Files
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
Flatten subprojects into a single base monorepo
2026-09-02 16:34:47 -07:00

1491 lines
54 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv32 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32I
; RUN: llc -mtriple=riscv64 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV64I
; RUN: llc -mtriple=riscv32 -mattr=+f -target-abi=ilp32 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32IF
; RUN: llc -mtriple=riscv64 -mattr=+f -target-abi=lp64 -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV64IF
; RUN: llc -mtriple=riscv32 -mattr=+f -target-abi=ilp32f -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32-ILP32F
; RUN: llc -mtriple=riscv64 -mattr=+f -target-abi=lp64f -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV64-LP64F
; RUN: llc -mtriple=riscv32 -mattr=+f,+zfhmin -target-abi=ilp32f -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV32-ILP32ZFHMIN
; RUN: llc -mtriple=riscv64 -mattr=+f,+zfhmin -target-abi=lp64f -verify-machineinstrs < %s | FileCheck %s -check-prefix=RV64-LP64ZFHMIN
; RUN: llc -mtriple=riscv32 -mattr=+f -mattr=+zfh -verify-machineinstrs -target-abi ilp32 -O1 < %s \
; RUN: | FileCheck %s -check-prefix=RV32-ZFH-ILP32
; RUN: llc -mtriple=riscv32 -mattr=+f -mattr=+zfh -verify-machineinstrs -target-abi ilp32f -O1 < %s \
; RUN: | FileCheck %s -check-prefix=RV32-ZFH-ILP32F
; RUN: llc -mtriple=riscv64 -mattr=+f -mattr=+zfh -verify-machineinstrs -target-abi lp64 -O1 < %s \
; RUN: | FileCheck %s -check-prefix=RV64-ZFH-LP64
; RUN: llc -mtriple=riscv64 -mattr=+f -mattr=+zfh -verify-machineinstrs -target-abi lp64f -O1 < %s \
; RUN: | FileCheck %s -check-prefix=RV64-ZFH-LP64F
; Tests passing half arguments and returns.
; Covers with and without F extension, with and without Zfhmin/Zfh extensions,
; and ilp32f/ilp64f calling conventions.
define i32 @callee_half_in_regs(i32 %a, half %b) nounwind {
; RV32I-LABEL: callee_half_in_regs:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32I-NEXT: mv s0, a0
; RV32I-NEXT: slli a0, a1, 16
; RV32I-NEXT: srli a0, a0, 16
; RV32I-NEXT: call __extendhfsf2
; RV32I-NEXT: call __fixsfsi
; RV32I-NEXT: add a0, s0, a0
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: callee_half_in_regs:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64I-NEXT: mv s0, a0
; RV64I-NEXT: slli a0, a1, 48
; RV64I-NEXT: srli a0, a0, 48
; RV64I-NEXT: call __extendhfsf2
; RV64I-NEXT: call __fixsfdi
; RV64I-NEXT: addw a0, s0, a0
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: callee_half_in_regs:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32IF-NEXT: mv s0, a0
; RV32IF-NEXT: mv a0, a1
; RV32IF-NEXT: call __extendhfsf2
; RV32IF-NEXT: fmv.w.x fa5, a0
; RV32IF-NEXT: fcvt.w.s a0, fa5, rtz
; RV32IF-NEXT: add a0, s0, a0
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: callee_half_in_regs:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64IF-NEXT: mv s0, a0
; RV64IF-NEXT: mv a0, a1
; RV64IF-NEXT: call __extendhfsf2
; RV64IF-NEXT: fmv.w.x fa5, a0
; RV64IF-NEXT: fcvt.l.s a0, fa5, rtz
; RV64IF-NEXT: addw a0, s0, a0
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: callee_half_in_regs:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: mv s0, a0
; RV32-ILP32F-NEXT: call __extendhfsf2
; RV32-ILP32F-NEXT: fcvt.w.s a0, fa0, rtz
; RV32-ILP32F-NEXT: add a0, s0, a0
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: callee_half_in_regs:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: mv s0, a0
; RV64-LP64F-NEXT: call __extendhfsf2
; RV64-LP64F-NEXT: fcvt.l.s a0, fa0, rtz
; RV64-LP64F-NEXT: addw a0, s0, a0
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: callee_half_in_regs:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV32-ILP32ZFHMIN-NEXT: fcvt.w.s a1, fa5, rtz
; RV32-ILP32ZFHMIN-NEXT: add a0, a0, a1
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: callee_half_in_regs:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV64-LP64ZFHMIN-NEXT: fcvt.w.s a1, fa5, rtz
; RV64-LP64ZFHMIN-NEXT: addw a0, a0, a1
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: callee_half_in_regs:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: fmv.h.x fa5, a1
; RV32-ZFH-ILP32-NEXT: fcvt.w.h a1, fa5, rtz
; RV32-ZFH-ILP32-NEXT: add a0, a0, a1
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: callee_half_in_regs:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: fcvt.w.h a1, fa0, rtz
; RV32-ZFH-ILP32F-NEXT: add a0, a0, a1
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: callee_half_in_regs:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: fmv.h.x fa5, a1
; RV64-ZFH-LP64-NEXT: fcvt.w.h a1, fa5, rtz
; RV64-ZFH-LP64-NEXT: addw a0, a0, a1
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: callee_half_in_regs:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: fcvt.w.h a1, fa0, rtz
; RV64-ZFH-LP64F-NEXT: addw a0, a0, a1
; RV64-ZFH-LP64F-NEXT: ret
%b_fptosi = fptosi half %b to i32
%1 = add i32 %a, %b_fptosi
ret i32 %1
}
define i32 @caller_half_in_regs() nounwind {
; RV32I-LABEL: caller_half_in_regs:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: li a0, 1
; RV32I-NEXT: lui a1, 4
; RV32I-NEXT: call callee_half_in_regs
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: caller_half_in_regs:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: li a0, 1
; RV64I-NEXT: lui a1, 4
; RV64I-NEXT: call callee_half_in_regs
