derek.holloway
7502bf64a4
CodeQL / Analyze (actions) (push) Canceled after 0s
CodeQL / Analyze (python) (push) Canceled after 0s
pre-commit / pre-commit (push) Canceled after 0s
Multi-Arch CI / setup (push) Canceled after 0s
Publish build_manylinux_x86_64 images / publish_build_manylinux_x86_64 (push) Canceled after 0s
Setup (Multi-Arch) / setup (push) Canceled after 0s
Multi-Arch CI / Linux::${{ fromJSON(needs.setup.outputs.linux_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubi10 image / publish_no_rocm_image_ubi10 (push) Canceled after 0s
Multi-Arch CI / Windows::${{ fromJSON(needs.setup.outputs.windows_build_config || '{}').build_variant_label || 'skip' }} (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Multi-Arch CI / Manifest Diff (push) Canceled after 0s
TheRock Manifest Diff Report / Generate Manifest Diff Report (push) Canceled after 0s
Multi-Arch CI / CI Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_ocl_rt images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04_rocgdb images / publish_no_rocm_image_ubuntu24_04_rocgdb (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_ocl_rt (push) Canceled after 0s
Publish TheRock Docker image / build-and-push-image (push) Canceled after 0s
Unit Tests / Unit Tests :: ubuntu-24.04 (push) Canceled after 0s
Unit Tests / Unit Tests :: windows-2022 (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_media (push) Canceled after 0s
Unit Tests / Unit Tests Summary (push) Canceled after 0s
Publish no_rocm_image_ubuntu24_04 images / publish_no_rocm_image_ubuntu24_04_openmpi (push) Canceled after 0s
291 lines
8.0 KiB
LLVM
291 lines
8.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu | FileCheck %s --check-prefixes=CHECK,Z10
|
|
; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z16 \
|
|
; RUN: | FileCheck %s --check-prefixes=CHECK,Z16
|
|
;
|
|
; Test copysign intrinsics.
|
|
|
|
declare half @llvm.copysign.f16(half, half)
|
|
declare float @llvm.copysign.f32(float, float)
|
|
declare double @llvm.copysign.f64(double, double)
|
|
declare fp128 @llvm.copysign.f128(fp128, fp128)
|
|
|
|
; Test copysign with an f16 result and f16 sign argument.
|
|
define half @f0(half %a, half %b) {
|
|
; CHECK-LABEL: f0:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%res = call half @llvm.copysign.f16(half %a, half %b)
|
|
ret half %res
|
|
}
|
|
|
|
; Test copysign with an f16 result and f32 sign argument.
|
|
define half @f1(half %a, float %b) {
|
|
; CHECK-LABEL: f1:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bh = fptrunc float %b to half
|
|
%res = call half @llvm.copysign.f16(half %a, half %bh)
|
|
ret half %res
|
|
}
|
|
|
|
; Test copysign with an f16 result and f64 sign argument.
|
|
define half @f2(half %a, double %b) {
|
|
; CHECK-LABEL: f2:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bh = fptrunc double %b to half
|
|
%res = call half @llvm.copysign.f16(half %a, half %bh)
|
|
ret half %res
|
|
}
|
|
|
|
; Test copysign with an f16 result and f128 sign argument.
|
|
define half @f3(half %a, fp128 %b) {
|
|
; Z10-LABEL: f3:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r2)
|
|
; Z10-NEXT: ld %f3, 8(%r2)
|
|
; Z10-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f3:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: vl %v1, 0(%r2), 3
|
|
; Z16-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z16-NEXT: br %r14
|
|
%bh = fptrunc fp128 %b to half
|
|
%res = call half @llvm.copysign.f16(half %a, half %bh)
|
|
ret half %res
|
|
}
|
|
|
|
; Test copysign with an f32 result and f16 sign argument.
|
|
define float @f4(float %a, half %b) {
|
|
; CHECK-LABEL: f4:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bf = fpext half %b to float
|
|
%res = call float @llvm.copysign.f32(float %a, float %bf)
|
|
ret float %res
|
|
}
|
|
|
|
; Test copysign with an f32 result and f32 sign argument.
