From 52f1b6dbede268c5cf2ffffb0cc0facd7dbb44f1 Mon Sep 17 00:00:00 2001
From: Chandler Carruth <chandlerc@gmail.com>
Date: Sun, 15 Feb 2015 10:34:52 +0000
Subject: [PATCH] [x86] Stop shuffling zero vectors. =]

I was somewhat surprised this pattern really came up, but it does. It
seems better to just directly handle it than try to special case every
place where we end up forming a shuffle that devolves to a shuffle of
a zero vector.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@229301 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp         | 7 +++++++
 test/CodeGen/X86/vector-shuffle-256-v16.ll | 7 +++----
 test/CodeGen/X86/vector-shuffle-256-v32.ll | 7 +++----
 3 files changed, 13 insertions(+), 8 deletions(-)

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index e1e03c2e5ea..b047e1684d2 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -11285,6 +11285,13 @@ static SDValue lowerVectorShuffle(SDValue Op, const X86Subtarget *Subtarget,
         return DAG.getVectorShuffle(VT, dl, V1, V2, NewMask);
       }
 
+  // We actually see shuffles that are entirely re-arrangements of a set of
+  // zero inputs. This mostly happens while decomposing complex shuffles into
+  // simple ones. Directly lower these as a buildvector of zeros.
+  SmallBitVector Zeroable = computeZeroableShuffleElements(Mask, V1, V2);
+  if (Zeroable.all())
+    return getZeroVector(VT, Subtarget, DAG, dl);
+
   // Try to collapse shuffles into using a vector type with fewer elements but
   // wider element types. We cap this to not form integers or floating point
   // elements wider than 64 bits, but it might be interesting to form i128
diff --git a/test/CodeGen/X86/vector-shuffle-256-v16.ll b/test/CodeGen/X86/vector-shuffle-256-v16.ll
index f53466ff19a..c2e06f5fcfa 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v16.ll
@@ -1457,10 +1457,9 @@ define <16 x i16> @shuffle_v16i16_16_zz_zz_zz_17_zz_zz_zz_18_zz_zz_zz_19_zz_zz_z
 define <16 x i16> @shuffle_v16i16_16_zz_17_zz_18_zz_19_zz_20_zz_21_zz_22_zz_22_zz(<16 x i16> %a) {
 ; AVX1-LABEL: shuffle_v16i16_16_zz_17_zz_18_zz_19_zz_20_zz_21_zz_22_zz_22_zz:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpmovzxwq {{.*#+}} xmm1 = xmm1[0],zero,zero,zero,xmm1[1],zero,zero,zero
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
 ; AVX1-NEXT:    vpmovzxwd {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index c0c5f1b21de..9905dfdac28 100644
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -1802,10 +1802,9 @@ define <32 x i8> @shuffle_v32i8_32_zz_zz_zz_33_zz_zz_zz_34_zz_zz_zz_35_zz_zz_zz_
 define <32 x i8> @shuffle_v32i8_32_zz_33_zz_34_zz_35_zz_36_zz_37_zz_38_zz_39_zz_40_zz_41_zz_42_zz_43_zz_44_zz_45_zz_46_zz_47_zz(<32 x i8> %a) {
 ; AVX1-LABEL: shuffle_v32i8_32_zz_33_zz_34_zz_35_zz_36_zz_37_zz_38_zz_39_zz_40_zz_41_zz_42_zz_43_zz_44_zz_45_zz_46_zz_47_zz:
 ; AVX1:       # BB#0:
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpsllq $56, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufd {{.*#+}} xmm2 = xmm0[2,3,0,1]
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3],xmm2[4],xmm1[4],xmm2[5],xmm1[5],xmm2[6],xmm1[6],xmm2[7],xmm1[7]
+; AVX1-NEXT:    vpshufd {{.*#+}} xmm1 = xmm0[2,3,0,1]
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3],xmm1[4],xmm2[4],xmm1[5],xmm2[5],xmm1[6],xmm2[6],xmm1[7],xmm2[7]
 ; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
-- 
2.34.1