[X86] SimplifyDemandedBitsForTargetNode - add X86ISD::BLENDI handling (#133102)

RKSimon · web-flow · commit f7a333401658 · 2025-03-27T09:32:40.000Z
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -44227,6 +44227,35 @@ bool X86TargetLowering::SimplifyDemandedBitsForTargetNode(
     }
     return false;
   }
+  case X86ISD::BLENDI: {
+    SDValue LHS = Op.getOperand(0);
+    SDValue RHS = Op.getOperand(1);
+    APInt Mask = getBLENDIBlendMask(Op);
+
+    APInt DemandedEltsLHS = OriginalDemandedElts & ~Mask;
+    if (SimplifyDemandedBits(LHS, OriginalDemandedBits, DemandedEltsLHS, Known,
+                             TLO, Depth + 1))
+      return true;
+
+    APInt DemandedEltsRHS = OriginalDemandedElts & Mask;
+    if (SimplifyDemandedBits(RHS, OriginalDemandedBits, DemandedEltsRHS, Known,
+                             TLO, Depth + 1))
+      return true;
+
+    // Attempt to avoid multi-use ops if we don't need anything from them.
+    SDValue NewLHS = SimplifyMultipleUseDemandedBits(
+        LHS, OriginalDemandedBits, DemandedEltsLHS, TLO.DAG, Depth + 1);
+    SDValue NewRHS = SimplifyMultipleUseDemandedBits(
+        RHS, OriginalDemandedBits, DemandedEltsRHS, TLO.DAG, Depth + 1);
+    if (NewLHS || NewRHS) {
+      NewLHS = NewLHS ? NewLHS : LHS;
+      NewRHS = NewRHS ? NewRHS : RHS;
+      return TLO.CombineTo(Op,
+                           TLO.DAG.getNode(Op.getOpcode(), SDLoc(Op), VT,
+                                           NewLHS, NewRHS, Op.getOperand(2)));
+    }
+    break;
+  }
   case X86ISD::BLENDV: {
     SDValue Sel = Op.getOperand(0);
     SDValue LHS = Op.getOperand(1);
diff --git a/llvm/test/CodeGen/X86/combine-movmsk-avx.ll b/llvm/test/CodeGen/X86/combine-movmsk-avx.ll
@@ -75,39 +75,15 @@ define i1 @movmskps_allof_bitcast_v4f64(<4 x double> %a0) {
 }
 
 ;
-; TODO - Avoid sign extension ops when just extracting the sign bits.
+; Avoid sign extension ops when just extracting the sign bits.
 ;
 
 define i32 @movmskpd_cmpgt_v4i64(<4 x i64> %a0) {
-; VTEST-AVX1-LABEL: movmskpd_cmpgt_v4i64:
-; VTEST-AVX1:       # %bb.0:
-; VTEST-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; VTEST-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm1
-; VTEST-AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; VTEST-AVX1-NEXT:    vmovmskpd %ymm0, %eax
-; VTEST-AVX1-NEXT:    vzeroupper
-; VTEST-AVX1-NEXT:    retq
-;
-; VTEST-AVX2-LABEL: movmskpd_cmpgt_v4i64:
-; VTEST-AVX2:       # %bb.0:
-; VTEST-AVX2-NEXT:    vmovmskpd %ymm0, %eax
-; VTEST-AVX2-NEXT:    vzeroupper
-; VTEST-AVX2-NEXT:    retq
-;
-; MOVMSK-AVX1-LABEL: movmskpd_cmpgt_v4i64:
-; MOVMSK-AVX1:       # %bb.0:
-; MOVMSK-AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; MOVMSK-AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm1
-; MOVMSK-AVX1-NEXT:    vblendpd {{.*#+}} ymm0 = ymm1[0,1],ymm0[2,3]
-; MOVMSK-AVX1-NEXT:    vmovmskpd %ymm0, %eax
-; MOVMSK-AVX1-NEXT:    vzeroupper
-; MOVMSK-AVX1-NEXT:    retq
-;
-; MOVMSK-AVX2-LABEL: movmskpd_cmpgt_v4i64:
-; MOVMSK-AVX2:       # %bb.0:
-; MOVMSK-AVX2-NEXT:    vmovmskpd %ymm0, %eax
-; MOVMSK-AVX2-NEXT:    vzeroupper
