From: notaz Date: Wed, 28 Oct 2009 15:40:38 +0000 (+0000) Subject: 32x: drc: one manual page worth of opcodes implemented (x86 and arm) X-Git-Tag: v1.85~223 X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=picodrive.git;a=commitdiff_plain;h=3863edbd9d8245f11238819a8edeef30e79ef4ce 32x: drc: one manual page worth of opcodes implemented (x86 and arm) git-svn-id: file:///home/notaz/opt/svn/PicoDrive@827 be3aeb3a-fb24-0410-a615-afba39da0efa --- diff --git a/cpu/drc/emit_arm.c b/cpu/drc/emit_arm.c index d208779..bdef48b 100644 --- a/cpu/drc/emit_arm.c +++ b/cpu/drc/emit_arm.c @@ -28,9 +28,17 @@ #define A_COND_AL 0xe #define A_COND_EQ 0x0 #define A_COND_NE 0x1 +#define A_COND_HS 0x2 +#define A_COND_LO 0x3 #define A_COND_MI 0x4 #define A_COND_PL 0x5 +#define A_COND_VS 0x6 +#define A_COND_VC 0x7 +#define A_COND_HI 0x8 #define A_COND_LS 0x9 +#define A_COND_GE 0xa +#define A_COND_LT 0xb +#define A_COND_GT 0xc #define A_COND_LE 0xd /* unified conditions */ @@ -38,6 +46,18 @@ #define DCOND_NE A_COND_NE #define DCOND_MI A_COND_MI #define DCOND_PL A_COND_PL +#define DCOND_HI A_COND_HI +#define DCOND_HS A_COND_HS +#define DCOND_LO A_COND_LO +#define DCOND_GE A_COND_GE +#define DCOND_GT A_COND_GT +#define DCOND_LT A_COND_LT +#define DCOND_LS A_COND_LS +#define DCOND_LE A_COND_LE +#define DCOND_VS A_COND_VS +#define DCOND_VC A_COND_VC +#define DCOND_CS A_COND_HS +#define DCOND_CC A_COND_LO /* addressing mode 1 */ #define A_AM1_LSL 0 @@ -55,12 +75,15 @@ #define A_OP_SUB 0x2 #define A_OP_RSB 0x3 #define A_OP_ADD 0x4 +#define A_OP_ADC 0x5 +#define A_OP_SBC 0x6 #define A_OP_TST 0x8 #define A_OP_TEQ 0x9 #define A_OP_CMP 0xa #define A_OP_ORR 0xc #define A_OP_MOV 0xd #define A_OP_BIC 0xe +#define A_OP_MVN 0xf #define EOP_C_DOP_X(cond,op,s,rn,rd,shifter_op) \ EMIT(((cond)<<28) | ((op)<< 21) | ((s)<<20) | ((rn)<<16) | ((rd)<<12) | (shifter_op)) @@ -71,6 +94,7 @@ #define EOP_MOV_IMM(rd, ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_MOV,0, 0,rd,ror2,imm8) #define EOP_ORR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ORR,0,rn,rd,ror2,imm8) +#define EOP_EOR_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_EOR,0,rn,rd,ror2,imm8) #define EOP_ADD_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_ADD,0,rn,rd,ror2,imm8) #define EOP_BIC_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_BIC,0,rn,rd,ror2,imm8) #define EOP_AND_IMM(rd,rn,ror2,imm8) EOP_C_DOP_IMM(A_COND_AL,A_OP_AND,0,rn,rd,ror2,imm8) @@ -86,7 +110,12 @@ #define EOP_MOV_REG(cond,s,rd, rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_MOV,s, 0,rd,shift_imm,shift_op,rm) #define EOP_ORR_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ORR,s,rn,rd,shift_imm,shift_op,rm) #define EOP_ADD_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ADD,s,rn,rd,shift_imm,shift_op,rm) +#define EOP_ADC_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_ADC,s,rn,rd,shift_imm,shift_op,rm) #define EOP_SUB_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_SUB,s,rn,rd,shift_imm,shift_op,rm) +#define EOP_SBC_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_SBC,s,rn,rd,shift_imm,shift_op,rm) +#define