RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-10-18 18:42:46 +02:00
Code Issues Projects Releases Wiki Activity

[RISCV] Support Constant Pools in Load/Store Peephole

Browse Source 
Summary:
RISC-V uses a post-select peephole pass to optimise
`(load/store (ADDI $reg, %lo(addr)), 0)` into `(load/store $reg, %lo(addr))`.
This peephole wasn't firing for accesses to constant pools, which is how we
materialise most floating point constants.

This adds support for the constantpool case, which improves code generation for
lots of small FP loading examples. I have not added any tests because this
structure is well-covered by the `fp-imm.ll` testcases, as well as almost
all other uses of floating point constants in the RISC-V backend tests.

Reviewed By: luismarques, asb

Differential Revision: https://reviews.llvm.org/D79523
This commit is contained in:
Sam Elliott 2020-05-11 18:52:42 +01:00
parent 6c9d2b0609
commit 1b981ebf72
9 changed files with 72 additions and 131 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Target/RISCV/RISCVISelDAGToDAG.cpp
View File

@ -223,6 +223,10 @@ void RISCVDAGToDAGISel::doPeepholeLoadStoreADDI() {
ImmOperand = CurDAG->getTargetGlobalAddress(
GA->getGlobal(), SDLoc(ImmOperand), ImmOperand.getValueType(),
GA->getOffset(), GA->getTargetFlags());
} else if (auto CP = dyn_cast<ConstantPoolSDNode>(ImmOperand)) {
ImmOperand = CurDAG->getTargetConstantPool(
CP->getConstVal(), ImmOperand.getValueType(), CP->getAlign(),
CP->getOffset(), CP->getTargetFlags());
} else {
continue;
}

