5a24dfa339
This is a followup to https://github.com/llvm/llvm-project/pull/171114, removing the handling for most libcalls that are already canonicalized to intrinsics in the middle-end. The only remaining one is fabs, which has more test coverage than the others.
313 lines
9.8 KiB
LLVM
313 lines
9.8 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; Test vector maximum on z14.
|
|
;
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z14 | FileCheck %s
|
|
|
|
declare double @fmax(double, double)
|
|
declare double @llvm.maxnum.f64(double, double)
|
|
declare <2 x double> @llvm.maxnum.v2f64(<2 x double>, <2 x double>)
|
|
declare double @llvm.maximum.f64(double, double)
|
|
declare <2 x double> @llvm.maximum.v2f64(<2 x double>, <2 x double>)
|
|
|
|
declare float @fmaxf(float, float)
|
|
declare float @llvm.maxnum.f32(float, float)
|
|
declare <4 x float> @llvm.maxnum.v4f32(<4 x float>, <4 x float>)
|
|
declare float @llvm.maximum.f32(float, float)
|
|
declare <4 x float> @llvm.maximum.v4f32(<4 x float>, <4 x float>)
|
|
|
|
declare half @llvm.maxnum.f16(half, half)
|
|
|
|
declare fp128 @fmaxl(fp128, fp128)
|
|
declare fp128 @llvm.maxnum.f128(fp128, fp128)
|
|
declare fp128 @llvm.maximum.f128(fp128, fp128)
|
|
|
|
; Test the fmax library function.
|
|
define double @f1(double %dummy, double %val1, double %val2) nounwind {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: stmg %r14, %r15, 112(%r15)
|
|
; CHECK-NEXT: aghi %r15, -160
|
|
; CHECK-NEXT: ldr %f0, %f2
|
|
; CHECK-NEXT: ldr %f2, %f4
|
|
; CHECK-NEXT: brasl %r14, fmax@PLT
|
|
; CHECK-NEXT: lmg %r14, %r15, 272(%r15)
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call double @fmax(double %val1, double %val2) readnone
|
|
ret double %ret
|
|
}
|
|
|
|
; Test the f64 maxnum intrinsic.
|
|
define double @f2(double %dummy, double %val1, double %val2) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: wfmaxdb %f0, %f2, %f4, 4
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call double @llvm.maxnum.f64(double %val1, double %val2)
|
|
ret double %ret
|
|
}
|
|
|
|
; Test the f64 maximum intrinsic.
|
|
define double @f3(double %dummy, double %val1, double %val2) {
|
|
; CHECK-LABEL: f3:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: wfmaxdb %f0, %f2, %f4, 1
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call double @llvm.maximum.f64(double %val1, double %val2)
|
|
ret double %ret
|
|
}
|
|
|
|
; Test a f64 constant compare/select resulting in maxnum.
|
|
define double @f4(double %dummy, double %val) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: lzdr %f0
|
|
; CHECK-NEXT: wfmaxdb %f0, %f2, %f0, 4
|
|
; CHECK-NEXT: br %r14
|
|
%cmp = fcmp ogt double %val, 0.0
|
|
%ret = select i1 %cmp, double %val, double 0.0
|
|
ret double %ret
|
|
}
|
|
|
|
; Test a f64 constant compare/select resulting in maximum.
|
|
define double @f5(double %dummy, double %val) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ltdbr %f1, %f2
|
|
; CHECK-NEXT: ldr %f0, %f2
|
|
; CHECK-NEXT: bnler %r14
|
|
; CHECK-NEXT: .LBB4_1:
|
|
; CHECK-NEXT: lzdr %f0
|
|
; CHECK-NEXT: br %r14
|
|
%cmp = fcmp ugt double %val, 0.0
|
|
%ret = select i1 %cmp, double %val, double 0.0
|
|
ret double %ret
|
|
}
|
|
|
|
; Test the v2f64 maxnum intrinsic.
|
|
define <2 x double> @f6(<2 x double> %dummy, <2 x double> %val1, <2 x double> %val2) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfmaxdb %v24, %v26, %v28, 4
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call <2 x double> @llvm.maxnum.v2f64(<2 x double> %val1, <2 x double> %val2)
|
|
ret <2 x double> %ret
|
|
}
|
|
|
|
; Test the v2f64 maximum intrinsic.
|
|
define <2 x double> @f7(<2 x double> %dummy, <2 x double> %val1, <2 x double> %val2) {
|
|
; CHECK-LABEL: f7:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfmaxdb %v24, %v26, %v28, 1
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call <2 x double> @llvm.maximum.v2f64(<2 x double> %val1, <2 x double> %val2)
|
|
ret <2 x double> %ret
|
|
}
|
|
|
|
; Test the fmaxf library function.
