From 881bcbe09471c9e6ae70228cf9754d568ac8e350 Mon Sep 17 00:00:00 2001
From: Sanjay Patel <spatel@rotateright.com>
Date: Fri, 29 Mar 2019 13:34:48 +0000
Subject: [PATCH] [x86] add tests for decrement+sext; NFC

llvm-svn: 357251
---
 llvm/test/CodeGen/X86/sext-i1.ll     | 19 ++++++++
 llvm/test/CodeGen/X86/vector-sext.ll | 86 ++++++++++++++++++++++++++++++++++++
 2 files changed, 105 insertions(+)

diff --git a/llvm/test/CodeGen/X86/sext-i1.ll b/llvm/test/CodeGen/X86/sext-i1.ll
index f2ce001..d14fd75 100644
--- a/llvm/test/CodeGen/X86/sext-i1.ll
+++ b/llvm/test/CodeGen/X86/sext-i1.ll
@@ -173,3 +173,22 @@ define i32 @select_0_or_1s_signext(i1 signext %cond) {
   ret i32 %sext
 }
 
+define i32 @zext_decrement_sext(i8 %x) {
+; X32-LABEL: zext_decrement_sext:
+; X32:       # %bb.0:
+; X32-NEXT:    movzbl {{[0-9]+}}(%esp), %eax
+; X32-NEXT:    decl %eax
+; X32-NEXT:    retl
+;
+; X64-LABEL: zext_decrement_sext:
+; X64:       # %bb.0:
+; X64-NEXT:    movzbl %dil, %eax
+; X64-NEXT:    decl %eax
+; X64-NEXT:    cwtl
+; X64-NEXT:    retq
+  %z = zext i8 %x to i16
+  %dec = add i16 %z, -1
+  %r = sext i16 %dec to i32
+  ret i32 %r
+}
+
diff --git a/llvm/test/CodeGen/X86/vector-sext.ll b/llvm/test/CodeGen/X86/vector-sext.ll
index f43fc99..60a58cc 100644
--- a/llvm/test/CodeGen/X86/vector-sext.ll
+++ b/llvm/test/CodeGen/X86/vector-sext.ll
@@ -6421,3 +6421,89 @@ define <8 x i32> @zext_negate_sext(<8 x i8> %x) {
   %r = sext <8 x i16> %neg to <8 x i32>
   ret <8 x i32> %r
 }
+
+define <8 x i32> @zext_decremenet_sext(<8 x i8> %x) {
+; SSE2-LABEL: zext_decremenet_sext:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT:    paddw %xmm0, %xmm1
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSE2-NEXT:    psrad $16, %xmm0
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    psrad $16, %xmm1
+; SSE2-NEXT:    retq
+;
+; SSSE3-LABEL: zext_decremenet_sext:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSSE3-NEXT:    paddw %xmm0, %xmm1
+; SSSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; SSSE3-NEXT:    psrad $16, %xmm0
+; SSSE3-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; SSSE3-NEXT:    psrad $16, %xmm1
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: zext_decremenet_sext:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    pand {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE41-NEXT:    paddw %xmm0, %xmm1
+; SSE41-NEXT:    pmovsxwd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; SSE41-NEXT:    pmovsxwd %xmm1, %xmm1
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: zext_decremenet_sext:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpmovsxwd %xmm0, %xmm1
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpmovsxwd %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: zext_decremenet_sext:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: zext_decremenet_sext:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; AVX512-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512-NEXT:    vpaddw %xmm1, %xmm0, %xmm0
+; AVX512-NEXT:    vpmovsxwd %xmm0, %ymm0
+; AVX512-NEXT:    retq
+;
+; X32-SSE2-LABEL: zext_decremenet_sext:
+; X32-SSE2:       # %bb.0:
+; X32-SSE2-NEXT:    pand {{\.LCPI.*}}, %xmm0
+; X32-SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
+; X32-SSE2-NEXT:    paddw %xmm0, %xmm1
+; X32-SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; X32-SSE2-NEXT:    psrad $16, %xmm0
+; X32-SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; X32-SSE2-NEXT:    psrad $16, %xmm1
+; X32-SSE2-NEXT:    retl
+;
+; X32-SSE41-LABEL: zext_decremenet_sext:
+; X32-SSE41:       # %bb.0:
+; X32-SSE41-NEXT:    pand {{\.LCPI.*}}, %xmm0
+; X32-SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
+; X32-SSE41-NEXT:    paddw %xmm0, %xmm1
+; X32-SSE41-NEXT:    pmovsxwd %xmm1, %xmm0
+; X32-SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,0,1]
+; X32-SSE41-NEXT:    pmovsxwd %xmm1, %xmm1
+; X32-SSE41-NEXT:    retl
+  %z = zext <8 x i8> %x to <8 x i16>
+  %dec = add <8 x i16> %z, <i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1, i16 -1>
+  %r = sext <8 x i16> %dec to <8 x i32>
+  ret <8 x i32> %r
+}
-- 
2.7.4