-; MOVMSK-AVX2-NEXT:    retq
+; CHECK-LABEL: movmskpd_cmpgt_v4i64:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmovmskpd %ymm0, %eax
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %1 = icmp sgt <4 x i64> zeroinitializer, %a0
   %2 = sext <4 x i1> %1 to <4 x i64>
   %3 = bitcast <4 x i64> %2 to <4 x double>
@@ -116,33 +92,11 @@ define i32 @movmskpd_cmpgt_v4i64(<4 x i64> %a0) {
 }
 
 define i32 @movmskps_ashr_v8i32(<8 x i32> %a0)  {
-; VTEST-AVX1-LABEL: movmskps_ashr_v8i32:
-; VTEST-AVX1:       # %bb.0:
-; VTEST-AVX1-NEXT:    vpsrad $31, %xmm0, %xmm1
-; VTEST-AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; VTEST-AVX1-NEXT:    vmovmskps %ymm0, %eax
-; VTEST-AVX1-NEXT:    vzeroupper
-; VTEST-AVX1-NEXT:    retq
-;
-; VTEST-AVX2-LABEL: movmskps_ashr_v8i32:
-; VTEST-AVX2:       # %bb.0:
-; VTEST-AVX2-NEXT:    vmovmskps %ymm0, %eax
-; VTEST-AVX2-NEXT:    vzeroupper
-; VTEST-AVX2-NEXT:    retq
-;
-; MOVMSK-AVX1-LABEL: movmskps_ashr_v8i32:
-; MOVMSK-AVX1:       # %bb.0:
-; MOVMSK-AVX1-NEXT:    vpsrad $31, %xmm0, %xmm1
-; MOVMSK-AVX1-NEXT:    vblendps {{.*#+}} ymm0 = ymm1[0,1,2,3],ymm0[4,5,6,7]
-; MOVMSK-AVX1-NEXT:    vmovmskps %ymm0, %eax
-; MOVMSK-AVX1-NEXT:    vzeroupper
-; MOVMSK-AVX1-NEXT:    retq
-;
-; MOVMSK-AVX2-LABEL: movmskps_ashr_v8i32:
-; MOVMSK-AVX2:       # %bb.0:
-; MOVMSK-AVX2-NEXT:    vmovmskps %ymm0, %eax
-; MOVMSK-AVX2-NEXT:    vzeroupper
-; MOVMSK-AVX2-NEXT:    retq
+; CHECK-LABEL: movmskps_ashr_v8i32:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vmovmskps %ymm0, %eax
+; CHECK-NEXT:    vzeroupper
+; CHECK-NEXT:    retq
   %1 = ashr <8 x i32> %a0, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
   %2 = bitcast <8 x i32> %1 to <8 x float>
   %3 = tail call i32 @llvm.x86.avx.movmsk.ps.256(<8 x float> %2)
diff --git a/llvm/test/CodeGen/X86/combine-sdiv.ll b/llvm/test/CodeGen/X86/combine-sdiv.ll
@@ -2187,15 +2187,14 @@ define <16 x i8> @non_splat_minus_one_divisor_1(<16 x i8> %A) {
 ; SSE41-NEXT:    pxor %xmm0, %xmm0
 ; SSE41-NEXT:    pxor %xmm3, %xmm3
 ; SSE41-NEXT:    pcmpgtb %xmm1, %xmm3
-; SSE41-NEXT:    pxor %xmm4, %xmm4
-; SSE41-NEXT:    punpcklbw {{.*#+}} xmm4 = xmm4[0],xmm3[0],xmm4[1],xmm3[1],xmm4[2],xmm3[2],xmm4[3],xmm3[3],xmm4[4],xmm3[4],xmm4[5],xmm3[5],xmm4[6],xmm3[6],xmm4[7],xmm3[7]
-; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm2 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
-; SSE41-NEXT:    paddw %xmm2, %xmm2
-; SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm4[0,1],xmm2[2],xmm4[3,4,5],xmm2[6],xmm4[7]
-; SSE41-NEXT:    psrlw $8, %xmm2
+; SSE41-NEXT:    pmovzxbw {{.*#+}} xmm4 = xmm3[0],zero,xmm3[1],zero,xmm3[2],zero,xmm3[3],zero,xmm3[4],zero,xmm3[5],zero,xmm3[6],zero,xmm3[7],zero
 ; SSE41-NEXT:    punpckhbw {{.*#+}} xmm3 = xmm3[8],xmm0[8],xmm3[9],xmm0[9],xmm3[10],xmm0[10],xmm3[11],xmm0[11],xmm3[12],xmm0[12],xmm3[13],xmm0[13],xmm3[14],xmm0[14],xmm3[15],xmm0[15]
 ; SSE41-NEXT:    pmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3 # [256,2,2,2,2,128,2,128]
 ; SSE41-NEXT:    psrlw $8, %xmm3
+; SSE41-NEXT:    paddw %xmm4, %xmm4
+; SSE41-NEXT:    pmovsxbw %xmm1, %xmm2
