RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-31 20:51:52 +01:00
Code Issues Projects Releases Wiki Activity
llvm-mirror/test/CodeGen/AArch64/div-rem-pair-recomposition-unsigned.ll
Roman Lebedev 3e5d42babb [NFC][CodeGen][X86][AArch64] div-rem pair reconstruction tests (PR42673) 
As discussed in https://bugs.llvm.org/show_bug.cgi?id=42673
there is a TTI hook hasDivRemOp() that matters here.
While -div-rem-pairs will decompose 'rem' if that hook returns false,
nothing does the opposite transform.

We can't to this in InstCombine, because it does not currently
access TTI, and i'm not sure we should change that.

We may be able to teach DivRemPairs to do this, but this really is a
per-target perf optimization, and we seem to do the opposite transform
in backend if hasDivRemOp() returned false: https://godbolt.org/z/ttt4HZ
I think it makes sense to be consistent.

https://bugs.llvm.org/show_bug.cgi?id=42673

llvm-svn: 367034
2019-07-25 16:39:57 +00:00

320 lines
9.4 KiB
LLVM
Raw Blame History

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=aarch64-unknown-unknown | FileCheck %s --check-prefixes=ALL,X64
; If the target does not have a single div/rem operation,
; -div-rem-pairs pass will decompose the remainder calculation as:
; X % Y --> X - ((X / Y) * Y)
; But if the target does have a single div/rem operation,
; the opposite transform is likely beneficial.
define i8 @scalar_i8(i8 %x, i8 %y, i8* %divdst) nounwind {
; ALL-LABEL: scalar_i8:
; ALL: // %bb.0:
; ALL-NEXT: and w8, w1, #0xff
; ALL-NEXT: and w9, w0, #0xff
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: msub w0, w8, w1, w0
; ALL-NEXT: strb w8, [x2]
; ALL-NEXT: ret
%div = udiv i8 %x, %y
store i8 %div, i8* %divdst, align 4
%t1 = mul i8 %div, %y
%t2 = sub i8 %x, %t1
ret i8 %t2
}
define i16 @scalar_i16(i16 %x, i16 %y, i16* %divdst) nounwind {
; ALL-LABEL: scalar_i16:
; ALL: // %bb.0:
; ALL-NEXT: and w8, w1, #0xffff
; ALL-NEXT: and w9, w0, #0xffff
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: msub w0, w8, w1, w0
; ALL-NEXT: strh w8, [x2]
; ALL-NEXT: ret
%div = udiv i16 %x, %y
store i16 %div, i16* %divdst, align 4
%t1 = mul i16 %div, %y
%t2 = sub i16 %x, %t1
ret i16 %t2
}
define i32 @scalar_i32(i32 %x, i32 %y, i32* %divdst) nounwind {
; ALL-LABEL: scalar_i32:
; ALL: // %bb.0:
; ALL-NEXT: udiv w8, w0, w1
; ALL-NEXT: msub w0, w8, w1, w0
; ALL-NEXT: str w8, [x2]
; ALL-NEXT: ret
%div = udiv i32 %x, %y
store i32 %div, i32* %divdst, align 4
%t1 = mul i32 %div, %y
%t2 = sub i32 %x, %t1
ret i32 %t2
}
define i64 @scalar_i64(i64 %x, i64 %y, i64* %divdst) nounwind {
; ALL-LABEL: scalar_i64:
; ALL: // %bb.0:
; ALL-NEXT: udiv x8, x0, x1
; ALL-NEXT: msub x0, x8, x1, x0
; ALL-NEXT: str x8, [x2]
; ALL-NEXT: ret
%div = udiv i64 %x, %y
store i64 %div, i64* %divdst, align 4
%t1 = mul i64 %div, %y
%t2 = sub i64 %x, %t1
ret i64 %t2
}
define <16 x i8> @vector_i128_i8(<16 x i8> %x, <16 x i8> %y, <16 x i8>* %divdst) nounwind {
; ALL-LABEL: vector_i128_i8:
; ALL: // %bb.0:
; ALL-NEXT: umov w10, v1.b[0]
; ALL-NEXT: umov w11, v0.b[0]
; ALL-NEXT: umov w8, v1.b[1]
; ALL-NEXT: umov w9, v0.b[1]
; ALL-NEXT: udiv w10, w11, w10
; ALL-NEXT: umov w12, v1.b[2]
; ALL-NEXT: umov w13, v0.b[2]
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: fmov s2, w10
; ALL-NEXT: umov w14, v1.b[3]
; ALL-NEXT: umov w15, v0.b[3]
; ALL-NEXT: udiv w12, w13, w12
; ALL-NEXT: mov v2.b[1], w8
; ALL-NEXT: umov w16, v1.b[4]
; ALL-NEXT: umov w17, v0.b[4]
; ALL-NEXT: udiv w14, w15, w14
; ALL-NEXT: mov v2.b[2], w12
; ALL-NEXT: umov w18, v1.b[5]
; ALL-NEXT: umov w1, v0.b[5]
; ALL-NEXT: udiv w16, w17, w16
; ALL-NEXT: mov v2.b[3], w14
; ALL-NEXT: umov w2, v1.b[6]
; ALL-NEXT: umov w3, v0.b[6]
; ALL-NEXT: udiv w18, w1, w18
; ALL-NEXT: mov v2.b[4], w16
; ALL-NEXT: umov w4, v1.b[7]
; ALL-NEXT: umov w5, v0.b[7]
; ALL-NEXT: udiv w2, w3, w2
; ALL-NEXT: mov v2.b[5], w18
; ALL-NEXT: umov w9, v1.b[8]
; ALL-NEXT: umov w11, v0.b[8]
; ALL-NEXT: udiv w4, w5, w4
; ALL-NEXT: mov v2.b[6], w2
; ALL-NEXT: umov w13, v1.b[9]
; ALL-NEXT: umov w15, v0.b[9]
; ALL-NEXT: udiv w9, w11, w9
; ALL-NEXT: mov v2.b[7], w4
; ALL-NEXT: umov w17, v1.b[10]
; ALL-NEXT: umov w1, v0.b[10]
; ALL-NEXT: udiv w13, w15, w13
; ALL-NEXT: mov v2.b[8], w9
; ALL-NEXT: umov w3, v1.b[11]
; ALL-NEXT: umov w5, v0.b[11]
; ALL-NEXT: udiv w17, w1, w17
; ALL-NEXT: mov v2.b[9], w13
; ALL-NEXT: umov w11, v1.b[12]
; ALL-NEXT: umov w15, v0.b[12]
; ALL-NEXT: udiv w3, w5, w3
; ALL-NEXT: mov v2.b[10], w17
; ALL-NEXT: umov w1, v1.b[13]
; ALL-NEXT: umov w5, v0.b[13]
; ALL-NEXT: udiv w11, w15, w11
; ALL-NEXT: mov v2.b[11], w3
; ALL-NEXT: umov w15, v1.b[14]
; ALL-NEXT: udiv w1, w5, w1
; ALL-NEXT: umov w5, v0.b[14]
; ALL-NEXT: mov v2.b[12], w11
; ALL-NEXT: udiv w15, w5, w15
; ALL-NEXT: umov w8, v1.b[15]
; ALL-NEXT: mov v2.b[13], w1
; ALL-NEXT: umov w9, v0.b[15]
; ALL-NEXT: mov v2.b[14], w15
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: mov v2.b[15], w8
; ALL-NEXT: mls v0.16b, v2.16b, v1.16b
; ALL-NEXT: str q2, [x0]
; ALL-NEXT: ret
%div = udiv <16 x i8> %x, %y
store <16 x i8> %div, <16 x i8>* %divdst, align 16
%t1 = mul <16 x i8> %div, %y
%t2 = sub <16 x i8> %x, %t1
ret <16 x i8> %t2
}
define <8 x i16> @vector_i128_i16(<8 x i16> %x, <8 x i16> %y, <8 x i16>* %divdst) nounwind {
; ALL-LABEL: vector_i128_i16:
; ALL: // %bb.0:
; ALL-NEXT: umov w10, v1.h[0]
; ALL-NEXT: umov w11, v0.h[0]
