; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-- | FileCheck %s
;
; 32-bit float to unsigned integer
;
declare <4 x i1> @llvm.fptoui.sat.v4i1.v4f32(<4 x float>)
declare <4 x i8> @llvm.fptoui.sat.v4i8.v4f32(<4 x float>)
declare <4 x i16> @llvm.fptoui.sat.v4i16.v4f32(<4 x float>)
declare <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float>)
declare <4 x i64> @llvm.fptoui.sat.v4i64.v4f32(<4 x float>)
declare <4 x i128> @llvm.fptoui.sat.v4i128.v4f32(<4 x float>)
define <4 x i1> @test_unsigned_v4i1_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i1_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT: xorps %xmm2, %xmm2
; CHECK-NEXT: maxss %xmm2, %xmm1
; CHECK-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; CHECK-NEXT: minss %xmm3, %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %eax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: movaps %xmm0, %xmm4
; CHECK-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
; CHECK-NEXT: maxss %xmm2, %xmm4
; CHECK-NEXT: minss %xmm3, %xmm4
; CHECK-NEXT: cvttss2si %xmm4, %eax
; CHECK-NEXT: movd %eax, %xmm4
; CHECK-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: maxss %xmm2, %xmm1
; CHECK-NEXT: minss %xmm3, %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %eax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: maxss %xmm2, %xmm0
; CHECK-NEXT: minss %xmm3, %xmm0
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <4 x i1> @llvm.fptoui.sat.v4i1.v4f32(<4 x float> %f)
ret <4 x i1> %x
}
define <4 x i8> @test_unsigned_v4i8_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i8_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT: xorps %xmm2, %xmm2
; CHECK-NEXT: xorps %xmm3, %xmm3
; CHECK-NEXT: maxss %xmm1, %xmm3
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: movaps %xmm1, %xmm4
; CHECK-NEXT: minss %xmm3, %xmm4
; CHECK-NEXT: cvttss2si %xmm4, %eax
; CHECK-NEXT: shll $8, %eax
; CHECK-NEXT: movaps %xmm0, %xmm3
; CHECK-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
; CHECK-NEXT: xorps %xmm4, %xmm4
; CHECK-NEXT: maxss %xmm3, %xmm4
; CHECK-NEXT: movaps %xmm1, %xmm3
; CHECK-NEXT: minss %xmm4, %xmm3
; CHECK-NEXT: cvttss2si %xmm3, %ecx
; CHECK-NEXT: movzbl %cl, %ecx
; CHECK-NEXT: orl %eax, %ecx
; CHECK-NEXT: xorps %xmm3, %xmm3
; CHECK-NEXT: maxss %xmm0, %xmm3
; CHECK-NEXT: movaps %xmm1, %xmm4
; CHECK-NEXT: minss %xmm3, %xmm4
; CHECK-NEXT: cvttss2si %xmm4, %eax
; CHECK-NEXT: movzbl %al, %eax
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: maxss %xmm0, %xmm2
; CHECK-NEXT: minss %xmm2, %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %edx
; CHECK-NEXT: shll $8, %edx
; CHECK-NEXT: orl %eax, %edx
; CHECK-NEXT: movd %edx, %xmm0
; CHECK-NEXT: pinsrw $1, %ecx, %xmm0
; CHECK-NEXT: retq
%x = call <4 x i8> @llvm.fptoui.sat.v4i8.v4f32(<4 x float> %f)
ret <4 x i8> %x
}
define <4 x i16> @test_unsigned_v4i16_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i16_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[1,1],xmm0[1,1]
; CHECK-NEXT: xorps %xmm2, %xmm2
; CHECK-NEXT: xorps %xmm3, %xmm3
; CHECK-NEXT: maxss %xmm1, %xmm3
; CHECK-NEXT: movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
; CHECK-NEXT: movaps %xmm4, %xmm1
; CHECK-NEXT: minss %xmm3, %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %eax
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: maxss %xmm0, %xmm1
; CHECK-NEXT: movaps %xmm4, %xmm3
; CHECK-NEXT: minss %xmm1, %xmm3
; CHECK-NEXT: cvttss2si %xmm3, %ecx
; CHECK-NEXT: movd %ecx, %xmm1
; CHECK-NEXT: pinsrw $1, %eax, %xmm1
; CHECK-NEXT: movaps %xmm0, %xmm3
; CHECK-NEXT: unpckhpd {{.*#+}} xmm3 = xmm3[1],xmm0[1]
; CHECK-NEXT: xorps %xmm5, %xmm5
; CHECK-NEXT: maxss %xmm3, %xmm5
; CHECK-NEXT: movaps %xmm4, %xmm3
; CHECK-NEXT: minss %xmm5, %xmm3
; CHECK-NEXT: cvttss2si %xmm3, %eax
; CHECK-NEXT: pinsrw $2, %eax, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: maxss %xmm0, %xmm2
; CHECK-NEXT: minss %xmm2, %xmm4
; CHECK-NEXT: cvttss2si %xmm4, %eax
; CHECK-NEXT: pinsrw $3, %eax, %xmm1
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <4 x i16> @llvm.fptoui.sat.v4i16.v4f32(<4 x float> %f)
ret <4 x i16> %x
}
define <4 x i32> @test_unsigned_v4i32_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i32_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[3,3],xmm0[3,3]
; CHECK-NEXT: cvttss2si %xmm1, %rdx
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: xorps %xmm2, %xmm2
; CHECK-NEXT: ucomiss %xmm2, %xmm1
; CHECK-NEXT: cmovbl %eax, %edx
; CHECK-NEXT: movss {{.*#+}} xmm3 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss %xmm3, %xmm1
; CHECK-NEXT: movl $-1, %ecx
; CHECK-NEXT: cmoval %ecx, %edx
; CHECK-NEXT: movd %edx, %xmm1
; CHECK-NEXT: movaps %xmm0, %xmm4
; CHECK-NEXT: unpckhpd {{.*#+}} xmm4 = xmm4[1],xmm0[1]
; CHECK-NEXT: cvttss2si %xmm4, %rdx
; CHECK-NEXT: ucomiss %xmm2, %xmm4
; CHECK-NEXT: cmovbl %eax, %edx
; CHECK-NEXT: ucomiss %xmm3, %xmm4
; CHECK-NEXT: cmoval %ecx, %edx
; CHECK-NEXT: movd %edx, %xmm4
; CHECK-NEXT: punpckldq {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
; CHECK-NEXT: cvttss2si %xmm0, %rdx
; CHECK-NEXT: ucomiss %xmm2, %xmm0
; CHECK-NEXT: cmovbl %eax, %edx
; CHECK-NEXT: ucomiss %xmm3, %xmm0
; CHECK-NEXT: cmoval %ecx, %edx
; CHECK-NEXT: movd %edx, %xmm1
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: cvttss2si %xmm0, %rdx
; CHECK-NEXT: ucomiss %xmm2, %xmm0
; CHECK-NEXT: cmovbl %eax, %edx
; CHECK-NEXT: ucomiss %xmm3, %xmm0
; CHECK-NEXT: cmoval %ecx, %edx
; CHECK-NEXT: movd %edx, %xmm0
; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm4[0]
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float> %f)
ret <4 x i32> %x
}
define <4 x i64> @test_unsigned_v4i64_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i64_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: movaps %xmm0, %xmm2
; CHECK-NEXT: subss %xmm1, %xmm2
; CHECK-NEXT: cvttss2si %xmm2, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: xorps %xmm3, %xmm3
; CHECK-NEXT: ucomiss %xmm3, %xmm0
; CHECK-NEXT: cmovbq %rax, %rdx
; CHECK-NEXT: movss {{.*#+}} xmm4 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss %xmm4, %xmm0
; CHECK-NEXT: movq $-1, %rcx
; CHECK-NEXT: cmovaq %rcx, %rdx
; CHECK-NEXT: movq %rdx, %xmm2
; CHECK-NEXT: movaps %xmm0, %xmm5
; CHECK-NEXT: shufps {{.*#+}} xmm5 = xmm5[1,1],xmm0[1,1]
; CHECK-NEXT: movaps %xmm5, %xmm6
; CHECK-NEXT: subss %xmm1, %xmm6
; CHECK-NEXT: cvttss2si %xmm6, %rdx
; CHECK-NEXT: cvttss2si %xmm5, %rsi
; CHECK-NEXT: movq %rsi, %rdi
; CHECK-NEXT: sarq $63, %rdi
; CHECK-NEXT: andq %rdx, %rdi
; CHECK-NEXT: orq %rsi, %rdi
; CHECK-NEXT: ucomiss %xmm3, %xmm5
; CHECK-NEXT: cmovbq %rax, %rdi
; CHECK-NEXT: ucomiss %xmm4, %xmm5
; CHECK-NEXT: cmovaq %rcx, %rdi
; CHECK-NEXT: movq %rdi, %xmm5
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm2 = xmm2[0],xmm5[0]
; CHECK-NEXT: movaps %xmm0, %xmm5
; CHECK-NEXT: shufps {{.*#+}} xmm5 = xmm5[3,3],xmm0[3,3]
