[x86] autoupgrade and remove AVX2 integer min/max intrinsics

This will (hopefully very temporarily) break clang.
The clang side of this should be the next commit.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@272932 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td
index ca169b1c208e358817c7458a06051af6cd360bd8..3f53159f8213042501f4ca33fcc257432ecec23b 100644 (file)
--- a/include/llvm/IR/IntrinsicsX86.td
+++ b/include/llvm/IR/IntrinsicsX86.td
@@ -1934,42 +1934,6 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 
 // Vector min, max
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
-  def int_x86_avx2_pmaxu_b : GCCBuiltin<"__builtin_ia32_pmaxub256">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v32i8_ty,
-                         llvm_v32i8_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmaxu_w : GCCBuiltin<"__builtin_ia32_pmaxuw256">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
-                         llvm_v16i16_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmaxu_d : GCCBuiltin<"__builtin_ia32_pmaxud256">,
-              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty,
-                         llvm_v8i32_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmaxs_b : GCCBuiltin<"__builtin_ia32_pmaxsb256">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v32i8_ty,
-                         llvm_v32i8_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmaxs_w : GCCBuiltin<"__builtin_ia32_pmaxsw256">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
-                         llvm_v16i16_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmaxs_d : GCCBuiltin<"__builtin_ia32_pmaxsd256">,
-              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty,
-                         llvm_v8i32_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pminu_b : GCCBuiltin<"__builtin_ia32_pminub256">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v32i8_ty,
-                         llvm_v32i8_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pminu_w : GCCBuiltin<"__builtin_ia32_pminuw256">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
-                         llvm_v16i16_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pminu_d : GCCBuiltin<"__builtin_ia32_pminud256">,
-              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty,
-                         llvm_v8i32_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmins_b : GCCBuiltin<"__builtin_ia32_pminsb256">,
-              Intrinsic<[llvm_v32i8_ty], [llvm_v32i8_ty,
-                         llvm_v32i8_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmins_w : GCCBuiltin<"__builtin_ia32_pminsw256">,
-              Intrinsic<[llvm_v16i16_ty], [llvm_v16i16_ty,
-                         llvm_v16i16_ty], [IntrNoMem, Commutative]>;
-  def int_x86_avx2_pmins_d : GCCBuiltin<"__builtin_ia32_pminsd256">,
-              Intrinsic<[llvm_v8i32_ty], [llvm_v8i32_ty,
-                         llvm_v8i32_ty], [IntrNoMem, Commutative]>;
   def int_x86_avx512_mask_pmaxs_b_128 : GCCBuiltin<"__builtin_ia32_pmaxsb128_mask">,
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty, llvm_v16i8_ty,
                          llvm_v16i8_ty, llvm_i16_ty], [IntrNoMem]>;

diff --git a/lib/IR/AutoUpgrade.cpp b/lib/IR/AutoUpgrade.cpp
index f13009a4b4e40c286d2064c987b36a58a760dfc3..d3ee5e8e5dbbc8ed0a12765d129ddc0e5cfe19e9 100644 (file)
--- a/lib/IR/AutoUpgrade.cpp
+++ b/lib/IR/AutoUpgrade.cpp
@@ -186,6 +186,8 @@ static bool UpgradeIntrinsicFunction1(Function *F, Function *&NewFn) {
         Name == "x86.sse2.pminu.b" ||
         Name == "x86.sse41.pminuw" ||
         Name == "x86.sse41.pminud" ||
+        Name.startswith("x86.avx2.pmax") ||
+        Name.startswith("x86.avx2.pmin") ||
         Name.startswith("x86.avx2.vbroadcast") ||
         Name.startswith("x86.avx2.pbroadcast") ||
         Name.startswith("x86.avx.vpermil.") ||
@@ -566,19 +568,23 @@ void llvm::UpgradeIntrinsicCall(CallInst *CI, Function *NewFn) {
       Rep = Builder.CreateSExt(Rep, CI->getType(), "");
     } else if (Name == "llvm.x86.sse41.pmaxsb" ||
                Name == "llvm.x86.sse2.pmaxs.w" ||
-               Name == "llvm.x86.sse41.pmaxsd") {
+               Name == "llvm.x86.sse41.pmaxsd" ||
+               Name.startswith("llvm.x86.avx2.pmaxs")) {
       Rep = upgradeIntMinMax(Builder, *CI, ICmpInst::ICMP_SGT);
     } else if (Name == "llvm.x86.sse2.pmaxu.b" ||
                Name == "llvm.x86.sse41.pmaxuw" ||
-               Name == "llvm.x86.sse41.pmaxud") {
+               Name == "llvm.x86.sse41.pmaxud" ||
+               Name.startswith("llvm.x86.avx2.pmaxu")) {
       Rep = upgradeIntMinMax(Builder, *CI, ICmpInst::ICMP_UGT);
     } else if (Name == "llvm.x86.sse41.pminsb" ||
                Name == "llvm.x86.sse2.pmins.w" ||
-               Name == "llvm.x86.sse41.pminsd") {
+               Name == "llvm.x86.sse41.pminsd" ||
+               Name.startswith("llvm.x86.avx2.pmins")) {
       Rep = upgradeIntMinMax(Builder, *CI, ICmpInst::ICMP_SLT);
     } else if (Name == "llvm.x86.sse2.pminu.b" ||
                Name == "llvm.x86.sse41.pminuw" ||
-               Name == "llvm.x86.sse41.pminud") {
+               Name == "llvm.x86.sse41.pminud" ||
+               Name.startswith("llvm.x86.avx2.pminu")) {
       Rep = upgradeIntMinMax(Builder, *CI, ICmpInst::ICMP_ULT);
     } else if (Name == "llvm.x86.sse2.cvtdq2pd" ||
                Name == "llvm.x86.sse2.cvtps2pd" ||

