Emulator-Archive/ppsspp
1
0
Fork 0
mirror of https://github.com/hrydgard/ppsspp.git synced 2024-11-26 23:10:38 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

riscv: Add vector bitmanip encoding.

Browse Source
This commit is contained in:
Unknown W. Brackets 2023-12-01 23:49:51 -08:00
parent d6ce03e1db
commit 52db66e7ad
4 changed files with 137 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Common/CPUDetect.h
View File

@ -119,6 +119,8 @@ struct CPUInfo {
bool RiscV_Zfh;
bool RiscV_Zfhmin;
bool RiscV_Zicond;
bool RiscV_Zvbb;
bool RiscV_Zvkb;
// LoongArch specific extension flags.
bool LOONGARCH_CPUCFG;

2
Common/RiscVCPUDetect.cpp
View File

@ -241,6 +241,8 @@ std::vector<std::string> CPUInfo::Features() {
{ RiscV_D, "Double" },
{ RiscV_C, "Compressed" },
{ RiscV_V, "Vector" },
{ RiscV_Zvbb, "Vector Basic Bitmanip" },
{ RiscV_Zvkb, "Vector Crypto Bitmanip" },
{ RiscV_Zba, "Bitmanip Zba" },
{ RiscV_Zbb, "Bitmanip Zbb" },
{ RiscV_Zbc, "Bitmanip Zbc" },

118
Common/RiscVEmitter.cpp
View File

@ -67,6 +67,14 @@ static inline bool SupportsVector() {
return cpu_info.RiscV_V;
}
static inline bool SupportsVectorBitmanip(char zvxb) {
switch (zvxb) {
case 'b': return cpu_info.RiscV_Zvbb;
case 'k': return cpu_info.RiscV_Zvkb;
default: return false;
}
}
static inline bool SupportsBitmanip(char zbx) {
switch (zbx) {
case 'a': return cpu_info.RiscV_Zba;
@ -357,7 +365,13 @@ enum class Funct5 {
VFNCVT_RTZ_X_F = 0b10111,
VMV_S = 0b00000,
VPOPC = 0b10000,
VBREV8 = 0b01000,
VREV8 = 0b01001,
VBREV = 0b01010,
VCLZ = 0b01100,
VCTZ = 0b01101,
VCPOP_V = 0b01110,
VCPOP = 0b10000,
VFIRST = 0b10001,
VMSBF = 0b00001,
@ -397,6 +411,7 @@ enum class Funct6 {
C_SH = 0b100011,
VADD = 0b000000,
VANDN = 0b000001,
VSUB = 0b000010,
VRSUB = 0b000011,
VMINU = 0b000100,
@ -410,6 +425,9 @@ enum class Funct6 {
VSLIDEUP = 0b001110,
VRGATHEREI16 = 0b001110,
VSLIDEDOWN = 0b001111,
VROR = 0b010100,
VROL = 0b010101,
VWSLL = 0b110101,
VREDSUM = 0b000000,
VREDAND = 0b000001,
@ -3832,7 +3850,7 @@ void RiscVEmitter::VMXNOR_MM(RiscVReg vd, RiscVReg vs2, RiscVReg vs1) {
void RiscVEmitter::VCPOP_M(RiscVReg rd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(IsGPR(rd), "%s instruction rd must be GPR", __func__);
_assert_msg_(rd != R_ZERO, "%s should avoid write to zero", __func__);
Write32(EncodeV(rd, Funct3::OPMVV, (RiscVReg)Funct5::VPOPC, vs2, vm, Funct6::VRWUNARY0));
Write32(EncodeV(rd, Funct3::OPMVV, (RiscVReg)Funct5::VCPOP, vs2, vm, Funct6::VRWUNARY0));
}
void RiscVEmitter::VFIRST_M(RiscVReg rd, RiscVReg vs2, VUseMask vm) {
@ -3979,6 +3997,102 @@ void RiscVEmitter::VMVR_V(int regs, RiscVReg vd, RiscVReg vs2) {
Write32(EncodeIVI(vd, regs - 1, vs2, VUseMask::NONE, Funct6::VSMUL_VMVR));
}
void RiscVEmitter::VANDN_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVV(vd, vs1, vs2, vm, Funct6::VANDN));
}
void RiscVEmitter::VANDN_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVX(vd, rs1, vs2, vm, Funct6::VANDN));
}
void RiscVEmitter::VBREV_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VBREV, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VBREV8_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VBREV8, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VREV8_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VREV8, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VCLZ_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VCLZ, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VCTZ_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VCTZ, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VCPOP_V(RiscVReg vd, RiscVReg vs2, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
_assert_msg_(IsVPR(vd), "%s instruction vd must be VPR", __func__);
_assert_msg_(vm != VUseMask::V0_T || vd != V0, "%s instruction vd overlap with mask", __func__);
Write32(EncodeV(vd, Funct3::OPMVV, (RiscVReg)Funct5::VCPOP_V, vs2, vm, Funct6::VFXUNARY0));
}
void RiscVEmitter::VROL_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVV(vd, vs1, vs2, vm, Funct6::VROL));
}
void RiscVEmitter::VROL_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVX(vd, rs1, vs2, vm, Funct6::VROL));
}
void RiscVEmitter::VROR_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVV(vd, vs1, vs2, vm, Funct6::VROR));
}
void RiscVEmitter::VROR_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
Write32(EncodeIVX(vd, rs1, vs2, vm, Funct6::VROR));
}
void RiscVEmitter::VROR_VI(RiscVReg vd, RiscVReg vs2, u8 uimm6, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b') || SupportsVectorBitmanip('k'), "%s instruction requires Zvbb or Zvkb", __func__);
_assert_msg_(uimm6 < 64, "%s immediate must be 0-63", __func__);
// From an encoding perspective, easier to think of this as vror and vror32.
Funct6 variant = uimm6 >= 32 ? Funct6::VROL : Funct6::VROR;
Write32(EncodeIVI(vd, SignReduce32(uimm6, 5), vs2, vm, variant));
}
void RiscVEmitter::VWSLL_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
Write32(EncodeIVV(vd, vs1, vs2, vm, Funct6::VWSLL));
}
void RiscVEmitter::VWSLL_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
Write32(EncodeIVX(vd, rs1, vs2, vm, Funct6::VWSLL));
}
void RiscVEmitter::VWSLL_VI(RiscVReg vd, RiscVReg vs2, u8 uimm5, VUseMask vm) {
_assert_msg_(SupportsVectorBitmanip('b'), "%s instruction requires Zvbb", __func__);
_assert_msg_(uimm5 < 32, "%s immediate must be 0-31", __func__);
Write32(EncodeIVI(vd, SignReduce32(uimm5, 5), vs2, vm, Funct6::VWSLL));
}
void RiscVEmitter::ADD_UW(RiscVReg rd, RiscVReg rs1, RiscVReg rs2) {
if (BitsSupported() == 32) {
ADD(rd, rs1, rs2);