|
|
define float @f11(float %dummy, float %val1, float %val2) nounwind {
|
|
; CHECK-LABEL: f11:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: stmg %r14, %r15, 112(%r15)
|
|
; CHECK-NEXT: aghi %r15, -160
|
|
; CHECK-NEXT: ldr %f0, %f2
|
|
; CHECK-NEXT: ldr %f2, %f4
|
|
; CHECK-NEXT: brasl %r14, fmaxf@PLT
|
|
; CHECK-NEXT: lmg %r14, %r15, 272(%r15)
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call float @fmaxf(float %val1, float %val2) readnone
|
|
ret float %ret
|
|
}
|
|
|
|
; Test the f16 maxnum intrinsic.
|
|
define half @f12_half(half %dummy, half %val1, half %val2) {
|
|
; CHECK-LABEL: f12_half:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: stmg %r14, %r15, 112(%r15)
|
|
; CHECK-NEXT: .cfi_offset %r14, -48
|
|
; CHECK-NEXT: .cfi_offset %r15, -40
|
|
; CHECK-NEXT: aghi %r15, -176
|
|
; CHECK-NEXT: .cfi_def_cfa_offset 336
|
|
; CHECK-NEXT: std %f8, 168(%r15) # 8-byte Spill
|
|
; CHECK-NEXT: std %f9, 160(%r15) # 8-byte Spill
|
|
; CHECK-NEXT: .cfi_offset %f8, -168
|
|
; CHECK-NEXT: .cfi_offset %f9, -176
|
|
; CHECK-NEXT: ldr %f0, %f4
|
|
; CHECK-NEXT: ldr %f8, %f2
|
|
; CHECK-NEXT: brasl %r14, __extendhfsf2@PLT
|
|
; CHECK-NEXT: ldr %f9, %f0
|
|
; CHECK-NEXT: ldr %f0, %f8
|
|
; CHECK-NEXT: brasl %r14, __extendhfsf2@PLT
|
|
; CHECK-NEXT: wfmaxsb %f0, %f0, %f9, 4
|
|
; CHECK-NEXT: brasl %r14, __truncsfhf2@PLT
|
|
; CHECK-NEXT: ld %f8, 168(%r15) # 8-byte Reload
|
|
; CHECK-NEXT: ld %f9, 160(%r15) # 8-byte Reload
|
|
; CHECK-NEXT: lmg %r14, %r15, 288(%r15)
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call half @llvm.maxnum.f16(half %val1, half %val2)
|
|
ret half %ret
|
|
}
|
|
|
|
; Test the f32 maxnum intrinsic.
|
|
define float @f12(float %dummy, float %val1, float %val2) {
|
|
; CHECK-LABEL: f12:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: wfmaxsb %f0, %f2, %f4, 4
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call float @llvm.maxnum.f32(float %val1, float %val2)
|
|
ret float %ret
|
|
}
|
|
|
|
; Test the f32 maximum intrinsic.
|
|
define float @f13(float %dummy, float %val1, float %val2) {
|
|
; CHECK-LABEL: f13:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: wfmaxsb %f0, %f2, %f4, 1
|
|
; CHECK-NEXT: br %r14
|
|
%ret = call float @llvm.maximum.f32(float %val1, float %val2)
|
|
ret float %ret
|
|
}
|
|
|
|
; Test a f32 constant compare/select resulting in maxnum.
|
|
define float @f14(float %dummy, float %val) {
|
|
; CHECK-LABEL: f14:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: lzer %f0
|
|
; CHECK-NEXT: wfmaxsb %f0, %f2, %f0, 4
|
|
; CHECK-NEXT: br %r14
|
|
%cmp = fcmp ogt float %val, 0.0
|
|
%ret = select i1 %cmp, float %val, float 0.0
|
|
ret float %ret
|
|
}
|
|
|
|
; Test a f32 constant compare/select resulting in maximum.
|
|
define float @f15(float %dummy, float %val) {
|
|
; CHECK-LABEL: f15:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: ltebr %f1, %f2
|
|
; CHECK-NEXT: ldr %f0, %f2
|
|
; CHECK-NEXT: bnler %r14
|
|
; CHECK-NEXT: .LBB12_1:
|
|
; CHECK-NEXT: lzer %f0
|
|
; CHECK-NEXT: br %r14
|
|
%cmp = fcmp ugt float %val, 0.0
|
|
%ret = select i1 %cmp, float %val, float 0.0
|
|
ret float %ret
|
|
}
|
|
|
|
; Test the v4f32 maxnum intrinsic.
|
|
define <4 x float> @f16(<4 x float> %dummy, <4 x float> %val1,
|
|
; CHECK-LABEL: f16:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfmaxsb %v24, %v26, %v28, 4
|
|
; CHECK-NEXT: br %r14
|
|
<4 x float> %val2) {
|
|
%ret = call <4 x float> @llvm.maxnum.v4f32(<4 x float> %val1, <4 x float> %val2)
|
|
ret <4 x float> %ret
|
|
}
|
|
|
|
; Test the v4f32 maximum intrinsic.