; CHECK-NEXT: movaps %xmm5, %xmm6
; CHECK-NEXT: subss %xmm1, %xmm6
; CHECK-NEXT: cvttss2si %xmm6, %rdx
; CHECK-NEXT: cvttss2si %xmm5, %rsi
; CHECK-NEXT: movq %rsi, %rdi
; CHECK-NEXT: sarq $63, %rdi
; CHECK-NEXT: andq %rdx, %rdi
; CHECK-NEXT: orq %rsi, %rdi
; CHECK-NEXT: ucomiss %xmm3, %xmm5
; CHECK-NEXT: cmovbq %rax, %rdi
; CHECK-NEXT: ucomiss %xmm4, %xmm5
; CHECK-NEXT: cmovaq %rcx, %rdi
; CHECK-NEXT: movq %rdi, %xmm5
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: movaps %xmm0, %xmm6
; CHECK-NEXT: subss %xmm1, %xmm6
; CHECK-NEXT: cvttss2si %xmm6, %rdx
; CHECK-NEXT: cvttss2si %xmm0, %rsi
; CHECK-NEXT: movq %rsi, %rdi
; CHECK-NEXT: sarq $63, %rdi
; CHECK-NEXT: andq %rdx, %rdi
; CHECK-NEXT: orq %rsi, %rdi
; CHECK-NEXT: ucomiss %xmm3, %xmm0
; CHECK-NEXT: cmovbq %rax, %rdi
; CHECK-NEXT: ucomiss %xmm4, %xmm0
; CHECK-NEXT: cmovaq %rcx, %rdi
; CHECK-NEXT: movq %rdi, %xmm1
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm5[0]
; CHECK-NEXT: movdqa %xmm2, %xmm0
; CHECK-NEXT: retq
%x = call <4 x i64> @llvm.fptoui.sat.v4i64.v4f32(<4 x float> %f)
ret <4 x i64> %x
}
define <4 x i128> @test_unsigned_v4i128_v4f32(<4 x float> %f) nounwind {
; CHECK-LABEL: test_unsigned_v4i128_v4f32:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $56, %rsp
; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rdx, %r15
; CHECK-NEXT: xorl %r14d, %r14d
; CHECK-NEXT: xorps %xmm0, %xmm0
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: ucomiss %xmm0, %xmm1
; CHECK-NEXT: cmovbq %r14, %r15
; CHECK-NEXT: cmovbq %r14, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: movq $-1, %rbp
; CHECK-NEXT: cmovaq %rbp, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %rbp, %r15
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rax, %r12
; CHECK-NEXT: movq %rdx, %r13
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %r13
; CHECK-NEXT: cmovbq %r14, %r12
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbp, %r12
; CHECK-NEXT: cmovaq %rbp, %r13
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: movaps %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rax, %rbp
; CHECK-NEXT: movq %rdx, %r14
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $0, %eax
; CHECK-NEXT: cmovbq %rax, %r14
; CHECK-NEXT: cmovbq %rax, %rbp
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movq $-1, %rax
; CHECK-NEXT: cmovaq %rax, %rbp
; CHECK-NEXT: cmovaq %rax, %r14
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $0, %ecx
; CHECK-NEXT: cmovbq %rcx, %rdx
; CHECK-NEXT: cmovbq %rcx, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movq $-1, %rcx
; CHECK-NEXT: cmovaq %rcx, %rax
; CHECK-NEXT: cmovaq %rcx, %rdx
; CHECK-NEXT: movq %rdx, 8(%rbx)
; CHECK-NEXT: movq %rax, (%rbx)
; CHECK-NEXT: movq %r14, 56(%rbx)
; CHECK-NEXT: movq %rbp, 48(%rbx)
; CHECK-NEXT: movq %r13, 40(%rbx)
; CHECK-NEXT: movq %r12, 32(%rbx)
; CHECK-NEXT: movq %r15, 24(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 16(%rbx)
; CHECK-NEXT: movq %rbx, %rax
; CHECK-NEXT: addq $56, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r12
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <4 x i128> @llvm.fptoui.sat.v4i128.v4f32(<4 x float> %f)
ret <4 x i128> %x
}
;
; 64-bit float to unsigned integer
;
declare <2 x i1> @llvm.fptoui.sat.v2i1.v2f64(<2 x double>)
declare <2 x i8> @llvm.fptoui.sat.v2i8.v2f64(<2 x double>)
declare <2 x i16> @llvm.fptoui.sat.v2i16.v2f64(<2 x double>)
