[X86][SSE] Use bitmasks instead of shuffles where possible.

VPAND is a lot faster than VPSHUFB and VPBLENDVB - this patch ensures we attempt to lower to a basic bitmask before lowering to the slower byte shuffle/blend instructions.

Split off from D11518.

Differential Revision: http://reviews.llvm.org/D11541

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@243395 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 109047751791be80a6644935a70ebfa8a84ad375..b3672c3216dc43b4362bc8c1482a35d2f54197d0 100644 (file)
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -6663,6 +6663,10 @@ static SDValue lowerVectorShuffleAsBlend(SDLoc DL, MVT VT, SDValue V1,
     assert((VT.getSizeInBits() == 128 || Subtarget->hasAVX2()) &&
            "256-bit byte-blends require AVX2 support!");
 
+    // Attempt to lower to a bitmask if we can. VPAND is faster than VPBLENDVB.
+    if (SDValue Masked = lowerVectorShuffleAsBitMask(DL, VT, V1, V2, Mask, DAG))
+      return Masked;
+
     // Scale the blend by the number of bytes per element.
     int Scale = VT.getScalarSizeInBits() / 8;
 
@@ -9071,6 +9075,10 @@ static SDValue lowerV16I8VectorShuffle(SDValue Op, SDValue V1, SDValue V2,
       return V;
   }
 
+  if (SDValue Masked =
+          lowerVectorShuffleAsBitMask(DL, MVT::v16i8, V1, V2, Mask, DAG))
+    return Masked;
+
   // Use dedicated unpack instructions for masks that match their pattern.
   if (isShuffleEquivalent(V1, V2, Mask, {// Low half.
                                          0, 16, 1, 17, 2, 18, 3, 19,

diff --git a/test/CodeGen/X86/vector-shuffle-128-v16.ll b/test/CodeGen/X86/vector-shuffle-128-v16.ll
index ff07c4b01c384bfb86975eea45d2d3ba7b704c0e..0d949f6a76c2cf392e2adf1cfac7c36730a01553 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-128-v16.ll
+++ b/test/CodeGen/X86/vector-shuffle-128-v16.ll
@@ -470,28 +470,14 @@ define <16 x i8> @shuffle_v16i8_00_01_02_19_04_05_06_23_08_09_10_27_12_13_14_31(
 }
 
 define <16 x i8> @shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz(<16 x i8> %a) {
-; SSE2-LABEL: shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz:
-; SSE2:       # BB#0:
-; SSE2-NEXT:    xorps %xmm1, %xmm1
-; SSE2-NEXT:    movaps {{.*#+}} xmm2 = [255,255,255,0,255,255,255,0,255,255,255,0,255,255,255,0]
-; SSE2-NEXT:    andps %xmm2, %xmm0
-; SSE2-NEXT:    andnps %xmm1, %xmm2
-; SSE2-NEXT:    orps %xmm2, %xmm0
-; SSE2-NEXT:    retq
-;
-; SSSE3-LABEL: shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz:
-; SSSE3:       # BB#0:
-; SSSE3-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,2],zero,xmm0[4,5,6],zero,xmm0[8,9,10],zero,xmm0[12,13,14],zero
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz:
-; SSE41:       # BB#0:
-; SSE41-NEXT:    pshufb {{.*#+}} xmm0 = xmm0[0,1,2],zero,xmm0[4,5,6],zero,xmm0[8,9,10],zero,xmm0[12,13,14],zero
-; SSE41-NEXT:    retq
+; SSE-LABEL: shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz:
+; SSE:       # BB#0:
+; SSE-NEXT:    andps {{.*}}(%rip), %xmm0
+; SSE-NEXT:    retq
 ;
 ; AVX-LABEL: shuffle_v16i8_00_01_02_zz_04_05_06_zz_08_09_10_zz_12_13_14_zz:
 ; AVX:       # BB#0:
-; AVX-NEXT:    vpshufb {{.*#+}} xmm0 = xmm0[0,1,2],zero,xmm0[4,5,6],zero,xmm0[8,9,10],zero,xmm0[12,13,14],zero
+; AVX-NEXT:    vandps {{.*}}(%rip), %xmm0, %xmm0
 ; AVX-NEXT:    retq
   %shuffle = shufflevector <16 x i8> %a, <16 x i8> zeroinitializer, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 4, i32 5, i32 6, i32 23, i32 8, i32 9, i32 10, i32 27, i32 12, i32 13, i32 14, i32 31>
   ret <16 x i8> %shuffle

