From 970d1e48f194693de7f7aa17aec1126c5ef7ad84 Mon Sep 17 00:00:00 2001 From: Roman Lebedev Date: Wed, 22 May 2019 14:42:41 +0000 Subject: [PATCH] [NFC][X86][AArch64] Rewrite sink-addsub-of-const.ll tests to have full permutation coverage Somehow missed some patterns initially.. While there, add comments. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@361390 91177308-0d34-0410-b5e6-96231b3b80d8 --- test/CodeGen/AArch64/sink-addsub-of-const.ll | 319 +++++++++++--- test/CodeGen/X86/sink-addsub-of-const.ll | 418 ++++++++++++++++--- 2 files changed, 607 insertions(+), 130 deletions(-) diff --git a/test/CodeGen/AArch64/sink-addsub-of-const.ll b/test/CodeGen/AArch64/sink-addsub-of-const.ll index 69d16af4e64..c571dac94b8 100644 --- a/test/CodeGen/AArch64/sink-addsub-of-const.ll +++ b/test/CodeGen/AArch64/sink-addsub-of-const.ll @@ -1,22 +1,41 @@ ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py ; RUN: llc < %s -mtriple=aarch64-unknown-unknown | FileCheck %s -; Scalar tests. Trying to avoid LEA here, so the output is actually readable.. +; Scalar tests. -define i32 @sink_add_of_const_to_add(i32 %a, i32 %b, i32 %c) { -; CHECK-LABEL: sink_add_of_const_to_add: +; add (add %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_add_of_const_to_add0(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_add_of_const_to_add0: ; CHECK: // %bb.0: ; CHECK-NEXT: add w8, w0, w1 ; CHECK-NEXT: add w8, w8, w2 ; CHECK-NEXT: add w0, w8, #32 // =32 ; CHECK-NEXT: ret %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 + %t1 = add i32 %t0, 32 ; constant always on RHS %r = add i32 %t1, %c ret i32 %r } -define i32 @sink_sub_of_const_to_add(i32 %a, i32 %b, i32 %c) { -; CHECK-LABEL: sink_sub_of_const_to_add: +define i32 @sink_add_of_const_to_add1(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_add_of_const_to_add1: +; CHECK: // %bb.0: +; CHECK-NEXT: add w8, w0, w1 +; CHECK-NEXT: add w8, w8, w2 +; CHECK-NEXT: add w0, w8, #32 // =32 +; CHECK-NEXT: ret + %t0 = add i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS + %r = add i32 %c, %t1 + ret i32 %r +} + +; add (sub %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_sub_of_const_to_add0(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_of_const_to_add0: ; CHECK: // %bb.0: ; CHECK-NEXT: add w8, w0, w1 ; CHECK-NEXT: add w8, w8, w2 @@ -27,21 +46,37 @@ define i32 @sink_sub_of_const_to_add(i32 %a, i32 %b, i32 %c) { %r = add i32 %t1, %c ret i32 %r } +define i32 @sink_sub_of_const_to_add1(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_of_const_to_add1: +; CHECK: // %bb.0: +; CHECK-NEXT: add w8, w0, w1 +; CHECK-NEXT: add w8, w8, w2 +; CHECK-NEXT: sub w0, w8, #32 // =32 +; CHECK-NEXT: ret + %t0 = add i32 %a, %b + %t1 = sub i32 %t0, 32 + %r = add i32 %c, %t1 + ret i32 %r +} -define i32 @sink_add_of_const_to_sub(i32 %a, i32 %b, i32 %c) { -; CHECK-LABEL: sink_add_of_const_to_sub: +; add (sub C, %x), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_sub_from_const_to_add0(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_from_const_to_add0: ; CHECK: // %bb.0: ; CHECK-NEXT: add w8, w0, w1 -; CHECK-NEXT: add w8, w8, #32 // =32 -; CHECK-NEXT: sub w0, w8, w2 +; CHECK-NEXT: mov w9, #32 +; CHECK-NEXT: sub w8, w9, w8 +; CHECK-NEXT: add w0, w8, w2 ; CHECK-NEXT: ret %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 - %r = sub i32 %t1, %c + %t1 = sub i32 32, %t0 + %r = add i32 %t1, %c ret i32 %r } -define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { -; CHECK-LABEL: sink_sub_of_const_to_sub2: +define i32 @sink_sub_from_const_to_add1(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_from_const_to_add1: ; CHECK: // %bb.0: ; CHECK-NEXT: add w8, w0, w1 ; CHECK-NEXT: mov w9, #32 @@ -49,129 +84,287 @@ define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; CHECK-NEXT: add w0, w2, w8 ; CHECK-NEXT: ret %t0 = add i32 %a, %b - %t1 = sub i32 %t0, 32 - %r = sub i32 %c, %t1 + %t1 = sub i32 32, %t0 + %r = add i32 %c, %t1 ret i32 %r } +; sub (add %x, C), %y +; sub %y, (add %x, C) + +define i32 @sink_add_of_const_to_sub(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_add_of_const_to_sub: +; CHECK: // %bb.0: +; CHECK-NEXT: sub w8, w0, w1 +; CHECK-NEXT: add w8, w8, #32 // =32 +; CHECK-NEXT: sub w0, w8, w2 +; CHECK-NEXT: ret + %t0 = sub i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS + %r = sub i32 %t1, %c + ret i32 %r +} define i32 @sink_add_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; CHECK-LABEL: sink_add_of_const_to_sub2: ; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, w1 +; CHECK-NEXT: sub w8, w0, w1 ; CHECK-NEXT: add w8, w8, #32 // =32 ; CHECK-NEXT: sub w0, w2, w8 ; CHECK-NEXT: ret - %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 + %t0 = sub i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS %r = sub i32 %c, %t1 ret i32 %r } + +; sub (sub %x, C), %y +; sub %y, (sub %x, C) + define i32 @sink_sub_of_const_to_sub(i32 %a, i32 %b, i32 %c) { ; CHECK-LABEL: sink_sub_of_const_to_sub: ; CHECK: // %bb.0: -; CHECK-NEXT: add w8, w0, w1 +; CHECK-NEXT: sub w8, w0, w1 ; CHECK-NEXT: sub w8, w8, #32 // =32 ; CHECK-NEXT: sub w0, w8, w2 ; CHECK-NEXT: ret - %t0 = add i32 %a, %b + %t0 = sub i32 %a, %b %t1 = sub i32 %t0, 32 %r = sub i32 %t1, %c ret i32 %r } +define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_of_const_to_sub2: +; CHECK: // %bb.0: +; CHECK-NEXT: sub w8, w1, w0 +; CHECK-NEXT: add w8, w8, w2 +; CHECK-NEXT: add w0, w8, #32 // =32 +; CHECK-NEXT: ret + %t0 = sub i32 %a, %b + %t1 = sub i32 %t0, 32 + %r = sub i32 %c, %t1 + ret i32 %r +} +; sub (sub C, %x), %y +; sub %y, (sub C, %x) + +define i32 @sink_sub_from_const_to_sub(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_from_const_to_sub: +; CHECK: // %bb.0: +; CHECK-NEXT: sub w8, w1, w0 +; CHECK-NEXT: add w8, w8, #32 // =32 +; CHECK-NEXT: sub w0, w8, w2 +; CHECK-NEXT: ret + %t0 = sub i32 %a, %b + %t1 = sub i32 32, %t0 + %r = sub i32 %t1, %c + ret i32 %r +} +define i32 @sink_sub_from_const_to_sub2(i32 %a, i32 %b, i32 %c) { +; CHECK-LABEL: sink_sub_from_const_to_sub2: +; CHECK: // %bb.0: +; CHECK-NEXT: sub w8, w0, w1 +; CHECK-NEXT: add w8, w8, w2 +; CHECK-NEXT: sub w0, w8, #32 // =32 +; CHECK-NEXT: ret + %t0 = sub i32 %a, %b + %t1 = sub i32 32, %t0 + %r = sub i32 %c, %t1 + ret i32 %r +} + +;------------------------------------------------------------------------------; ; Basic vector tests. Here it is easier to see where the constant operand is. +;------------------------------------------------------------------------------; + +; add (add %x, C), %y +; Outer 'add' is commutative - 2 variants. -define <4 x i32> @vec_sink_add_of_const_to_add(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; CHECK-LABEL: vec_sink_add_of_const_to_add: +define <4 x i32> @vec_sink_add_of_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_add_of_const_to_add0: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI6_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI6_0] +; CHECK-NEXT: adrp x8, .LCPI12_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI12_0] ; CHECK-NEXT: add v0.4s, v0.4s, v1.4s ; CHECK-NEXT: add v0.4s, v0.4s, v2.4s ; CHECK-NEXT: add v0.4s, v0.4s, v3.4s ; CHECK-NEXT: ret %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, + %t1 = add <4 x i32> %t0, ; constant always on RHS %r = add <4 x i32> %t1, %c ret <4 x i32> %r } -define <4 x i32> @vec_sink_sub_of_const_to_add(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; CHECK-LABEL: vec_sink_sub_of_const_to_add: +define <4 x i32> @vec_sink_add_of_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_add_of_const_to_add1: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI7_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI7_0] +; CHECK-NEXT: adrp x8, .LCPI13_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI13_0] +; CHECK-NEXT: add v0.4s, v0.4s, v1.4s +; CHECK-NEXT: add v0.4s, v0.4s, v2.4s +; CHECK-NEXT: add v0.4s, v0.4s, v3.4s +; CHECK-NEXT: ret + %t0 = add <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS + %r = add <4 x i32> %c, %t1 + ret <4 x i32> %r +} + +; add (sub %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define <4 x