diff --git a/lib/Target/X86/X86IntrinsicsInfo.h b/lib/Target/X86/X86IntrinsicsInfo.h
index 8fab49cf4492560918b1d361b927a88440a44a63..d81412510a72e23c87998aa7f6a463d943fb2ffb 100644 (file)
--- a/lib/Target/X86/X86IntrinsicsInfo.h
+++ b/lib/Target/X86/X86IntrinsicsInfo.h
@@ -291,18 +291,6 @@ static const IntrinsicData  IntrinsicsWithoutChain[] = {
   X86_INTRINSIC_DATA(avx2_phadd_w, INTR_TYPE_2OP, X86ISD::HADD, 0),
   X86_INTRINSIC_DATA(avx2_phsub_d, INTR_TYPE_2OP, X86ISD::HSUB, 0),
   X86_INTRINSIC_DATA(avx2_phsub_w, INTR_TYPE_2OP, X86ISD::HSUB, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxs_b, INTR_TYPE_2OP, ISD::SMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxs_d, INTR_TYPE_2OP, ISD::SMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxs_w, INTR_TYPE_2OP, ISD::SMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxu_b, INTR_TYPE_2OP, ISD::UMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxu_d, INTR_TYPE_2OP, ISD::UMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmaxu_w, INTR_TYPE_2OP, ISD::UMAX, 0),
-  X86_INTRINSIC_DATA(avx2_pmins_b, INTR_TYPE_2OP, ISD::SMIN, 0),
-  X86_INTRINSIC_DATA(avx2_pmins_d, INTR_TYPE_2OP, ISD::SMIN, 0),
-  X86_INTRINSIC_DATA(avx2_pmins_w, INTR_TYPE_2OP, ISD::SMIN, 0),
-  X86_INTRINSIC_DATA(avx2_pminu_b, INTR_TYPE_2OP, ISD::UMIN, 0),
-  X86_INTRINSIC_DATA(avx2_pminu_d, INTR_TYPE_2OP, ISD::UMIN, 0),
-  X86_INTRINSIC_DATA(avx2_pminu_w, INTR_TYPE_2OP, ISD::UMIN, 0),
   X86_INTRINSIC_DATA(avx2_pmovmskb, INTR_TYPE_1OP, X86ISD::MOVMSK, 0),
   X86_INTRINSIC_DATA(avx2_pmul_dq, INTR_TYPE_2OP, X86ISD::PMULDQ, 0),
   X86_INTRINSIC_DATA(avx2_pmulh_w, INTR_TYPE_2OP, ISD::MULHS, 0),

diff --git a/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll b/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
index b9f1f2283f265d7a20a7f2b7071ec5584a9bde6c..4f410b415c085be6df21bc64ef13815155d287ab 100644 (file)
--- a/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
+++ b/test/CodeGen/X86/avx2-intrinsics-fast-isel.ll
@@ -1931,11 +1931,11 @@ define <4 x i64> @test_mm256_max_epi8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <32 x i8>
   %arg1 = bitcast <4 x i64> %a1 to <32 x i8>
-  %res = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %arg0, <32 x i8> %arg1)
-  %bc = bitcast <32 x i8> %res to <4 x i64>
+  %cmp = icmp sgt <32 x i8> %arg0, %arg1
+  %sel = select <32 x i1> %cmp, <32 x i8> %arg0, <32 x i8> %arg1
+  %bc = bitcast <32 x i8> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8>, <32 x i8>) nounwind readnone
 
