[X86][AVX] Add PACKSSDW/PACKUSDW support for truncation of clamped values

SSE and shorter vector sizes will have to wait until we can add support for general SMIN/SMAX matching.

llvm-svn: 324485

3 files changed, 20 insertions, 60 deletions

diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 789ac890184..a9b672e29b7 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -34067,13 +34067,15 @@ static SDValue combineTruncateWithSat(SDValue In, EVT VT, const SDLoc &DL,
     if (auto USatVal = detectUSatPattern(In, VT))
       return DAG.getNode(X86ISD::VTRUNCUS, DL, VT, USatVal);
   }
-  if (VT.getScalarType() == MVT::i8 && InVT.getScalarType() == MVT::i16) {
+  if ((VT.getScalarType() == MVT::i8 && InVT.getScalarType() == MVT::i16) ||
+      (VT.getScalarType() == MVT::i16 && InVT.getScalarType() == MVT::i32)) {
     if (auto SSatVal = detectSSatPattern(In, VT))
       return truncateVectorWithPACK(X86ISD::PACKSS, VT, SSatVal, DL, DAG,
                                     Subtarget);
-    if (auto USatVal = detectSSatPattern(In, VT, true))
-      return truncateVectorWithPACK(X86ISD::PACKUS, VT, USatVal, DL, DAG,
-                                    Subtarget);
+    if (Subtarget.hasSSE41() || VT.getScalarType() == MVT::i8)
+      if (auto USatVal = detectSSatPattern(In, VT, true))
+        return truncateVectorWithPACK(X86ISD::PACKUS, VT, USatVal, DL, DAG,
+                                      Subtarget);
   }
   return SDValue();
 }
diff --git a/llvm/test/CodeGen/X86/vector-trunc-packus.ll b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
index f466808b436..68aa9a03e38 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-packus.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-packus.ll
@@ -1185,28 +1185,14 @@ define <8 x i16> @trunc_packus_v8i32_v8i16(<8 x i32> %a0) {
 ; AVX1-LABEL: trunc_packus_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [65535,65535,65535,65535]
-; AVX1-NEXT:    vpminsd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpminsd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpmaxsd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpmaxsd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_packus_v8i32_v8i16:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX2-NEXT:    vpminsd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpackusdw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1400,17 +1386,10 @@ define <16 x i16> @trunc_packus_v16i32_v16i16(<16 x i32> %a0) {
 ;
 ; AVX2-LABEL: trunc_packus_v16i32_v16i16:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [65535,65535,65535,65535,65535,65535,65535,65535]
-; AVX2-NEXT:    vpminsd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpminsd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpmaxsd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpmaxsd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpackusdw %xmm2, %xmm1, %xmm1
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpackusdw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;
diff --git a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
index 2dbfe76a6e8..c47bfe908b4 100644
--- a/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
+++ b/llvm/test/CodeGen/X86/vector-trunc-ssat.ll
@@ -1256,28 +1256,14 @@ define <8 x i16> @trunc_ssat_v8i32_v8i16(<8 x i32> %a0) {
 ; AVX1-LABEL: trunc_ssat_v8i32_v8i16:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [32767,32767,32767,32767]
-; AVX1-NEXT:    vpminsd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpminsd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [4294934528,4294934528,4294934528,4294934528]
-; AVX1-NEXT:    vpmaxsd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpmaxsd %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vmovdqa {{.*#+}} xmm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15]
-; AVX1-NEXT:    vpshufb %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpshufb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpunpcklqdq {{.*#+}} xmm0 = xmm0[0],xmm1[0]
+; AVX1-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    vzeroupper
 ; AVX1-NEXT:    retq
 ;
 ; AVX2-LABEL: trunc_ssat_v8i32_v8i16:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32767,32767,32767,32767,32767,32767,32767,32767]
-; AVX2-NEXT:    vpminsd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4294934528,4294934528,4294934528,4294934528,4294934528,4294934528,4294934528,4294934528]
-; AVX2-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpshufb {{.*#+}} ymm0 = ymm0[0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    # kill: def $xmm0 killed $xmm0 killed $ymm0
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX2-NEXT:    vpackssdw %xmm1, %xmm0, %xmm0
 ; AVX2-NEXT:    vzeroupper
 ; AVX2-NEXT:    retq
 ;
@@ -1483,17 +1469,10 @@ define <16 x i16> @trunc_ssat_v16i32_v16i16(<16 x i32> %a0) {
 ;
 ; AVX2-LABEL: trunc_ssat_v16i32_v16i16:
 ; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [32767,32767,32767,32767,32767,32767,32767,32767]
-; AVX2-NEXT:    vpminsd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpminsd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm2 = [4294934528,4294934528,4294934528,4294934528,4294934528,4294934528,4294934528,4294934528]
-; AVX2-NEXT:    vpmaxsd %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpmaxsd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vmovdqa {{.*#+}} ymm2 = [0,1,4,5,8,9,12,13,8,9,12,13,12,13,14,15,16,17,20,21,24,25,28,29,24,25,28,29,28,29,30,31]
-; AVX2-NEXT:    vpshufb %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpermq {{.*#+}} ymm0 = ymm0[0,2,2,3]
-; AVX2-NEXT:    vpshufb %ymm2, %ymm1, %ymm1
-; AVX2-NEXT:    vpermq {{.*#+}} ymm1 = ymm1[0,2,2,3]
+; AVX2-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-NEXT:    vpackssdw %xmm2, %xmm1, %xmm1
+; AVX2-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-NEXT:    vpackssdw %xmm2, %xmm0, %xmm0
 ; AVX2-NEXT:    vinserti128 $1, %xmm1, %ymm0, %ymm0
 ; AVX2-NEXT:    retq
 ;