i32> @vec_sink_sub_of_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_of_const_to_add0: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI14_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI14_0] ; CHECK-NEXT: add v0.4s, v0.4s, v1.4s ; CHECK-NEXT: sub v0.4s, v0.4s, v3.4s ; CHECK-NEXT: add v0.4s, v0.4s, v2.4s ; CHECK-NEXT: ret %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, + %t1 = sub <4 x i32> %t0, %r = add <4 x i32> %t1, %c ret <4 x i32> %r } +define <4 x i32> @vec_sink_sub_of_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_of_const_to_add1: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI15_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI15_0] +; CHECK-NEXT: add v0.4s, v0.4s, v1.4s +; CHECK-NEXT: sub v0.4s, v0.4s, v3.4s +; CHECK-NEXT: add v0.4s, v2.4s, v0.4s +; CHECK-NEXT: ret + %t0 = add <4 x i32> %a, %b + %t1 = sub <4 x i32> %t0, + %r = add <4 x i32> %c, %t1 + ret <4 x i32> %r +} -define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; CHECK-LABEL: vec_sink_add_of_const_to_sub: +; add (sub C, %x), %y +; Outer 'add' is commutative - 2 variants. + +define <4 x i32> @vec_sink_sub_from_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_from_const_to_add0: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI8_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI8_0] +; CHECK-NEXT: adrp x8, .LCPI16_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI16_0] ; CHECK-NEXT: add v0.4s, v0.4s, v1.4s -; CHECK-NEXT: add v0.4s, v0.4s, v3.4s -; CHECK-NEXT: sub v0.4s, v0.4s, v2.4s +; CHECK-NEXT: sub v0.4s, v3.4s, v0.4s +; CHECK-NEXT: add v0.4s, v0.4s, v2.4s ; CHECK-NEXT: ret %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, - %r = sub <4 x i32> %t1, %c + %t1 = sub <4 x i32> , %t0 + %r = add <4 x i32> %t1, %c ret <4 x i32> %r } -define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; ALL-LABEL: vec_sink_sub_of_const_to_sub2: -; ALL: # %bb.0: -; ALL-NEXT: paddd %xmm1, %xmm0 -; ALL-NEXT: movdqa {{.*#+}} xmm1 = <93,u,45,81> -; ALL-NEXT: psubd %xmm0, %xmm1 -; ALL-NEXT: paddd %xmm2, %xmm1 -; ALL-NEXT: movdqa %xmm1, %xmm0 -; ALL-NEXT: ret{{[l|q]}} -; CHECK-LABEL: vec_sink_sub_of_const_to_sub2: +define <4 x i32> @vec_sink_sub_from_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_from_const_to_add1: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI9_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI9_0] +; CHECK-NEXT: adrp x8, .LCPI17_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI17_0] ; CHECK-NEXT: add v0.4s, v0.4s, v1.4s ; CHECK-NEXT: sub v0.4s, v3.4s, v0.4s ; CHECK-NEXT: add v0.4s, v2.4s, v0.4s ; CHECK-NEXT: ret %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, - %r = sub <4 x i32> %c, %t1 + %t1 = sub <4 x i32> , %t0 + %r = add <4 x i32> %c, %t1 ret <4 x i32> %r } +; sub (add %x, C), %y +; sub %y, (add %x, C) + +define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_add_of_const_to_sub: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI18_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI18_0] +; CHECK-NEXT: sub v0.4s, v0.4s, v1.4s +; CHECK-NEXT: add v0.4s, v0.4s, v3.4s +; CHECK-NEXT: sub v0.4s, v0.4s, v2.4s +; CHECK-NEXT: ret + %t0 = sub <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS + %r = sub <4 x i32> %t1, %c + ret <4 x i32> %r +} define <4 x i32> @vec_sink_add_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { ; CHECK-LABEL: vec_sink_add_of_const_to_sub2: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI10_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI10_0] -; CHECK-NEXT: add v0.4s, v0.4s, v1.4s +; CHECK-NEXT: adrp x8, .LCPI19_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI19_0] +; CHECK-NEXT: sub v0.4s, v0.4s, v1.4s ; CHECK-NEXT: add v0.4s, v0.4s, v3.4s ; CHECK-NEXT: sub v0.4s, v2.4s, v0.4s ; CHECK-NEXT: ret - %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, + %t0 = sub <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS %r = sub <4 x i32> %c, %t1 ret <4 x i32> %r } + +; sub (sub %x, C), %y +; sub %y, (sub %x, C) + define <4 x i32> @vec_sink_sub_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { ; CHECK-LABEL: vec_sink_sub_of_const_to_sub: ; CHECK: // %bb.0: -; CHECK-NEXT: adrp x8, .LCPI11_0 -; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI11_0] -; CHECK-NEXT: add v0.4s, v0.4s, v1.4s +; CHECK-NEXT: adrp x8, .LCPI20_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI20_0] +; CHECK-NEXT: sub v0.4s, v0.4s, v1.4s ; CHECK-NEXT: sub v0.4s, v0.4s, v3.4s ; CHECK-NEXT: sub v0.4s, v0.4s, v2.4s ; CHECK-NEXT: ret - %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> %t0, + %r = sub <4 x i32> %t1, %c + ret <4 x i32> %r +} +define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_of_const_to_sub2: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI21_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI21_0] +; CHECK-NEXT: sub v0.4s, v1.4s, v0.4s +; CHECK-NEXT: add v0.4s, v0.4s, v2.4s +; CHECK-NEXT: add v0.4s, v0.4s, v3.4s +; CHECK-NEXT: ret + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> %t0, + %r = sub <4 x i32> %c, %t1 + ret <4 x i32> %r +} + +; sub (sub C, %x), %y +; sub %y, (sub C, %x) + +define <4 x i32> @vec_sink_sub_from_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_from_const_to_sub: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI22_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI22_0] +; CHECK-NEXT: sub v0.4s, v1.4s, v0.4s +; CHECK-NEXT: add v0.4s, v0.4s, v3.4s +; CHECK-NEXT: sub v0.4s, v0.4s, v2.4s +; CHECK-NEXT: ret + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> , %t0 %r = sub <4 x i32> %t1, %c ret <4 x i32> %r } +define <4 x i32> @vec_sink_sub_from_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; CHECK-LABEL: vec_sink_sub_from_const_to_sub2: +; CHECK: // %bb.0: +; CHECK-NEXT: adrp x8, .LCPI23_0 +; CHECK-NEXT: ldr q3, [x8, :lo12:.LCPI23_0] +; CHECK-NEXT: sub v0.4s, v0.4s, v1.4s +; CHECK-NEXT: sub v0.4s, v0.4s, v3.4s +; CHECK-NEXT: add v0.4s, v2.4s, v0.4s +; CHECK-NEXT: ret + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> , %t0 + %r = sub <4 x i32> %c, %t1 + ret <4 x i32> %r +} diff --git a/test/CodeGen/X86/sink-addsub-of-const.ll b/test/CodeGen/X86/sink-addsub-of-const.ll index a74d2e63bab..2ffbfcb56b2 100644 --- a/test/CodeGen/X86/sink-addsub-of-const.ll +++ b/test/CodeGen/X86/sink-addsub-of-const.ll @@ -4,8 +4,11 @@ ; Scalar tests. Trying to avoid LEA here, so the output is actually readable.. -define i32 @sink_add_of_const_to_add(i32 %a, i32 %b, i32 %c) { -; X32-LABEL: sink_add_of_const_to_add: +; add (add %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_add_of_const_to_add0(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_add_of_const_to_add0: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx @@ -14,7 +17,7 @@ define i32 @sink_add_of_const_to_add(i32 %a, i32 %b, i32 %c) { ; X32-NEXT: addl $32, %eax ; X32-NEXT: retl ; -; X64-LABEL: sink_add_of_const_to_add: +; X64-LABEL: sink_add_of_const_to_add0: ; X64: # %bb.0: ; X64-NEXT: # kill: def $edx killed $edx def $rdx ; X64-NEXT: # kill: def $edi killed $edi def $rdi @@ -22,12 +25,38 @@ define i32 @sink_add_of_const_to_add(i32 %a, i32 %b, i32 %c) { ; X64-NEXT: leal 32(%rdx,%rdi), %eax ; X64-NEXT: retq %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 + %t1 = add i32 %t0, 32 ; constant always on RHS %r = add i32 %t1, %c ret i32 %r } -define i32 @sink_sub_of_const_to_add(i32 %a, i32 %b, i32 %c) { -; X32-LABEL: sink_sub_of_const_to_add: +define i32 @sink_add_of_const_to_add1(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_add_of_const_to_add1: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %eax +; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl %ecx, %eax +; X32-NEXT: addl $32, %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_add_of_const_to_add1: +; X64: # %bb.0: +; X64-NEXT: # kill: def $edx killed $edx def $rdx +; X64-NEXT: # kill: def $edi killed $edi def $rdi +; X64-NEXT: addl %esi, %edi +; X64-NEXT: leal 32(%rdx,%rdi), %eax +; X64-NEXT: retq + %t0 = add i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS + %r = add i32 %c, %t1 + ret i32 %r +} + +; add (sub %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_sub_of_const_to_add0(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_of_const_to_add0: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx @@ -36,7 +65,7 @@ define i32 @sink_sub_of_const_to_add(i32 %a, i32 %b, i32 %c) { ; X32-NEXT: addl $-32, %eax ; X32-NEXT: retl ; -; X64-LABEL: sink_sub_of_const_to_add: +; X64-LABEL: sink_sub_of_const_to_add0: ; X64: # %bb.0: ; X64-NEXT: # kill: def $edx killed $edx def $rdx ; X64-NEXT: # kill: def $edi killed $edi def $rdi @@ -48,31 +77,34 @@ define i32 @sink_sub_of_const_to_add(i32 %a, i32 %b, i32 %c) { %r = add i32 %t1, %c ret i32 %r } - -define i32 @sink_add_of_const_to_sub(i32 %a, i32 %b, i32 %c) { -; X32-LABEL: sink_add_of_const_to_sub: +define i32 @sink_sub_of_const_to_add1(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_of_const_to_add1: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl {{[0-9]+}}(%esp), %ecx ; X32-NEXT: addl %ecx, %eax -; X32-NEXT: addl $32, %eax -; X32-NEXT: subl {{[0-9]+}}(%esp), %eax +; X32-NEXT: addl $-32, %eax ; X32-NEXT: retl ; -; X64-LABEL: sink_add_of_const_to_sub: +; X64-LABEL: sink_sub_of_const_to_add1: ; X64: # %bb.0: -; X64-NEXT: # kill: def $esi killed $esi def $rsi +; X64-NEXT: # kill: def $edx killed $edx def $rdx ; X64-NEXT: # kill: def $edi killed $edi def $rdi -; X64-NEXT: leal 32(%rdi,%rsi), %eax -; X64-NEXT: subl %edx, %eax +; X64-NEXT: addl %esi, %edi +; X64-NEXT: leal -32(%rdx,%rdi), %eax ; X64-NEXT: retq %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 - %r = sub i32 %t1, %c + %t1 = sub i32 %t0, 32 + %r = add i32 %c, %t1 ret i32 %r } -define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { -; X32-LABEL: sink_sub_of_const_to_sub2: + +; add (sub C, %x), %y +; Outer 'add' is commutative - 2 variants. + +define i32 @sink_sub_from_const_to_add0(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_from_const_to_add0: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx ; X32-NEXT: addl {{[0-9]+}}(%esp), %ecx @@ -81,7 +113,7 @@ define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; X32-NEXT: addl {{[0-9]+}}(%esp), %eax ; X32-NEXT: retl ; -; X64-LABEL: sink_sub_of_const_to_sub2: +; X64-LABEL: sink_sub_from_const_to_add0: ; X64: # %bb.0: ; X64-NEXT: addl %esi, %edi ; X64-NEXT: movl $32, %eax @@ -89,18 +121,63 @@ define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; X64-NEXT: addl %edx, %eax ; X64-NEXT: retq %t0 = add i32 %a, %b - %t1 = sub i32 %t0, 32 - %r = sub i32 %c, %t1 + %t1 = sub i32 32, %t0 + %r = add i32 %t1, %c + ret i32 %r +} +define i32 @sink_sub_from_const_to_add1(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_from_const_to_add1: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: movl $32, %eax +; X32-NEXT: subl %ecx, %eax +; X32-NEXT: addl {{[0-9]+}}(%esp), %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_sub_from_const_to_add1: +; X64: # %bb.0: +; X64-NEXT: addl %esi, %edi +; X64-NEXT: movl $32, %eax +; X64-NEXT: subl %edi, %eax +; X64-NEXT: addl %edx, %eax +; X64-NEXT: retq + %t0 = add i32 %a, %b + %t1 = sub i32 32, %t0 + %r = add i32 %c, %t1 ret i32 %r } +; sub (add %x, C), %y +; sub %y, (add %x, C) + +define i32 @sink_add_of_const_to_sub(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_add_of_const_to_sub: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %eax +; X32-NEXT: subl {{[0-9]+}}(%esp), %eax +; X32-NEXT: addl $32, %eax +; X32-NEXT: subl {{[0-9]+}}(%esp), %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_add_of_const_to_sub: +; X64: # %bb.0: +; X64-NEXT: # kill: def $edi killed $edi def $rdi +; X64-NEXT: subl %esi, %edi +; X64-NEXT: leal 32(%rdi), %eax +; X64-NEXT: subl %edx, %eax +; X64-NEXT: retq + %t0 = sub i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS + %r = sub i32 %t1, %c + ret i32 %r +} define i32 @sink_add_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; X32-LABEL: sink_add_of_const_to_sub2: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax ; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx -; X32-NEXT: movl {{[0-9]+}}(%esp), %edx -; X32-NEXT: addl %edx, %ecx +; X32-NEXT: subl {{[0-9]+}}(%esp), %ecx ; X32-NEXT: addl $32, %ecx ; X32-NEXT: subl %ecx, %eax ; X32-NEXT: retl @@ -108,118 +185,256 @@ define i32 @sink_add_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { ; X64-LABEL: sink_add_of_const_to_sub2: ; X64: # %bb.0: ; X64-NEXT: movl %edx, %eax -; X64-NEXT: # kill: def $esi killed $esi def $rsi -; X64-NEXT: # kill: def $edi killed $edi def $rdi -; X64-NEXT: leal 32(%rdi,%rsi), %ecx -; X64-NEXT: subl %ecx, %eax +; X64-NEXT: subl %esi, %edi +; X64-NEXT: addl $32, %edi +; X64-NEXT: subl %edi, %eax ; X64-NEXT: retq - %t0 = add i32 %a, %b - %t1 = add i32 %t0, 32 + %t0 = sub i32 %a, %b + %t1 = add i32 %t0, 32 ; constant always on RHS %r = sub i32 %c, %t1 ret i32 %r } + +; sub (sub %x, C), %y +; sub %y, (sub %x, C) + define i32 @sink_sub_of_const_to_sub(i32 %a, i32 %b, i32 %c) { ; X32-LABEL: sink_sub_of_const_to_sub: ; X32: # %bb.0: ; X32-NEXT: movl {{[0-9]+}}(%esp), %eax -; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx -; X32-NEXT: addl %ecx, %eax +; X32-NEXT: subl {{[0-9]+}}(%esp), %eax ; X32-NEXT: addl $-32, %eax ; X32-NEXT: subl {{[0-9]+}}(%esp), %eax ; X32-NEXT: retl ; ; X64-LABEL: sink_sub_of_const_to_sub: ; X64: # %bb.0: -; X64-NEXT: # kill: def $esi killed $esi def $rsi ; X64-NEXT: # kill: def $edi killed $edi def $rdi -; X64-NEXT: leal -32(%rdi,%rsi), %eax +; X64-NEXT: subl %esi, %edi +; X64-NEXT: leal -32(%rdi), %eax ; X64-NEXT: subl %edx, %eax ; X64-NEXT: retq - %t0 = add i32 %a, %b + %t0 = sub i32 %a, %b %t1 = sub i32 %t0, 32 %r = sub i32 %t1, %c ret i32 %r } +define i32 @sink_sub_of_const_to_sub2(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_of_const_to_sub2: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %eax +; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: subl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl %ecx, %eax +; X32-NEXT: addl $32, %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_sub_of_const_to_sub2: +; X64: # %bb.0: +; X64-NEXT: # kill: def $edx killed $edx def $rdx +; X64-NEXT: # kill: def $esi killed $esi def $rsi +; X64-NEXT: subl %edi, %esi +; X64-NEXT: leal 32(%rsi,%rdx), %eax +; X64-NEXT: retq + %t0 = sub i32 %a, %b + %t1 = sub i32 %t0, 32 + %r = sub i32 %c, %t1 + ret i32 %r +} + +; sub (sub C, %x), %y +; sub %y, (sub C, %x) + +define i32 @sink_sub_from_const_to_sub(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_from_const_to_sub: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %eax +; X32-NEXT: subl {{[0-9]+}}(%esp), %eax +; X32-NEXT: addl $32, %eax +; X32-NEXT: subl {{[0-9]+}}(%esp), %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_sub_from_const_to_sub: +; X64: # %bb.0: +; X64-NEXT: # kill: def $esi killed $esi def $rsi +; X64-NEXT: subl %edi, %esi +; X64-NEXT: leal 32(%rsi), %eax +; X64-NEXT: subl %edx, %eax +; X64-NEXT: retq + %t0 = sub i32 %a, %b + %t1 = sub i32 32, %t0 + %r = sub i32 %t1, %c + ret i32 %r +} +define i32 @sink_sub_from_const_to_sub2(i32 %a, i32 %b, i32 %c) { +; X32-LABEL: sink_sub_from_const_to_sub2: +; X32: # %bb.0: +; X32-NEXT: movl {{[0-9]+}}(%esp), %eax +; X32-NEXT: movl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: subl {{[0-9]+}}(%esp), %ecx +; X32-NEXT: addl %ecx, %eax +; X32-NEXT: addl $-32, %eax +; X32-NEXT: retl +; +; X64-LABEL: sink_sub_from_const_to_sub2: +; X64: # %bb.0: +; X64-NEXT: # kill: def $edx killed $edx def $rdx +; X64-NEXT: # kill: def $edi killed $edi def $rdi +; X64-NEXT: subl %esi, %edi +; X64-NEXT: leal -32(%rdi,%rdx), %eax +; X64-NEXT: retq + %t0 = sub i32 %a, %b + %t1 = sub i32 32, %t0 + %r = sub i32 %c, %t1 + ret i32 %r +} +;------------------------------------------------------------------------------; ; Basic vector tests. Here it is easier to see where the constant operand is. +;------------------------------------------------------------------------------; + +; add (add %x, C), %y +; Outer 'add' is commutative - 2 variants. -define <4 x i32> @vec_sink_add_of_const_to_add(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; X32-LABEL: vec_sink_add_of_const_to_add: +define <4 x i32> @vec_sink_add_of_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_add_of_const_to_add0: ; X32: # %bb.0: ; X32-NEXT: paddd %xmm2, %xmm1 ; X32-NEXT: paddd %xmm1, %xmm0 ; X32-NEXT: paddd {{\.LCPI.