The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
; RUN: llc < %s -mtriple=armv7-apple-ios | FileCheck %s
|
|
|
|
|
|
|
|
; Test signed conversion.
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t0
|
|
|
|
; CHECK: vcvt.s32.f32 d{{[0-9]+}}, d{{[0-9]+}}, #2
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <2 x i32> @t0(<2 x float> %in) {
|
|
|
|
%mul.i = fmul <2 x float> %in, <float 4.0, float 4.0>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <2 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
; Test unsigned conversion.
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t1
|
|
|
|
; CHECK: vcvt.u32.f32 d{{[0-9]+}}, d{{[0-9]+}}, #3
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <2 x i32> @t1(<2 x float> %in) {
|
|
|
|
%mul.i = fmul <2 x float> %in, <float 8.0, float 8.0>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptoui <2 x float> %mul.i to <2 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <2 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
; Test which should not fold due to non-power of 2.
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t2
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
; CHECK: vmul
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK: vcvt.s32.f32 d{{[0-9]+}}, d{{[0-9]+}}
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <2 x i32> @t2(<2 x float> %in) {
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
entry:
|
2015-10-07 01:54:12 +08:00
|
|
|
%mul.i = fmul <2 x float> %in, <float 0x401B333340000000, float 0x401B333340000000>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <2 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
; Test which should not fold due to power of 2 out of range.
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t3
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
; CHECK: vmul
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK: vcvt.s32.f32 d{{[0-9]+}}, d{{[0-9]+}}
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <2 x i32> @t3(<2 x float> %in) {
|
|
|
|
%mul.i = fmul <2 x float> %in, <float 0x4200000000000000, float 0x4200000000000000>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <2 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
; Test which case where const is max power of 2 (i.e., 2^32).
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t4
|
|
|
|
; CHECK: vcvt.s32.f32 d{{[0-9]+}}, d{{[0-9]+}}, #32
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <2 x i32> @t4(<2 x float> %in) {
|
|
|
|
%mul.i = fmul <2 x float> %in, <float 0x41F0000000000000, float 0x41F0000000000000>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptosi <2 x float> %mul.i to <2 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <2 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
|
|
|
|
|
|
|
; Test quadword.
|
2015-10-07 01:54:12 +08:00
|
|
|
; CHECK-LABEL: @t5
|
|
|
|
; CHECK: vcvt.s32.f32 q{{[0-9]+}}, q{{[0-9]+}}, #3
|
|
|
|
; CHECK: bx lr
|
|
|
|
define <4 x i32> @t5(<4 x float> %in) {
|
|
|
|
%mul.i = fmul <4 x float> %in, <float 8.0, float 8.0, float 8.0, float 8.0>
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
%vcvt.i = fptosi <4 x float> %mul.i to <4 x i32>
|
2015-10-07 01:54:12 +08:00
|
|
|
ret <4 x i32> %vcvt.i
|
The Neon VCVT (between floating-point and fixed-point, Advanced SIMD)
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
2011-06-25 03:23:04 +08:00
|
|
|
}
|
2016-03-18 04:10:28 +08:00
|
|
|
|
|
|
|
; CHECK-LABEL: test_illegal_fp_to_int:
|
|
|
|
; CHECK: vcvt.s32.f32 {{q[0-9]+}}, {{q[0-9]+}}, #2
|
|
|
|
define <3 x i32> @test_illegal_fp_to_int(<3 x float> %in) {
|
|
|
|
%scale = fmul <3 x float> %in, <float 4.0, float 4.0, float 4.0>
|
|
|
|
%val = fptosi <3 x float> %scale to <3 x i32>
|
|
|
|
ret <3 x i32> %val
|
|
|
|
}
|