; ALL-NEXT: umov w8, v1.h[1]
; ALL-NEXT: umov w9, v0.h[1]
; ALL-NEXT: udiv w10, w11, w10
; ALL-NEXT: umov w12, v1.h[2]
; ALL-NEXT: umov w13, v0.h[2]
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: fmov s2, w10
; ALL-NEXT: umov w14, v1.h[3]
; ALL-NEXT: umov w15, v0.h[3]
; ALL-NEXT: udiv w12, w13, w12
; ALL-NEXT: mov v2.h[1], w8
; ALL-NEXT: umov w9, v1.h[4]
; ALL-NEXT: umov w11, v0.h[4]
; ALL-NEXT: udiv w14, w15, w14
; ALL-NEXT: mov v2.h[2], w12
; ALL-NEXT: umov w13, v1.h[5]
; ALL-NEXT: umov w15, v0.h[5]
; ALL-NEXT: udiv w9, w11, w9
; ALL-NEXT: mov v2.h[3], w14
; ALL-NEXT: umov w11, v1.h[6]
; ALL-NEXT: udiv w13, w15, w13
; ALL-NEXT: umov w15, v0.h[6]
; ALL-NEXT: mov v2.h[4], w9
; ALL-NEXT: udiv w11, w15, w11
; ALL-NEXT: umov w8, v1.h[7]
; ALL-NEXT: mov v2.h[5], w13
; ALL-NEXT: umov w9, v0.h[7]
; ALL-NEXT: mov v2.h[6], w11
; ALL-NEXT: udiv w8, w9, w8
; ALL-NEXT: mov v2.h[7], w8
; ALL-NEXT: mls v0.8h, v2.8h, v1.8h
; ALL-NEXT: str q2, [x0]
; ALL-NEXT: ret
%div = udiv <8 x i16> %x, %y
store <8 x i16> %div, <8 x i16>* %divdst, align 16
%t1 = mul <8 x i16> %div, %y
%t2 = sub <8 x i16> %x, %t1
ret <8 x i16> %t2
}
define <4 x i32> @vector_i128_i32(<4 x i32> %x, <4 x i32> %y, <4 x i32>* %divdst) nounwind {
; ALL-LABEL: vector_i128_i32:
; ALL: // %bb.0:
; ALL-NEXT: fmov w9, s1
; ALL-NEXT: fmov w10, s0
; ALL-NEXT: mov w8, v1.s[1]
; ALL-NEXT: udiv w9, w10, w9
; ALL-NEXT: mov w10, v0.s[1]
; ALL-NEXT: udiv w8, w10, w8
; ALL-NEXT: mov w10, v1.s[2]
; ALL-NEXT: fmov s2, w9
; ALL-NEXT: mov w9, v0.s[2]
; ALL-NEXT: udiv w9, w9, w10
; ALL-NEXT: mov w10, v1.s[3]
; ALL-NEXT: mov v2.s[1], w8
; ALL-NEXT: mov w8, v0.s[3]
; ALL-NEXT: mov v2.s[2], w9
; ALL-NEXT: udiv w8, w8, w10
; ALL-NEXT: mov v2.s[3], w8
; ALL-NEXT: mls v0.4s, v2.4s, v1.4s
; ALL-NEXT: str q2, [x0]
; ALL-NEXT: ret
%div = udiv <4 x i32> %x, %y
store <4 x i32> %div, <4 x i32>* %divdst, align 16
%t1 = mul <4 x i32> %div, %y
%t2 = sub <4 x i32> %x, %t1
ret <4 x i32> %t2
}
define <2 x i64> @vector_i128_i64(<2 x i64> %x, <2 x i64> %y, <2 x i64>* %divdst) nounwind {
; ALL-LABEL: vector_i128_i64:
; ALL: // %bb.0:
; ALL-NEXT: fmov x10, d1
; ALL-NEXT: fmov x11, d0
; ALL-NEXT: mov x8, v1.d[1]
; ALL-NEXT: mov x9, v0.d[1]
; ALL-NEXT: udiv x11, x11, x10
; ALL-NEXT: udiv x9, x9, x8
; ALL-NEXT: mul x10, x11, x10
; ALL-NEXT: mul x8, x9, x8
; ALL-NEXT: fmov d1, x10
; ALL-NEXT: mov v1.d[1], x8
; ALL-NEXT: sub v0.2d, v0.2d, v1.2d
; ALL-NEXT: fmov d1, x11
; ALL-NEXT: mov v1.d[1], x9
; ALL-NEXT: str q1, [x0]
; ALL-NEXT: ret
%div = udiv <2 x i64> %x, %y
store <2 x i64> %div, <2 x i64>* %divdst, align 16
%t1 = mul <2 x i64> %div, %y
%t2 = sub <2 x i64> %x, %t1
ret <2 x i64> %t2
}
; Special tests.