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: caller_half_in_regs:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: li a0, 1
; RV32IF-NEXT: lui a1, 1048564
; RV32IF-NEXT: call callee_half_in_regs
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: caller_half_in_regs:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: li a0, 1
; RV64IF-NEXT: lui a1, 1048564
; RV64IF-NEXT: call callee_half_in_regs
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: caller_half_in_regs:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: lui a0, 1048564
; RV32-ILP32F-NEXT: fmv.w.x fa0, a0
; RV32-ILP32F-NEXT: li a0, 1
; RV32-ILP32F-NEXT: call callee_half_in_regs
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: caller_half_in_regs:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: lui a0, 1048564
; RV64-LP64F-NEXT: fmv.w.x fa0, a0
; RV64-LP64F-NEXT: li a0, 1
; RV64-LP64F-NEXT: call callee_half_in_regs
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: caller_half_in_regs:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, -16
; RV32-ILP32ZFHMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32ZFHMIN-NEXT: lui a0, 4
; RV32-ILP32ZFHMIN-NEXT: fmv.h.x fa0, a0
; RV32-ILP32ZFHMIN-NEXT: li a0, 1
; RV32-ILP32ZFHMIN-NEXT: call callee_half_in_regs
; RV32-ILP32ZFHMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, 16
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: caller_half_in_regs:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, -16
; RV64-LP64ZFHMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64ZFHMIN-NEXT: lui a0, 4
; RV64-LP64ZFHMIN-NEXT: fmv.h.x fa0, a0
; RV64-LP64ZFHMIN-NEXT: li a0, 1
; RV64-LP64ZFHMIN-NEXT: call callee_half_in_regs
; RV64-LP64ZFHMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, 16
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: caller_half_in_regs:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32-NEXT: li a0, 1
; RV32-ZFH-ILP32-NEXT: lui a1, 4
; RV32-ZFH-ILP32-NEXT: call callee_half_in_regs
; RV32-ZFH-ILP32-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: caller_half_in_regs:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32F-NEXT: lui a0, 4
; RV32-ZFH-ILP32F-NEXT: fmv.h.x fa0, a0
; RV32-ZFH-ILP32F-NEXT: li a0, 1
; RV32-ZFH-ILP32F-NEXT: call callee_half_in_regs
; RV32-ZFH-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: caller_half_in_regs:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64-NEXT: li a0, 1
; RV64-ZFH-LP64-NEXT: lui a1, 4
; RV64-ZFH-LP64-NEXT: call callee_half_in_regs
; RV64-ZFH-LP64-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: caller_half_in_regs:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64F-NEXT: lui a0, 4
; RV64-ZFH-LP64F-NEXT: fmv.h.x fa0, a0
; RV64-ZFH-LP64F-NEXT: li a0, 1
; RV64-ZFH-LP64F-NEXT: call callee_half_in_regs
; RV64-ZFH-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64F-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64F-NEXT: ret
%1 = call i32 @callee_half_in_regs(i32 1, half 2.0)
ret i32 %1
}
define i32 @callee_half_on_stack(i32 %a, i32 %b, i32 %c, i32 %d, i32 %e, i32 %f, i32 %g, i32 %h, half %i) nounwind {
; RV32I-LABEL: callee_half_on_stack:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32I-NEXT: lhu a0, 16(sp)
; RV32I-NEXT: mv s0, a7
; RV32I-NEXT: call __extendhfsf2
; RV32I-NEXT: call __fixsfsi
; RV32I-NEXT: add a0, s0, a0
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: callee_half_on_stack:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64I-NEXT: lhu a0, 16(sp)
; RV64I-NEXT: mv s0, a7
; RV64I-NEXT: call __extendhfsf2
; RV64I-NEXT: call __fixsfdi
; RV64I-NEXT: addw a0, s0, a0
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: callee_half_on_stack:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32IF-NEXT: lhu a0, 16(sp)
; RV32IF-NEXT: mv s0, a7
; RV32IF-NEXT: call __extendhfsf2
; RV32IF-NEXT: fmv.w.x fa5, a0
; RV32IF-NEXT: fcvt.w.s a0, fa5, rtz
; RV32IF-NEXT: add a0, s0, a0
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: callee_half_on_stack:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64IF-NEXT: lhu a0, 16(sp)
; RV64IF-NEXT: mv s0, a7
; RV64IF-NEXT: call __extendhfsf2
; RV64IF-NEXT: fmv.w.x fa5, a0
; RV64IF-NEXT: fcvt.l.s a0, fa5, rtz
; RV64IF-NEXT: addw a0, s0, a0
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: callee_half_on_stack:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: mv s0, a7
; RV32-ILP32F-NEXT: call __extendhfsf2
; RV32-ILP32F-NEXT: fcvt.w.s a0, fa0, rtz
; RV32-ILP32F-NEXT: add a0, s0, a0
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: callee_half_on_stack:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: mv s0, a7
; RV64-LP64F-NEXT: call __extendhfsf2
; RV64-LP64F-NEXT: fcvt.l.s a0, fa0, rtz
; RV64-LP64F-NEXT: addw a0, s0, a0
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: callee_half_on_stack:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV32-ILP32ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV32-ILP32ZFHMIN-NEXT: add a0, a7, a0
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: callee_half_on_stack:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV64-LP64ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV64-LP64ZFHMIN-NEXT: addw a0, a7, a0
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: callee_half_on_stack:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: flh fa5, 0(sp)