81
test/CodeGen/RISCV/calling-conv-ilp32d.ll
View File

@ -23,8 +23,7 @@ define i32 @caller_double_in_fpr() nounwind {
; RV32-ILP32D-NEXT: addi sp, sp, -16
; RV32-ILP32D-NEXT: sw ra, 12(sp)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI1_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI1_0)(a0)
; RV32-ILP32D-NEXT: addi a0, zero, 1
; RV32-ILP32D-NEXT: call callee_double_in_fpr
; RV32-ILP32D-NEXT: lw ra, 12(sp)
@ -54,8 +53,7 @@ define i32 @caller_double_in_fpr_exhausted_gprs() nounwind {
; RV32-ILP32D-NEXT: sw ra, 12(sp)
; RV32-ILP32D-NEXT: addi a1, zero, 5
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI3_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI3_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI3_0)(a0)
; RV32-ILP32D-NEXT: addi a0, zero, 1
; RV32-ILP32D-NEXT: addi a2, zero, 2
; RV32-ILP32D-NEXT: addi a4, zero, 3
@ -99,29 +97,21 @@ define i32 @caller_double_in_gpr_exhausted_fprs() nounwind {
; RV32-ILP32D-NEXT: addi sp, sp, -16
; RV32-ILP32D-NEXT: sw ra, 12(sp)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI5_0)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_1)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_1)
; RV32-ILP32D-NEXT: fld fa1, 0(a0)
; RV32-ILP32D-NEXT: fld fa1, %lo(.LCPI5_1)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_2)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_2)
; RV32-ILP32D-NEXT: fld fa2, 0(a0)
; RV32-ILP32D-NEXT: fld fa2, %lo(.LCPI5_2)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_3)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_3)
; RV32-ILP32D-NEXT: fld fa3, 0(a0)
; RV32-ILP32D-NEXT: fld fa3, %lo(.LCPI5_3)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_4)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_4)
; RV32-ILP32D-NEXT: fld fa4, 0(a0)
; RV32-ILP32D-NEXT: fld fa4, %lo(.LCPI5_4)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_5)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_5)
; RV32-ILP32D-NEXT: fld fa5, 0(a0)
; RV32-ILP32D-NEXT: fld fa5, %lo(.LCPI5_5)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_6)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_6)
; RV32-ILP32D-NEXT: fld fa6, 0(a0)
; RV32-ILP32D-NEXT: fld fa6, %lo(.LCPI5_6)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI5_7)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI5_7)
; RV32-ILP32D-NEXT: fld fa7, 0(a0)
; RV32-ILP32D-NEXT: fld fa7, %lo(.LCPI5_7)(a0)
; RV32-ILP32D-NEXT: lui a1, 262688
; RV32-ILP32D-NEXT: mv a0, zero
; RV32-ILP32D-NEXT: call callee_double_in_gpr_exhausted_fprs
@ -159,29 +149,21 @@ define i32 @caller_double_in_gpr_and_stack_almost_exhausted_gprs_fprs() nounwind
; RV32-ILP32D-NEXT: sw ra, 12(sp)
; RV32-ILP32D-NEXT: lui a1, 262816
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI7_0)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_1)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_1)
; RV32-ILP32D-NEXT: fld fa1, 0(a0)
; RV32-ILP32D-NEXT: fld fa1, %lo(.LCPI7_1)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_2)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_2)
; RV32-ILP32D-NEXT: fld fa2, 0(a0)
; RV32-ILP32D-NEXT: fld fa2, %lo(.LCPI7_2)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_3)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_3)
; RV32-ILP32D-NEXT: fld fa3, 0(a0)
; RV32-ILP32D-NEXT: fld fa3, %lo(.LCPI7_3)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_4)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_4)
; RV32-ILP32D-NEXT: fld fa4, 0(a0)
; RV32-ILP32D-NEXT: fld fa4, %lo(.LCPI7_4)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_5)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_5)
; RV32-ILP32D-NEXT: fld fa5, 0(a0)
; RV32-ILP32D-NEXT: fld fa5, %lo(.LCPI7_5)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_6)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_6)
; RV32-ILP32D-NEXT: fld fa6, 0(a0)
; RV32-ILP32D-NEXT: fld fa6, %lo(.LCPI7_6)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI7_7)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI7_7)
; RV32-ILP32D-NEXT: fld fa7, 0(a0)
; RV32-ILP32D-NEXT: fld fa7, %lo(.LCPI7_7)(a0)
; RV32-ILP32D-NEXT: addi a0, zero, 1
; RV32-ILP32D-NEXT: addi a2, zero, 3
; RV32-ILP32D-NEXT: addi a4, zero, 5
@ -224,29 +206,21 @@ define i32 @caller_double_on_stack_exhausted_gprs_fprs() nounwind {
; RV32-ILP32D-NEXT: lui a0, 262816
; RV32-ILP32D-NEXT: sw a0, 4(sp)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI9_0)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_1)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_1)
; RV32-ILP32D-NEXT: fld fa1, 0(a0)
; RV32-ILP32D-NEXT: fld fa1, %lo(.LCPI9_1)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_2)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_2)
; RV32-ILP32D-NEXT: fld fa2, 0(a0)
; RV32-ILP32D-NEXT: fld fa2, %lo(.LCPI9_2)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_3)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_3)
; RV32-ILP32D-NEXT: fld fa3, 0(a0)
; RV32-ILP32D-NEXT: fld fa3, %lo(.LCPI9_3)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_4)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_4)
; RV32-ILP32D-NEXT: fld fa4, 0(a0)
; RV32-ILP32D-NEXT: fld fa4, %lo(.LCPI9_4)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_5)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_5)
; RV32-ILP32D-NEXT: fld fa5, 0(a0)
; RV32-ILP32D-NEXT: fld fa5, %lo(.LCPI9_5)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_6)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_6)
; RV32-ILP32D-NEXT: fld fa6, 0(a0)
; RV32-ILP32D-NEXT: fld fa6, %lo(.LCPI9_6)(a0)
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI9_7)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI9_7)
; RV32-ILP32D-NEXT: fld fa7, 0(a0)
; RV32-ILP32D-NEXT: fld fa7, %lo(.LCPI9_7)(a0)
; RV32-ILP32D-NEXT: addi a0, zero, 1
; RV32-ILP32D-NEXT: addi a2, zero, 3
; RV32-ILP32D-NEXT: addi a4, zero, 5
@ -270,8 +244,7 @@ define double @callee_double_ret() nounwind {
; RV32-ILP32D-LABEL: callee_double_ret:
; RV32-ILP32D: # %bb.0:
; RV32-ILP32D-NEXT: lui a0, %hi(.LCPI10_0)
; RV32-ILP32D-NEXT: addi a0, a0, %lo(.LCPI10_0)
; RV32-ILP32D-NEXT: fld fa0, 0(a0)
; RV32-ILP32D-NEXT: fld fa0, %lo(.LCPI10_0)(a0)
; RV32-ILP32D-NEXT: ret
ret double 1.0
}

