From 4864d89affbad659862b5fc41451ed20a7b03406 Mon Sep 17 00:00:00 2001
From: Simon Pilgrim <llvm-dev@redking.me.uk>
Date: Tue, 25 Oct 2016 21:14:11 +0000
Subject: [PATCH] [X86][SSE] Added vector srem combine tests

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@285121 91177308-0d34-0410-b5e6-96231b3b80d8
---
 test/CodeGen/X86/combine-srem.ll | 123 +++++++++++++++++++++++++++++++
 1 file changed, 123 insertions(+)
 create mode 100644 test/CodeGen/X86/combine-srem.ll

diff --git a/test/CodeGen/X86/combine-srem.ll b/test/CodeGen/X86/combine-srem.ll
new file mode 100644
index 00000000000..dc7373ce519
--- /dev/null
+++ b/test/CodeGen/X86/combine-srem.ll
@@ -0,0 +1,123 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=SSE
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=AVX
+
+; fold (srem undef, x) -> 0
+define <4 x i32> @combine_vec_srem_undef0(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_srem_undef0:
+; SSE:       # BB#0:
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_srem_undef0:
+; AVX:       # BB#0:
+; AVX-NEXT:    retq
+  %1 = srem <4 x i32> undef, %x
+  ret <4 x i32> %1
+}
+
+; fold (srem x, undef) -> undef
+define <4 x i32> @combine_vec_srem_undef1(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_srem_undef1:
+; SSE:       # BB#0:
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_srem_undef1:
+; AVX:       # BB#0:
+; AVX-NEXT:    retq
+  %1 = srem <4 x i32> %x, undef
+  ret <4 x i32> %1
+}
+
+; fold (srem x, y) -> (urem x, y) iff x and y are positive
+define <4 x i32> @combine_vec_srem_by_pos0(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_srem_by_pos0:
+; SSE:       # BB#0:
+; SSE-NEXT:    movdqa {{.*#+}} xmm1 = [255,0,0,0,255,0,0,0,255,0,0,0,255,0,0,0]
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE-NEXT:    psubd %xmm0, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_srem_by_pos0:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm1
+; AVX-NEXT:    vpbroadcastd {{.*}}(%rip), %xmm2
+; AVX-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX-NEXT:    vpsubd %xmm0, %xmm1, %xmm0
+; AVX-NEXT:    retq
+  %1 = and <4 x i32> %x, <i32 255, i32 255, i32 255, i32 255>
+  %2 = srem <4 x i32> %1, <i32 4, i32 4, i32 4, i32 4>
+  ret <4 x i32> %2
+}
+
+define <4 x i32> @combine_vec_srem_by_pos1(<4 x i32> %x) {
+; SSE-LABEL: combine_vec_srem_by_pos1:
+; SSE:       # BB#0:
+; SSE-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE-NEXT:    pextrd $3, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $28, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    andl $-16, %ecx
+; SSE-NEXT:    subl %ecx, %eax
+; SSE-NEXT:    movd %eax, %xmm1
+; SSE-NEXT:    pextrd $2, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $29, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    andl $-8, %ecx
+; SSE-NEXT:    subl %ecx, %eax
+; SSE-NEXT:    movd %eax, %xmm2
+; SSE-NEXT:    punpckldq {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; SSE-NEXT:    pshufd {{.*#+}} xmm1 = xmm2[0,1,0,1]
+; SSE-NEXT:    pextrd $1, %xmm0, %eax
+; SSE-NEXT:    movl %eax, %ecx
+; SSE-NEXT:    sarl $31, %ecx
+; SSE-NEXT:    shrl $30, %ecx
+; SSE-NEXT:    addl %eax, %ecx
+; SSE-NEXT:    andl $-4, %ecx
+; SSE-NEXT:    subl %ecx, %eax
+; SSE-NEXT:    movd %eax, %xmm0
+; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; SSE-NEXT:    pblendw {{.*#+}} xmm0 = xmm0[0,1,2,3],xmm1[4,5,6,7]
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: combine_vec_srem_by_pos1:
+; AVX:       # BB#0:
+; AVX-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX-NEXT:    vpextrd $3, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $28, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    andl $-16, %ecx
+; AVX-NEXT:    subl %ecx, %eax
+; AVX-NEXT:    vmovd %eax, %xmm1
+; AVX-NEXT:    vpextrd $2, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $29, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    andl $-8, %ecx
+; AVX-NEXT:    subl %ecx, %eax
+; AVX-NEXT:    vmovd %eax, %xmm2
+; AVX-NEXT:    vpunpckldq {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1]
+; AVX-NEXT:    vpbroadcastq %xmm1, %xmm1
+; AVX-NEXT:    vpextrd $1, %xmm0, %eax
+; AVX-NEXT:    movl %eax, %ecx
+; AVX-NEXT:    sarl $31, %ecx
+; AVX-NEXT:    shrl $30, %ecx
+; AVX-NEXT:    addl %eax, %ecx
+; AVX-NEXT:    andl $-4, %ecx
+; AVX-NEXT:    subl %ecx, %eax
+; AVX-NEXT:    vmovd %eax, %xmm0
+; AVX-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[1,0,2,3]
+; AVX-NEXT:    vpblendd {{.*#+}} xmm0 = xmm0[0,1],xmm1[2,3]
+; AVX-NEXT:    retq
+  %1 = and <4 x i32> %x, <i32 255, i32 255, i32 255, i32 255>
+  %2 = srem <4 x i32> %1, <i32 1, i32 4, i32 8, i32 16>
+  ret <4 x i32> %2
+}
-- 
2.40.0