declare <2 x i32> @llvm.fptoui.sat.v2i32.v2f64(<2 x double>)
declare <2 x i64> @llvm.fptoui.sat.v2i64.v2f64(<2 x double>)
declare <2 x i128> @llvm.fptoui.sat.v2i128.v2f64(<2 x double>)
define <2 x i1> @test_unsigned_v2i1_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i1_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: xorpd %xmm2, %xmm2
; CHECK-NEXT: movapd %xmm0, %xmm1
; CHECK-NEXT: maxsd %xmm2, %xmm1
; CHECK-NEXT: movsd {{.*#+}} xmm3 = mem[0],zero
; CHECK-NEXT: minsd %xmm3, %xmm1
; CHECK-NEXT: cvttsd2si %xmm1, %rax
; CHECK-NEXT: movq %rax, %xmm1
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: maxsd %xmm2, %xmm0
; CHECK-NEXT: minsd %xmm3, %xmm0
; CHECK-NEXT: cvttsd2si %xmm0, %rax
; CHECK-NEXT: movq %rax, %xmm0
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <2 x i1> @llvm.fptoui.sat.v2i1.v2f64(<2 x double> %f)
ret <2 x i1> %x
}
define <2 x i8> @test_unsigned_v2i8_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i8_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: xorpd %xmm1, %xmm1
; CHECK-NEXT: xorpd %xmm2, %xmm2
; CHECK-NEXT: maxsd %xmm0, %xmm2
; CHECK-NEXT: movsd {{.*#+}} xmm3 = mem[0],zero
; CHECK-NEXT: movapd %xmm3, %xmm4
; CHECK-NEXT: minsd %xmm2, %xmm4
; CHECK-NEXT: cvttsd2si %xmm4, %eax
; CHECK-NEXT: movzbl %al, %eax
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: maxsd %xmm0, %xmm1
; CHECK-NEXT: minsd %xmm1, %xmm3
; CHECK-NEXT: cvttsd2si %xmm3, %ecx
; CHECK-NEXT: shll $8, %ecx
; CHECK-NEXT: orl %eax, %ecx
; CHECK-NEXT: movd %ecx, %xmm0
; CHECK-NEXT: retq
%x = call <2 x i8> @llvm.fptoui.sat.v2i8.v2f64(<2 x double> %f)
ret <2 x i8> %x
}
define <2 x i16> @test_unsigned_v2i16_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i16_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: xorpd %xmm1, %xmm1
; CHECK-NEXT: maxsd %xmm0, %xmm1
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: xorpd %xmm2, %xmm2
; CHECK-NEXT: maxsd %xmm0, %xmm2
; CHECK-NEXT: movsd {{.*#+}} xmm0 = mem[0],zero
; CHECK-NEXT: movapd %xmm0, %xmm3
; CHECK-NEXT: minsd %xmm2, %xmm3
; CHECK-NEXT: cvttsd2si %xmm3, %eax
; CHECK-NEXT: minsd %xmm1, %xmm0
; CHECK-NEXT: cvttsd2si %xmm0, %ecx
; CHECK-NEXT: movd %ecx, %xmm0
; CHECK-NEXT: pinsrw $1, %eax, %xmm0
; CHECK-NEXT: retq
%x = call <2 x i16> @llvm.fptoui.sat.v2i16.v2f64(<2 x double> %f)
ret <2 x i16> %x
}
define <2 x i32> @test_unsigned_v2i32_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i32_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: xorpd %xmm2, %xmm2
; CHECK-NEXT: xorpd %xmm1, %xmm1
; CHECK-NEXT: maxsd %xmm0, %xmm1
; CHECK-NEXT: movsd {{.*#+}} xmm3 = mem[0],zero
; CHECK-NEXT: movapd %xmm3, %xmm4
; CHECK-NEXT: minsd %xmm1, %xmm4
; CHECK-NEXT: cvttsd2si %xmm4, %rax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: maxsd %xmm0, %xmm2
; CHECK-NEXT: minsd %xmm2, %xmm3
; CHECK-NEXT: cvttsd2si %xmm3, %rax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <2 x i32> @llvm.fptoui.sat.v2i32.v2f64(<2 x double> %f)
ret <2 x i32> %x
}
define <2 x i64> @test_unsigned_v2i64_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i64_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: movsd {{.*#+}} xmm2 = mem[0],zero
; CHECK-NEXT: movapd %xmm0, %xmm1
; CHECK-NEXT: subsd %xmm2, %xmm1
; CHECK-NEXT: cvttsd2si %xmm1, %rax
; CHECK-NEXT: cvttsd2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: xorl %eax, %eax
; CHECK-NEXT: xorpd %xmm3, %xmm3
; CHECK-NEXT: ucomisd %xmm3, %xmm0
