[X86] Replace VPCOM/VPCOMU with generic integer comparisons (llvm)

These intrinsics can always be replaced with generic integer comparisons without any regression in codegen, even for -O0/-fast-isel cases.

Noticed while cleaning up vector integer comparison costs for PR40376.

A future commit will remove/autoupgrade the existing VPCOM/VPCOMU llvm intrinsics.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@351688 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/IR/IntrinsicsX86.td b/include/llvm/IR/IntrinsicsX86.td
index d0617820aa7c803373d9960d2c16a5abcea16113..ea0b03b68a72651a8bd6faf470739ecfbe946de1 100644 (file)
--- a/include/llvm/IR/IntrinsicsX86.td
+++ b/include/llvm/IR/IntrinsicsX86.td
@@ -1908,28 +1908,28 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_xop_vfrcz_ps_256 : GCCBuiltin<"__builtin_ia32_vfrczps256">,
               Intrinsic<[llvm_v8f32_ty], [llvm_v8f32_ty], [IntrNoMem]>;
 
-  def int_x86_xop_vpcomb : GCCBuiltin<"__builtin_ia32_vpcomb">,
+  def int_x86_xop_vpcomb :
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty, llvm_v16i8_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomw : GCCBuiltin<"__builtin_ia32_vpcomw">,
+  def int_x86_xop_vpcomw :
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomd : GCCBuiltin<"__builtin_ia32_vpcomd">,
+  def int_x86_xop_vpcomd :
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomq : GCCBuiltin<"__builtin_ia32_vpcomq">,
+  def int_x86_xop_vpcomq :
               Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomub : GCCBuiltin<"__builtin_ia32_vpcomub">,
+  def int_x86_xop_vpcomub :
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty, llvm_v16i8_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomuw : GCCBuiltin<"__builtin_ia32_vpcomuw">,
+  def int_x86_xop_vpcomuw :
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty, llvm_v8i16_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomud : GCCBuiltin<"__builtin_ia32_vpcomud">,
+  def int_x86_xop_vpcomud :
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty, llvm_v4i32_ty,
                          llvm_i8_ty], [IntrNoMem]>;
-  def int_x86_xop_vpcomuq : GCCBuiltin<"__builtin_ia32_vpcomuq">,
+  def int_x86_xop_vpcomuq :
               Intrinsic<[llvm_v2i64_ty], [llvm_v2i64_ty, llvm_v2i64_ty,
                          llvm_i8_ty], [IntrNoMem]>;
 

diff --git a/test/CodeGen/X86/xop-intrinsics-fast-isel.ll b/test/CodeGen/X86/xop-intrinsics-fast-isel.ll
index 43f59708db52fb722dbaec47c7b2a17418fefabd..037c8f123aad8cb16e13e18e4dd0d58d87303332 100644 (file)
--- a/test/CodeGen/X86/xop-intrinsics-fast-isel.ll
+++ b/test/CodeGen/X86/xop-intrinsics-fast-isel.ll
@@ -577,11 +577,11 @@ define <2 x i64> @test_mm_com_epu8(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <16 x i8>
   %arg1 = bitcast <2 x i64> %a1 to <16 x i8>
-  %res = call <16 x i8> @llvm.x86.xop.vpcomub(<16 x i8> %arg0, <16 x i8> %arg1, i8 0)
+  %cmp = icmp ult <16 x i8> %arg0, %arg1
+  %res = sext <16 x i1> %cmp to <16 x i8>
   %bc = bitcast <16 x i8> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <16 x i8> @llvm.x86.xop.vpcomub(<16 x i8>, <16 x i8>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epu16(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epu16:
@@ -590,11 +590,11 @@ define <2 x i64> @test_mm_com_epu16(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <8 x i16>
   %arg1 = bitcast <2 x i64> %a1 to <8 x i16>
-  %res = call <8 x i16> @llvm.x86.xop.vpcomuw(<8 x i16> %arg0, <8 x i16> %arg1, i8 0)
+  %cmp = icmp ult <8 x i16> %arg0, %arg1
+  %res = sext <8 x i1> %cmp to <8 x i16>
   %bc = bitcast <8 x i16> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <8 x i16> @llvm.x86.xop.vpcomuw(<8 x i16>, <8 x i16>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epu32(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epu32:
@@ -603,21 +603,21 @@ define <2 x i64> @test_mm_com_epu32(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <4 x i32>
   %arg1 = bitcast <2 x i64> %a1 to <4 x i32>
-  %res = call <4 x i32> @llvm.x86.xop.vpcomud(<4 x i32> %arg0, <4 x i32> %arg1, i8 0)
+  %cmp = icmp ult <4 x i32> %arg0, %arg1
+  %res = sext <4 x i1> %cmp to <4 x i32>
   %bc = bitcast <4 x i32> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <4 x i32> @llvm.x86.xop.vpcomud(<4 x i32>, <4 x i32>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epu64(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epu64:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vpcomltuq %xmm1, %xmm0, %xmm0
 ; ALL-NEXT:    ret{{[l|q]}}
-  %res = call <2 x i64> @llvm.x86.xop.vpcomuq(<2 x i64> %a0, <2 x i64> %a1, i8 0)
+  %cmp = icmp ult <2 x i64> %a0, %a1
+  %res = sext <2 x i1> %cmp to <2 x i64>
   ret <2 x i64> %res
 }
-declare <2 x i64> @llvm.x86.xop.vpcomuq(<2 x i64>, <2 x i64>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epi8(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epi8:
@@ -626,11 +626,11 @@ define <2 x i64> @test_mm_com_epi8(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <16 x i8>
   %arg1 = bitcast <2 x i64> %a1 to <16 x i8>
-  %res = call <16 x i8> @llvm.x86.xop.vpcomb(<16 x i8> %arg0, <16 x i8> %arg1, i8 0)
+  %cmp = icmp slt <16 x i8> %arg0, %arg1
+  %res = sext <16 x i1> %cmp to <16 x i8>
   %bc = bitcast <16 x i8> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <16 x i8> @llvm.x86.xop.vpcomb(<16 x i8>, <16 x i8>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epi16(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epi16:
@@ -639,11 +639,11 @@ define <2 x i64> @test_mm_com_epi16(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <8 x i16>
   %arg1 = bitcast <2 x i64> %a1 to <8 x i16>
-  %res = call <8 x i16> @llvm.x86.xop.vpcomw(<8 x i16> %arg0, <8 x i16> %arg1, i8 0)
+  %cmp = icmp slt <8 x i16> %arg0, %arg1
+  %res = sext <8 x i1> %cmp to <8 x i16>
   %bc = bitcast <8 x i16> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <8 x i16> @llvm.x86.xop.vpcomw(<8 x i16>, <8 x i16>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epi32(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epi32:
@@ -652,21 +652,21 @@ define <2 x i64> @test_mm_com_epi32(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-NEXT:    ret{{[l|q]}}
   %arg0 = bitcast <2 x i64> %a0 to <4 x i32>
   %arg1 = bitcast <2 x i64> %a1 to <4 x i32>
-  %res = call <4 x i32> @llvm.x86.xop.vpcomd(<4 x i32> %arg0, <4 x i32> %arg1, i8 0)
+  %cmp = icmp slt <4 x i32> %arg0, %arg1
+  %res = sext <4 x i1> %cmp to <4 x i32>
   %bc = bitcast <4 x i32> %res to <2 x i64>
   ret <2 x i64> %bc
 }
-declare <4 x i32> @llvm.x86.xop.vpcomd(<4 x i32>, <4 x i32>, i8) nounwind readnone
 
 define <2 x i64> @test_mm_com_epi64(<2 x i64> %a0, <2 x i64> %a1) {
 ; ALL-LABEL: test_mm_com_epi64:
 ; ALL:       # %bb.0:
 ; ALL-NEXT:    vpcomltq %xmm1, %xmm0, %xmm0
 ; ALL-NEXT:    ret{{[l|q]}}
-  %res = call <2 x i64> @llvm.x86.xop.vpcomq(<2 x i64> %a0, <2 x i64> %a1, i8 0)
+  %cmp = icmp slt <2 x i64> %a0, %a1
+  %res = sext <2 x i1> %cmp to <2 x i64>
   ret <2 x i64> %res
 }
-declare <2 x i64> @llvm.x86.xop.vpcomq(<2 x i64>, <2 x i64>, i8) nounwind readnone
 
 define <2 x double> @test_mm_permute2_pd(<2 x double> %a0, <2 x double> %a1, <2 x i64> %a2) {
 ; ALL-LABEL: test_mm_permute2_pd: