[X86][AVX512] add comi with Sae

add builtin_ia32_vcomisd and builtin_ia32_vcomisd

Differential Revision: http://reviews.llvm.org/D14331

llvm-svn: 252153

2 files changed, 202 insertions, 0 deletions

diff --git a/llvm/test/CodeGen/X86/avx512-intrinsics.ll b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
index 6ce1da55aa4..923ad4d069f 100644
--- a/llvm/test/CodeGen/X86/avx512-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/avx512-intrinsics.ll
@@ -4676,3 +4676,78 @@ define <8 x i64>@test_int_x86_avx512_maskz_pternlog_q_512(<8 x i64> %x0, <8 x i6
   ret <8 x i64> %res2
 }
 
+define i32 @test_x86_avx512_comi_sd_eq_sae(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_comi_sd_eq_sae
+; CHECK: vcomisd {sae}, %xmm1, %xmm0
+; CHECK-NEXT: sete	%al
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 0, i32 8) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_ucomi_sd_eq_sae(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_ucomi_sd_eq_sae
+; CHECK: vucomisd {sae}, %xmm1, %xmm0
+; CHECK-NEXT: sete %al
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 8, i32 8) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_comi_sd_eq(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_comi_sd_eq
+; CHECK: vcomisd %xmm1, %xmm0
+; CHECK-NEXT: sete %al
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 0, i32 4) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_ucomi_sd_eq(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_ucomi_sd_eq
+; CHECK: vucomisd %xmm1, %xmm0
+; CHECK-NEXT: sete %al
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 8, i32 4) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_comi_sd_lt_sae(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_comi_sd_lt_sae
+; CHECK: vcomisd {sae}, %xmm1, %xmm0
+; CHECK-NEXT: sbbl %eax, %eax
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 1, i32 8) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_ucomi_sd_lt_sae(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_ucomi_sd_lt_sae
+; CHECK: vucomisd {sae}, %xmm1, %xmm0
+; CHECK-NEXT: sbbl %eax, %eax
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 9, i32 8) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_comi_sd_lt(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_comi_sd_lt
+; CHECK: vcomisd %xmm1, %xmm0
+; CHECK-NEXT: sbbl %eax, %eax
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 1, i32 4) 
+  ret i32 %res
+}
+
+define i32 @test_x86_avx512_ucomi_sd_lt(<2 x double> %a0, <2 x double> %a1) {
+; CHECK-LABEL: test_x86_avx512_ucomi_sd_lt
+; CHECK: vucomisd %xmm1, %xmm0
+; CHECK-NEXT: sbbl %eax, %eax
+  %res = call i32 @llvm.x86.avx512.vcomi.sd(<2 x double> %a0, <2 x double> %a1, i32 9, i32 4) 
+  ret i32 %res
+}
+
+declare i32 @llvm.x86.avx512.vcomi.sd(<2 x double>, <2 x double>, i32, i32) 
+
+define i32 @test_x86_avx512_ucomi_ss_lt(<4 x float> %a0, <4 x float> %a1) {
+; CHECK-LABEL: test_x86_avx512_ucomi_ss_lt
+; CHECK: vucomiss %xmm1, %xmm0
+; CHECK-NEXT: sbbl %eax, %eax
+  %res = call i32 @llvm.x86.avx512.vcomi.ss(<4 x float> %a0, <4 x float> %a1, i32 9, i32 4) 
+  ret i32 %res
+}
+
+declare i32 @llvm.x86.avx512.vcomi.ss(<4 x float>, <4 x float>, i32, i32)
diff --git a/llvm/test/MC/X86/avx512-encodings.s b/llvm/test/MC/X86/avx512-encodings.s
index dc0ee7e671b..98ec9ff297c 100644
--- a/llvm/test/MC/X86/avx512-encodings.s
+++ b/llvm/test/MC/X86/avx512-encodings.s
@@ -18297,3 +18297,130 @@ vpermilpd $0x23, 0x400(%rbx), %zmm2
 // CHECK:  encoding: [0xc5,0xf9,0x7e,0xaa,0xfc,0xfd,0xff,0xff]
           vmovd  %xmm5, -516(%rdx)
 