 define <4 x i64> @test_mm256_max_epi16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_max_epi16:
@@ -1949,11 +1949,11 @@ define <4 x i64> @test_mm256_max_epi16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <16 x i16>
   %arg1 = bitcast <4 x i64> %a1 to <16 x i16>
-  %res = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %arg0, <16 x i16> %arg1)
-  %bc = bitcast <16 x i16> %res to <4 x i64>
+  %cmp = icmp sgt <16 x i16> %arg0, %arg1
+  %sel = select <16 x i1> %cmp, <16 x i16> %arg0, <16 x i16> %arg1
+  %bc = bitcast <16 x i16> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16>, <16 x i16>) nounwind readnone
 
 define <4 x i64> @test_mm256_max_epi32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_max_epi32:
@@ -1967,11 +1967,11 @@ define <4 x i64> @test_mm256_max_epi32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <8 x i32>
   %arg1 = bitcast <4 x i64> %a1 to <8 x i32>
-  %res = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %arg0, <8 x i32> %arg1)
-  %bc = bitcast <8 x i32> %res to <4 x i64>
+  %cmp = icmp sgt <8 x i32> %arg0, %arg1
+  %sel = select <8 x i1> %cmp, <8 x i32> %arg0, <8 x i32> %arg1
+  %bc = bitcast <8 x i32> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32>, <8 x i32>) nounwind readnone
 
 define <4 x i64> @test_mm256_max_epu8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_max_epu8:
@@ -1985,11 +1985,11 @@ define <4 x i64> @test_mm256_max_epu8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <32 x i8>
   %arg1 = bitcast <4 x i64> %a1 to <32 x i8>
-  %res = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %arg0, <32 x i8> %arg1)
-  %bc = bitcast <32 x i8> %res to <4 x i64>
+  %cmp = icmp ugt <32 x i8> %arg0, %arg1
+  %sel = select <32 x i1> %cmp, <32 x i8> %arg0, <32 x i8> %arg1
+  %bc = bitcast <32 x i8> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8>, <32 x i8>) nounwind readnone
 
 define <4 x i64> @test_mm256_max_epu16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_max_epu16:
@@ -2003,11 +2003,11 @@ define <4 x i64> @test_mm256_max_epu16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <16 x i16>
   %arg1 = bitcast <4 x i64> %a1 to <16 x i16>
-  %res = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %arg0, <16 x i16> %arg1)
-  %bc = bitcast <16 x i16> %res to <4 x i64>
+  %cmp = icmp ugt <16 x i16> %arg0, %arg1
+  %sel = select <16 x i1> %cmp, <16 x i16> %arg0, <16 x i16> %arg1
+  %bc = bitcast <16 x i16> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16>, <16 x i16>) nounwind readnone
 