57
test/CodeGen/RISCV/calling-conv-ilp32f-ilp32d-common.ll
View File

@ -26,8 +26,7 @@ define i32 @caller_float_in_fpr() nounwind {
; RV32-ILP32FD-NEXT: addi sp, sp, -16
; RV32-ILP32FD-NEXT: sw ra, 12(sp)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI1_0)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
; RV32-ILP32FD-NEXT: flw fa0, %lo(.LCPI1_0)(a0)
; RV32-ILP32FD-NEXT: addi a0, zero, 1
; RV32-ILP32FD-NEXT: call callee_float_in_fpr
; RV32-ILP32FD-NEXT: lw ra, 12(sp)
@ -57,8 +56,7 @@ define i32 @caller_float_in_fpr_exhausted_gprs() nounwind {
; RV32-ILP32FD-NEXT: sw ra, 12(sp)
; RV32-ILP32FD-NEXT: addi a1, zero, 5
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI3_0)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI3_0)
; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
; RV32-ILP32FD-NEXT: flw fa0, %lo(.LCPI3_0)(a0)
; RV32-ILP32FD-NEXT: addi a0, zero, 1
; RV32-ILP32FD-NEXT: addi a2, zero, 2
; RV32-ILP32FD-NEXT: addi a4, zero, 3
@ -98,29 +96,21 @@ define i32 @caller_float_in_gpr_exhausted_fprs() nounwind {
; RV32-ILP32FD-NEXT: addi sp, sp, -16
; RV32-ILP32FD-NEXT: sw ra, 12(sp)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_0)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_0)
; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
; RV32-ILP32FD-NEXT: flw fa0, %lo(.LCPI5_0)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_1)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_1)
; RV32-ILP32FD-NEXT: flw fa1, 0(a0)
; RV32-ILP32FD-NEXT: flw fa1, %lo(.LCPI5_1)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_2)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_2)
; RV32-ILP32FD-NEXT: flw fa2, 0(a0)
; RV32-ILP32FD-NEXT: flw fa2, %lo(.LCPI5_2)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_3)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_3)
; RV32-ILP32FD-NEXT: flw fa3, 0(a0)
; RV32-ILP32FD-NEXT: flw fa3, %lo(.LCPI5_3)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_4)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_4)
; RV32-ILP32FD-NEXT: flw fa4, 0(a0)
; RV32-ILP32FD-NEXT: flw fa4, %lo(.LCPI5_4)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_5)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_5)
; RV32-ILP32FD-NEXT: flw fa5, 0(a0)
; RV32-ILP32FD-NEXT: flw fa5, %lo(.LCPI5_5)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_6)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_6)
; RV32-ILP32FD-NEXT: flw fa6, 0(a0)
; RV32-ILP32FD-NEXT: flw fa6, %lo(.LCPI5_6)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI5_7)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI5_7)
; RV32-ILP32FD-NEXT: flw fa7, 0(a0)
; RV32-ILP32FD-NEXT: flw fa7, %lo(.LCPI5_7)(a0)
; RV32-ILP32FD-NEXT: lui a0, 266496
; RV32-ILP32FD-NEXT: call callee_float_in_gpr_exhausted_fprs
; RV32-ILP32FD-NEXT: lw ra, 12(sp)
@ -153,29 +143,21 @@ define i32 @caller_float_on_stack_exhausted_gprs_fprs() nounwind {
; RV32-ILP32FD-NEXT: sw ra, 12(sp)
; RV32-ILP32FD-NEXT: lui a1, 267520
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_0)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_0)
; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
; RV32-ILP32FD-NEXT: flw fa0, %lo(.LCPI7_0)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_1)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_1)
; RV32-ILP32FD-NEXT: flw fa1, 0(a0)
; RV32-ILP32FD-NEXT: flw fa1, %lo(.LCPI7_1)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_2)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_2)
; RV32-ILP32FD-NEXT: flw fa2, 0(a0)
; RV32-ILP32FD-NEXT: flw fa2, %lo(.LCPI7_2)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_3)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_3)
; RV32-ILP32FD-NEXT: flw fa3, 0(a0)
; RV32-ILP32FD-NEXT: flw fa3, %lo(.LCPI7_3)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_4)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_4)
; RV32-ILP32FD-NEXT: flw fa4, 0(a0)
; RV32-ILP32FD-NEXT: flw fa4, %lo(.LCPI7_4)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_5)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_5)
; RV32-ILP32FD-NEXT: flw fa5, 0(a0)
; RV32-ILP32FD-NEXT: flw fa5, %lo(.LCPI7_5)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_6)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_6)
; RV32-ILP32FD-NEXT: flw fa6, 0(a0)
; RV32-ILP32FD-NEXT: flw fa6, %lo(.LCPI7_6)(a0)
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI7_7)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI7_7)
; RV32-ILP32FD-NEXT: flw fa7, 0(a0)
; RV32-ILP32FD-NEXT: flw fa7, %lo(.LCPI7_7)(a0)
; RV32-ILP32FD-NEXT: addi a0, zero, 1
; RV32-ILP32FD-NEXT: addi a2, zero, 3
; RV32-ILP32FD-NEXT: addi a4, zero, 5
@ -199,8 +181,7 @@ define float @callee_float_ret() nounwind {
; RV32-ILP32FD-LABEL: callee_float_ret:
; RV32-ILP32FD: # %bb.0:
; RV32-ILP32FD-NEXT: lui a0, %hi(.LCPI8_0)
; RV32-ILP32FD-NEXT: addi a0, a0, %lo(.LCPI8_0)
; RV32-ILP32FD-NEXT: flw fa0, 0(a0)
; RV32-ILP32FD-NEXT: flw fa0, %lo(.LCPI8_0)(a0)
; RV32-ILP32FD-NEXT: ret
ret float 1.0
}