+// CHECK: vcomisd %xmm21, %xmm23
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x08,0x2f,0xfd]
+          vcomisd %xmm21, %xmm23
+
+// CHECK: vcomisd {sae}, %xmm21, %xmm23
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x18,0x2f,0xfd]
+          vcomisd {sae}, %xmm21, %xmm23
+
+// CHECK: vcomisd (%rcx), %xmm23
+// CHECK:  encoding: [0x62,0xe1,0xfd,0x08,0x2f,0x39]
+          vcomisd (%rcx), %xmm23
+
+// CHECK: vcomisd 291(%rax,%r14,8), %xmm23
+// CHECK:  encoding: [0x62,0xa1,0xfd,0x08,0x2f,0xbc,0xf0,0x23,0x01,0x00,0x00]
+          vcomisd 291(%rax,%r14,8), %xmm23
+
+// CHECK: vcomisd 1016(%rdx), %xmm23
+// CHECK:  encoding: [0x62,0xe1,0xfd,0x08,0x2f,0x7a,0x7f]
+          vcomisd 1016(%rdx), %xmm23
+
+// CHECK: vcomisd 1024(%rdx), %xmm23
+// CHECK:  encoding: [0x62,0xe1,0xfd,0x08,0x2f,0xba,0x00,0x04,0x00,0x00]
+          vcomisd 1024(%rdx), %xmm23
+
+// CHECK: vcomisd -1024(%rdx), %xmm23
+// CHECK:  encoding: [0x62,0xe1,0xfd,0x08,0x2f,0x7a,0x80]
+          vcomisd -1024(%rdx), %xmm23
+
+// CHECK: vcomisd -1032(%rdx), %xmm23
+// CHECK:  encoding: [0x62,0xe1,0xfd,0x08,0x2f,0xba,0xf8,0xfb,0xff,0xff]
+          vcomisd -1032(%rdx), %xmm23
+
+// CHECK: vcomiss %xmm28, %xmm14
+// CHECK:  encoding: [0x62,0x11,0x7c,0x08,0x2f,0xf4]
+          vcomiss %xmm28, %xmm14
+
+// CHECK: vcomiss {sae}, %xmm28, %xmm14
+// CHECK:  encoding: [0x62,0x11,0x7c,0x18,0x2f,0xf4]
+          vcomiss {sae}, %xmm28, %xmm14
+
+// CHECK: vcomiss (%rcx), %xmm14
+// CHECK:  encoding: [0xc5,0x78,0x2f,0x31]
+          vcomiss (%rcx), %xmm14
+
+// CHECK: vcomiss 291(%rax,%r14,8), %xmm14
+// CHECK:  encoding: [0xc4,0x21,0x78,0x2f,0xb4,0xf0,0x23,0x01,0x00,0x00]
+          vcomiss 291(%rax,%r14,8), %xmm14
+
+// CHECK: vcomiss 508(%rdx), %xmm14
+// CHECK:  encoding: [0xc5,0x78,0x2f,0xb2,0xfc,0x01,0x00,0x00]
+          vcomiss 508(%rdx), %xmm14
+
+// CHECK: vcomiss 512(%rdx), %xmm14
+// CHECK:  encoding: [0xc5,0x78,0x2f,0xb2,0x00,0x02,0x00,0x00]
+          vcomiss 512(%rdx), %xmm14
+
+// CHECK: vcomiss -512(%rdx), %xmm14
+// CHECK:  encoding: [0xc5,0x78,0x2f,0xb2,0x00,0xfe,0xff,0xff]
+          vcomiss -512(%rdx), %xmm14
+
+// CHECK: vcomiss -516(%rdx), %xmm14
+// CHECK:  encoding: [0xc5,0x78,0x2f,0xb2,0xfc,0xfd,0xff,0xff]
+          vcomiss -516(%rdx), %xmm14