; RV32-ZFH-ILP32-NEXT: fcvt.w.h a0, fa5, rtz
; RV32-ZFH-ILP32-NEXT: add a0, a7, a0
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: callee_half_on_stack:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: fcvt.w.h a0, fa0, rtz
; RV32-ZFH-ILP32F-NEXT: add a0, a7, a0
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: callee_half_on_stack:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: flh fa5, 0(sp)
; RV64-ZFH-LP64-NEXT: fcvt.w.h a0, fa5, rtz
; RV64-ZFH-LP64-NEXT: addw a0, a7, a0
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: callee_half_on_stack:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: fcvt.w.h a0, fa0, rtz
; RV64-ZFH-LP64F-NEXT: addw a0, a7, a0
; RV64-ZFH-LP64F-NEXT: ret
%1 = fptosi half %i to i32
%2 = add i32 %h, %1
ret i32 %2
}
define i32 @caller_half_on_stack() nounwind {
; RV32I-LABEL: caller_half_on_stack:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: lui a7, 5
; RV32I-NEXT: li a0, 1
; RV32I-NEXT: li a1, 2
; RV32I-NEXT: li a2, 3
; RV32I-NEXT: li a3, 4
; RV32I-NEXT: li a4, 5
; RV32I-NEXT: li a5, 6
; RV32I-NEXT: li a6, 7
; RV32I-NEXT: addi t0, a7, -1792
; RV32I-NEXT: li a7, 8
; RV32I-NEXT: sw t0, 0(sp)
; RV32I-NEXT: call callee_half_on_stack
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: caller_half_on_stack:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: lui a7, 5
; RV64I-NEXT: li a0, 1
; RV64I-NEXT: li a1, 2
; RV64I-NEXT: li a2, 3
; RV64I-NEXT: li a3, 4
; RV64I-NEXT: li a4, 5
; RV64I-NEXT: li a5, 6
; RV64I-NEXT: li a6, 7
; RV64I-NEXT: addi t0, a7, -1792
; RV64I-NEXT: li a7, 8
; RV64I-NEXT: sd t0, 0(sp)
; RV64I-NEXT: call callee_half_on_stack
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: caller_half_on_stack:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: lui a7, 1048565
; RV32IF-NEXT: li a0, 1
; RV32IF-NEXT: li a1, 2
; RV32IF-NEXT: li a2, 3
; RV32IF-NEXT: li a3, 4
; RV32IF-NEXT: li a4, 5
; RV32IF-NEXT: li a5, 6
; RV32IF-NEXT: li a6, 7
; RV32IF-NEXT: addi t0, a7, -1792
; RV32IF-NEXT: li a7, 8
; RV32IF-NEXT: sw t0, 0(sp)
; RV32IF-NEXT: call callee_half_on_stack
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: caller_half_on_stack:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: lui a7, 1048565
; RV64IF-NEXT: li a0, 1
; RV64IF-NEXT: li a1, 2
; RV64IF-NEXT: li a2, 3
; RV64IF-NEXT: li a3, 4
; RV64IF-NEXT: li a4, 5
; RV64IF-NEXT: li a5, 6
; RV64IF-NEXT: li a6, 7
; RV64IF-NEXT: addi t0, a7, -1792
; RV64IF-NEXT: li a7, 8
; RV64IF-NEXT: sw t0, 0(sp)
; RV64IF-NEXT: call callee_half_on_stack
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: caller_half_on_stack:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: lui a7, 1048565
; RV32-ILP32F-NEXT: li a0, 1
; RV32-ILP32F-NEXT: li a1, 2
; RV32-ILP32F-NEXT: li a2, 3
; RV32-ILP32F-NEXT: li a3, 4
; RV32-ILP32F-NEXT: li a4, 5
; RV32-ILP32F-NEXT: li a5, 6
; RV32-ILP32F-NEXT: li a6, 7
; RV32-ILP32F-NEXT: addi a7, a7, -1792
; RV32-ILP32F-NEXT: fmv.w.x fa0, a7
; RV32-ILP32F-NEXT: li a7, 8
; RV32-ILP32F-NEXT: call callee_half_on_stack
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: caller_half_on_stack:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: lui a7, 1048565
; RV64-LP64F-NEXT: li a0, 1
; RV64-LP64F-NEXT: li a1, 2
; RV64-LP64F-NEXT: li a2, 3
; RV64-LP64F-NEXT: li a3, 4
; RV64-LP64F-NEXT: li a4, 5
; RV64-LP64F-NEXT: li a5, 6
; RV64-LP64F-NEXT: li a6, 7
; RV64-LP64F-NEXT: addi a7, a7, -1792
; RV64-LP64F-NEXT: fmv.w.x fa0, a7
; RV64-LP64F-NEXT: li a7, 8
; RV64-LP64F-NEXT: call callee_half_on_stack
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: caller_half_on_stack:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, -16
; RV32-ILP32ZFHMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32ZFHMIN-NEXT: lui a7, 5
; RV32-ILP32ZFHMIN-NEXT: li a0, 1
; RV32-ILP32ZFHMIN-NEXT: li a1, 2
; RV32-ILP32ZFHMIN-NEXT: li a2, 3
; RV32-ILP32ZFHMIN-NEXT: li a3, 4
; RV32-ILP32ZFHMIN-NEXT: li a4, 5
; RV32-ILP32ZFHMIN-NEXT: li a5, 6
; RV32-ILP32ZFHMIN-NEXT: li a6, 7
; RV32-ILP32ZFHMIN-NEXT: addi a7, a7, -1792
; RV32-ILP32ZFHMIN-NEXT: fmv.h.x fa0, a7
; RV32-ILP32ZFHMIN-NEXT: li a7, 8
; RV32-ILP32ZFHMIN-NEXT: call callee_half_on_stack
; RV32-ILP32ZFHMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, 16
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: caller_half_on_stack:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, -16
; RV64-LP64ZFHMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64ZFHMIN-NEXT: lui a7, 5
; RV64-LP64ZFHMIN-NEXT: li a0, 1
; RV64-LP64ZFHMIN-NEXT: li a1, 2
; RV64-LP64ZFHMIN-NEXT: li a2, 3
; RV64-LP64ZFHMIN-NEXT: li a3, 4
; RV64-LP64ZFHMIN-NEXT: li a4, 5
; RV64-LP64ZFHMIN-NEXT: li a5, 6
; RV64-LP64ZFHMIN-NEXT: li a6, 7
; RV64-LP64ZFHMIN-NEXT: addi a7, a7, -1792
; RV64-LP64ZFHMIN-NEXT: fmv.h.x fa0, a7
; RV64-LP64ZFHMIN-NEXT: li a7, 8
; RV64-LP64ZFHMIN-NEXT: call callee_half_on_stack
; RV64-LP64ZFHMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, 16
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: caller_half_on_stack:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32-NEXT: lui a7, 5
; RV32-ZFH-ILP32-NEXT: li a0, 1
; RV32-ZFH-ILP32-NEXT: li a1, 2
; RV32-ZFH-ILP32-NEXT: li a2, 3
; RV32-ZFH-ILP32-NEXT: li a3, 4
; RV32-ZFH-ILP32-NEXT: li a4, 5
; RV32-ZFH-ILP32-NEXT: li a5, 6
; RV32-ZFH-ILP32-NEXT: li a6, 7
; RV32-ZFH-ILP32-NEXT: addi a7, a7, -1792