3
test/CodeGen/RISCV/codemodel-lowering.ll
View File

@ -132,8 +132,7 @@ define float @lower_constantpool(float %a) nounwind {
; RV32I-SMALL-LABEL: lower_constantpool:
; RV32I-SMALL: # %bb.0:
; RV32I-SMALL-NEXT: lui a1, %hi(.LCPI3_0)
; RV32I-SMALL-NEXT: addi a1, a1, %lo(.LCPI3_0)
; RV32I-SMALL-NEXT: flw ft0, 0(a1)
; RV32I-SMALL-NEXT: flw ft0, %lo(.LCPI3_0)(a1)
; RV32I-SMALL-NEXT: fmv.w.x ft1, a0
; RV32I-SMALL-NEXT: fadd.s ft0, ft1, ft0
; RV32I-SMALL-NEXT: fmv.x.w a0, ft0

6
test/CodeGen/RISCV/double-imm.ll
View File

@ -38,8 +38,7 @@ define double @double_imm_op(double %a) nounwind {
; RV32IFD-NEXT: sw a1, 12(sp)
; RV32IFD-NEXT: fld ft0, 8(sp)
; RV32IFD-NEXT: lui a0, %hi(.LCPI1_0)
; RV32IFD-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32IFD-NEXT: fld ft1, 0(a0)
; RV32IFD-NEXT: fld ft1, %lo(.LCPI1_0)(a0)
; RV32IFD-NEXT: fadd.d ft0, ft0, ft1
; RV32IFD-NEXT: fsd ft0, 8(sp)
; RV32IFD-NEXT: lw a0, 8(sp)
@ -50,8 +49,7 @@ define double @double_imm_op(double %a) nounwind {
; RV64IFD-LABEL: double_imm_op:
; RV64IFD: # %bb.0:
; RV64IFD-NEXT: lui a1, %hi(.LCPI1_0)
; RV64IFD-NEXT: addi a1, a1, %lo(.LCPI1_0)
; RV64IFD-NEXT: fld ft0, 0(a1)
; RV64IFD-NEXT: fld ft0, %lo(.LCPI1_0)(a1)
; RV64IFD-NEXT: fmv.d.x ft1, a0
; RV64IFD-NEXT: fadd.d ft0, ft1, ft0
; RV64IFD-NEXT: fmv.x.d a0, ft0

