[X86] Add knownbits vector UREM/SREM tests

In preparation for demandedelts support

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@286577 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/test/CodeGen/X86/known-bits-vector.ll b/test/CodeGen/X86/known-bits-vector.ll
index a1efb9b176e1b8405b2caa96bfd9bb1c34bfdbfd..cb50d5f9611c148b6e9d000c686b3f4fe7aee282 100644 (file)
--- a/test/CodeGen/X86/known-bits-vector.ll
+++ b/test/CodeGen/X86/known-bits-vector.ll
@@ -255,3 +255,119 @@ define <4 x i32> @knownbits_mask_udiv_shuffle_lshr(<4 x i32> %a0, <4 x i32> %a1)
   %4 = lshr <4 x i32> %3, <i32 22, i32 22, i32 22, i32 22>
   ret <4 x i32> %4
 }
+
+define <4 x i32> @knownbits_urem_lshr(<4 x i32> %a0) nounwind {
+; X32-LABEL: knownbits_urem_lshr:
+; X32:       # BB#0:
+; X32-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: knownbits_urem_lshr:
+; X64:       # BB#0:
+; X64-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; X64-NEXT:    retq
+  %1 = urem <4 x i32> %a0, <i32 16, i32 16, i32 16, i32 16>
+  %2 = lshr <4 x i32> %1, <i32 22, i32 22, i32 22, i32 22>
+  ret <4 x i32> %2
+}
+
+define <4 x i32> @knownbits_mask_urem_shuffle_lshr(<4 x i32> %a0, <4 x i32> %a1) nounwind {
+; X32-LABEL: knownbits_mask_urem_shuffle_lshr:
+; X32:       # BB#0:
+; X32-NEXT:    pushl %esi
+; X32-NEXT:    vmovdqa {{.*#+}} xmm2 = [32767,4294967295,4294967295,32767]
+; X32-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; X32-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; X32-NEXT:    vpextrd $1, %xmm0, %eax
+; X32-NEXT:    vpextrd $1, %xmm1, %ecx
+; X32-NEXT:    xorl %edx, %edx
+; X32-NEXT:    divl %ecx
+; X32-NEXT:    movl %edx, %ecx
+; X32-NEXT:    vmovd %xmm0, %eax
+; X32-NEXT:    vmovd %xmm1, %esi
+; X32-NEXT:    xorl %edx, %edx
+; X32-NEXT:    divl %esi
+; X32-NEXT:    vmovd %edx, %xmm2
+; X32-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm2
+; X32-NEXT:    vpextrd $2, %xmm0, %eax
+; X32-NEXT:    vpextrd $2, %xmm1, %ecx
+; X32-NEXT:    xorl %edx, %edx
+; X32-NEXT:    divl %ecx
+; X32-NEXT:    vpinsrd $2, %edx, %xmm2, %xmm2
+; X32-NEXT:    vpextrd $3, %xmm0, %eax
+; X32-NEXT:    vpextrd $3, %xmm1, %ecx
+; X32-NEXT:    xorl %edx, %edx
+; X32-NEXT:    divl %ecx
+; X32-NEXT:    vpinsrd $3, %edx, %xmm2, %xmm0
+; X32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,3,3]
+; X32-NEXT:    vpsrld $22, %xmm0, %xmm0
+; X32-NEXT:    popl %esi
+; X32-NEXT:    retl
+;
+; X64-LABEL: knownbits_mask_urem_shuffle_lshr:
+; X64:       # BB#0:
+; X64-NEXT:    vmovdqa {{.*#+}} xmm2 = [32767,4294967295,4294967295,32767]
+; X64-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; X64-NEXT:    vpand %xmm2, %xmm1, %xmm1
+; X64-NEXT:    vpextrd $1, %xmm0, %eax
+; X64-NEXT:    vpextrd $1, %xmm1, %ecx
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    movl %edx, %ecx
+; X64-NEXT:    vmovd %xmm0, %eax
+; X64-NEXT:    vmovd %xmm1, %esi
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %esi
+; X64-NEXT:    vmovd %edx, %xmm2
+; X64-NEXT:    vpinsrd $1, %ecx, %xmm2, %xmm2
+; X64-NEXT:    vpextrd $2, %xmm0, %eax
+; X64-NEXT:    vpextrd $2, %xmm1, %ecx
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    vpinsrd $2, %edx, %xmm2, %xmm2
+; X64-NEXT:    vpextrd $3, %xmm0, %eax
+; X64-NEXT:    vpextrd $3, %xmm1, %ecx
+; X64-NEXT:    xorl %edx, %edx
+; X64-NEXT:    divl %ecx
+; X64-NEXT:    vpinsrd $3, %edx, %xmm2, %xmm0
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,3,3]
+; X64-NEXT:    vpsrld $22, %xmm0, %xmm0
+; X64-NEXT:    retq
+  %1 = and <4 x i32> %a0, <i32 32767, i32 -1, i32 -1, i32 32767>
+  %2 = and <4 x i32> %a1, <i32 32767, i32 -1, i32 -1, i32 32767>
+  %3 = urem <4 x i32> %1, %2
+  %4 = shufflevector <4 x i32> %3, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 3, i32 3>
+  %5 = lshr <4 x i32> %4, <i32 22, i32 22, i32 22, i32 22>
+  ret <4 x i32> %5
+}
+
+define <4 x i32> @knownbits_mask_srem_shuffle_lshr(<4 x i32> %a0) nounwind {
+; X32-LABEL: knownbits_mask_srem_shuffle_lshr:
+; X32:       # BB#0:
+; X32-NEXT:    vpand {{\.LCPI.*}}, %xmm0, %xmm0
+; X32-NEXT:    vpsrad $31, %xmm0, %xmm1
+; X32-NEXT:    vpsrld $28, %xmm1, %xmm1
+; X32-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; X32-NEXT:    vpand {{\.LCPI.*}}, %xmm1, %xmm1
+; X32-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; X32-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,3,3]
+; X32-NEXT:    vpsrld $22, %xmm0, %xmm0
+; X32-NEXT:    retl
+;
+; X64-LABEL: knownbits_mask_srem_shuffle_lshr:
+; X64:       # BB#0:
+; X64-NEXT:    vpand {{.*}}(%rip), %xmm0, %xmm0
+; X64-NEXT:    vpsrad $31, %xmm0, %xmm1
+; X64-NEXT:    vpsrld $28, %xmm1, %xmm1
+; X64-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; X64-NEXT:    vpand {{.*}}(%rip), %xmm1, %xmm1
+; X64-NEXT:    vpsubd %xmm1, %xmm0, %xmm0
+; X64-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,3,3]
+; X64-NEXT:    vpsrld $22, %xmm0, %xmm0
+; X64-NEXT:    retq
+  %1 = and <4 x i32> %a0, <i32 -32768, i32 -1, i32 -1, i32 -32768>
+  %2 = srem <4 x i32> %1, <i32 16, i32 16, i32 16, i32 16>
+  %3 = shufflevector <4 x i32> %2, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 3, i32 3>
+  %4 = lshr <4 x i32> %3, <i32 22, i32 22, i32 22, i32 22>
+  ret <4 x i32> %4
+}