 define <4 x i64> @test_mm256_max_epu32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_max_epu32:
@@ -2021,11 +2021,11 @@ define <4 x i64> @test_mm256_max_epu32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <8 x i32>
   %arg1 = bitcast <4 x i64> %a1 to <8 x i32>
-  %res = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %arg0, <8 x i32> %arg1)
-  %bc = bitcast <8 x i32> %res to <4 x i64>
+  %cmp = icmp ugt <8 x i32> %arg0, %arg1
+  %sel = select <8 x i1> %cmp, <8 x i32> %arg0, <8 x i32> %arg1
+  %bc = bitcast <8 x i32> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32>, <8 x i32>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epi8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epi8:
@@ -2039,11 +2039,11 @@ define <4 x i64> @test_mm256_min_epi8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <32 x i8>
   %arg1 = bitcast <4 x i64> %a1 to <32 x i8>
-  %res = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %arg0, <32 x i8> %arg1)
-  %bc = bitcast <32 x i8> %res to <4 x i64>
+  %cmp = icmp slt <32 x i8> %arg0, %arg1
+  %sel = select <32 x i1> %cmp, <32 x i8> %arg0, <32 x i8> %arg1
+  %bc = bitcast <32 x i8> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8>, <32 x i8>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epi16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epi16:
@@ -2057,11 +2057,11 @@ define <4 x i64> @test_mm256_min_epi16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <16 x i16>
   %arg1 = bitcast <4 x i64> %a1 to <16 x i16>
-  %res = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %arg0, <16 x i16> %arg1)
-  %bc = bitcast <16 x i16> %res to <4 x i64>
+  %cmp = icmp slt <16 x i16> %arg0, %arg1
+  %sel = select <16 x i1> %cmp, <16 x i16> %arg0, <16 x i16> %arg1
+  %bc = bitcast <16 x i16> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16>, <16 x i16>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epi32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epi32:
@@ -2075,11 +2075,11 @@ define <4 x i64> @test_mm256_min_epi32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <8 x i32>
   %arg1 = bitcast <4 x i64> %a1 to <8 x i32>
-  %res = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %arg0, <8 x i32> %arg1)
-  %bc = bitcast <8 x i32> %res to <4 x i64>
+  %cmp = icmp slt <8 x i32> %arg0, %arg1
+  %sel = select <8 x i1> %cmp, <8 x i32> %arg0, <8 x i32> %arg1
+  %bc = bitcast <8 x i32> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32>, <8 x i32>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epu8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epu8:
@@ -2093,11 +2093,11 @@ define <4 x i64> @test_mm256_min_epu8(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <32 x i8>
   %arg1 = bitcast <4 x i64> %a1 to <32 x i8>
-  %res = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %arg0, <32 x i8> %arg1)
-  %bc = bitcast <32 x i8> %res to <4 x i64>
+  %cmp = icmp ult <32 x i8> %arg0, %arg1
+  %sel = select <32 x i1> %cmp, <32 x i8> %arg0, <32 x i8> %arg1
+  %bc = bitcast <32 x i8> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8>, <32 x i8>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epu16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epu16:
@@ -2111,11 +2111,11 @@ define <4 x i64> @test_mm256_min_epu16(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <16 x i16>
   %arg1 = bitcast <4 x i64> %a1 to <16 x i16>
-  %res = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %arg0, <16 x i16> %arg1)
-  %bc = bitcast <16 x i16> %res to <4 x i64>
+  %cmp = icmp ult <16 x i16> %arg0, %arg1
+  %sel = select <16 x i1> %cmp, <16 x i16> %arg0, <16 x i16> %arg1
+  %bc = bitcast <16 x i16> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16>, <16 x i16>) nounwind readnone
 
 define <4 x i64> @test_mm256_min_epu32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X32-LABEL: test_mm256_min_epu32:
@@ -2129,11 +2129,11 @@ define <4 x i64> @test_mm256_min_epu32(<4 x i64> %a0, <4 x i64> %a1) {
 ; X64-NEXT:    retq
   %arg0 = bitcast <4 x i64> %a0 to <8 x i32>
   %arg1 = bitcast <4 x i64> %a1 to <8 x i32>
-  %res = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %arg0, <8 x i32> %arg1)
-  %bc = bitcast <8 x i32> %res to <4 x i64>
+  %cmp = icmp ult <8 x i32> %arg0, %arg1
+  %sel = select <8 x i1> %cmp, <8 x i32> %arg0, <8 x i32> %arg1
+  %bc = bitcast <8 x i32> %sel to <4 x i64>
   ret <4 x i64> %bc
 }
-declare <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32>, <8 x i32>) nounwind readnone
 
 define i32 @test_mm256_movemask_epi8(<4 x i64> %a0) nounwind {
 ; X32-LABEL: test_mm256_movemask_epi8:

diff --git a/test/CodeGen/X86/avx2-intrinsics-x86-upgrade.ll b/test/CodeGen/X86/avx2-intrinsics-x86-upgrade.ll
index 2c7d055fe2a0fc6edbd6235224f21ff27991f4b6..b6b8447beda1b4c4b2c2483babfc0384cbf17390 100644 (file)
--- a/test/CodeGen/X86/avx2-intrinsics-x86-upgrade.ll
+++ b/test/CodeGen/X86/avx2-intrinsics-x86-upgrade.ll
@@ -1,4 +1,4 @@
-; NOTE: Assertions have been autogenerated by update_llc_test_checks.py
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: llc < %s -mtriple=x86_64-apple-darwin -march=x86 -mattr=avx2 | FileCheck %s
 