6
test/CodeGen/RISCV/double-previous-failure.ll
View File

@ -24,14 +24,12 @@ define i32 @main() nounwind {
; RV32IFD-NEXT: sw a1, 4(sp)
; RV32IFD-NEXT: fld ft0, 0(sp)
; RV32IFD-NEXT: lui a0, %hi(.LCPI1_0)
; RV32IFD-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32IFD-NEXT: fld ft1, 0(a0)
; RV32IFD-NEXT: fld ft1, %lo(.LCPI1_0)(a0)
; RV32IFD-NEXT: flt.d a0, ft0, ft1
; RV32IFD-NEXT: bnez a0, .LBB1_3
; RV32IFD-NEXT: # %bb.1: # %entry
; RV32IFD-NEXT: lui a0, %hi(.LCPI1_1)
; RV32IFD-NEXT: addi a0, a0, %lo(.LCPI1_1)
; RV32IFD-NEXT: fld ft1, 0(a0)
; RV32IFD-NEXT: fld ft1, %lo(.LCPI1_1)(a0)
; RV32IFD-NEXT: flt.d a0, ft1, ft0
; RV32IFD-NEXT: xori a0, a0, 1
; RV32IFD-NEXT: beqz a0, .LBB1_3

10
test/CodeGen/RISCV/float-imm.ll
View File

@ -15,20 +15,17 @@ define float @float_imm() nounwind {
; RV64IF-LABEL: float_imm:
; RV64IF: # %bb.0:
; RV64IF-NEXT: lui a0, %hi(.LCPI0_0)
; RV64IF-NEXT: addi a0, a0, %lo(.LCPI0_0)
; RV64IF-NEXT: flw ft0, 0(a0)
; RV64IF-NEXT: flw ft0, %lo(.LCPI0_0)(a0)
; RV64IF-NEXT: fmv.x.w a0, ft0
; RV64IF-NEXT: ret
ret float 3.14159274101257324218750
}
define float @float_imm_op(float %a) nounwind {
; TODO: addi should be folded in to the flw
; RV32IF-LABEL: float_imm_op:
; RV32IF: # %bb.0:
; RV32IF-NEXT: lui a1, %hi(.LCPI1_0)
; RV32IF-NEXT: addi a1, a1, %lo(.LCPI1_0)
; RV32IF-NEXT: flw ft0, 0(a1)
; RV32IF-NEXT: flw ft0, %lo(.LCPI1_0)(a1)
; RV32IF-NEXT: fmv.w.x ft1, a0
; RV32IF-NEXT: fadd.s ft0, ft1, ft0
; RV32IF-NEXT: fmv.x.w a0, ft0
@ -37,8 +34,7 @@ define float @float_imm_op(float %a) nounwind {
; RV64IF-LABEL: float_imm_op:
; RV64IF: # %bb.0:
; RV64IF-NEXT: lui a1, %hi(.LCPI1_0)
; RV64IF-NEXT: addi a1, a1, %lo(.LCPI1_0)
; RV64IF-NEXT: flw ft0, 0(a1)
; RV64IF-NEXT: flw ft0, %lo(.LCPI1_0)(a1)
; RV64IF-NEXT: fmv.w.x ft1, a0
; RV64IF-NEXT: fadd.s ft0, ft1, ft0
; RV64IF-NEXT: fmv.x.w a0, ft0