+
+// CHECK: vucomisd %xmm10, %xmm11
+// CHECK:  encoding: [0xc4,0x41,0x79,0x2e,0xda]
+          vucomisd %xmm10, %xmm11
+
+// CHECK: vucomisd {sae}, %xmm10, %xmm11
+// CHECK:  encoding: [0x62,0x51,0xfd,0x18,0x2e,0xda]
+          vucomisd {sae}, %xmm10, %xmm11
+
+// CHECK: vucomisd (%rcx), %xmm11
+// CHECK:  encoding: [0xc5,0x79,0x2e,0x19]
+          vucomisd (%rcx), %xmm11
+
+// CHECK: vucomisd 291(%rax,%r14,8), %xmm11
+// CHECK:  encoding: [0xc4,0x21,0x79,0x2e,0x9c,0xf0,0x23,0x01,0x00,0x00]
+          vucomisd 291(%rax,%r14,8), %xmm11
+
+// CHECK: vucomisd 1016(%rdx), %xmm11
+// CHECK:  encoding: [0xc5,0x79,0x2e,0x9a,0xf8,0x03,0x00,0x00]
+          vucomisd 1016(%rdx), %xmm11
+
+// CHECK: vucomisd 1024(%rdx), %xmm11
+// CHECK:  encoding: [0xc5,0x79,0x2e,0x9a,0x00,0x04,0x00,0x00]
+          vucomisd 1024(%rdx), %xmm11
+
+// CHECK: vucomisd -1024(%rdx), %xmm11
+// CHECK:  encoding: [0xc5,0x79,0x2e,0x9a,0x00,0xfc,0xff,0xff]
+          vucomisd -1024(%rdx), %xmm11
+
+// CHECK: vucomisd -1032(%rdx), %xmm11
+// CHECK:  encoding: [0xc5,0x79,0x2e,0x9a,0xf8,0xfb,0xff,0xff]
+          vucomisd -1032(%rdx), %xmm11
+
+// CHECK: vucomiss %xmm11, %xmm22
+// CHECK:  encoding: [0x62,0xc1,0x7c,0x08,0x2e,0xf3]
+          vucomiss %xmm11, %xmm22
+
+// CHECK: vucomiss {sae}, %xmm11, %xmm22
+// CHECK:  encoding: [0x62,0xc1,0x7c,0x18,0x2e,0xf3]
+          vucomiss {sae}, %xmm11, %xmm22
+
+// CHECK: vucomiss (%rcx), %xmm22
+// CHECK:  encoding: [0x62,0xe1,0x7c,0x08,0x2e,0x31]
+          vucomiss (%rcx), %xmm22
+
+// CHECK: vucomiss 291(%rax,%r14,8), %xmm22
+// CHECK:  encoding: [0x62,0xa1,0x7c,0x08,0x2e,0xb4,0xf0,0x23,0x01,0x00,0x00]
+          vucomiss 291(%rax,%r14,8), %xmm22
+
+// CHECK: vucomiss 508(%rdx), %xmm22
+// CHECK:  encoding: [0x62,0xe1,0x7c,0x08,0x2e,0x72,0x7f]
+          vucomiss 508(%rdx), %xmm22
+
+// CHECK: vucomiss 512(%rdx), %xmm22
+// CHECK:  encoding: [0x62,0xe1,0x7c,0x08,0x2e,0xb2,0x00,0x02,0x00,0x00]
+          vucomiss 512(%rdx), %xmm22
+
+// CHECK: vucomiss -512(%rdx), %xmm22
+// CHECK:  encoding: [0x62,0xe1,0x7c,0x08,0x2e,0x72,0x80]
+          vucomiss -512(%rdx), %xmm22
+
+// CHECK: vucomiss -516(%rdx), %xmm22
+// CHECK:  encoding: [0x62,0xe1,0x7c,0x08,0x2e,0xb2,0xfc,0xfd,0xff,0xff]
+          vucomiss -516(%rdx), %xmm22