 define <16 x i16> @test_x86_avx2_pblendw(<16 x i16> %a0, <16 x i16> %a1) {
@@ -95,7 +95,6 @@ define <2 x i64> @test_x86_avx2_vextracti128(<4 x i64> %a0) {
 ; CHECK-NEXT:    vextractf128 $1, %ymm0, %xmm0
 ; CHECK-NEXT:    vzeroupper
 ; CHECK-NEXT:    retl
-
   %res = call <2 x i64> @llvm.x86.avx2.vextracti128(<4 x i64> %a0, i8 7)
   ret <2 x i64> %res
 }
@@ -107,7 +106,6 @@ define <4 x i64> @test_x86_avx2_vinserti128(<4 x i64> %a0, <2 x i64> %a1) {
 ; CHECK:       ## BB#0:
 ; CHECK-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; CHECK-NEXT:    retl
-
   %res = call <4 x i64> @llvm.x86.avx2.vinserti128(<4 x i64> %a0, <2 x i64> %a1, i8 7)
   ret <4 x i64> %res
 }
@@ -381,3 +379,136 @@ define void @test_x86_avx_storeu_dq_256(i8* %a0, <32 x i8> %a1) {
   ret void
 }
 declare void @llvm.x86.avx.storeu.dq.256(i8*, <32 x i8>) nounwind
+
+define <32 x i8> @mm256_max_epi8(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: mm256_max_epi8:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxsb %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8> %a0, <32 x i8> %a1)
+  ret <32 x i8> %res
+}
+declare <32 x i8> @llvm.x86.avx2.pmaxs.b(<32 x i8>, <32 x i8>) nounwind readnone
+
+define <16 x i16> @mm256_max_epi16(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: mm256_max_epi16:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxsw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16> %a0, <16 x i16> %a1)
+  ret <16 x i16> %res
+}
+declare <16 x i16> @llvm.x86.avx2.pmaxs.w(<16 x i16>, <16 x i16>) nounwind readnone
+
+define <8 x i32> @mm256_max_epi32(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: mm256_max_epi32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxsd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32> %a0, <8 x i32> %a1)
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.pmaxs.d(<8 x i32>, <8 x i32>) nounwind readnone
+
+define <32 x i8> @mm256_max_epu8(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: mm256_max_epu8:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxub %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8> %a0, <32 x i8> %a1)
+  ret <32 x i8> %res
+}
+declare <32 x i8> @llvm.x86.avx2.pmaxu.b(<32 x i8>, <32 x i8>) nounwind readnone
+
+define <16 x i16> @mm256_max_epu16(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: mm256_max_epu16:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxuw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16> %a0, <16 x i16> %a1)
+  ret <16 x i16> %res
+}
+declare <16 x i16> @llvm.x86.avx2.pmaxu.w(<16 x i16>, <16 x i16>) nounwind readnone
+
+define <8 x i32> @mm256_max_epu32(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: mm256_max_epu32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpmaxud %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32> %a0, <8 x i32> %a1)
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.pmaxu.d(<8 x i32>, <8 x i32>) nounwind readnone
+
+define <32 x i8> @mm256_min_epi8(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: mm256_min_epi8:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminsb %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8> %a0, <32 x i8> %a1)
+  ret <32 x i8> %res
+}
+declare <32 x i8> @llvm.x86.avx2.pmins.b(<32 x i8>, <32 x i8>) nounwind readnone
+
+define <16 x i16> @mm256_min_epi16(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: mm256_min_epi16:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminsw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16> %a0, <16 x i16> %a1)
+  ret <16 x i16> %res
+}
+declare <16 x i16> @llvm.x86.avx2.pmins.w(<16 x i16>, <16 x i16>) nounwind readnone
+
+define <8 x i32> @mm256_min_epi32(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: mm256_min_epi32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminsd %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32> %a0, <8 x i32> %a1)
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.pmins.d(<8 x i32>, <8 x i32>) nounwind readnone
+
+define <32 x i8> @mm256_min_epu8(<32 x i8> %a0, <32 x i8> %a1) {
+; CHECK-LABEL: mm256_min_epu8:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminub %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8> %a0, <32 x i8> %a1)
+  ret <32 x i8> %res
+}
+declare <32 x i8> @llvm.x86.avx2.pminu.b(<32 x i8>, <32 x i8>) nounwind readnone
+
+define <16 x i16> @mm256_min_epu16(<16 x i16> %a0, <16 x i16> %a1) {
+; CHECK-LABEL: mm256_min_epu16:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminuw %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16> %a0, <16 x i16> %a1)
+  ret <16 x i16> %res
+}
+declare <16 x i16> @llvm.x86.avx2.pminu.w(<16 x i16>, <16 x i16>) nounwind readnone
+
+define <8 x i32> @mm256_min_epu32(<8 x i32> %a0, <8 x i32> %a1) {
+; CHECK-LABEL: mm256_min_epu32:
+; CHECK:       ## BB#0:
+; CHECK-NEXT:    vpminud %ymm1, %ymm0, %ymm0
+; CHECK-NEXT:    retl
+;
+  %res = call <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32> %a0, <8 x i32> %a1)
+  ret <8 x i32> %res
+}
+declare <8 x i32> @llvm.x86.avx2.pminu.d(<8 x i32>, <8 x i32>) nounwind readnone
+