+; SSE41-NEXT:    pblendw {{.*#+}} xmm2 = xmm2[0,1],xmm4[2],xmm2[3,4,5],xmm4[6],xmm2[7]
+; SSE41-NEXT:    psrlw $8, %xmm2
 ; SSE41-NEXT:    packuswb %xmm3, %xmm2
 ; SSE41-NEXT:    paddb %xmm1, %xmm2
 ; SSE41-NEXT:    movdqa %xmm2, %xmm0
@@ -2223,15 +2222,15 @@ define <16 x i8> @non_splat_minus_one_divisor_1(<16 x i8> %A) {
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm2
-; AVX1-NEXT:    vpunpcklbw {{.*#+}} xmm3 = 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 {{.*#+}} xmm4 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
-; AVX1-NEXT:    vpaddw %xmm4, %xmm4, %xmm4
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1],xmm4[2],xmm3[3,4,5],xmm4[6],xmm3[7]
-; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm1 = xmm2[8],xmm1[8],xmm2[9],xmm1[9],xmm2[10],xmm1[10],xmm2[11],xmm1[11],xmm2[12],xmm1[12],xmm2[13],xmm1[13],xmm2[14],xmm1[14],xmm2[15],xmm1[15]
 ; AVX1-NEXT:    vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1, %xmm1 # [256,2,2,2,2,128,2,128]
 ; AVX1-NEXT:    vpsrlw $8, %xmm1, %xmm1
-; AVX1-NEXT:    vpackuswb %xmm1, %xmm3, %xmm1
+; AVX1-NEXT:    vpmovzxbw {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero,xmm2[2],zero,xmm2[3],zero,xmm2[4],zero,xmm2[5],zero,xmm2[6],zero,xmm2[7],zero
+; AVX1-NEXT:    vpaddw %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovsxbw %xmm0, %xmm3
+; AVX1-NEXT:    vpblendw {{.*#+}} xmm2 = xmm3[0,1],xmm2[2],xmm3[3,4,5],xmm2[6],xmm3[7]
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm2, %xmm1
 ; AVX1-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm2 = xmm1[8,8,9,9,10,10,11,11,12,12,13,13,14,14,15,15]
 ; AVX1-NEXT:    vpsraw $8, %xmm2, %xmm2
diff --git a/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll b/llvm/test/CodeGen/X86/vector-idiv-udiv-256.ll
@@ -273,8 +273,7 @@ define <32 x i8> @test_divconstant_32i8(<32 x i8> %a) nounwind {
 ; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
 ; AVX1-NEXT:    vpackuswb %xmm3, %xmm2, %xmm2
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm3 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm4
-; AVX1-NEXT:    vpsllw $7, %xmm4, %xmm4
+; AVX1-NEXT:    vpsrlw $1, %xmm0, %xmm4
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm3 = xmm3[0,1,2,3,4,5,6],xmm4[7]
 ; AVX1-NEXT:    vpsrlw $8, %xmm3, %xmm3
 ; AVX1-NEXT:    vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm3, %xmm3 # [137,16,241,57,27,205,135,187]
@@ -711,8 +710,7 @@ define <32 x i8> @test_remconstant_32i8(<32 x i8> %a) nounwind {
 ; AVX1-NEXT:    vpor %xmm4, %xmm5, %xmm4
 ; AVX1-NEXT:    vpsubb %xmm4, %xmm2, %xmm2
 ; AVX1-NEXT:    vpunpckhbw {{.*#+}} xmm4 = xmm1[8],xmm0[8],xmm1[9],xmm0[9],xmm1[10],xmm0[10],xmm1[11],xmm0[11],xmm1[12],xmm0[12],xmm1[13],xmm0[13],xmm1[14],xmm0[14],xmm1[15],xmm0[15]
-; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm5
-; AVX1-NEXT:    vpsllw $7, %xmm5, %xmm5
+; AVX1-NEXT:    vpsrlw $1, %xmm0, %xmm5
 ; AVX1-NEXT:    vpblendw {{.*#+}} xmm4 = xmm4[0,1,2,3,4,5,6],xmm5[7]
 ; AVX1-NEXT:    vpsrlw $8, %xmm4, %xmm4
 ; AVX1-NEXT:    vpmullw {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm4, %xmm4 # [137,16,241,57,27,205,135,187]