|
|
define <4 x float> @f17(<4 x float> %dummy, <4 x float> %val1,
|
|
; CHECK-LABEL: f17:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vfmaxsb %v24, %v26, %v28, 1
|
|
; CHECK-NEXT: br %r14
|
|
<4 x float> %val2) {
|
|
%ret = call <4 x float> @llvm.maximum.v4f32(<4 x float> %val1, <4 x float> %val2)
|
|
ret <4 x float> %ret
|
|
}
|
|
|
|
; Test the fmaxl library function.
|
|
define void @f21(ptr %ptr1, ptr %ptr2, ptr %dst) nounwind {
|
|
; CHECK-LABEL: f21:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: stmg %r13, %r15, 104(%r15)
|
|
; CHECK-NEXT: aghi %r15, -208
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: lgr %r13, %r4
|
|
; CHECK-NEXT: la %r2, 192(%r15)
|
|
; CHECK-NEXT: la %r3, 176(%r15)
|
|
; CHECK-NEXT: la %r4, 160(%r15)
|
|
; CHECK-NEXT: vst %v1, 160(%r15), 3
|
|
; CHECK-NEXT: vst %v0, 176(%r15), 3
|
|
; CHECK-NEXT: brasl %r14, fmaxl@PLT
|
|
; CHECK-NEXT: vl %v0, 192(%r15), 3
|
|
; CHECK-NEXT: vst %v0, 0(%r13), 3
|
|
; CHECK-NEXT: lmg %r13, %r15, 312(%r15)
|
|
; CHECK-NEXT: br %r14
|
|
%val1 = load fp128, ptr %ptr1
|
|
%val2 = load fp128, ptr %ptr2
|
|
%res = call fp128 @fmaxl(fp128 %val1, fp128 %val2) readnone
|
|
store fp128 %res, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
; Test the f128 maxnum intrinsic.
|
|
define void @f22(ptr %ptr1, ptr %ptr2, ptr %dst) {
|
|
; CHECK-LABEL: f22:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: wfmaxxb %v0, %v0, %v1, 4
|
|
; CHECK-NEXT: vst %v0, 0(%r4), 3
|
|
; CHECK-NEXT: br %r14
|
|
%val1 = load fp128, ptr %ptr1
|
|
%val2 = load fp128, ptr %ptr2
|
|
%res = call fp128 @llvm.maxnum.f128(fp128 %val1, fp128 %val2)
|
|
store fp128 %res, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
; Test the f128 maximum intrinsic.
|
|
define void @f23(ptr %ptr1, ptr %ptr2, ptr %dst) {
|
|
; CHECK-LABEL: f23:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vl %v1, 0(%r3), 3
|
|
; CHECK-NEXT: wfmaxxb %v0, %v0, %v1, 1
|
|
; CHECK-NEXT: vst %v0, 0(%r4), 3
|
|
; CHECK-NEXT: br %r14
|
|
%val1 = load fp128, ptr %ptr1
|
|
%val2 = load fp128, ptr %ptr2
|
|
%res = call fp128 @llvm.maximum.f128(fp128 %val1, fp128 %val2)
|
|
store fp128 %res, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
; Test a f128 constant compare/select resulting in maxnum.
|
|
define void @f24(ptr %ptr, ptr %dst) {
|
|
; CHECK-LABEL: f24:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vzero %v1
|
|
; CHECK-NEXT: wfmaxxb %v0, %v0, %v1, 4
|
|
; CHECK-NEXT: vst %v0, 0(%r3), 3
|
|
; CHECK-NEXT: br %r14
|
|
%val = load fp128, ptr %ptr
|
|
%cmp = fcmp ogt fp128 %val, 0xL00000000000000000000000000000000
|
|
%res = select i1 %cmp, fp128 %val, fp128 0xL00000000000000000000000000000000
|
|
store fp128 %res, ptr %dst
|
|
ret void
|
|
}
|
|
|
|
; Test a f128 constant compare/select resulting in maximum.
|
|
define void @f25(ptr %ptr, ptr %dst) {
|
|
; CHECK-LABEL: f25:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: vl %v0, 0(%r2), 3
|
|
; CHECK-NEXT: vzero %v1
|
|
; CHECK-NEXT: wfcxb %v0, %v1
|
|
; CHECK-NEXT: jnle .LBB19_2
|
|
; CHECK-NEXT: # %bb.1:
|
|
; CHECK-NEXT: vzero %v0
|
|
; CHECK-NEXT: .LBB19_2:
|
|
; CHECK-NEXT: vst %v0, 0(%r3), 3
|
|
; CHECK-NEXT: br %r14
|
|
%val = load fp128, ptr %ptr
|
|
%cmp = fcmp ugt fp128 %val, 0xL00000000000000000000000000000000
|
|
%res = select i1 %cmp, fp128 %val, fp128 0xL00000000000000000000000000000000
|
|
store fp128 %res, ptr %dst
|
|
ret void
|
|
}
|
|
|