; CHECK-NEXT: cmovbq %rax, %rdx
; CHECK-NEXT: movsd {{.*#+}} xmm4 = mem[0],zero
; CHECK-NEXT: ucomisd %xmm4, %xmm0
; CHECK-NEXT: movq $-1, %rcx
; CHECK-NEXT: cmovaq %rcx, %rdx
; CHECK-NEXT: movq %rdx, %xmm1
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: movapd %xmm0, %xmm5
; CHECK-NEXT: subsd %xmm2, %xmm5
; CHECK-NEXT: cvttsd2si %xmm5, %rdx
; CHECK-NEXT: cvttsd2si %xmm0, %rsi
; CHECK-NEXT: movq %rsi, %rdi
; CHECK-NEXT: sarq $63, %rdi
; CHECK-NEXT: andq %rdx, %rdi
; CHECK-NEXT: orq %rsi, %rdi
; CHECK-NEXT: ucomisd %xmm3, %xmm0
; CHECK-NEXT: cmovbq %rax, %rdi
; CHECK-NEXT: ucomisd %xmm4, %xmm0
; CHECK-NEXT: cmovaq %rcx, %rdi
; CHECK-NEXT: movq %rdi, %xmm0
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT: movdqa %xmm1, %xmm0
; CHECK-NEXT: retq
%x = call <2 x i64> @llvm.fptoui.sat.v2i64.v2f64(<2 x double> %f)
ret <2 x i64> %x
}
define <2 x i128> @test_unsigned_v2i128_v2f64(<2 x double> %f) nounwind {
; CHECK-LABEL: test_unsigned_v2i128_v2f64:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $32, %rsp
; CHECK-NEXT: movapd %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: movq %rdi, %r14
; CHECK-NEXT: unpckhpd {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: movapd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: callq __fixunsdfti@PLT
; CHECK-NEXT: movq %rax, %r15
; CHECK-NEXT: movq %rdx, %rbx
; CHECK-NEXT: xorl %r12d, %r12d
; CHECK-NEXT: xorpd %xmm0, %xmm0
; CHECK-NEXT: movapd {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: ucomisd %xmm0, %xmm1
; CHECK-NEXT: cmovbq %r12, %rbx
; CHECK-NEXT: cmovbq %r12, %r15
; CHECK-NEXT: ucomisd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: movq $-1, %r13
; CHECK-NEXT: cmovaq %r13, %r15
; CHECK-NEXT: cmovaq %r13, %rbx
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __fixunsdfti@PLT
; CHECK-NEXT: movapd (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: ucomisd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %rdx
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomisd {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: cmovaq %r13, %rdx
; CHECK-NEXT: movq %rdx, 8(%r14)
; CHECK-NEXT: movq %rax, (%r14)
; CHECK-NEXT: movq %rbx, 24(%r14)
; CHECK-NEXT: movq %r15, 16(%r14)
; CHECK-NEXT: movq %r14, %rax
; CHECK-NEXT: addq $32, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r12
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: retq
%x = call <2 x i128> @llvm.fptoui.sat.v2i128.v2f64(<2 x double> %f)
ret <2 x i128> %x
}
;
; 16-bit float to unsigned integer
;
declare <8 x i1> @llvm.fptoui.sat.v8i1.v8f16(<8 x half>)
declare <8 x i8> @llvm.fptoui.sat.v8i8.v8f16(<8 x half>)
declare <8 x i16> @llvm.fptoui.sat.v8i16.v8f16(<8 x half>)
declare <8 x i32> @llvm.fptoui.sat.v8i32.v8f16(<8 x half>)
declare <8 x i64> @llvm.fptoui.sat.v8i64.v8f16(<8 x half>)
declare <8 x i128> @llvm.fptoui.sat.v8i128.v8f16(<8 x half>)
define <8 x i1> @test_unsigned_v8i1_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i1_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $72, %rsp
; CHECK-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: xorl %ebx, %ebx
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: ucomiss %xmm1, %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $1, %ebp
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT: addq $72, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <8 x i1> @llvm.fptoui.sat.v8i1.v8f16(<8 x half> %f)
ret <8 x i1> %x
}