; RV32-ZFH-ILP32-NEXT: fmv.h.x fa5, a7
; RV32-ZFH-ILP32-NEXT: li a7, 8
; RV32-ZFH-ILP32-NEXT: fsh fa5, 0(sp)
; RV32-ZFH-ILP32-NEXT: call callee_half_on_stack
; RV32-ZFH-ILP32-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: caller_half_on_stack:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32F-NEXT: lui a7, 5
; RV32-ZFH-ILP32F-NEXT: li a0, 1
; RV32-ZFH-ILP32F-NEXT: li a1, 2
; RV32-ZFH-ILP32F-NEXT: li a2, 3
; RV32-ZFH-ILP32F-NEXT: li a3, 4
; RV32-ZFH-ILP32F-NEXT: li a4, 5
; RV32-ZFH-ILP32F-NEXT: li a5, 6
; RV32-ZFH-ILP32F-NEXT: li a6, 7
; RV32-ZFH-ILP32F-NEXT: addi a7, a7, -1792
; RV32-ZFH-ILP32F-NEXT: fmv.h.x fa0, a7
; RV32-ZFH-ILP32F-NEXT: li a7, 8
; RV32-ZFH-ILP32F-NEXT: call callee_half_on_stack
; RV32-ZFH-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: caller_half_on_stack:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64-NEXT: lui a7, 5
; RV64-ZFH-LP64-NEXT: li a0, 1
; RV64-ZFH-LP64-NEXT: li a1, 2
; RV64-ZFH-LP64-NEXT: li a2, 3
; RV64-ZFH-LP64-NEXT: li a3, 4
; RV64-ZFH-LP64-NEXT: li a4, 5
; RV64-ZFH-LP64-NEXT: li a5, 6
; RV64-ZFH-LP64-NEXT: li a6, 7
; RV64-ZFH-LP64-NEXT: addi a7, a7, -1792
; RV64-ZFH-LP64-NEXT: fmv.h.x fa5, a7
; RV64-ZFH-LP64-NEXT: li a7, 8
; RV64-ZFH-LP64-NEXT: fsh fa5, 0(sp)
; RV64-ZFH-LP64-NEXT: call callee_half_on_stack
; RV64-ZFH-LP64-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: caller_half_on_stack:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64F-NEXT: lui a7, 5
; RV64-ZFH-LP64F-NEXT: li a0, 1
; RV64-ZFH-LP64F-NEXT: li a1, 2
; RV64-ZFH-LP64F-NEXT: li a2, 3
; RV64-ZFH-LP64F-NEXT: li a3, 4
; RV64-ZFH-LP64F-NEXT: li a4, 5
; RV64-ZFH-LP64F-NEXT: li a5, 6
; RV64-ZFH-LP64F-NEXT: li a6, 7
; RV64-ZFH-LP64F-NEXT: addi a7, a7, -1792
; RV64-ZFH-LP64F-NEXT: fmv.h.x fa0, a7
; RV64-ZFH-LP64F-NEXT: li a7, 8
; RV64-ZFH-LP64F-NEXT: call callee_half_on_stack
; RV64-ZFH-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64F-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64F-NEXT: ret
%1 = call i32 @callee_half_on_stack(i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, half 10.0)
ret i32 %1
}
define i32 @callee_half_on_stack_exhausted_gprs_fprs(i32 %a, float %fa, i32 %b, float %fb, i32 %c, float %fc, i32 %d, float %fd, i32 %e, float %fe, i32 %f, float %ff, i32 %g, float %fg, i32 %h, float %fh, half %i) nounwind {
; RV32I-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32I-NEXT: lw s0, 40(sp)
; RV32I-NEXT: lhu a0, 48(sp)
; RV32I-NEXT: call __extendhfsf2
; RV32I-NEXT: call __fixsfsi
; RV32I-NEXT: add a0, s0, a0
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64I-NEXT: lw s0, 64(sp)
; RV64I-NEXT: lhu a0, 80(sp)
; RV64I-NEXT: call __extendhfsf2
; RV64I-NEXT: call __fixsfdi
; RV64I-NEXT: addw a0, s0, a0
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32IF-NEXT: lw s0, 40(sp)
; RV32IF-NEXT: lhu a0, 48(sp)
; RV32IF-NEXT: call __extendhfsf2
; RV32IF-NEXT: fmv.w.x fa5, a0
; RV32IF-NEXT: fcvt.w.s a0, fa5, rtz
; RV32IF-NEXT: add a0, s0, a0
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64IF-NEXT: lw s0, 64(sp)
; RV64IF-NEXT: lhu a0, 80(sp)
; RV64IF-NEXT: call __extendhfsf2
; RV64IF-NEXT: fmv.w.x fa5, a0
; RV64IF-NEXT: fcvt.l.s a0, fa5, rtz
; RV64IF-NEXT: addw a0, s0, a0
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: sw s0, 8(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: mv s0, a7
; RV32-ILP32F-NEXT: lhu a0, 16(sp)
; RV32-ILP32F-NEXT: fmv.w.x fa0, a0
; RV32-ILP32F-NEXT: call __extendhfsf2
; RV32-ILP32F-NEXT: fcvt.w.s a0, fa0, rtz
; RV32-ILP32F-NEXT: add a0, s0, a0
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: lw s0, 8(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: sd s0, 0(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: lhu a0, 16(sp)
; RV64-LP64F-NEXT: mv s0, a7
; RV64-LP64F-NEXT: fmv.w.x fa0, a0
; RV64-LP64F-NEXT: call __extendhfsf2
; RV64-LP64F-NEXT: fcvt.l.s a0, fa0, rtz
; RV64-LP64F-NEXT: addw a0, s0, a0
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: ld s0, 0(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: flh fa5, 0(sp)
; RV32-ILP32ZFHMIN-NEXT: fcvt.s.h fa5, fa5
; RV32-ILP32ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV32-ILP32ZFHMIN-NEXT: add a0, a7, a0
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: flh fa5, 0(sp)
; RV64-LP64ZFHMIN-NEXT: fcvt.s.h fa5, fa5
; RV64-LP64ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV64-LP64ZFHMIN-NEXT: addw a0, a7, a0
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: flh fa5, 32(sp)
; RV32-ZFH-ILP32-NEXT: lw a0, 24(sp)
; RV32-ZFH-ILP32-NEXT: fcvt.w.h a1, fa5, rtz
; RV32-ZFH-ILP32-NEXT: add a0, a0, a1
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: flh fa5, 0(sp)
; RV32-ZFH-ILP32F-NEXT: fcvt.w.h a0, fa5, rtz
; RV32-ZFH-ILP32F-NEXT: add a0, a7, a0
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: flh fa5, 64(sp)
; RV64-ZFH-LP64-NEXT: lw a0, 48(sp)
; RV64-ZFH-LP64-NEXT: fcvt.w.h a1, fa5, rtz
; RV64-ZFH-LP64-NEXT: addw a0, a0, a1
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: callee_half_on_stack_exhausted_gprs_fprs:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: flh fa5, 0(sp)
; RV64-ZFH-LP64F-NEXT: fcvt.w.h a0, fa5, rtz
; RV64-ZFH-LP64F-NEXT: addw a0, a7, a0
; RV64-ZFH-LP64F-NEXT: ret
%1 = fptosi half %i to i32
%2 = add i32 %h, %1
ret i32 %2
}
define i32 @caller_half_on_stack_exhausted_gprs_fprs() nounwind {
; RV32I-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -48
; RV32I-NEXT: sw ra, 44(sp) # 4-byte Folded Spill
; RV32I-NEXT: lui a7, 5
; RV32I-NEXT: lui a6, 266240
; RV32I-NEXT: li t0, 8
; RV32I-NEXT: lui t1, 265728
; RV32I-NEXT: li t2, 7
; RV32I-NEXT: lui t3, 265216
; RV32I-NEXT: li t4, 6
; RV32I-NEXT: lui t5, 264704
; RV32I-NEXT: li t6, 5
; RV32I-NEXT: li a0, 1
; RV32I-NEXT: lui a1, 260096
; RV32I-NEXT: li a2, 2
; RV32I-NEXT: lui a3, 262144
; RV32I-NEXT: li a4, 3
; RV32I-NEXT: lui a5, 263168
; RV32I-NEXT: sw t2, 16(sp)
; RV32I-NEXT: sw t1, 20(sp)
; RV32I-NEXT: sw t0, 24(sp)
; RV32I-NEXT: sw a6, 28(sp)
; RV32I-NEXT: li a6, 4
; RV32I-NEXT: addi a7, a7, -1792
; RV32I-NEXT: sw a7, 32(sp)
; RV32I-NEXT: lui a7, 264192
; RV32I-NEXT: sw t6, 0(sp)
; RV32I-NEXT: sw t5, 4(sp)
; RV32I-NEXT: sw t4, 8(sp)
; RV32I-NEXT: sw t3, 12(sp)
; RV32I-NEXT: call callee_half_on_stack
; RV32I-NEXT: lw ra, 44(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 48
; RV32I-NEXT: ret
;
; RV64I-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -80
; RV64I-NEXT: sd ra, 72(sp) # 8-byte Folded Spill
; RV64I-NEXT: lui a7, 5
; RV64I-NEXT: lui a6, 266240
; RV64I-NEXT: li t0, 8
; RV64I-NEXT: lui t1, 265728
; RV64I-NEXT: li t2, 7
; RV64I-NEXT: lui t3, 265216
; RV64I-NEXT: li t4, 6
; RV64I-NEXT: lui t5, 264704
; RV64I-NEXT: li t6, 5
; RV64I-NEXT: li a0, 1
; RV64I-NEXT: lui a1, 260096
; RV64I-NEXT: li a2, 2
; RV64I-NEXT: lui a3, 262144
; RV64I-NEXT: li a4, 3
; RV64I-NEXT: lui a5, 263168
; RV64I-NEXT: sd t2, 32(sp)
; RV64I-NEXT: sd t1, 40(sp)
; RV64I-NEXT: sd t0, 48(sp)
; RV64I-NEXT: sd a6, 56(sp)
; RV64I-NEXT: li a6, 4
; RV64I-NEXT: addi a7, a7, -1792
; RV64I-NEXT: sd a7, 64(sp)
; RV64I-NEXT: lui a7, 264192
; RV64I-NEXT: sd t6, 0(sp)
; RV64I-NEXT: sd t5, 8(sp)
; RV64I-NEXT: sd t4, 16(sp)
; RV64I-NEXT: sd t3, 24(sp)
; RV64I-NEXT: call callee_half_on_stack
; RV64I-NEXT: ld ra, 72(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 80
; RV64I-NEXT: ret
;
; RV32IF-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -48
; RV32IF-NEXT: sw ra, 44(sp) # 4-byte Folded Spill
; RV32IF-NEXT: lui a7, 1048565
; RV32IF-NEXT: lui a6, 266240
; RV32IF-NEXT: li t0, 8
; RV32IF-NEXT: lui t1, 265728
; RV32IF-NEXT: li t2, 7
; RV32IF-NEXT: lui t3, 265216
; RV32IF-NEXT: li t4, 6
; RV32IF-NEXT: lui t5, 264704
; RV32IF-NEXT: li t6, 5
; RV32IF-NEXT: li a0, 1
; RV32IF-NEXT: lui a1, 260096
; RV32IF-NEXT: li a2, 2
; RV32IF-NEXT: lui a3, 262144
; RV32IF-NEXT: li a4, 3
; RV32IF-NEXT: lui a5, 263168
; RV32IF-NEXT: sw t2, 16(sp)
; RV32IF-NEXT: sw t1, 20(sp)
; RV32IF-NEXT: sw t0, 24(sp)
; RV32IF-NEXT: sw a6, 28(sp)
; RV32IF-NEXT: li a6, 4
; RV32IF-NEXT: addi a7, a7, -1792
; RV32IF-NEXT: sw a7, 32(sp)
; RV32IF-NEXT: lui a7, 264192
; RV32IF-NEXT: sw t6, 0(sp)
; RV32IF-NEXT: sw t5, 4(sp)
; RV32IF-NEXT: sw t4, 8(sp)
; RV32IF-NEXT: sw t3, 12(sp)
; RV32IF-NEXT: call callee_half_on_stack
; RV32IF-NEXT: lw ra, 44(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 48
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -80
; RV64IF-NEXT: sd ra, 72(sp) # 8-byte Folded Spill
; RV64IF-NEXT: lui a7, 1048565
; RV64IF-NEXT: lui a6, 266240
; RV64IF-NEXT: li t0, 8
; RV64IF-NEXT: lui t1, 265728
; RV64IF-NEXT: li t2, 7
; RV64IF-NEXT: lui t3, 265216
; RV64IF-NEXT: li t4, 6
; RV64IF-NEXT: lui t5, 264704
; RV64IF-NEXT: li t6, 5
; RV64IF-NEXT: li a0, 1
; RV64IF-NEXT: lui a1, 260096
; RV64IF-NEXT: li a2, 2
; RV64IF-NEXT: lui a3, 262144
; RV64IF-NEXT: li a4, 3
; RV64IF-NEXT: lui a5, 263168
; RV64IF-NEXT: sd t2, 32(sp)
; RV64IF-NEXT: sw t1, 40(sp)
; RV64IF-NEXT: sd t0, 48(sp)
; RV64IF-NEXT: sw a6, 56(sp)
; RV64IF-NEXT: li a6, 4
; RV64IF-NEXT: addi a7, a7, -1792
; RV64IF-NEXT: sw a7, 64(sp)
; RV64IF-NEXT: lui a7, 264192
; RV64IF-NEXT: sd t6, 0(sp)
; RV64IF-NEXT: sw t5, 8(sp)
; RV64IF-NEXT: sd t4, 16(sp)
; RV64IF-NEXT: sw t3, 24(sp)
; RV64IF-NEXT: call callee_half_on_stack
; RV64IF-NEXT: ld ra, 72(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 80
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: lui a7, 1048565
; RV32-ILP32F-NEXT: lui a0, 260096
; RV32-ILP32F-NEXT: lui a1, 262144
; RV32-ILP32F-NEXT: lui a2, 263168
; RV32-ILP32F-NEXT: lui a3, 264192
; RV32-ILP32F-NEXT: lui a4, 264704
; RV32-ILP32F-NEXT: lui a5, 265216
; RV32-ILP32F-NEXT: lui a6, 265728
; RV32-ILP32F-NEXT: lui t0, 266240
; RV32-ILP32F-NEXT: fmv.w.x fa0, a0
; RV32-ILP32F-NEXT: li a0, 1
; RV32-ILP32F-NEXT: fmv.w.x fa1, a1
; RV32-ILP32F-NEXT: li a1, 2
; RV32-ILP32F-NEXT: fmv.w.x fa2, a2
; RV32-ILP32F-NEXT: li a2, 3
; RV32-ILP32F-NEXT: fmv.w.x fa3, a3
; RV32-ILP32F-NEXT: li a3, 4
; RV32-ILP32F-NEXT: fmv.w.x fa4, a4
; RV32-ILP32F-NEXT: li a4, 5
; RV32-ILP32F-NEXT: fmv.w.x fa5, a5
; RV32-ILP32F-NEXT: li a5, 6
; RV32-ILP32F-NEXT: fmv.w.x fa6, a6
; RV32-ILP32F-NEXT: li a6, 7
; RV32-ILP32F-NEXT: addi t1, a7, -1792
; RV32-ILP32F-NEXT: fmv.w.x fa7, t0
; RV32-ILP32F-NEXT: li a7, 8
; RV32-ILP32F-NEXT: sw t1, 0(sp)
; RV32-ILP32F-NEXT: call callee_half_on_stack
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: lui a7, 1048565
; RV64-LP64F-NEXT: lui a0, 260096
; RV64-LP64F-NEXT: lui a1, 262144
; RV64-LP64F-NEXT: lui a2, 263168
; RV64-LP64F-NEXT: lui a3, 264192
; RV64-LP64F-NEXT: lui a4, 264704
; RV64-LP64F-NEXT: lui a5, 265216
; RV64-LP64F-NEXT: lui a6, 265728
; RV64-LP64F-NEXT: lui t0, 266240
; RV64-LP64F-NEXT: fmv.w.x fa0, a0
; RV64-LP64F-NEXT: li a0, 1
; RV64-LP64F-NEXT: fmv.w.x fa1, a1
; RV64-LP64F-NEXT: li a1, 2
; RV64-LP64F-NEXT: fmv.w.x fa2, a2
; RV64-LP64F-NEXT: li a2, 3
; RV64-LP64F-NEXT: fmv.w.x fa3, a3
; RV64-LP64F-NEXT: li a3, 4
; RV64-LP64F-NEXT: fmv.w.x fa4, a4
; RV64-LP64F-NEXT: li a4, 5
; RV64-LP64F-NEXT: fmv.w.x fa5, a5
; RV64-LP64F-NEXT: li a5, 6
; RV64-LP64F-NEXT: fmv.w.x fa6, a6
; RV64-LP64F-NEXT: li a6, 7
; RV64-LP64F-NEXT: addi t1, a7, -1792
; RV64-LP64F-NEXT: fmv.w.x fa7, t0
; RV64-LP64F-NEXT: li a7, 8
; RV64-LP64F-NEXT: sw t1, 0(sp)
; RV64-LP64F-NEXT: call callee_half_on_stack
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, -16
; RV32-ILP32ZFHMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32ZFHMIN-NEXT: lui a7, 5
; RV32-ILP32ZFHMIN-NEXT: lui a0, 260096
; RV32-ILP32ZFHMIN-NEXT: lui a1, 262144
; RV32-ILP32ZFHMIN-NEXT: lui a2, 263168
; RV32-ILP32ZFHMIN-NEXT: lui a3, 264192
; RV32-ILP32ZFHMIN-NEXT: lui a4, 264704
; RV32-ILP32ZFHMIN-NEXT: lui a5, 265216
; RV32-ILP32ZFHMIN-NEXT: lui a6, 265728
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa0, a0
; RV32-ILP32ZFHMIN-NEXT: lui t0, 266240
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa1, a1
; RV32-ILP32ZFHMIN-NEXT: li a0, 1
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa2, a2
; RV32-ILP32ZFHMIN-NEXT: li a1, 2
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa3, a3
; RV32-ILP32ZFHMIN-NEXT: li a2, 3
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa4, a4
; RV32-ILP32ZFHMIN-NEXT: li a3, 4
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa5, a5
; RV32-ILP32ZFHMIN-NEXT: li a4, 5
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa6, a6
; RV32-ILP32ZFHMIN-NEXT: li a5, 6
; RV32-ILP32ZFHMIN-NEXT: fmv.w.x fa7, t0
; RV32-ILP32ZFHMIN-NEXT: li a6, 7
; RV32-ILP32ZFHMIN-NEXT: addi a7, a7, -1792
; RV32-ILP32ZFHMIN-NEXT: fmv.h.x ft0, a7
; RV32-ILP32ZFHMIN-NEXT: li a7, 8
; RV32-ILP32ZFHMIN-NEXT: fsh ft0, 0(sp)
; RV32-ILP32ZFHMIN-NEXT: call callee_half_on_stack
; RV32-ILP32ZFHMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, 16
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, -16
; RV64-LP64ZFHMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64ZFHMIN-NEXT: lui a7, 5
; RV64-LP64ZFHMIN-NEXT: lui a0, 260096
; RV64-LP64ZFHMIN-NEXT: lui a1, 262144
; RV64-LP64ZFHMIN-NEXT: lui a2, 263168
; RV64-LP64ZFHMIN-NEXT: lui a3, 264192
; RV64-LP64ZFHMIN-NEXT: lui a4, 264704
; RV64-LP64ZFHMIN-NEXT: lui a5, 265216
; RV64-LP64ZFHMIN-NEXT: lui a6, 265728
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa0, a0
; RV64-LP64ZFHMIN-NEXT: lui t0, 266240
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa1, a1
; RV64-LP64ZFHMIN-NEXT: li a0, 1
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa2, a2
; RV64-LP64ZFHMIN-NEXT: li a1, 2
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa3, a3
; RV64-LP64ZFHMIN-NEXT: li a2, 3
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa4, a4
; RV64-LP64ZFHMIN-NEXT: li a3, 4
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa5, a5
; RV64-LP64ZFHMIN-NEXT: li a4, 5
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa6, a6
; RV64-LP64ZFHMIN-NEXT: li a5, 6
; RV64-LP64ZFHMIN-NEXT: fmv.w.x fa7, t0
; RV64-LP64ZFHMIN-NEXT: li a6, 7
; RV64-LP64ZFHMIN-NEXT: addi a7, a7, -1792
; RV64-LP64ZFHMIN-NEXT: fmv.h.x ft0, a7
; RV64-LP64ZFHMIN-NEXT: li a7, 8
; RV64-LP64ZFHMIN-NEXT: fsh ft0, 0(sp)
; RV64-LP64ZFHMIN-NEXT: call callee_half_on_stack
; RV64-LP64ZFHMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, 16
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: addi sp, sp, -48
; RV32-ZFH-ILP32-NEXT: sw ra, 44(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32-NEXT: lui a5, 266240
; RV32-ZFH-ILP32-NEXT: li a6, 8
; RV32-ZFH-ILP32-NEXT: lui a7, 265728
; RV32-ZFH-ILP32-NEXT: li t0, 7
; RV32-ZFH-ILP32-NEXT: lui t1, 265216
; RV32-ZFH-ILP32-NEXT: li t2, 6
; RV32-ZFH-ILP32-NEXT: lui t3, 264704
; RV32-ZFH-ILP32-NEXT: li t4, 5
; RV32-ZFH-ILP32-NEXT: lui t5, 5
; RV32-ZFH-ILP32-NEXT: li a0, 1
; RV32-ZFH-ILP32-NEXT: lui a1, 260096
; RV32-ZFH-ILP32-NEXT: li a2, 2
; RV32-ZFH-ILP32-NEXT: lui a3, 262144
; RV32-ZFH-ILP32-NEXT: li a4, 3
; RV32-ZFH-ILP32-NEXT: sw t0, 16(sp)
; RV32-ZFH-ILP32-NEXT: sw a7, 20(sp)
; RV32-ZFH-ILP32-NEXT: sw a6, 24(sp)
; RV32-ZFH-ILP32-NEXT: sw a5, 28(sp)
; RV32-ZFH-ILP32-NEXT: lui a5, 263168
; RV32-ZFH-ILP32-NEXT: sw t4, 0(sp)
; RV32-ZFH-ILP32-NEXT: sw t3, 4(sp)
; RV32-ZFH-ILP32-NEXT: sw t2, 8(sp)
; RV32-ZFH-ILP32-NEXT: sw t1, 12(sp)
; RV32-ZFH-ILP32-NEXT: li a6, 4
; RV32-ZFH-ILP32-NEXT: addi a7, t5, -1792
; RV32-ZFH-ILP32-NEXT: fmv.h.x fa5, a7
; RV32-ZFH-ILP32-NEXT: lui a7, 264192
; RV32-ZFH-ILP32-NEXT: fsh fa5, 32(sp)
; RV32-ZFH-ILP32-NEXT: call callee_half_on_stack
; RV32-ZFH-ILP32-NEXT: lw ra, 44(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32-NEXT: addi sp, sp, 48
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32F-NEXT: lui a7, 5
; RV32-ZFH-ILP32F-NEXT: lui a0, 260096
; RV32-ZFH-ILP32F-NEXT: lui a1, 262144
; RV32-ZFH-ILP32F-NEXT: lui a2, 263168
; RV32-ZFH-ILP32F-NEXT: lui a3, 264192
; RV32-ZFH-ILP32F-NEXT: lui a4, 264704
; RV32-ZFH-ILP32F-NEXT: lui a5, 265216
; RV32-ZFH-ILP32F-NEXT: lui a6, 265728
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa0, a0
; RV32-ZFH-ILP32F-NEXT: lui t0, 266240
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa1, a1
; RV32-ZFH-ILP32F-NEXT: li a0, 1
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa2, a2
; RV32-ZFH-ILP32F-NEXT: li a1, 2
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa3, a3
; RV32-ZFH-ILP32F-NEXT: li a2, 3
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa4, a4
; RV32-ZFH-ILP32F-NEXT: li a3, 4
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa5, a5
; RV32-ZFH-ILP32F-NEXT: li a4, 5
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa6, a6
; RV32-ZFH-ILP32F-NEXT: li a5, 6
; RV32-ZFH-ILP32F-NEXT: fmv.w.x fa7, t0
; RV32-ZFH-ILP32F-NEXT: li a6, 7
; RV32-ZFH-ILP32F-NEXT: addi a7, a7, -1792
; RV32-ZFH-ILP32F-NEXT: fmv.h.x ft0, a7
; RV32-ZFH-ILP32F-NEXT: li a7, 8
; RV32-ZFH-ILP32F-NEXT: fsh ft0, 0(sp)
; RV32-ZFH-ILP32F-NEXT: call callee_half_on_stack
; RV32-ZFH-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: addi sp, sp, -80
; RV64-ZFH-LP64-NEXT: sd ra, 72(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64-NEXT: lui a5, 266240
; RV64-ZFH-LP64-NEXT: li a6, 8
; RV64-ZFH-LP64-NEXT: lui a7, 265728
; RV64-ZFH-LP64-NEXT: li t0, 7
; RV64-ZFH-LP64-NEXT: lui t1, 265216
; RV64-ZFH-LP64-NEXT: li t2, 6
; RV64-ZFH-LP64-NEXT: lui t3, 264704
; RV64-ZFH-LP64-NEXT: li t4, 5
; RV64-ZFH-LP64-NEXT: lui t5, 5
; RV64-ZFH-LP64-NEXT: li a0, 1
; RV64-ZFH-LP64-NEXT: lui a1, 260096
; RV64-ZFH-LP64-NEXT: li a2, 2
; RV64-ZFH-LP64-NEXT: lui a3, 262144
; RV64-ZFH-LP64-NEXT: li a4, 3
; RV64-ZFH-LP64-NEXT: sd t0, 32(sp)
; RV64-ZFH-LP64-NEXT: sw a7, 40(sp)
; RV64-ZFH-LP64-NEXT: sd a6, 48(sp)
; RV64-ZFH-LP64-NEXT: sw a5, 56(sp)
; RV64-ZFH-LP64-NEXT: lui a5, 263168
; RV64-ZFH-LP64-NEXT: sd t4, 0(sp)
; RV64-ZFH-LP64-NEXT: sw t3, 8(sp)
; RV64-ZFH-LP64-NEXT: sd t2, 16(sp)
; RV64-ZFH-LP64-NEXT: sw t1, 24(sp)
; RV64-ZFH-LP64-NEXT: li a6, 4
; RV64-ZFH-LP64-NEXT: addi a7, t5, -1792
; RV64-ZFH-LP64-NEXT: fmv.h.x fa5, a7
; RV64-ZFH-LP64-NEXT: lui a7, 264192
; RV64-ZFH-LP64-NEXT: fsh fa5, 64(sp)
; RV64-ZFH-LP64-NEXT: call callee_half_on_stack
; RV64-ZFH-LP64-NEXT: ld ra, 72(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64-NEXT: addi sp, sp, 80
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: caller_half_on_stack_exhausted_gprs_fprs:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64F-NEXT: lui a7, 5
; RV64-ZFH-LP64F-NEXT: lui a0, 260096
; RV64-ZFH-LP64F-NEXT: lui a1, 262144
; RV64-ZFH-LP64F-NEXT: lui a2, 263168
; RV64-ZFH-LP64F-NEXT: lui a3, 264192
; RV64-ZFH-LP64F-NEXT: lui a4, 264704
; RV64-ZFH-LP64F-NEXT: lui a5, 265216
; RV64-ZFH-LP64F-NEXT: lui a6, 265728
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa0, a0
; RV64-ZFH-LP64F-NEXT: lui t0, 266240
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa1, a1
; RV64-ZFH-LP64F-NEXT: li a0, 1
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa2, a2
; RV64-ZFH-LP64F-NEXT: li a1, 2
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa3, a3
; RV64-ZFH-LP64F-NEXT: li a2, 3
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa4, a4
; RV64-ZFH-LP64F-NEXT: li a3, 4
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa5, a5
; RV64-ZFH-LP64F-NEXT: li a4, 5
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa6, a6
; RV64-ZFH-LP64F-NEXT: li a5, 6
; RV64-ZFH-LP64F-NEXT: fmv.w.x fa7, t0
; RV64-ZFH-LP64F-NEXT: li a6, 7
; RV64-ZFH-LP64F-NEXT: addi a7, a7, -1792
; RV64-ZFH-LP64F-NEXT: fmv.h.x ft0, a7
; RV64-ZFH-LP64F-NEXT: li a7, 8
; RV64-ZFH-LP64F-NEXT: fsh ft0, 0(sp)
; RV64-ZFH-LP64F-NEXT: call callee_half_on_stack
; RV64-ZFH-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64F-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64F-NEXT: ret
%1 = call i32 @callee_half_on_stack(i32 1, float 1.0, i32 2, float 2.0, i32 3, float 3.0, i32 4, float 4.0, i32 5, float 5.0, i32 6, float 6.0, i32 7, float 7.0, i32 8, float 8.0, half 10.0)
ret i32 %1
}
define half @callee_half_ret() nounwind {
; RV32I-LABEL: callee_half_ret:
; RV32I: # %bb.0:
; RV32I-NEXT: li a0, 15
; RV32I-NEXT: slli a0, a0, 10
; RV32I-NEXT: ret
;
; RV64I-LABEL: callee_half_ret:
; RV64I: # %bb.0:
; RV64I-NEXT: li a0, 15
; RV64I-NEXT: slli a0, a0, 10
; RV64I-NEXT: ret
;
; RV32IF-LABEL: callee_half_ret:
; RV32IF: # %bb.0:
; RV32IF-NEXT: lui a0, 1048564
; RV32IF-NEXT: addi a0, a0, -1024
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: callee_half_ret:
; RV64IF: # %bb.0:
; RV64IF-NEXT: lui a0, 1048564
; RV64IF-NEXT: addi a0, a0, -1024
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: callee_half_ret:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: lui a0, 1048564
; RV32-ILP32F-NEXT: addi a0, a0, -1024
; RV32-ILP32F-NEXT: fmv.w.x fa0, a0
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: callee_half_ret:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: lui a0, 1048564
; RV64-LP64F-NEXT: addi a0, a0, -1024
; RV64-LP64F-NEXT: fmv.w.x fa0, a0
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: callee_half_ret:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: li a0, 15
; RV32-ILP32ZFHMIN-NEXT: slli a0, a0, 10
; RV32-ILP32ZFHMIN-NEXT: fmv.h.x fa0, a0
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: callee_half_ret:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: li a0, 15
; RV64-LP64ZFHMIN-NEXT: slli a0, a0, 10
; RV64-LP64ZFHMIN-NEXT: fmv.h.x fa0, a0
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: callee_half_ret:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: li a0, 15
; RV32-ZFH-ILP32-NEXT: slli a0, a0, 10
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: callee_half_ret:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: li a0, 15
; RV32-ZFH-ILP32F-NEXT: slli a0, a0, 10
; RV32-ZFH-ILP32F-NEXT: fmv.h.x fa0, a0
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: callee_half_ret:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: li a0, 15
; RV64-ZFH-LP64-NEXT: slli a0, a0, 10
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: callee_half_ret:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: li a0, 15
; RV64-ZFH-LP64F-NEXT: slli a0, a0, 10
; RV64-ZFH-LP64F-NEXT: fmv.h.x fa0, a0
; RV64-ZFH-LP64F-NEXT: ret
ret half 1.0
}
define i32 @caller_half_ret() nounwind {
; RV32I-LABEL: caller_half_ret:
; RV32I: # %bb.0:
; RV32I-NEXT: addi sp, sp, -16
; RV32I-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32I-NEXT: call callee_half_ret
; RV32I-NEXT: slli a0, a0, 16
; RV32I-NEXT: srli a0, a0, 16
; RV32I-NEXT: call __extendhfsf2
; RV32I-NEXT: call __fixsfsi
; RV32I-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32I-NEXT: addi sp, sp, 16
; RV32I-NEXT: ret
;
; RV64I-LABEL: caller_half_ret:
; RV64I: # %bb.0:
; RV64I-NEXT: addi sp, sp, -16
; RV64I-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64I-NEXT: call callee_half_ret
; RV64I-NEXT: slli a0, a0, 48
; RV64I-NEXT: srli a0, a0, 48
; RV64I-NEXT: call __extendhfsf2
; RV64I-NEXT: call __fixsfdi
; RV64I-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64I-NEXT: addi sp, sp, 16
; RV64I-NEXT: ret
;
; RV32IF-LABEL: caller_half_ret:
; RV32IF: # %bb.0:
; RV32IF-NEXT: addi sp, sp, -16
; RV32IF-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32IF-NEXT: call callee_half_ret
; RV32IF-NEXT: call __extendhfsf2
; RV32IF-NEXT: fmv.w.x fa5, a0
; RV32IF-NEXT: fcvt.w.s a0, fa5, rtz
; RV32IF-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32IF-NEXT: addi sp, sp, 16
; RV32IF-NEXT: ret
;
; RV64IF-LABEL: caller_half_ret:
; RV64IF: # %bb.0:
; RV64IF-NEXT: addi sp, sp, -16
; RV64IF-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64IF-NEXT: call callee_half_ret
; RV64IF-NEXT: call __extendhfsf2
; RV64IF-NEXT: fmv.w.x fa5, a0
; RV64IF-NEXT: fcvt.l.s a0, fa5, rtz
; RV64IF-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64IF-NEXT: addi sp, sp, 16
; RV64IF-NEXT: ret
;
; RV32-ILP32F-LABEL: caller_half_ret:
; RV32-ILP32F: # %bb.0:
; RV32-ILP32F-NEXT: addi sp, sp, -16
; RV32-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32F-NEXT: call callee_half_ret
; RV32-ILP32F-NEXT: call __extendhfsf2
; RV32-ILP32F-NEXT: fcvt.w.s a0, fa0, rtz
; RV32-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32F-NEXT: addi sp, sp, 16
; RV32-ILP32F-NEXT: ret
;
; RV64-LP64F-LABEL: caller_half_ret:
; RV64-LP64F: # %bb.0:
; RV64-LP64F-NEXT: addi sp, sp, -16
; RV64-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64F-NEXT: call callee_half_ret
; RV64-LP64F-NEXT: call __extendhfsf2
; RV64-LP64F-NEXT: fcvt.l.s a0, fa0, rtz
; RV64-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64F-NEXT: addi sp, sp, 16
; RV64-LP64F-NEXT: ret
;
; RV32-ILP32ZFHMIN-LABEL: caller_half_ret:
; RV32-ILP32ZFHMIN: # %bb.0:
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, -16
; RV32-ILP32ZFHMIN-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ILP32ZFHMIN-NEXT: call callee_half_ret
; RV32-ILP32ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV32-ILP32ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV32-ILP32ZFHMIN-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ILP32ZFHMIN-NEXT: addi sp, sp, 16
; RV32-ILP32ZFHMIN-NEXT: ret
;
; RV64-LP64ZFHMIN-LABEL: caller_half_ret:
; RV64-LP64ZFHMIN: # %bb.0:
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, -16
; RV64-LP64ZFHMIN-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-LP64ZFHMIN-NEXT: call callee_half_ret
; RV64-LP64ZFHMIN-NEXT: fcvt.s.h fa5, fa0
; RV64-LP64ZFHMIN-NEXT: fcvt.w.s a0, fa5, rtz
; RV64-LP64ZFHMIN-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-LP64ZFHMIN-NEXT: addi sp, sp, 16
; RV64-LP64ZFHMIN-NEXT: ret
;
; RV32-ZFH-ILP32-LABEL: caller_half_ret:
; RV32-ZFH-ILP32: # %bb.0:
; RV32-ZFH-ILP32-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32-NEXT: call callee_half_ret
; RV32-ZFH-ILP32-NEXT: fmv.h.x fa5, a0
; RV32-ZFH-ILP32-NEXT: fcvt.w.h a0, fa5, rtz
; RV32-ZFH-ILP32-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32-NEXT: ret
;
; RV32-ZFH-ILP32F-LABEL: caller_half_ret:
; RV32-ZFH-ILP32F: # %bb.0:
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, -16
; RV32-ZFH-ILP32F-NEXT: sw ra, 12(sp) # 4-byte Folded Spill
; RV32-ZFH-ILP32F-NEXT: call callee_half_ret
; RV32-ZFH-ILP32F-NEXT: fcvt.w.h a0, fa0, rtz
; RV32-ZFH-ILP32F-NEXT: lw ra, 12(sp) # 4-byte Folded Reload
; RV32-ZFH-ILP32F-NEXT: addi sp, sp, 16
; RV32-ZFH-ILP32F-NEXT: ret
;
; RV64-ZFH-LP64-LABEL: caller_half_ret:
; RV64-ZFH-LP64: # %bb.0:
; RV64-ZFH-LP64-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64-NEXT: call callee_half_ret
; RV64-ZFH-LP64-NEXT: fmv.h.x fa5, a0
; RV64-ZFH-LP64-NEXT: fcvt.w.h a0, fa5, rtz
; RV64-ZFH-LP64-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64-NEXT: ret
;
; RV64-ZFH-LP64F-LABEL: caller_half_ret:
; RV64-ZFH-LP64F: # %bb.0:
; RV64-ZFH-LP64F-NEXT: addi sp, sp, -16
; RV64-ZFH-LP64F-NEXT: sd ra, 8(sp) # 8-byte Folded Spill
; RV64-ZFH-LP64F-NEXT: call callee_half_ret
; RV64-ZFH-LP64F-NEXT: fcvt.w.h a0, fa0, rtz
; RV64-ZFH-LP64F-NEXT: ld ra, 8(sp) # 8-byte Folded Reload
; RV64-ZFH-LP64F-NEXT: addi sp, sp, 16
; RV64-ZFH-LP64F-NEXT: ret
%1 = call half @callee_half_ret()
%2 = fptosi half %1 to i32
ret i32 %2
}