i to <2 x i32>
|
|
tail call void @foo_uint32x2_t(<2 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
declare void @foo_uint32x2_t(<2 x i32>)
|
|
|
|
; Test which should not fold due to non-power of 2.
|
|
; CHECK: t2
|
|
; CHECK: vmul
|
|
define void @t2() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <2 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <2 x float> %vecinit.i, float %tmp, i32 1
|
|
%mul.i = fmul <2 x float> %vecinit2.i, <float 0x401B333340000000, float 0x401B333340000000>
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
|
tail call void @foo_int32x2_t(<2 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
; Test which should not fold due to power of 2 out of range.
|
|
; CHECK: t3
|
|
; CHECK: vmul
|
|
define void @t3() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <2 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <2 x float> %vecinit.i, float %tmp, i32 1
|
|
%mul.i = fmul <2 x float> %vecinit2.i, <float 0x4200000000000000, float 0x4200000000000000>
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
|
tail call void @foo_int32x2_t(<2 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
; Test which case where const is max power of 2 (i.e., 2^32).
|
|
; CHECK: t4
|
|
; CHECK-NOT: vmul
|
|
define void @t4() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <2 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <2 x float> %vecinit.i, float %tmp, i32 1
|
|
%mul.i = fmul <2 x float> %vecinit2.i, <float 0x41F0000000000000, float 0x41F0000000000000>
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
|
tail call void @foo_int32x2_t(<2 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
; Test quadword.
|
|
; CHECK: t5
|
|
; CHECK-NOT: vmul
|
|
define void @t5() nounwind {
|
|
entry:
|
|
%tmp = load float* @in, align 4, !tbaa !0
|
|
%vecinit.i = insertelement <4 x float> undef, float %tmp, i32 0
|
|
%vecinit2.i = insertelement <4 x float> %vecinit.i, float %tmp, i32 1
|
|
%vecinit4.i = insertelement <4 x float> %vecinit2.i, float %tmp, i32 2
|
|
%vecinit6.i = insertelement <4 x float> %vecinit4.i, float %tmp, i32 3
|
|
%mul.i = fmul <4 x float> %vecinit6.i, <float 8.000000e+00, float 8.000000e+00, float 8.000000e+00, float 8.000000e+00>
|
|
%vcvt.i = fptosi <4 x float> %mul.i to <4 x i32>
|
|
tail call void @foo_int32x4_t(<4 x i32> %vcvt.i) nounwind
|
|
ret void
|
|
}
|
|
|
|
declare void @foo_int32x4_t(<4 x i32>)
|
|
|
|
!0 = metadata !{metadata !"float", metadata !1}
|
|
!1 = metadata !{metadata !"omnipotent char", metadata !2}
|
|
!2 = metadata !{metadata !"Simple C/C++ TBAA", null}
|