define <8 x i8> @test_unsigned_v8i8_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i8_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $40, %rsp
; CHECK-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %ebp
; CHECK-NEXT: xorl %r14d, %r14d
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: ucomiss %xmm1, %xmm0
; CHECK-NEXT: cmovbl %r14d, %ebp
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $255, %r15d
; CHECK-NEXT: cmoval %r15d, %ebp
; CHECK-NEXT: shll $8, %ebp
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %eax
; CHECK-NEXT: movzbl %al, %ebx
; CHECK-NEXT: orl %ebp, %ebx
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %eax
; CHECK-NEXT: movzbl %al, %ebp
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %eax
; CHECK-NEXT: shll $8, %eax
; CHECK-NEXT: orl %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: pinsrw $1, %ebx, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %ebx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %ebx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %ebx
; CHECK-NEXT: shll $8, %ebx
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %eax
; CHECK-NEXT: movzbl %al, %eax
; CHECK-NEXT: orl %ebx, %eax
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: pinsrw $2, %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %ebx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %ebx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %ebx
; CHECK-NEXT: shll $8, %ebx
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %r14d, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %r15d, %eax
; CHECK-NEXT: movzbl %al, %eax
; CHECK-NEXT: orl %ebx, %eax
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: pinsrw $3, %eax, %xmm0
; CHECK-NEXT: addq $40, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <8 x i8> @llvm.fptoui.sat.v8i8.v8f16(<8 x half> %f)
ret <8 x i8> %x
}
define <8 x i16> @test_unsigned_v8i16_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i16_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $72, %rsp
; CHECK-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: xorl %ebx, %ebx
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: ucomiss %xmm1, %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $65535, %ebp # imm = 0xFFFF
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpcklwd {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1],xmm0[2],mem[2],xmm0[3],mem[3]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT: addq $72, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <8 x i16> @llvm.fptoui.sat.v8i16.v8f16(<8 x half> %f)
ret <8 x i16> %x
}
define <8 x i32> @test_unsigned_v8i32_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i32_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $72, %rsp
; CHECK-NEXT: movdqa %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: xorl %ebx, %ebx
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: ucomiss %xmm1, %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $-1, %ebp
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; CHECK-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: cvttss2si %xmm0, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbl %ebx, %eax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmoval %ebp, %eax
; CHECK-NEXT: movd %eax, %xmm1
; CHECK-NEXT: punpckldq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT: # xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Folded Reload
; CHECK-NEXT: # xmm1 = xmm1[0],mem[0]
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: addq $72, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <8 x i32> @llvm.fptoui.sat.v8i32.v8f16(<8 x half> %f)
ret <8 x i32> %x
}
define <8 x i64> @test_unsigned_v8i64_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i64_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $88, %rsp
; CHECK-NEXT: movaps %xmm0, (%rsp) # 16-byte Spill
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: xorl %r14d, %r14d
; CHECK-NEXT: xorps %xmm1, %xmm1
; CHECK-NEXT: ucomiss %xmm1, %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movq $-1, %rbx
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movdqa %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: punpcklqdq {{.*#+}} xmm1 = xmm1[0],xmm0[0]
; CHECK-NEXT: movdqa %xmm1, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movdqa %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movdqa %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Folded Reload
; CHECK-NEXT: # xmm0 = xmm0[0],mem[0]
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movdqa (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movdqa %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm0
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movaps (%rsp), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movaps %xmm0, %xmm1
; CHECK-NEXT: subss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: cvttss2si %xmm1, %rax
; CHECK-NEXT: cvttss2si %xmm0, %rcx
; CHECK-NEXT: movq %rcx, %rdx
; CHECK-NEXT: sarq $63, %rdx
; CHECK-NEXT: andq %rax, %rdx
; CHECK-NEXT: orq %rcx, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r14, %rdx
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %rbx, %rdx
; CHECK-NEXT: movq %rdx, %xmm3
; CHECK-NEXT: punpcklqdq {{[-0-9]+}}(%r{{[sb]}}p), %xmm3 # 16-byte Folded Reload
; CHECK-NEXT: # xmm3 = xmm3[0],mem[0]
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 16-byte Reload
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm2 # 16-byte Reload
; CHECK-NEXT: addq $88, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r14
; CHECK-NEXT: retq
%x = call <8 x i64> @llvm.fptoui.sat.v8i64.v8f16(<8 x half> %f)
ret <8 x i64> %x
}
define <8 x i128> @test_unsigned_v8i128_v8f16(<8 x half> %f) nounwind {
; CHECK-LABEL: test_unsigned_v8i128_v8f16:
; CHECK: # %bb.0:
; CHECK-NEXT: pushq %rbp
; CHECK-NEXT: pushq %r15
; CHECK-NEXT: pushq %r14
; CHECK-NEXT: pushq %r13
; CHECK-NEXT: pushq %r12
; CHECK-NEXT: pushq %rbx
; CHECK-NEXT: subq $104, %rsp
; CHECK-NEXT: movdqa %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 16-byte Spill
; CHECK-NEXT: movq %rdi, %rbx
; CHECK-NEXT: psrld $16, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: xorl %r12d, %r12d
; CHECK-NEXT: pxor %xmm0, %xmm0
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm1 # 4-byte Reload
; CHECK-NEXT: # xmm1 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss %xmm0, %xmm1
; CHECK-NEXT: cmovbq %r12, %rdx
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm1
; CHECK-NEXT: movq $-1, %r13
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %r13, %rdx
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[1,1,1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %rdx
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %r13, %rdx
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrlq $48, %xmm0
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %rdx
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %r13, %rdx
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: movhlps {{.*#+}} xmm0 = xmm0[1,1]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %rdx
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %r13, %rdx
; CHECK-NEXT: movq %rdx, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[10,11,12,13,14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rdx, %rbp
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %rbp
; CHECK-NEXT: cmovbq %r12, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %rax
; CHECK-NEXT: movq %rax, {{[-0-9]+}}(%r{{[sb]}}p) # 8-byte Spill
; CHECK-NEXT: cmovaq %r13, %rbp
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: shufps {{.*#+}} xmm0 = xmm0[3,3,3,3]
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rax, %r14
; CHECK-NEXT: movq %rdx, %r15
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovbq %r12, %r15
; CHECK-NEXT: cmovbq %r12, %r14
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: cmovaq %r13, %r14
; CHECK-NEXT: cmovaq %r13, %r15
; CHECK-NEXT: movdqa {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: psrldq {{.*#+}} xmm0 = xmm0[14,15],zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movd %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Folded Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movq %rax, %r12
; CHECK-NEXT: movq %rdx, %r13
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $0, %eax
; CHECK-NEXT: cmovbq %rax, %r13
; CHECK-NEXT: cmovbq %rax, %r12
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movq $-1, %rax
; CHECK-NEXT: cmovaq %rax, %r12
; CHECK-NEXT: cmovaq %rax, %r13
; CHECK-NEXT: movaps {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 16-byte Reload
; CHECK-NEXT: callq __extendhfsf2@PLT
; CHECK-NEXT: movss %xmm0, {{[-0-9]+}}(%r{{[sb]}}p) # 4-byte Spill
; CHECK-NEXT: callq __fixunssfti@PLT
; CHECK-NEXT: movss {{[-0-9]+}}(%r{{[sb]}}p), %xmm0 # 4-byte Reload
; CHECK-NEXT: # xmm0 = mem[0],zero,zero,zero
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movl $0, %ecx
; CHECK-NEXT: cmovbq %rcx, %rdx
; CHECK-NEXT: cmovbq %rcx, %rax
; CHECK-NEXT: ucomiss {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
; CHECK-NEXT: movq $-1, %rcx
; CHECK-NEXT: cmovaq %rcx, %rax
; CHECK-NEXT: cmovaq %rcx, %rdx
; CHECK-NEXT: movq %rdx, 8(%rbx)
; CHECK-NEXT: movq %rax, (%rbx)
; CHECK-NEXT: movq %r13, 120(%rbx)
; CHECK-NEXT: movq %r12, 112(%rbx)
; CHECK-NEXT: movq %r15, 104(%rbx)
; CHECK-NEXT: movq %r14, 96(%rbx)
; CHECK-NEXT: movq %rbp, 88(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 80(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 72(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 64(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 56(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 48(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 40(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 32(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 24(%rbx)
; CHECK-NEXT: movq {{[-0-9]+}}(%r{{[sb]}}p), %rax # 8-byte Reload
; CHECK-NEXT: movq %rax, 16(%rbx)
; CHECK-NEXT: movq %rbx, %rax
; CHECK-NEXT: addq $104, %rsp
; CHECK-NEXT: popq %rbx
; CHECK-NEXT: popq %r12
; CHECK-NEXT: popq %r13
; CHECK-NEXT: popq %r14
; CHECK-NEXT: popq %r15
; CHECK-NEXT: popq %rbp
; CHECK-NEXT: retq
%x = call <8 x i128> @llvm.fptoui.sat.v8i128.v8f16(<8 x half> %f)
ret <8 x i128> %x
}