//=== ---- PPCMacroFusion.def - PowerPC MacroFuson Candidates -v-*- C++ -*-===// // // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. // See https)//llvm.org/LICENSE.txt for license information. // SPDX-License-Identifier) Apache-2.0 WITH LLVM-exception // //===----------------------------------------------------------------------===// // // This file contains descriptions of the macro-fusion pair for PowerPC. // //===----------------------------------------------------------------------===// // NOTE: NO INCLUDE GUARD DESIRED! #ifndef FUSION_FEATURE // Each FUSION_FEATURE is assigned with one TYPE, and can be enabled/disabled // by HAS_FEATURE. The instructions pair is fusable only when the opcode // of the first instruction is in OPSET1, and the second instruction opcode is // in OPSET2. And if DEP_OP_IDX >=0, we will check the result of first OP is // the operand of the second op with DEP_OP_IDX as its operand index. We assume // that the result of the first op is its operand zero. #define FUSION_FEATURE(TYPE, HAS_FEATURE, DEP_OP_IDX, OPSET1, OPSET2) #endif #ifndef FUSION_OP_SET #define FUSION_OP_SET(...) __VA_ARGS__ #endif // Power8 User Manual Section 10.1.12, Instruction Fusion // {addi} followed by one of these {lxvd2x, lxvw4x, lxvdsx, lvebx, lvehx, // lvewx, lvx, lxsdx} FUSION_FEATURE(AddiLoad, hasAddiLoadFusion, 2, \ FUSION_OP_SET(ADDI, ADDI8, ADDItocL, ADDItocL8), \ FUSION_OP_SET(LXVD2X, LXVW4X, LXVDSX, LVEBX, LVEHX, LVEWX, \ LVX, LXSDX)) // {addis) followed by one of these {ld, lbz, lhz, lwz} FUSION_FEATURE(AddisLoad, hasAddisLoadFusion, 2, \ FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8), \ FUSION_OP_SET(LD, LBZ, LBZ8, LHZ, LHZ8, LWZ, LWZ8)) // Power10 User Manual Section 19.1.5.4, Fusion // {add, mulld} - add FUSION_FEATURE(ArithAdd, hasArithAddFusion, -1, FUSION_OP_SET(ADD4, ADD8, MULLD), FUSION_OP_SET(ADD4, ADD8)) // {add, subf} - {and, nand, nor, or} FUSION_FEATURE(ArithLogical, hasAddLogicalFusion, -1, FUSION_OP_SET(ADD4, ADD8, SUBF, SUBF8), FUSION_OP_SET(AND, AND8, OR, OR8, NAND, NAND8, NOR, NOR8)) // {and, andc, eqv, nand, nor, or, orc, xor} - {add, subf} FUSION_FEATURE(LogicalArith, hasLogicalAddFusion, -1, FUSION_OP_SET(AND, ANDC, EQV, NAND, NOR, OR, ORC, XOR, AND8, ANDC8, EQV8, NAND8, NOR8, OR8, ORC8, XOR8), FUSION_OP_SET(ADD4, ADD8, SUBF, SUBF8)) // Either of {and, andc, eqv, nand, nor, or, orc, xor} FUSION_FEATURE(Logical, hasLogicalFusion, -1, FUSION_OP_SET(AND, ANDC, EQV, NAND, NOR, OR, ORC, XOR, AND8, ANDC8, EQV8, NAND8, NOR8, OR8, ORC8, XOR8), FUSION_OP_SET(AND, ANDC, EQV, NAND, NOR, OR, ORC, XOR, AND8, ANDC8, EQV8, NAND8, NOR8, OR8, ORC8, XOR8)) // vaddudm - vaddudm FUSION_FEATURE(VecAdd, hasArithAddFusion, -1, FUSION_OP_SET(VADDUDM), FUSION_OP_SET(VADDUDM)) // Either of {vand, vandc, veqv, vnand, vnor, vor, vorc, vxor} FUSION_FEATURE(VecLogical, hasLogicalFusion, -1, FUSION_OP_SET(VAND, VANDC, VEQV, VNAND, VNOR, VOR, VORC, VXOR), FUSION_OP_SET(VAND, VANDC, VEQV, VNAND, VNOR, VOR, VORC, VXOR)) // sldi rx, ra, {3, 6} - {add, subf} // sldi rx, ra n is alias of rldicr rx, ra, n, 63-n FUSION_FEATURE(SldiAdd, hasArithAddFusion, -1, FUSION_OP_SET(RLDICR, RLDICR_32), FUSION_OP_SET(ADD4, ADD8, SUBF, SUBF8)) // rldicl rx, ra, 1, 0 - xor FUSION_FEATURE(RotateLeftXor, hasSha3Fusion, 1, FUSION_OP_SET(RLDICL, RLDICL_32, RLDICL_32_64), FUSION_OP_SET(XOR, XOR8)) // rldicr rx, ra, 1, 63 - xor FUSION_FEATURE(RotateRightXor, hasSha3Fusion, 1, FUSION_OP_SET(RLDICR, RLDICR_32), FUSION_OP_SET(XOR, XOR8)) // There're two special cases in 'load-compare' series, so we have to split // them into several pattern groups to fit into current framework. This can // be clearer once we switched to a more expressive approach. // { lbz,lbzx,lhz,lhzx,lwz,lwzx } - cmpi 0,1,rx,{ 0,1,-1 } // { lbz,lbzx,lhz,lhzx,lwz,lwzx } - cmpli 0,L,rx,{ 0,1 } FUSION_FEATURE(LoadCmp1, hasCompareFusion, 1, FUSION_OP_SET(LBZ, LBZ8, LBZX, LBZX8, LBZXTLS, LBZXTLS_, LBZXTLS_32, LHZ, LHZ8, LHZX, LHZX8, LHZXTLS, LHZXTLS_, LHZXTLS_32, LWZ, LWZ8, LWZX, LWZX8, LWZXTLS, LWZXTLS_, LWZXTLS_32), FUSION_OP_SET(CMPDI, CMPLDI, CMPLWI)) // { ld,ldx } - cmpi 0,1,rx,{ 0,1,-1 } // { ld,ldx } - cmpli 0,1,rx,{ 0,1 } FUSION_FEATURE(LoadCmp2, hasCompareFusion, 1, FUSION_OP_SET(LD, LDX, LDXTLS, LDXTLS_), FUSION_OP_SET(CMPDI, CMPLDI)) // { lha,lhax,lwa,lwax } - cmpi 0,L,rx,{ 0,1,-1 } FUSION_FEATURE(LoadCmp3, hasCompareFusion, 1, FUSION_OP_SET(LHA, LHA8, LHAX, LHAX8, LWA, LWA_32, LWAX, LWAX_32), FUSION_OP_SET(CMPLDI, CMPLWI)) // ori - oris FUSION_FEATURE(OriOris, hasWideImmFusion, 1, FUSION_OP_SET(ORI, ORI8), FUSION_OP_SET(ORIS, ORIS8)) // lis - ori FUSION_FEATURE(LisOri, hasWideImmFusion, 1, FUSION_OP_SET(LIS, LIS8), FUSION_OP_SET(ORI, ORI8)) // oris - ori FUSION_FEATURE(OrisOri, hasWideImmFusion, 1, FUSION_OP_SET(ORIS, ORIS8), FUSION_OP_SET(ORI, ORI8)) // xori - xoris FUSION_FEATURE(XoriXoris, hasWideImmFusion, 1, FUSION_OP_SET(XORI, XORI8), FUSION_OP_SET(XORIS, XORIS8)) // xoris - xori FUSION_FEATURE(XorisXori, hasWideImmFusion, 1, FUSION_OP_SET(XORIS, XORIS8), FUSION_OP_SET(XORI, XORI8)) // addis rx,ra,si - addi rt,rx,SI, SI >= 0 FUSION_FEATURE(AddisAddi, hasWideImmFusion, 1, FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8, ADDIStocHA), FUSION_OP_SET(ADDI, ADDI8, ADDItocL8, ADDItocL)) // addi rx,ra,si - addis rt,rx,SI, ra > 0, SI >= 2 FUSION_FEATURE(AddiAddis, hasWideImmFusion, 1, FUSION_OP_SET(ADDI, ADDI8, ADDItocL8, ADDItocL), FUSION_OP_SET(ADDIS, ADDIS8, ADDIStocHA8, ADDIStocHA)) // mtctr - { bcctr,bcctrl } FUSION_FEATURE(ZeroMoveCTR, hasZeroMoveFusion, -1, FUSION_OP_SET(MTCTR, MTCTRloop, MTSPR8, MTSPR), FUSION_OP_SET(BCCTR, BCCTRn, BCCTR8, BCCTR8n, BCCTRL, BCCTRLn, BCCTRL8, BCCTRL8n, gBCCTR, gBCCTRL)) // mtlr - { bclr,bclrl } FUSION_FEATURE(ZeroMoveLR, hasZeroMoveFusion, -1, FUSION_OP_SET(MTLR8, MTLR, MTSPR8, MTSPR), FUSION_OP_SET(BCLR, BCLRn, gBCLR, BCLRL, BCLRLn, gBCLRL)) #include "PPCBack2BackFusion.def" #undef FUSION_FEATURE #undef FUSION_OP_SET