|
|
define float @f5(float %a, float %b) {
|
|
; CHECK-LABEL: f5:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%res = call float @llvm.copysign.f32(float %a, float %b)
|
|
ret float %res
|
|
}
|
|
|
|
; Test copysign with an f32 result and f64 sign argument.
|
|
define float @f6(float %a, double %b) {
|
|
; CHECK-LABEL: f6:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bf = fptrunc double %b to float
|
|
%res = call float @llvm.copysign.f32(float %a, float %bf)
|
|
ret float %res
|
|
}
|
|
|
|
; Test copysign with an f32 result and f128 sign argument.
|
|
define float @f7(float %a, fp128 %b) {
|
|
; Z10-LABEL: f7:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r2)
|
|
; Z10-NEXT: ld %f3, 8(%r2)
|
|
; Z10-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f7:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: vl %v1, 0(%r2), 3
|
|
; Z16-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z16-NEXT: br %r14
|
|
%bf = fptrunc fp128 %b to float
|
|
%res = call float @llvm.copysign.f32(float %a, float %bf)
|
|
ret float %res
|
|
}
|
|
|
|
; Test copysign with an f64 result and f16 sign argument.
|
|
define double @f8(double %a, half %b) {
|
|
; CHECK-LABEL: f8:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bd = fpext half %b to double
|
|
%res = call double @llvm.copysign.f64(double %a, double %bd)
|
|
ret double %res
|
|
}
|
|
|
|
; Test copysign with an f64 result and f32 sign argument.
|
|
define double @f9(double %a, float %b) {
|
|
; CHECK-LABEL: f9:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%bd = fpext float %b to double
|
|
%res = call double @llvm.copysign.f64(double %a, double %bd)
|
|
ret double %res
|
|
}
|
|
|
|
; Test copysign with an f64 result and f64 sign argument.
|
|
define double @f10(double %a, double %b) {
|
|
; CHECK-LABEL: f10:
|
|
; CHECK: # %bb.0:
|
|
; CHECK-NEXT: cpsdr %f0, %f2, %f0
|
|
; CHECK-NEXT: br %r14
|
|
%res = call double @llvm.copysign.f64(double %a, double %b)
|
|
ret double %res
|
|
}
|
|
|
|
; Test copysign with an f64 result and f128 sign argument.
|
|
define double @f11(double %a, fp128 %b) {
|
|
; Z10-LABEL: f11:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r2)
|
|
; Z10-NEXT: ld %f3, 8(%r2)
|
|
; Z10-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f11:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: vl %v1, 0(%r2), 3
|
|
; Z16-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z16-NEXT: br %r14
|
|
%bd = fptrunc fp128 %b to double
|
|
%res = call double @llvm.copysign.f64(double %a, double %bd)
|
|
ret double %res
|
|
}
|
|
|
|
; Test copysign with an f128 result and f16 sign argument.
|
|
define fp128 @f12(fp128 %a, half %b) {
|
|
; Z10-LABEL: f12:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r3)
|
|
; Z10-NEXT: ld %f3, 8(%r3)
|
|
; Z10-NEXT: cpsdr %f1, %f0, %f1
|
|
; Z10-NEXT: std %f1, 0(%r2)
|
|
; Z10-NEXT: std %f3, 8(%r2)
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f12:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: aghi %r15, -168
|
|
; Z16-NEXT: .cfi_def_cfa_offset 328
|
|
; Z16-NEXT: vl %v1, 0(%r3), 3
|
|
; Z16-NEXT: vsteh %v0, 164(%r15), 0
|
|
; Z16-NEXT: tm 164(%r15), 128
|
|
; Z16-NEXT: je .LBB12_2
|
|
; Z16-NEXT: # %bb.1:
|
|
; Z16-NEXT: wflnxb %v0, %v1
|
|
; Z16-NEXT: j .LBB12_3
|
|
; Z16-NEXT: .LBB12_2:
|
|
; Z16-NEXT: wflpxb %v0, %v1
|
|
; Z16-NEXT: .LBB12_3:
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: aghi %r15, 168
|
|
; Z16-NEXT: br %r14
|
|
%b128 = fpext half %b to fp128
|
|
%res = call fp128 @llvm.copysign.f128(fp128 %a, fp128 %b128)
|
|
ret fp128 %res
|
|
}
|
|
|
|
; Test copysign with an f128 result and f32 sign argument.
|
|
define fp128 @f13(fp128 %a, float %b) {
|
|
; Z10-LABEL: f13:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r3)
|
|
; Z10-NEXT: ld %f3, 8(%r3)
|
|
; Z10-NEXT: cpsdr %f1, %f0, %f1
|
|
; Z10-NEXT: std %f1, 0(%r2)
|
|
; Z10-NEXT: std %f3, 8(%r2)
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f13:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: vl %v1, 0(%r3), 3
|
|
; Z16-NEXT: vlgvf %r0, %v0, 0
|
|
; Z16-NEXT: tmlh %r0, 32768
|
|
; Z16-NEXT: je .LBB13_2
|
|
; Z16-NEXT: # %bb.1:
|
|
; Z16-NEXT: wflnxb %v0, %v1
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
; Z16-NEXT: .LBB13_2:
|
|
; Z16-NEXT: wflpxb %v0, %v1
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
%b128 = fpext float %b to fp128
|
|
%res = call fp128 @llvm.copysign.f128(fp128 %a, fp128 %b128)
|
|
ret fp128 %res
|
|
}
|
|
|
|
; Test copysign with an f128 result and f64 sign argument.
|
|
define fp128 @f14(fp128 %a, double %b) {
|
|
; Z10-LABEL: f14:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f1, 0(%r3)
|
|
; Z10-NEXT: ld %f3, 8(%r3)
|
|
; Z10-NEXT: cpsdr %f1, %f0, %f1
|
|
; Z10-NEXT: std %f1, 0(%r2)
|
|
; Z10-NEXT: std %f3, 8(%r2)
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f14:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: vl %v1, 0(%r3), 3
|
|
; Z16-NEXT: lgdr %r0, %f0
|
|
; Z16-NEXT: tmhh %r0, 32768
|
|
; Z16-NEXT: je .LBB14_2
|
|
; Z16-NEXT: # %bb.1:
|
|
; Z16-NEXT: wflnxb %v0, %v1
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
; Z16-NEXT: .LBB14_2:
|
|
; Z16-NEXT: wflpxb %v0, %v1
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
%b128 = fpext double %b to fp128
|
|
%res = call fp128 @llvm.copysign.f128(fp128 %a, fp128 %b128)
|
|
ret fp128 %res
|
|
}
|
|
|
|
; Test copysign with an f128 result and f128 sign argument.
|
|
define fp128 @f15(fp128 %a, fp128 %b) {
|
|
; Z10-LABEL: f15:
|
|
; Z10: # %bb.0:
|
|
; Z10-NEXT: ld %f0, 0(%r3)
|
|
; Z10-NEXT: ld %f2, 8(%r3)
|
|
; Z10-NEXT: ld %f1, 0(%r4)
|
|
; Z10-NEXT: ld %f3, 8(%r4)
|
|
; Z10-NEXT: cpsdr %f0, %f1, %f0
|
|
; Z10-NEXT: std %f0, 0(%r2)
|
|
; Z10-NEXT: std %f2, 8(%r2)
|
|
; Z10-NEXT: br %r14
|
|
;
|
|
; Z16-LABEL: f15:
|
|
; Z16: # %bb.0:
|
|
; Z16-NEXT: larl %r1, .LCPI15_0
|
|
; Z16-NEXT: vl %v1, 0(%r4), 3
|
|
; Z16-NEXT: vl %v2, 0(%r1), 3
|
|
; Z16-NEXT: vl %v0, 0(%r3), 3
|
|
; Z16-NEXT: vtm %v1, %v2
|
|
; Z16-NEXT: je .LBB15_2
|
|
; Z16-NEXT: # %bb.1:
|
|
; Z16-NEXT: wflnxb %v0, %v0
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
; Z16-NEXT: .LBB15_2:
|
|
; Z16-NEXT: wflpxb %v0, %v0
|
|
; Z16-NEXT: vst %v0, 0(%r2), 3
|
|
; Z16-NEXT: br %r14
|
|
%res = call fp128 @llvm.copysign.f128(fp128 %a, fp128 %b)
|
|
ret fp128 %res
|
|
}
|