*}}, %xmm0 ; X32-NEXT: retl ; -; X64-LABEL: vec_sink_add_of_const_to_add: +; X64-LABEL: vec_sink_add_of_const_to_add0: ; X64: # %bb.0: ; X64-NEXT: paddd %xmm2, %xmm1 ; X64-NEXT: paddd %xmm1, %xmm0 ; X64-NEXT: paddd {{.*}}(%rip), %xmm0 ; X64-NEXT: retq %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, + %t1 = add <4 x i32> %t0, ; constant always on RHS %r = add <4 x i32> %t1, %c ret <4 x i32> %r } -define <4 x i32> @vec_sink_sub_of_const_to_add(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; X32-LABEL: vec_sink_sub_of_const_to_add: +define <4 x i32> @vec_sink_add_of_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_add_of_const_to_add1: +; X32: # %bb.0: +; X32-NEXT: paddd %xmm2, %xmm1 +; X32-NEXT: paddd %xmm1, %xmm0 +; X32-NEXT: paddd {{\.LCPI.*}}, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: vec_sink_add_of_const_to_add1: +; X64: # %bb.0: +; X64-NEXT: paddd %xmm2, %xmm1 +; X64-NEXT: paddd %xmm1, %xmm0 +; X64-NEXT: paddd {{.*}}(%rip), %xmm0 +; X64-NEXT: retq + %t0 = add <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS + %r = add <4 x i32> %c, %t1 + ret <4 x i32> %r +} + +; add (sub %x, C), %y +; Outer 'add' is commutative - 2 variants. + +define <4 x i32> @vec_sink_sub_of_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_sub_of_const_to_add0: ; X32: # %bb.0: ; X32-NEXT: paddd %xmm1, %xmm0 ; X32-NEXT: psubd {{\.LCPI.*}}, %xmm0 ; X32-NEXT: paddd %xmm2, %xmm0 ; X32-NEXT: retl ; -; X64-LABEL: vec_sink_sub_of_const_to_add: +; X64-LABEL: vec_sink_sub_of_const_to_add0: ; X64: # %bb.0: ; X64-NEXT: paddd %xmm1, %xmm0 ; X64-NEXT: psubd {{.*}}(%rip), %xmm0 ; X64-NEXT: paddd %xmm2, %xmm0 ; X64-NEXT: retq %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, + %t1 = sub <4 x i32> %t0, %r = add <4 x i32> %t1, %c ret <4 x i32> %r } - -define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; X32-LABEL: vec_sink_add_of_const_to_sub: +define <4 x i32> @vec_sink_sub_of_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_sub_of_const_to_add1: ; X32: # %bb.0: ; X32-NEXT: paddd %xmm1, %xmm0 -; X32-NEXT: paddd {{\.LCPI.*}}, %xmm0 -; X32-NEXT: psubd %xmm2, %xmm0 +; X32-NEXT: psubd {{\.LCPI.*}}, %xmm0 +; X32-NEXT: paddd %xmm2, %xmm0 ; X32-NEXT: retl ; -; X64-LABEL: vec_sink_add_of_const_to_sub: +; X64-LABEL: vec_sink_sub_of_const_to_add1: ; X64: # %bb.0: ; X64-NEXT: paddd %xmm1, %xmm0 -; X64-NEXT: paddd {{.*}}(%rip), %xmm0 -; X64-NEXT: psubd %xmm2, %xmm0 +; X64-NEXT: psubd {{.*}}(%rip), %xmm0 +; X64-NEXT: paddd %xmm2, %xmm0 ; X64-NEXT: retq %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, - %r = sub <4 x i32> %t1, %c + %t1 = sub <4 x i32> %t0, + %r = add <4 x i32> %c, %t1 ret <4 x i32> %r } -define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { -; ALL-LABEL: vec_sink_sub_of_const_to_sub2: + +; add (sub C, %x), %y +; Outer 'add' is commutative - 2 variants. + +define <4 x i32> @vec_sink_sub_from_const_to_add0(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; ALL-LABEL: vec_sink_sub_from_const_to_add0: ; ALL: # %bb.0: ; ALL-NEXT: paddd %xmm1, %xmm0 -; ALL-NEXT: movdqa {{.*#+}} xmm1 = <93,u,45,81> +; ALL-NEXT: movdqa {{.*#+}} xmm1 = <42,24,u,46> ; ALL-NEXT: psubd %xmm0, %xmm1 ; ALL-NEXT: paddd %xmm2, %xmm1 ; ALL-NEXT: movdqa %xmm1, %xmm0 ; ALL-NEXT: ret{{[l|q]}} %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, - %r = sub <4 x i32> %c, %t1 + %t1 = sub <4 x i32> , %t0 + %r = add <4 x i32> %t1, %c + ret <4 x i32> %r +} +define <4 x i32> @vec_sink_sub_from_const_to_add1(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; ALL-LABEL: vec_sink_sub_from_const_to_add1: +; ALL: # %bb.0: +; ALL-NEXT: paddd %xmm1, %xmm0 +; ALL-NEXT: movdqa {{.*#+}} xmm1 = <42,24,u,46> +; ALL-NEXT: psubd %xmm0, %xmm1 +; ALL-NEXT: paddd %xmm2, %xmm1 +; ALL-NEXT: movdqa %xmm1, %xmm0 +; ALL-NEXT: ret{{[l|q]}} + %t0 = add <4 x i32> %a, %b + %t1 = sub <4 x i32> , %t0 + %r = add <4 x i32> %c, %t1 ret <4 x i32> %r } +; sub (add %x, C), %y +; sub %y, (add %x, C) + +define <4 x i32> @vec_sink_add_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_add_of_const_to_sub: +; X32: # %bb.0: +; X32-NEXT: psubd %xmm1, %xmm0 +; X32-NEXT: paddd {{\.LCPI.