piwalker/dolphin
1
0
Fork 0
mirror of https://github.com/dolphin-emu/dolphin.git synced 2025-07-24 06:39:46 -06:00
Code Issues Packages Projects Releases Wiki Activity

JitArm64: Optimize ps_mergeXX

Browse Source 
1. In some cases, ps_merge01 can be implemented using one instruction.
2. When we need two instructions for ps_merge01, it's best to start with
   a MOV to avoid false dependencies on the destination register.
3. ps_merge10 can be implemented using a single EXT instruction.
This commit is contained in:
JosJuice
2022-11-26 17:57:16 +01:00
parent 0ef6d30a0d
commit f45d3a6a2c
3 changed files with 35 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
Source/Core/Common/Arm64Emitter.cpp
View File

@ -2334,6 +2334,16 @@ void ARM64FloatEmitter::EmitPermute(u32 size, u32 op, ARM64Reg Rd, ARM64Reg Rn,
(1 << 11) | (DecodeReg(Rn) << 5) | DecodeReg(Rd));
}
void ARM64FloatEmitter::EmitExtract(u32 imm4, u32 op, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
{
ASSERT_MSG(DYNA_REC, !IsSingle(Rd), "Singles are not supported!");
bool quad = IsQuad(Rd);
Write32((quad << 30) | (23 << 25) | (op << 22) | (DecodeReg(Rm) << 16) | (imm4 << 11) |
(DecodeReg(Rn) << 5) | DecodeReg(Rd));
}
void ARM64FloatEmitter::EmitScalarImm(bool M, bool S, u32 type, u32 imm5, ARM64Reg Rd, u32 imm8)
{
ASSERT_MSG(DYNA_REC, !IsQuad(Rd), "Vector is not supported!");
@ -3540,6 +3550,12 @@ void ARM64FloatEmitter::ZIP2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm)
EmitPermute(size, 0b111, Rd, Rn, Rm);
}
// Extract
void ARM64FloatEmitter::EXT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, u32 index)
{
EmitExtract(index, 0, Rd, Rn, Rm);
}
// Scalar shift by immediate
void ARM64FloatEmitter::SHL(ARM64Reg Rd, ARM64Reg Rn, u32 shift)
{

4
Source/Core/Common/Arm64Emitter.h
View File

@ -1247,6 +1247,9 @@ public:
void TRN2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
void ZIP2(u8 size, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
// Extract
void EXT(ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm, u32 index);
// Scalar shift by immediate
void SHL(ARM64Reg Rd, ARM64Reg Rn, u32 shift);
void URSHR(ARM64Reg Rd, ARM64Reg Rn, u32 shift);
@ -1305,6 +1308,7 @@ private:
void EmitCompare(bool M, bool S, u32 op, u32 opcode2, ARM64Reg Rn, ARM64Reg Rm);
void EmitCondSelect(bool M, bool S, CCFlags cond, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
void EmitPermute(u32 size, u32 op, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
void EmitExtract(u32 imm4, u32 op, ARM64Reg Rd, ARM64Reg Rn, ARM64Reg Rm);
void EmitScalarImm(bool M, bool S, u32 type, u32 imm5, ARM64Reg Rd, u32 imm8);
void EmitShiftImm(bool Q, bool U, u32 imm, u32 opcode, ARM64Reg Rd, ARM64Reg Rn);
void EmitScalarShiftImm(bool U, u32 imm, u32 opcode, ARM64Reg Rd, ARM64Reg Rn);