diff --git a/test/CodeGen/X86/vector-shuffle-256-v32.ll b/test/CodeGen/X86/vector-shuffle-256-v32.ll
index b852cd0d6730cabcde1687190f64859f411c6fa2..bbbe0cd8dfce71b1acbfab0800708453c806eab2 100644 (file)
--- a/test/CodeGen/X86/vector-shuffle-256-v32.ll
+++ b/test/CodeGen/X86/vector-shuffle-256-v32.ll
@@ -951,17 +951,15 @@ define <32 x i8> @shuffle_v32i8_zz_01_zz_03_zz_05_zz_07_zz_09_zz_11_zz_13_zz_15_
 ; AVX1-LABEL: shuffle_v32i8_zz_01_zz_03_zz_05_zz_07_zz_09_zz_11_zz_13_zz_15_zz_17_zz_19_zz_21_zz_23_zz_25_zz_27_zz_29_zz_31:
 ; AVX1:       # BB#0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [128,1,128,3,128,5,128,7,128,9,128,11,128,13,128,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovaps {{.*#+}} xmm2 = [0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255]
+; AVX1-NEXT:    vandps %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vandps %xmm2, %xmm0, %xmm0
 ; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: shuffle_v32i8_zz_01_zz_03_zz_05_zz_07_zz_09_zz_11_zz_13_zz_15_zz_17_zz_19_zz_21_zz_23_zz_25_zz_27_zz_29_zz_31:
 ; AVX2:       # BB#0:
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0,255,0]
-; AVX2-NEXT:    vpxor %ymm2, %ymm2, %ymm2
-; AVX2-NEXT:    vpblendvb %ymm1, %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vandps {{.*}}(%rip), %ymm0, %ymm0
 ; AVX2-NEXT:    retq
   %shuffle = shufflevector <32 x i8> %a, <32 x i8> zeroinitializer, <32 x i32> <i32 32, i32 1, i32 34, i32 3, i32 36, i32 5, i32 38, i32 7, i32 40, i32 9, i32 42, i32 11, i32 44, i32 13, i32 46, i32 15, i32 48, i32 17, i32 50, i32 19, i32 52, i32 21, i32 54, i32 23, i32 56, i32 25, i32 58, i32 27, i32 60, i32 29, i32 62, i32 31>
   ret <32 x i8> %shuffle

diff --git a/test/CodeGen/X86/vector-zext.ll b/test/CodeGen/X86/vector-zext.ll
index 77435fd3cac14b174d62bd4c5dcac09a3dad8f3d..dca2d976b74e58263a9154147ac0a6876b7706a0 100644 (file)
--- a/test/CodeGen/X86/vector-zext.ll
+++ b/test/CodeGen/X86/vector-zext.ll
@@ -1092,17 +1092,16 @@ entry:
 define <8 x i32> @shuf_zext_8i8_to_8i32(<8 x i8> %A) {
 ; SSE2-LABEL: shuf_zext_8i8_to_8i32:
 ; SSE2:       # BB#0: # %entry
-; SSE2-NEXT:    pand {{.*}}(%rip), %xmm0
-; SSE2-NEXT:    packuswb %xmm0, %xmm0
-; SSE2-NEXT:    pxor %xmm1, %xmm1
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm2 = 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]
-; SSE2-NEXT:    punpcklwd {{.*#+}} xmm2 = xmm2[0],xmm1[0],xmm2[1],xmm1[1],xmm2[2],xmm1[2],xmm2[3],xmm1[3]
-; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    punpckhwd {{.*#+}} xmm0 = xmm0[4,4,5,5,6,6,7,7]
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,255,255,255,0,255,255,255,0,255,255,255,0,255,255,255]
-; SSE2-NEXT:    pandn %xmm0, %xmm1
-; SSE2-NEXT:    movdqa %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    packuswb %xmm1, %xmm1
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3],xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
+; SSE2-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm2[0],xmm0[1],xmm2[1],xmm0[2],xmm2[2],xmm0[3],xmm2[3]
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm1 = xmm1[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    punpckhwd {{.*#+}} xmm1 = xmm1[4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
 ; SSE2-NEXT:    retq
 ;
 ; SSSE3-LABEL: shuf_zext_8i8_to_8i32: