[X86] Remove type restrictions from WidenMaskArithmetic.

This can help AVX-512 code where mask types are legal allowing us to remove extends and truncates to/from mask types.

llvm-svn: 321408

3 files changed, 34 insertions, 87 deletions

diff --git a/llvm/test/CodeGen/X86/avx512-calling-conv.ll b/llvm/test/CodeGen/X86/avx512-calling-conv.ll
index 6e6d61f37d2..248462d0de5 100644
--- a/llvm/test/CodeGen/X86/avx512-calling-conv.ll
+++ b/llvm/test/CodeGen/X86/avx512-calling-conv.ll
@@ -17,78 +17,40 @@ define <16 x i1> @test1() {
 }
 
 define <16 x i1> @test2(<16 x i1>%a, <16 x i1>%b) {
-; KNL-LABEL: test2:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    vpmovsxbd %xmm1, %zmm1
-; KNL-NEXT:    vpslld $31, %zmm1, %zmm1
-; KNL-NEXT:    vpmovsxbd %xmm0, %zmm0
-; KNL-NEXT:    vpslld $31, %zmm0, %zmm0
-; KNL-NEXT:    vptestmd %zmm0, %zmm0, %k1
-; KNL-NEXT:    vptestmd %zmm1, %zmm1, %k1 {%k1}
-; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; KNL-NEXT:    vpmovdb %zmm0, %xmm0
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test2:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    vpsllw $7, %xmm1, %xmm1
-; SKX-NEXT:    vpmovb2m %xmm1, %k0
-; SKX-NEXT:    vpsllw $7, %xmm0, %xmm0
-; SKX-NEXT:    vpmovb2m %xmm0, %k1
-; SKX-NEXT:    kandw %k0, %k1, %k0
-; SKX-NEXT:    vpmovm2b %k0, %xmm0
-; SKX-NEXT:    retq
+; ALL_X64-LABEL: test2:
+; ALL_X64:       ## %bb.0:
+; ALL_X64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; ALL_X64-NEXT:    vpsllw $7, %xmm0, %xmm0
+; ALL_X64-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; ALL_X64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; ALL_X64-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; ALL_X64-NEXT:    retq
 ;
 ; KNL_X32-LABEL: test2:
 ; KNL_X32:       ## %bb.0:
-; KNL_X32-NEXT:    vpmovsxbd %xmm1, %zmm1
-; KNL_X32-NEXT:    vpslld $31, %zmm1, %zmm1
-; KNL_X32-NEXT:    vpmovsxbd %xmm0, %zmm0
-; KNL_X32-NEXT:    vpslld $31, %zmm0, %zmm0
-; KNL_X32-NEXT:    vptestmd %zmm0, %zmm0, %k1
-; KNL_X32-NEXT:    vptestmd %zmm1, %zmm1, %k1 {%k1}
-; KNL_X32-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; KNL_X32-NEXT:    vpmovdb %zmm0, %xmm0
+; KNL_X32-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; KNL_X32-NEXT:    vpsllw $7, %xmm0, %xmm0
+; KNL_X32-NEXT:    vpand LCPI1_0, %xmm0, %xmm0
+; KNL_X32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; KNL_X32-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
 ; KNL_X32-NEXT:    retl
   %c = and <16 x i1>%a, %b
   ret <16 x i1> %c
 }
 
 define <8 x i1> @test3(<8 x i1>%a, <8 x i1>%b) {
-; KNL-LABEL: test3:
-; KNL:       ## %bb.0:
-; KNL-NEXT:    vpmovsxwq %xmm1, %zmm1
-; KNL-NEXT:    vpsllq $63, %zmm1, %zmm1
-; KNL-NEXT:    vpmovsxwq %xmm0, %zmm0
-; KNL-NEXT:    vpsllq $63, %zmm0, %zmm0
-; KNL-NEXT:    vptestmq %zmm0, %zmm0, %k1
-; KNL-NEXT:    vptestmq %zmm1, %zmm1, %k1 {%k1}
-; KNL-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; KNL-NEXT:    vpmovdw %zmm0, %ymm0
-; KNL-NEXT:    ## kill: def %xmm0 killed %xmm0 killed %ymm0
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: test3:
-; SKX:       ## %bb.0:
-; SKX-NEXT:    vpsllw $15, %xmm1, %xmm1
-; SKX-NEXT:    vpmovw2m %xmm1, %k0
-; SKX-NEXT:    vpsllw $15, %xmm0, %xmm0
-; SKX-NEXT:    vpmovw2m %xmm0, %k1
-; SKX-NEXT:    kandb %k0, %k1, %k0
-; SKX-NEXT:    vpmovm2w %k0, %xmm0
-; SKX-NEXT:    retq
+; ALL_X64-LABEL: test3:
+; ALL_X64:       ## %bb.0:
+; ALL_X64-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; ALL_X64-NEXT:    vpsllw $15, %xmm0, %xmm0
+; ALL_X64-NEXT:    vpsraw $15, %xmm0, %xmm0
+; ALL_X64-NEXT:    retq
 ;
 ; KNL_X32-LABEL: test3:
 ; KNL_X32:       ## %bb.0:
-; KNL_X32-NEXT:    vpmovsxwq %xmm1, %zmm1
-; KNL_X32-NEXT:    vpsllq $63, %zmm1, %zmm1
-; KNL_X32-NEXT:    vpmovsxwq %xmm0, %zmm0
-; KNL_X32-NEXT:    vpsllq $63, %zmm0, %zmm0
-; KNL_X32-NEXT:    vptestmq %zmm0, %zmm0, %k1
-; KNL_X32-NEXT:    vptestmq %zmm1, %zmm1, %k1 {%k1}
-; KNL_X32-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; KNL_X32-NEXT:    vpmovdw %zmm0, %ymm0
-; KNL_X32-NEXT:    ## kill: def %xmm0 killed %xmm0 killed %ymm0
+; KNL_X32-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; KNL_X32-NEXT:    vpsllw $15, %xmm0, %xmm0
+; KNL_X32-NEXT:    vpsraw $15, %xmm0, %xmm0
 ; KNL_X32-NEXT:    retl
   %c = and <8 x i1>%a, %b
   ret <8 x i1> %c
@@ -102,11 +64,9 @@ define <4 x i1> @test4(<4 x i1>%a, <4 x i1>%b) {
 ;
 ; SKX-LABEL: test4:
 ; SKX:       ## %bb.0:
-; SKX-NEXT:    vpslld $31, %xmm1, %xmm1
+; SKX-NEXT:    vpand %xmm1, %xmm0, %xmm0
 ; SKX-NEXT:    vpslld $31, %xmm0, %xmm0
-; SKX-NEXT:    vptestmd %xmm0, %xmm0, %k1
-; SKX-NEXT:    vptestmd %xmm1, %xmm1, %k0 {%k1}
-; SKX-NEXT:    vpmovm2d %k0, %xmm0
+; SKX-NEXT:    vpsrad $31, %xmm0, %xmm0
 ; SKX-NEXT:    retq
 ;
 ; KNL_X32-LABEL: test4:
diff --git a/llvm/test/CodeGen/X86/avx512-ext.ll b/llvm/test/CodeGen/X86/avx512-ext.ll
index 97beff63811..8c794159121 100644
--- a/llvm/test/CodeGen/X86/avx512-ext.ll
+++ b/llvm/test/CodeGen/X86/avx512-ext.ll
@@ -1366,21 +1366,12 @@ define i16 @trunc_16i32_to_16i1(<16 x i32> %a) {
 }
 
 define <4 x i32> @trunc_4i32_to_4i1(<4 x i32> %a, <4 x i32> %b) {
-; KNL-LABEL: trunc_4i32_to_4i1:
-; KNL:       # %bb.0:
-; KNL-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; KNL-NEXT:    vpslld $31, %xmm0, %xmm0
-; KNL-NEXT:    vpsrad $31, %xmm0, %xmm0
-; KNL-NEXT:    retq
-;
-; SKX-LABEL: trunc_4i32_to_4i1:
-; SKX:       # %bb.0:
-; SKX-NEXT:    vpslld $31, %xmm0, %xmm0
-; SKX-NEXT:    vptestmd %xmm0, %xmm0, %k1
-; SKX-NEXT:    vpslld $31, %xmm1, %xmm0
-; SKX-NEXT:    vptestmd %xmm0, %xmm0, %k0 {%k1}
-; SKX-NEXT:    vpmovm2d %k0, %xmm0
-; SKX-NEXT:    retq
+; ALL-LABEL: trunc_4i32_to_4i1:
+; ALL:       # %bb.0:
+; ALL-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; ALL-NEXT:    vpslld $31, %xmm0, %xmm0
+; ALL-NEXT:    vpsrad $31, %xmm0, %xmm0
+; ALL-NEXT:    retq
   %mask_a = trunc <4 x i32>%a to <4 x i1>
   %mask_b = trunc <4 x i32>%b to <4 x i1>
   %a_and_b = and <4 x i1>%mask_a, %mask_b
diff --git a/llvm/test/CodeGen/X86/avx512-schedule.ll b/llvm/test/CodeGen/X86/avx512-schedule.ll
index 1b450b98a6d..78111874b58 100755
--- a/llvm/test/CodeGen/X86/avx512-schedule.ll
+++ b/llvm/test/CodeGen/X86/avx512-schedule.ll
@@ -4376,20 +4376,16 @@ define i16 @trunc_16i32_to_16i1(<16 x i32> %a) {
 define <4 x i32> @trunc_4i32_to_4i1(<4 x i32> %a, <4 x i32> %b) {
 ; GENERIC-LABEL: trunc_4i32_to_4i1:
 ; GENERIC:       # %bb.0:
+; GENERIC-NEXT:    vpand %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
 ; GENERIC-NEXT:    vpslld $31, %xmm0, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT:    vptestmd %xmm0, %xmm0, %k1 # sched: [1:1.00]
-; GENERIC-NEXT:    vpslld $31, %xmm1, %xmm0 # sched: [1:1.00]
-; GENERIC-NEXT:    vptestmd %xmm0, %xmm0, %k0 {%k1} # sched: [1:1.00]
-; GENERIC-NEXT:    vpmovm2d %k0, %xmm0 # sched: [1:0.33]
+; GENERIC-NEXT:    vpsrad $31, %xmm0, %xmm0 # sched: [1:1.00]
 ; GENERIC-NEXT:    retq # sched: [1:1.00]
 ;
 ; SKX-LABEL: trunc_4i32_to_4i1:
 ; SKX:       # %bb.0:
+; SKX-NEXT:    vpand %xmm1, %xmm0, %xmm0 # sched: [1:0.33]
 ; SKX-NEXT:    vpslld $31, %xmm0, %xmm0 # sched: [1:0.50]
-; SKX-NEXT:    vptestmd %xmm0, %xmm0, %k1 # sched: [3:1.00]
-; SKX-NEXT:    vpslld $31, %xmm1, %xmm0 # sched: [1:0.50]
-; SKX-NEXT:    vptestmd %xmm0, %xmm0, %k0 {%k1} # sched: [3:1.00]
-; SKX-NEXT:    vpmovm2d %k0, %xmm0 # sched: [1:0.25]
+; SKX-NEXT:    vpsrad $31, %xmm0, %xmm0 # sched: [1:0.50]
 ; SKX-NEXT:    retq # sched: [7:1.00]
   %mask_a = trunc <4 x i32>%a to <4 x i1>
   %mask_b = trunc <4 x i32>%b to <4 x i1>