18
test/CodeGen/RISCV/fp-imm.ll
View File

@ -35,29 +35,25 @@ define float @f32_negative_zero(float *%pf) nounwind {
; RV32F-LABEL: f32_negative_zero:
; RV32F: # %bb.0:
; RV32F-NEXT: lui a0, %hi(.LCPI1_0)
; RV32F-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32F-NEXT: flw fa0, 0(a0)
; RV32F-NEXT: flw fa0, %lo(.LCPI1_0)(a0)
; RV32F-NEXT: ret
;
; RV32D-LABEL: f32_negative_zero:
; RV32D: # %bb.0:
; RV32D-NEXT: lui a0, %hi(.LCPI1_0)
; RV32D-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV32D-NEXT: flw fa0, 0(a0)
; RV32D-NEXT: flw fa0, %lo(.LCPI1_0)(a0)
; RV32D-NEXT: ret
;
; RV64F-LABEL: f32_negative_zero:
; RV64F: # %bb.0:
; RV64F-NEXT: lui a0, %hi(.LCPI1_0)
; RV64F-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV64F-NEXT: flw fa0, 0(a0)
; RV64F-NEXT: flw fa0, %lo(.LCPI1_0)(a0)
; RV64F-NEXT: ret
;
; RV64D-LABEL: f32_negative_zero:
; RV64D: # %bb.0:
; RV64D-NEXT: lui a0, %hi(.LCPI1_0)
; RV64D-NEXT: addi a0, a0, %lo(.LCPI1_0)
; RV64D-NEXT: flw fa0, 0(a0)
; RV64D-NEXT: flw fa0, %lo(.LCPI1_0)(a0)
; RV64D-NEXT: ret
ret float -0.0
}
@ -96,8 +92,7 @@ define double @f64_negative_zero(double *%pd) nounwind {
; RV32D-LABEL: f64_negative_zero:
; RV32D: # %bb.0:
; RV32D-NEXT: lui a0, %hi(.LCPI3_0)
; RV32D-NEXT: addi a0, a0, %lo(.LCPI3_0)
; RV32D-NEXT: fld fa0, 0(a0)
; RV32D-NEXT: fld fa0, %lo(.LCPI3_0)(a0)
; RV32D-NEXT: ret
;
; RV64F-LABEL: f64_negative_zero:
@ -109,8 +104,7 @@ define double @f64_negative_zero(double *%pd) nounwind {
; RV64D-LABEL: f64_negative_zero:
; RV64D: # %bb.0:
; RV64D-NEXT: lui a0, %hi(.LCPI3_0)
; RV64D-NEXT: addi a0, a0, %lo(.LCPI3_0)
; RV64D-NEXT: fld fa0, 0(a0)
; RV64D-NEXT: fld fa0, %lo(.LCPI3_0)(a0)
; RV64D-NEXT: ret
ret double -0.0
}

18
test/CodeGen/RISCV/select-const.ll
View File

@ -148,13 +148,12 @@ define float @select_const_fp(i1 zeroext %a) nounwind {
; RV32IF-NEXT: bnez a0, .LBB4_2
; RV32IF-NEXT: # %bb.1:
; RV32IF-NEXT: lui a0, %hi(.LCPI4_0)
; RV32IF-NEXT: addi a0, a0, %lo(.LCPI4_0)
; RV32IF-NEXT: j .LBB4_3
; RV32IF-NEXT: flw ft0, %lo(.LCPI4_0)(a0)
; RV32IF-NEXT: fmv.x.w a0, ft0
; RV32IF-NEXT: ret
; RV32IF-NEXT: .LBB4_2:
; RV32IF-NEXT: lui a0, %hi(.LCPI4_1)
; RV32IF-NEXT: addi a0, a0, %lo(.LCPI4_1)
; RV32IF-NEXT: .LBB4_3:
; RV32IF-NEXT: flw ft0, 0(a0)
; RV32IF-NEXT: flw ft0, %lo(.LCPI4_1)(a0)
; RV32IF-NEXT: fmv.x.w a0, ft0
; RV32IF-NEXT: ret
;
@ -173,13 +172,12 @@ define float @select_const_fp(i1 zeroext %a) nounwind {
; RV64IFD-NEXT: bnez a0, .LBB4_2
; RV64IFD-NEXT: # %bb.1:
; RV64IFD-NEXT: lui a0, %hi(.LCPI4_0)
; RV64IFD-NEXT: addi a0, a0, %lo(.LCPI4_0)
; RV64IFD-NEXT: j .LBB4_3
; RV64IFD-NEXT: flw ft0, %lo(.LCPI4_0)(a0)
; RV64IFD-NEXT: fmv.x.w a0, ft0
; RV64IFD-NEXT: ret
; RV64IFD-NEXT: .LBB4_2:
; RV64IFD-NEXT: lui a0, %hi(.LCPI4_1)
; RV64IFD-NEXT: addi a0, a0, %lo(.LCPI4_1)
; RV64IFD-NEXT: .LBB4_3:
; RV64IFD-NEXT: flw ft0, 0(a0)
; RV64IFD-NEXT: flw ft0, %lo(.LCPI4_1)(a0)
; RV64IFD-NEXT: fmv.x.w a0, ft0
; RV64IFD-NEXT: ret
%1 = select i1 %a, float 3.0, float 4.0