*}}, %xmm0 +; X32-NEXT: psubd %xmm2, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: vec_sink_add_of_const_to_sub: +; X64: # %bb.0: +; X64-NEXT: psubd %xmm1, %xmm0 +; X64-NEXT: paddd {{.*}}(%rip), %xmm0 +; X64-NEXT: psubd %xmm2, %xmm0 +; X64-NEXT: retq + %t0 = sub <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS + %r = sub <4 x i32> %t1, %c + ret <4 x i32> %r +} define <4 x i32> @vec_sink_add_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { ; X32-LABEL: vec_sink_add_of_const_to_sub2: ; X32: # %bb.0: -; X32-NEXT: paddd %xmm1, %xmm0 +; X32-NEXT: psubd %xmm1, %xmm0 ; X32-NEXT: paddd {{\.LCPI.*}}, %xmm0 ; X32-NEXT: psubd %xmm0, %xmm2 ; X32-NEXT: movdqa %xmm2, %xmm0 @@ -227,32 +442,101 @@ define <4 x i32> @vec_sink_add_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x ; ; X64-LABEL: vec_sink_add_of_const_to_sub2: ; X64: # %bb.0: -; X64-NEXT: paddd %xmm1, %xmm0 +; X64-NEXT: psubd %xmm1, %xmm0 ; X64-NEXT: paddd {{.*}}(%rip), %xmm0 ; X64-NEXT: psubd %xmm0, %xmm2 ; X64-NEXT: movdqa %xmm2, %xmm0 ; X64-NEXT: retq - %t0 = add <4 x i32> %a, %b - %t1 = add <4 x i32> %t0, + %t0 = sub <4 x i32> %a, %b + %t1 = add <4 x i32> %t0, ; constant always on RHS %r = sub <4 x i32> %c, %t1 ret <4 x i32> %r } + +; sub (sub %x, C), %y +; sub %y, (sub %x, C) + define <4 x i32> @vec_sink_sub_of_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { ; X32-LABEL: vec_sink_sub_of_const_to_sub: ; X32: # %bb.0: -; X32-NEXT: paddd %xmm1, %xmm0 +; X32-NEXT: psubd %xmm1, %xmm0 ; X32-NEXT: psubd {{\.LCPI.*}}, %xmm0 ; X32-NEXT: psubd %xmm2, %xmm0 ; X32-NEXT: retl ; ; X64-LABEL: vec_sink_sub_of_const_to_sub: ; X64: # %bb.0: -; X64-NEXT: paddd %xmm1, %xmm0 +; X64-NEXT: psubd %xmm1, %xmm0 ; X64-NEXT: psubd {{.*}}(%rip), %xmm0 ; X64-NEXT: psubd %xmm2, %xmm0 ; X64-NEXT: retq - %t0 = add <4 x i32> %a, %b - %t1 = sub <4 x i32> %t0, + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> %t0, %r = sub <4 x i32> %t1, %c ret <4 x i32> %r } +define <4 x i32> @vec_sink_sub_of_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_sub_of_const_to_sub2: +; X32: # %bb.0: +; X32-NEXT: psubd %xmm0, %xmm1 +; X32-NEXT: paddd %xmm2, %xmm1 +; X32-NEXT: paddd {{\.LCPI.*}}, %xmm1 +; X32-NEXT: movdqa %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: vec_sink_sub_of_const_to_sub2: +; X64: # %bb.0: +; X64-NEXT: psubd %xmm0, %xmm1 +; X64-NEXT: paddd %xmm2, %xmm1 +; X64-NEXT: paddd {{.*}}(%rip), %xmm1 +; X64-NEXT: movdqa %xmm1, %xmm0 +; X64-NEXT: retq + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> %t0, + %r = sub <4 x i32> %c, %t1 + ret <4 x i32> %r +} + +; sub (sub C, %x), %y +; sub %y, (sub C, %x) + +define <4 x i32> @vec_sink_sub_from_const_to_sub(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_sub_from_const_to_sub: +; X32: # %bb.0: +; X32-NEXT: psubd %xmm0, %xmm1 +; X32-NEXT: paddd {{\.LCPI.*}}, %xmm1 +; X32-NEXT: psubd %xmm2, %xmm1 +; X32-NEXT: movdqa %xmm1, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: vec_sink_sub_from_const_to_sub: +; X64: # %bb.0: +; X64-NEXT: psubd %xmm0, %xmm1 +; X64-NEXT: paddd {{.*}}(%rip), %xmm1 +; X64-NEXT: psubd %xmm2, %xmm1 +; X64-NEXT: movdqa %xmm1, %xmm0 +; X64-NEXT: retq + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> , %t0 + %r = sub <4 x i32> %t1, %c + ret <4 x i32> %r +} +define <4 x i32> @vec_sink_sub_from_const_to_sub2(<4 x i32> %a, <4 x i32> %b, <4 x i32> %c) { +; X32-LABEL: vec_sink_sub_from_const_to_sub2: +; X32: # %bb.0: +; X32-NEXT: psubd %xmm1, %xmm0 +; X32-NEXT: psubd {{\.LCPI.*}}, %xmm0 +; X32-NEXT: paddd %xmm2, %xmm0 +; X32-NEXT: retl +; +; X64-LABEL: vec_sink_sub_from_const_to_sub2: +; X64: # %bb.0: +; X64-NEXT: psubd %xmm1, %xmm0 +; X64-NEXT: psubd {{.*}}(%rip), %xmm0 +; X64-NEXT: paddd %xmm2, %xmm0 +; X64-NEXT: retq + %t0 = sub <4 x i32> %a, %b + %t1 = sub <4 x i32> , %t0 + %r = sub <4 x i32> %c, %t1 + ret <4 x i32> %r +} -- 2.50.1