define i32 @scalar_i32_commutative(i32 %x, i32* %ysrc, i32* %divdst) nounwind {
; ALL-LABEL: scalar_i32_commutative:
; ALL: // %bb.0:
; ALL-NEXT: ldr w8, [x1]
; ALL-NEXT: udiv w9, w0, w8
; ALL-NEXT: msub w0, w8, w9, w0
; ALL-NEXT: str w9, [x2]
; ALL-NEXT: ret
%y = load i32, i32* %ysrc, align 4
%div = udiv i32 %x, %y
store i32 %div, i32* %divdst, align 4
%t1 = mul i32 %y, %div ; commutative
%t2 = sub i32 %x, %t1
ret i32 %t2
}
; We do not care about extra uses.
define i32 @extrause(i32 %x, i32 %y, i32* %divdst, i32* %t1dst) nounwind {
; ALL-LABEL: extrause:
; ALL: // %bb.0:
; ALL-NEXT: udiv w8, w0, w1
; ALL-NEXT: str w8, [x2]
; ALL-NEXT: mul w8, w8, w1
; ALL-NEXT: sub w0, w0, w8
; ALL-NEXT: str w8, [x3]
; ALL-NEXT: ret
%div = udiv i32 %x, %y
store i32 %div, i32* %divdst, align 4
%t1 = mul i32 %div, %y
store i32 %t1, i32* %t1dst, align 4
%t2 = sub i32 %x, %t1
ret i32 %t2
}
; 'rem' should appear next to 'div'.
define i32 @multiple_bb(i32 %x, i32 %y, i32* %divdst, i1 zeroext %store_urem, i32* %uremdst) nounwind {
; ALL-LABEL: multiple_bb:
; ALL: // %bb.0:
; ALL-NEXT: mov w8, w0
; ALL-NEXT: udiv w0, w0, w1
; ALL-NEXT: str w0, [x2]
; ALL-NEXT: cbz w3, .LBB10_2
; ALL-NEXT: // %bb.1: // %do_urem
; ALL-NEXT: msub w8, w0, w1, w8
; ALL-NEXT: str w8, [x4]
; ALL-NEXT: .LBB10_2: // %end
; ALL-NEXT: ret
%div = udiv i32 %x, %y
store i32 %div, i32* %divdst, align 4
br i1 %store_urem, label %do_urem, label %end
do_urem:
%t1 = mul i32 %div, %y
%t2 = sub i32 %x, %t1
store i32 %t2, i32* %uremdst, align 4
br label %end
end:
ret i32 %div
}
define i32 @negative_different_x(i32 %x0, i32 %x1, i32 %y, i32* %divdst) nounwind {
; ALL-LABEL: negative_different_x:
; ALL: // %bb.0:
; ALL-NEXT: udiv w8, w0, w2
; ALL-NEXT: msub w0, w8, w2, w1
; ALL-NEXT: str w8, [x3]
; ALL-NEXT: ret
%div = udiv i32 %x0, %y ; not %x1
store i32 %div, i32* %divdst, align 4
%t1 = mul i32 %div, %y
%t2 = sub i32 %x1, %t1 ; not %x0
ret i32 %t2
}
Reference in New Issue View Git Blame Copy Permalink