Files
clang-p2996/llvm/test/CodeGen/RISCV/calling-conv-vector-float.ll
Craig Topper 7b0c41841e [RISCV] Move compressible registers to the beginning of the FP allocation order.
We don't have very many compressible FP instructions, just load and store.
These instruction require the FP register to be f8-f15.

This patch changes the FP allocation order to prioritize f10-f15 first.
These are also the FP argument registers. So I allocated them in reverse
order starting at f15 to avoid taking the first argument registers.
This appears to match gcc allocation order.

Reviewed By: asb

Differential Revision: https://reviews.llvm.org/D146488
2023-03-27 17:29:28 -07:00

64 lines
2.0 KiB
LLVM

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc -mtriple=riscv64 -mattr=+f -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV64 %s
; RUN: llc -mtriple=riscv64 -mattr=+f -target-abi=lp64f -verify-machineinstrs < %s \
; RUN: | FileCheck -check-prefix=RV64LP64F %s
define <2 x float> @callee_v2f32(<2 x float> %x, <2 x float> %y) {
; RV64-LABEL: callee_v2f32:
; RV64: # %bb.0:
; RV64-NEXT: fmv.w.x fa5, a2
; RV64-NEXT: fmv.w.x fa4, a0
; RV64-NEXT: fmv.w.x fa3, a3
; RV64-NEXT: fmv.w.x fa2, a1
; RV64-NEXT: fadd.s fa3, fa2, fa3
; RV64-NEXT: fadd.s fa5, fa4, fa5
; RV64-NEXT: fmv.x.w a0, fa5
; RV64-NEXT: fmv.x.w a1, fa3
; RV64-NEXT: ret
;
; RV64LP64F-LABEL: callee_v2f32:
; RV64LP64F: # %bb.0:
; RV64LP64F-NEXT: fadd.s fa0, fa0, fa2
; RV64LP64F-NEXT: fadd.s fa1, fa1, fa3
; RV64LP64F-NEXT: ret
%z = fadd <2 x float> %x, %y
ret <2 x float> %z
}
define <4 x float> @callee_v4f32(<4 x float> %x, <4 x float> %y) {
; RV64-LABEL: callee_v4f32:
; RV64: # %bb.0:
; RV64-NEXT: fmv.w.x fa5, a4
; RV64-NEXT: fmv.w.x fa4, a7
; RV64-NEXT: fmv.w.x fa3, a3
; RV64-NEXT: fmv.w.x fa2, a6
; RV64-NEXT: fmv.w.x fa1, a2
; RV64-NEXT: fmv.w.x fa0, a5
; RV64-NEXT: fmv.w.x ft0, a1
; RV64-NEXT: flw ft1, 0(sp)
; RV64-NEXT: fadd.s fa0, ft0, fa0
; RV64-NEXT: fadd.s fa2, fa1, fa2
; RV64-NEXT: fadd.s fa4, fa3, fa4
; RV64-NEXT: fadd.s fa5, fa5, ft1
; RV64-NEXT: fsw fa5, 12(a0)
; RV64-NEXT: fsw fa4, 8(a0)
; RV64-NEXT: fsw fa2, 4(a0)
; RV64-NEXT: fsw fa0, 0(a0)
; RV64-NEXT: ret
;
; RV64LP64F-LABEL: callee_v4f32:
; RV64LP64F: # %bb.0:
; RV64LP64F-NEXT: fadd.s fa4, fa0, fa4
; RV64LP64F-NEXT: fadd.s fa5, fa1, fa5
; RV64LP64F-NEXT: fadd.s fa2, fa2, fa6
; RV64LP64F-NEXT: fadd.s fa3, fa3, fa7
; RV64LP64F-NEXT: fsw fa3, 12(a0)
; RV64LP64F-NEXT: fsw fa2, 8(a0)
; RV64LP64F-NEXT: fsw fa5, 4(a0)
; RV64LP64F-NEXT: fsw fa4, 0(a0)
; RV64LP64F-NEXT: ret
%z = fadd <4 x float> %x, %y
ret <4 x float> %z
}