EOP_AND_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_AND,s,rn,rd,shift_imm,shift_op,rm) +#define EOP_EOR_REG(cond,s,rd,rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_EOR,s,rn,rd,shift_imm,shift_op,rm) +#define EOP_CMP_REG(cond, rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_CMP,1,rn, 0,shift_imm,shift_op,rm) #define EOP_TST_REG(cond, rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_TST,1,rn, 0,shift_imm,shift_op,rm) #define EOP_TEQ_REG(cond, rn,rm,shift_op,shift_imm) EOP_C_DOP_REG_XIMM(cond,A_OP_TEQ,1,rn, 0,shift_imm,shift_op,rm) @@ -167,6 +196,12 @@ #define EOP_C_MUL(cond,s,rd,rs,rm) \ EMIT(((cond)<<28) | ((s)<<20) | ((rd)<<16) | ((rs)<<8) | 0x90 | (rm)) +#define EOP_C_UMULL(cond,s,rdhi,rdlo,rs,rm) \ + EMIT(((cond)<<28) | 0x00800000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm)) + +#define EOP_C_SMULL(cond,s,rdhi,rdlo,rs,rm) \ + EMIT(((cond)<<28) | 0x00c00000 | ((s)<<20) | ((rdhi)<<16) | ((rdlo)<<12) | ((rs)<<8) | 0x90 | (rm)) + #define EOP_MUL(rd,rm,rs) EOP_C_MUL(A_COND_AL,0,rd,rs,rm) // note: rd != rm #define EOP_C_MRS(cond,rd) \ @@ -249,31 +284,57 @@ static int emith_xbranch(int cond, void *target, int is_call) #define EMITH_SJMP_END(cond) \ (void)(cond) -#define EMITH_CONDITIONAL(code, is_nonzero) { \ - u32 val, cond, *ptr; \ - cond = (is_nonzero) ? A_COND_NE : A_COND_EQ; \ - ptr = (void *)tcache_ptr; \ - tcache_ptr = (void *)(ptr + 1); \ - code; \ - val = (u32 *)tcache_ptr - (ptr + 2); \ - EMIT_PTR(ptr, ((cond)<<28) | 0x0a000000 | (val & 0xffffff)); \ -} - #define emith_move_r_r(d, s) \ EOP_MOV_REG_SIMPLE(d, s) +#define emith_or_r_r_r_lsl(d, s1, s2, lslimm) \ + EOP_ORR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm) + +#define emith_eor_r_r_r_lsl(d, s1, s2, lslimm) \ + EOP_EOR_REG(A_COND_AL,0,d,s1,s2,A_AM1_LSL,lslimm) + +#define emith_or_r_r_r(d, s1, s2) \ + emith_or_r_r_r_lsl(d, s1, s2, 0) + +#define emith_eor_r_r_r(d, s1, s2) \ + emith_eor_r_r_r_lsl(d, s1, s2, 0) + #define emith_add_r_r(d, s) \ EOP_ADD_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0) #define emith_sub_r_r(d, s) \ EOP_SUB_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0) +#define emith_and_r_r(d, s) \ + EOP_AND_REG(A_COND_AL,0,d,d,s,A_AM1_LSL,0) + +#define emith_or_r_r(d, s) \ + emith_or_r_r_r(d, d, s) + +#define emith_eor_r_r(d, s) \ + emith_eor_r_r_r(d, d, s) + +#define emith_tst_r_r(d, s) \ + EOP_TST_REG(A_COND_AL,d,s,A_AM1_LSL,0) + #define emith_teq_r_r(d, s) \ EOP_TEQ_REG(A_COND_AL,d,s,A_AM1_LSL,0) +#define emith_cmp_r_r(d, s) \ + EOP_CMP_REG(A_COND_AL,d,s,A_AM1_LSL,0) + +#define emith_addf_r_r(d, s) \ + EOP_ADD_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0) + #define emith_subf_r_r(d, s) \ EOP_SUB_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0) +#define emith_adcf_r_r(d, s) \ + EOP_ADC_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0) + +#define emith_sbcf_r_r(d, s) \ + EOP_SBC_REG(A_COND_AL,1,d,d,s,A_AM1_LSL,0) + #define emith_move_r_imm(r, imm) \ emith_op_imm(A_COND_AL, 0, A_OP_MOV, r, imm) @@ -305,12 +366,18 @@ static int emith_xbranch(int cond, void *target, int is_call) #define emith_or_r_imm_c(cond, r, imm) \ emith_op_imm(cond, 0, A_OP_ORR, r, imm) +#define emith_bic_r_imm_c(cond, r, imm) \ + emith_op_imm(cond, 0, A_OP_BIC, r, imm) + #define emith_lsl(d, s, cnt) \ EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSL,cnt) #define emith_lsr(d, s, cnt) \ EOP_MOV_REG(A_COND_AL,0,d,s,A_AM1_LSR,cnt) +#define emith_lslf(d, s, cnt) \ + EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_LSL,cnt) + #define emith_asrf(d, s, cnt) \ EOP_MOV_REG(A_COND_AL,1,d,s,A_AM1_ASR,cnt) @@ -321,6 +388,13 @@ static int emith_xbranch(int cond, void *target, int is_call) EOP_MUL(d, s2, s1); \ } +#define emith_mul_u64(dlo, dhi, s1, s2) \ + EOP_C_UMULL(A_COND_AL,0,dhi,dlo,s1,s2) + +#define emith_mul_s64(dlo, dhi, s1, s2) \ + EOP_C_SMULL(A_COND_AL,0,dhi,dlo,s1,s2) + +// misc #define emith_ctx_read(r, offs) \ EOP_LDR_IMM(r, CONTEXT_REG, offs) @@ -348,6 +422,18 @@ static int emith_xbranch(int cond, void *target, int is_call) EOP_MOV_REG_ASR(d,d,32 - (bits)); \ } +// put bit0 of r0 to carry +#define emith_set_carry(r0) \ + EOP_TST_REG(A_COND_AL,r0,r0,A_AM1_LSR,1) /* shift out to carry */ \ + +// put bit0 of r0 to carry (for subtraction, inverted on ARM) +#define emith_set_carry_sub(r0) { \ + int t = rcache_get_tmp(); \ + EOP_EOR_IMM(t,r0,0,1); /* invert */ \ + EOP_MOV_REG(A_COND_AL,1,t,t,A_AM1_LSR,1); /* shift out to carry */ \ + rcache_free_tmp(t); \ +} + #define host_arg2reg(rd, arg) \ rd = arg diff --git a/cpu/drc/emit_x86.c b/cpu/drc/emit_x86.c index e94e685..6183d4e 100644 --- a/cpu/drc/emit_x86.c +++ b/cpu/drc/emit_x86.c @@ -1,22 +1,46 @@ +/* + * note about silly things like emith_eor_r_r_r_lsl: + * these are here because the compiler was designed + * for ARM as it's primary target. + */ #include enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; #define CONTEXT_REG xBP +#define IOP_JO 0x70 +#define IOP_JNO 0x71 +#define IOP_JB 0x72 +#define IOP_JAE 0x73 #define IOP_JE 0x74 #define IOP_JNE 0x75 #define IOP_JBE 0x76 #define IOP_JA 0x77 #define IOP_JS 0x78 #define IOP_JNS 0x79 +#define IOP_JL 0x7c +#define IOP_JGE 0x7d #define IOP_JLE 0x7e +#define IOP_JG 0x7f // unified conditions (we just use rel8 jump instructions for x86) #define DCOND_EQ IOP_JE #define DCOND_NE IOP_JNE #define DCOND_MI IOP_JS // MInus #define DCOND_PL IOP_JNS // PLus or zero +#define DCOND_HI IOP_JA // higher (unsigned) +#define DCOND_HS IOP_JAE // higher || same (unsigned) +#define DCOND_LO IOP_JB // lower (unsigned) +#define DCOND_LS IOP_JBE // lower || same (unsigned) +#define DCOND_GE IOP_JGE // greater || equal (signed) +#define DCOND_GT IOP_JG // greater (signed) +#define DCOND_LE IOP_JLE // less || equal (signed) +#define DCOND_LT IOP_JL // less (signed) +#define DCOND_VS IOP_JO // oVerflow Set +#define DCOND_VC IOP_JNO // oVerflow Clear +#define DCOND_CS IOP_JB // Carry Set +#define DCOND_CC IOP_JAE // Carry Clear #define EMIT_PTR(ptr, val, type) \ *(type *)(ptr) = val @@ -47,6 +71,7 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; EMIT_PTR(ptr, op, u8); \ EMIT_PTR(ptr + 1, (tcache_ptr - (ptr+2)), u8) +// _r_r #define emith_move_r_r(dst, src) \ EMIT_OP_MODRM(0x8b, 3, dst, src) @@ -56,11 +81,26 