17
Common/RiscVEmitter.h
View File

@ -915,6 +915,23 @@ public:
void VCOMPRESS_VM(RiscVReg vd, RiscVReg vs2, RiscVReg vs1);
void VMVR_V(int regs, RiscVReg vd, RiscVReg vs2);
void VANDN_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm = VUseMask::NONE);
void VANDN_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm = VUseMask::NONE);
void VBREV_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VBREV8_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VREV8_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VCLZ_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VCTZ_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VCPOP_V(RiscVReg vd, RiscVReg vs2, VUseMask vm = VUseMask::NONE);
void VROL_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm = VUseMask::NONE);
void VROL_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm = VUseMask::NONE);
void VROR_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm = VUseMask::NONE);
void VROR_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm = VUseMask::NONE);
void VROR_VI(RiscVReg vd, RiscVReg vs2, u8 uimm6, VUseMask vm = VUseMask::NONE);
void VWSLL_VV(RiscVReg vd, RiscVReg vs2, RiscVReg vs1, VUseMask vm = VUseMask::NONE);
void VWSLL_VX(RiscVReg vd, RiscVReg vs2, RiscVReg rs1, VUseMask vm = VUseMask::NONE);
void VWSLL_VI(RiscVReg vd, RiscVReg vs2, u8 uimm5, VUseMask vm = VUseMask::NONE);
// Bitmanip instructions.
void ADD_UW(RiscVReg rd, RiscVReg rs1, RiscVReg rs2);
void SH_ADD(int shift, RiscVReg rd, RiscVReg rs1, RiscVReg rs2);