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; #define emith_sub_r_r(d, s) \ EMIT_OP_MODRM(0x29, 3, s, d) +#define emith_adc_r_r(d, s) \ + EMIT_OP_MODRM(0x11, 3, s, d) + +#define emith_sbc_r_r(d, s) \ + EMIT_OP_MODRM(0x19, 3, s, d) /* SBB */ + #define emith_or_r_r(d, s) \ EMIT_OP_MODRM(0x09, 3, s, d) +#define emith_and_r_r(d, s) \ + EMIT_OP_MODRM(0x21, 3, s, d) + #define emith_eor_r_r(d, s) \ - EMIT_OP_MODRM(0x31, 3, s, d) + EMIT_OP_MODRM(0x31, 3, s, d) /* XOR */ + +#define emith_tst_r_r(d, s) \ + EMIT_OP_MODRM(0x85, 3, s, d) /* TEST */ + +#define emith_cmp_r_r(d, s) \ + EMIT_OP_MODRM(0x39, 3, s, d) // fake teq - test equivalence - get_flags(d ^ s) #define emith_teq_r_r(d, s) { \ @@ -69,6 +109,27 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; emith_pop(d); \ } +// _r_r_r +#define emith_eor_r_r_r(d, s1, s2) { \ + if (d != s1) \ + emith_move_r_r(d, s1); \ + emith_eor_r_r(d, s2); \ +} + +#define emith_or_r_r_r_lsl(d, s1, s2, lslimm) { \ + if (d != s2 && d != s1) { \ + emith_lsl(d, s2, lslimm); \ + emith_or_r_r(d, s1); \ + } else { \ + if (d != s1) \ + emith_move_r_r(d, s1); \ + emith_push(s2); \ + emith_lsl(s2, s2, lslimm); \ + emith_or_r_r(d, s2); \ + emith_pop(s2); \ + } \ +} + // _r_imm #define emith_move_r_imm(r, imm) { \ EMIT_OP(0xb8 + (r)); \ @@ -105,17 +166,22 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; // fake conditionals (using SJMP instead) #define emith_add_r_imm_c(cond, r, imm) { \ (void)(cond); \ - emith_arith_r_imm(0, r, imm); \ + emith_add_r_imm(r, imm); \ } #define emith_or_r_imm_c(cond, r, imm) { \ (void)(cond); \ - emith_arith_r_imm(1, r, imm); \ + emith_or_r_imm(r, imm); \ } #define emith_sub_r_imm_c(cond, r, imm) { \ (void)(cond); \ - emith_arith_r_imm(5, r, imm); \ + emith_sub_r_imm(r, imm); \ +} + +#define emith_bic_r_imm_c(cond, r, imm) { \ + (void)(cond); \ + emith_bic_r_imm(r, imm); \ } // shift @@ -126,12 +192,15 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; EMIT(cnt, u8); \ } -#define emith_asr(d, s, cnt) \ - emith_shift(7, d, s, cnt) - #define emith_lsl(d, s, cnt) \ emith_shift(4, d, s, cnt) +#define emith_lsr(d, s, cnt) \ + emith_shift(5, d, s, cnt) + +#define emith_asr(d, s, cnt) \ + emith_shift(7, d, s, cnt) + // misc #define emith_push(r) \ EMIT_OP(0x50 + (r)) @@ -155,11 +224,24 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; emith_asr(d, d, 32 - (bits)); \ } +// put bit0 of r0 to carry +#define emith_set_carry(r0) { \ + emith_tst_r_imm(r0, 1); /* clears C */ \ + EMITH_SJMP_START(DCOND_EQ); \ + EMIT_OP(0xf9); /* STC */ \ + EMITH_SJMP_END(DCOND_EQ); \ +} + +// put bit0 of r0 to carry (for subtraction) +#define emith_set_carry_sub emith_set_carry + // XXX: stupid mess -#define emith_mul(d, s1, s2) { \ +#define emith_mul_(op, dlo, dhi, s1, s2) { \ int rmr; \ - if (d != xAX) \ + if (dlo != xAX && dhi != xAX) \ emith_push(xAX); \ + if (dlo != xDX && dhi != xDX) \ + emith_push(xDX); \ if ((s1) == xAX) \ rmr = s2; \ else if ((s2) == xAX) \ @@ -168,18 +250,39 @@ enum { xAX = 0, xCX, xDX, xBX, xSP, xBP, xSI, xDI }; emith_move_r_r(xAX, s1); \ rmr = s2; \ } \ - emith_push(xDX); \ - EMIT_OP_MODRM(0xf7, 3, 4, rmr); /* MUL rmr */ \ - emith_pop(xDX); \ - if (d != xAX) { \ - emith_move_r_r(d, xAX); \ + EMIT_OP_MODRM(0xf7, 3, op, rmr); /* xMUL rmr */ \ + /* XXX: using push/pop for the case of edx->eax; eax->edx */ \ + if (dhi != xDX && dhi != -1) \ + emith_push(xDX); \ + if (dlo != xAX) \ + emith_move_r_r(dlo, xAX); \ + if (dhi != xDX && dhi != -1) \ + emith_pop(dhi); \ + if (dlo != xDX && dhi != xDX) \ + emith_pop(xDX); \ + if (dlo != xAX && dhi != xAX) \ emith_pop(xAX); \ - } \ } +#define emith_mul_u64(dlo, dhi, s1, s2) \ + emith_mul_(4, dlo, dhi, s1, s2) /* MUL */ + +#define emith_mul_s64(dlo, dhi, s1, s2) \ + emith_mul_(5, dlo, dhi, s1, s2) /* IMUL */ + +#define emith_mul(d, s1, s2) \ + emith_mul_(4, d, -1, s1, s2) + // "flag" instructions are the same #define emith_subf_r_imm emith_sub_r_imm +#define emith_addf_r_r emith_add_r_r #define emith_subf_r_r emith_sub_r_r +#define emith_adcf_r_r emith_adc_r_r +#define emith_sbcf_r_r emith_sbc_r_r + +#define emith_lslf emith_lsl +#define emith_lsrf emith_lsr +#define emith_asrf emith_asr // XXX: offs is 8bit only #define emith_ctx_read(r, offs) { \ diff --git a/cpu/sh2/compiler.c b/cpu/sh2/compiler.c index 43ca7b7..e2883bf 100644 --- a/cpu/sh2/compiler.c +++ b/cpu/sh2/compiler.c @@ -75,7 +75,7 @@ typedef struct { } temp_reg_t; // note: reg_temp[] must have at least the amount of -// registers used by handlers in worst case (currently 3?) +// registers used by handlers in worst case (currently 4) #ifdef ARM #include "../drc/emit_arm.c" @@ -109,9 +109,10 @@ static const int reg_map_g2h[] = { -1, -1, -1, -1, }; -// ax, cx, dx are usually temporaries +// ax, cx, dx are usually temporaries by convention static temp_reg_t reg_temp[] = { { xAX, }, + { xBX, }, { xCX, }, { xDX, }, }; @@ -509,56 +510,32 @@ MOV.L @(disp,GBR),R0 11000110dddddddd MOVA @(disp,PC),R0 11000111dddddddd SWAP.B Rm,Rn 0110nnnnmmmm1000 SWAP.W Rm,Rn 0110nnnnmmmm1001 -XTRCT Rm,Rn 0010nnnnmmmm1101 -ADD Rm,Rn 0011nnnnmmmm1100 ADD #imm,Rn 0111nnnniiiiiiii -ADDC Rm,Rn 0011nnnnmmmm1110 -ADDV Rm,Rn 0011nnnnmmmm1111 CMP/EQ #imm,R0 10001000iiiiiiii -CMP/EQ Rm,Rn 0011nnnnmmmm0000 -CMP/HS Rm,Rn 0011nnnnmmmm0010 -CMP/GE Rm,Rn 0011nnnnmmmm0011 -CMP/HI Rm,Rn 0011nnnnmmmm0110 -CMP/GT Rm,Rn 0011nnnnmmmm0111 CMP/PZ Rn 0100nnnn00010001 CMP/PL Rn 0100nnnn00010101 -CMP/ST Rm,Rn 0010nnnnmmmm1100 -DIV1 Rm,Rn 0011nnnnmmmm0100 -DMULS. Rm,Rn 0011nnnnmmmm1101 -DMULU.L Rm,Rn 0011nnnnmmmm0101 EXTS.B Rm,Rn 0110nnnnmmmm1110 EXTS.W Rm,Rn 0110nnnnmmmm1111 EXTU.B Rm,Rn 0110nnnnmmmm1100 EXTU.W Rm,Rn 0110nnnnmmmm1101 MAC @Rm+,@Rn+ 0100nnnnmmmm1111 -MULS.W Rm,Rn 0010nnnnmmmm1111 -MULU.W Rm,Rn 0010nnnnmmmm1110 NEG Rm,Rn 0110nnnnmmmm1011 NEGC Rm,Rn 0110nnnnmmmm1010 -SUB Rm,Rn 0011nnnnmmmm1000 -SUBC Rm,Rn 0011nnnnmmmm1010 -SUBV Rm,Rn 0011nnnnmmmm1011 -AND Rm,Rn 0010nnnnmmmm1001 AND #imm,R0 11001001iiiiiiii AND.B #imm,@(R0,GBR) 11001101iiiiiiii NOT Rm,Rn 0110nnnnmmmm0111 -OR Rm,Rn 0010nnnnmmmm1011 OR #imm,R0 11001011iiiiiiii OR.B #imm,@(R0,GBR) 11001111iiiiiiii TAS.B @Rn 0100nnnn00011011 -TST Rm,Rn 0010nnnnmmmm1000 TST #imm,R0 11001000iiiiiiii TST.B #imm,@(R0,GBR) 11001100iiiiiiii -XOR Rm,Rn 0010nnnnmmmm1010 XOR #imm,R0 11001010iiiiiiii XOR.B #imm,@(R0,GBR) 11001110iiiiiiii ROTL Rn 0100nnnn00000100 ROTR Rn 0100nnnn00000101 ROTCL Rn 0100nnnn00100100 ROTCR Rn 0100nnnn00100101 -SHAL Rn 0100nnnn00100000 SHAR Rn 0100nnnn00100001 -SHLL Rn 0100nnnn00000000 SHLR Rn 0100nnnn00000001 SHLL2 Rn 0100nnnn00001000 SHLR2 Rn 0100nnnn00001001 @@ -613,7 +590,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) int op, delayed_op = 0, test_irq = 0; int tcache_id = 0, blkid = 0; int cycles = 0; - u32 tmp, tmp2, tmp3; + u32 tmp, tmp2, tmp3, tmp4; // validate PC tmp = sh2->pc >> 29; @@ -673,6 +650,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) switch ((op >> 12) & 0x0f) { + ///////////////////////////////////////////// case 0x00: switch (op & 0x0f) { @@ -838,6 +816,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) } goto default_; + ///////////////////////////////////////////// case 0x01: // MOV.L Rm,@(disp,Rn) 0001nnnnmmmmdddd rcache_clean(); @@ -886,27 +865,225 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) emith_or_r_imm_c(DCOND_MI, tmp, T); EMITH_SJMP_END(DCOND_PL); goto end_op; + case 0x08: // TST Rm,Rn 0010nnnnmmmm1000 + tmp = rcache_get_reg(SHR_SR, RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp3 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_bic_r_imm(tmp, T); + emith_tst_r_r(tmp2, tmp3); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp, T); + EMITH_SJMP_END(DCOND_NE); + goto end_op; + case 0x09: // AND Rm,Rn 0010nnnnmmmm1001 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_and_r_r(tmp, tmp2); + goto end_op; + case 0x0a: // XOR Rm,Rn 0010nnnnmmmm1010 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_eor_r_r(tmp, tmp2); + goto end_op; + case 0x0b: // OR Rm,Rn 0010nnnnmmmm1011 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_or_r_r(tmp, tmp2); + goto end_op; + case 0x0c: // CMP/STR Rm,Rn 0010nnnnmmmm1100 + tmp = rcache_get_tmp(); + tmp2 = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp3 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_eor_r_r_r(tmp, tmp2, tmp3); + tmp2 = rcache_get_reg(SHR_SR, RC_GR_RMW); + emith_bic_r_imm(tmp2, T); + emith_tst_r_imm(tmp, 0x000000ff); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp2, T); + EMITH_SJMP_END(DCOND_NE); + emith_tst_r_imm(tmp, 0x0000ff00); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp2, T); + EMITH_SJMP_END(DCOND_NE); + emith_tst_r_imm(tmp, 0x00ff0000); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp2, T); + EMITH_SJMP_END(DCOND_NE); + emith_tst_r_imm(tmp, 0xff000000); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp2, T); + EMITH_SJMP_END(DCOND_NE); + rcache_free_tmp(tmp); + goto end_op; + case 0x0d: // XTRCT Rm,Rn 0010nnnnmmmm1101 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_lsr(tmp, tmp, 16); + emith_or_r_r_r_lsl(tmp, tmp, tmp2, 16); + goto end_op; + case 0x0e: // MULU.W Rm,Rn 0010nnnnmmmm1110 + case 0x0f: // MULS.W Rm,Rn 0010nnnnmmmm1111 + tmp2 = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp = rcache_get_reg(SHR_MACL, RC_GR_WRITE); + if (op & 1) { + emith_sext(tmp, tmp2, 16); + } else + emith_clear_msb(tmp, tmp2, 16); + tmp3 = rcache_get_reg(GET_Rm(), RC_GR_READ); + tmp2 = rcache_get_tmp(); + if (op & 1) { + emith_sext(tmp2, tmp3, 16); + } else + emith_clear_msb(tmp2, tmp3, 16); + emith_mul(tmp, tmp, tmp2); + rcache_free_tmp(tmp2); +// FIXME: causes timing issues in Doom? +// cycles++; + goto end_op; } goto default_; + ///////////////////////////////////////////// + case 0x03: + switch (op & 0x0f) + { + case 0x00: // CMP/EQ Rm,Rn 0011nnnnmmmm0000 + case 0x02: // CMP/HS Rm,Rn 0011nnnnmmmm0010 + case 0x03: // CMP/GE Rm,Rn 0011nnnnmmmm0011 + case 0x06: // CMP/HI Rm,Rn 0011nnnnmmmm0110 + case 0x07: // CMP/GT Rm,Rn 0011nnnnmmmm0111 + tmp = rcache_get_reg(SHR_SR, RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp3 = rcache_get_reg(GET_Rm(), RC_GR_READ); + emith_bic_r_imm(tmp, T); + emith_cmp_r_r(tmp2, tmp3); + switch (op & 0x07) + { + case 0x00: // CMP/EQ + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp, T); + EMITH_SJMP_END(DCOND_NE); + break; + case 0x02: // CMP/HS + EMITH_SJMP_START(DCOND_LO); + emith_or_r_imm_c(DCOND_HS, tmp, T); + EMITH_SJMP_END(DCOND_LO); + break; + case 0x03: // CMP/GE + EMITH_SJMP_START(DCOND_LT); + emith_or_r_imm_c(DCOND_GE, tmp, T); + EMITH_SJMP_END(DCOND_LT); + break; + case 0x06: // CMP/HI + EMITH_SJMP_START(DCOND_LS); + emith_or_r_imm_c(DCOND_HI, tmp, T); + EMITH_SJMP_END(DCOND_LS); + break; + case 0x07: // CMP/GT + EMITH_SJMP_START(DCOND_LE); + emith_or_r_imm_c(DCOND_GT, tmp, T); + EMITH_SJMP_END(DCOND_LE); + break; + } + goto end_op; + case 0x04: // DIV1 Rm,Rn 0011nnnnmmmm0100 + // TODO + break; + case 0x05: // DMULU.L Rm,Rn 0011nnnnmmmm0101 + tmp = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + tmp3 = rcache_get_reg(SHR_MACL, RC_GR_WRITE); + tmp4 = rcache_get_reg(SHR_MACH, RC_GR_WRITE); + emith_mul_u64(tmp3, tmp4, tmp, tmp2); + goto end_op; + case 0x08: // SUB Rm,Rn 0011nnnnmmmm1000 + case 0x0c: // ADD Rm,Rn 0011nnnnmmmm1100 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + if (op & 4) { + emith_add_r_r(tmp, tmp2); + } else + emith_sub_r_r(tmp, tmp2); + goto end_op; + case 0x0a: // SUBC Rm,Rn 0011nnnnmmmm1010 + case 0x0e: // ADDC Rm,Rn 0011nnnnmmmm1110 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + tmp3 = rcache_get_reg(SHR_SR, RC_GR_RMW); + if (op & 4) { // adc + emith_set_carry(tmp3); + emith_adcf_r_r(tmp, tmp2); + tmp = DCOND_CS; // set condition + tmp2 = DCOND_CC; // clear condition + } else { + emith_set_carry_sub(tmp3); + emith_sbcf_r_r(tmp, tmp2); + tmp = DCOND_LO; // using LO/HS instead of CS/CC + tmp2 = DCOND_HS; // due to ARM target.. + } + EMITH_SJMP_START(tmp); + emith_bic_r_imm_c(tmp2, tmp3, T); + EMITH_SJMP_END(tmp); + EMITH_SJMP_START(tmp2); + emith_or_r_imm_c(tmp, tmp3, T); + EMITH_SJMP_END(tmp2); + goto end_op; + case 0x0b: // SUBV Rm,Rn 0011nnnnmmmm1011 + case 0x0f: // ADDV Rm,Rn 0011nnnnmmmm1111 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + tmp3 = rcache_get_reg(SHR_SR, RC_GR_RMW); + emith_bic_r_imm(tmp3, T); + if (op & 4) { + emith_addf_r_r(tmp, tmp2); + } else + emith_subf_r_r(tmp, tmp2); + EMITH_SJMP_START(DCOND_VC); + emith_or_r_imm_c(DCOND_VS, tmp3, T); + EMITH_SJMP_END(DCOND_VC); + goto end_op; + case 0x0d: // DMULS.L Rm,Rn 0011nnnnmmmm1101 + tmp = rcache_get_reg(GET_Rn(), RC_GR_READ); + tmp2 = rcache_get_reg(GET_Rm(), RC_GR_READ); + tmp3 = rcache_get_reg(SHR_MACL, RC_GR_WRITE); + tmp4 = rcache_get_reg(SHR_MACH, RC_GR_WRITE); + emith_mul_s64(tmp3, tmp4, tmp, tmp2); + goto end_op; + } + goto default_; + + ///////////////////////////////////////////// case 0x04: - switch (op & 0x0f) { + switch (op & 0x0f) + { case 0x00: - if ((op & 0xf0) != 0x10) - goto default_; - // DT Rn 0100nnnn00010000 - if (p32x_sh2_read16(pc, sh2) == 0x8bfd) { // BF #-2 - emith_sh2_dtbf_loop(); + switch (GET_Fx()) + { + case 0: // SHLL Rn 0100nnnn00000000 + case 2: // SHAL Rn 0100nnnn00100000 + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(SHR_SR, RC_GR_RMW); + emith_bic_r_imm(tmp2, T); + emith_lslf(tmp, tmp, 1); + EMITH_SJMP_START(DCOND_CC); + emith_or_r_imm_c(DCOND_CS, tmp2, T); + EMITH_SJMP_END(DCOND_CC); + goto end_op; + case 1: // DT Rn 0100nnnn00010000 + if (p32x_sh2_read16(pc, sh2) == 0x8bfd) { // BF #-2 + emith_sh2_dtbf_loop(); + goto end_op; + } + tmp = rcache_get_reg(GET_Rn(), RC_GR_RMW); + tmp2 = rcache_get_reg(SHR_SR, RC_GR_RMW); + emith_bic_r_imm(tmp2, T); + emith_subf_r_imm(tmp, 1); + EMITH_SJMP_START(DCOND_NE); + emith_or_r_imm_c(DCOND_EQ, tmp2, T); + EMITH_SJMP_END(DCOND_NE); goto end_op; } - tmp = rcache_get_reg((op >> 8) & 0x0f, RC_GR_RMW); - tmp2 = rcache_get_reg(SHR_SR, RC_GR_RMW); - emith_bic_r_imm(tmp2, T); - emith_subf_r_imm(tmp, 1); - EMITH_SJMP_START(DCOND_NE); - emith_or_r_imm_c(DCOND_EQ, tmp2, T); - EMITH_SJMP_END(DCOND_NE); - goto end_op; + goto default_; case 0x07: if ((op & 0xf0) != 0) goto default_; @@ -933,6 +1110,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) } goto default_; + ///////////////////////////////////////////// case 0x08: switch (op & 0x0f00) { // BT/S label 10001101dddddddd @@ -968,6 +1146,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) }} goto default_; + ///////////////////////////////////////////// case 0x0a: // BRA label 1010dddddddddddd DELAYED_OP; @@ -977,6 +1156,7 @@ static void *sh2_translate(SH2 *sh2, block_desc *other_block) cycles++; break; + ///////////////////////////////////////////// case 0x0b: // BSR label 1011dddddddddddd DELAYED_OP; diff --git a/cpu/sh2/stub_x86.asm b/cpu/sh2/stub_x86.asm index 256d64a..e53e947 100644 --- a/cpu/sh2/stub_x86.asm +++ b/cpu/sh2/stub_x86.asm @@ -3,14 +3,16 @@ section .text global sh2_drc_entry ; SH2 *sh2, void *block sh2_drc_entry: + push ebx push ebp - mov ebp, [esp+4+4] ; context - mov eax, [esp+4+8] + mov ebp, [esp+8+4] ; context + mov eax, [esp+8+8] jmp eax global sh2_drc_exit sh2_drc_exit: pop ebp + pop ebx ret diff --git a/platform/linux/Makefile b/platform/linux/Makefile index 8d37fd4..3638286 100644 --- a/platform/linux/Makefile +++ b/platform/linux/Makefile @@ -141,6 +141,7 @@ include ../common/revision.mak pico/carthw/svp/compiler.o : ../../cpu/drc/emit_arm.c cpu/sh2/compiler.o : ../../cpu/drc/emit_x86.c +cpu/sh2/mame/sh2pico.o : ../../cpu/sh2/mame/sh2.c pico/pico.o pico/cd/pico.o : ../../pico/pico_cmn.c ../../pico/pico_int.h pico/memory.o pico/cd/memory.o : ../../pico/pico_int.h ../../pico/memory.h