[pcsx_rearmed.git] / libpcsxcore / psxdma.c

/***************************************************************************
 *   Copyright (C) 2007 Ryan Schultz, PCSX-df Team, PCSX team              *
 *                                                                         *
 *   This program is free software; you can redistribute it and/or modify  *
 *   it under the terms of the GNU General Public License as published by  *
 *   the Free Software Foundation; either version 2 of the License, or     *
 *   (at your option) any later version.                                   *
 *                                                                         *
 *   This program is distributed in the hope that it will be useful,       *
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of        *
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         *
 *   GNU General Public License for more details.                          *
 *                                                                         *
 *   You should have received a copy of the GNU General Public License     *
 *   along with this program; if not, write to the                         *
 *   Free Software Foundation, Inc.,                                       *
 *   51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA.           *
 ***************************************************************************/

/*
* Handles PSX DMA functions.
*/

#include "psxdma.h"
#include "gpu.h"

#ifndef min
#define min(a, b) ((b) < (a) ? (b) : (a))
#endif
#ifndef PSXDMA_LOG
#define PSXDMA_LOG(...)
#endif

// Dma0/1 in Mdec.c
// Dma3   in CdRom.c

void spuInterrupt() {
	if (HW_DMA4_CHCR & SWAP32(0x01000000))
	{
		HW_DMA4_CHCR &= SWAP32(~0x01000000);
		DMA_INTERRUPT(4);
	}
}

void psxDma4(u32 madr, u32 bcr, u32 chcr) { // SPU
	u32 words, words_max = 0, words_copy;
	u16 *ptr;

	madr &= ~3;
	ptr = getDmaRam(madr, &words_max);
	if (ptr == INVALID_PTR)
		log_unhandled("bad dma4 madr %x\n", madr);

	words = words_copy = (bcr >> 16) * (bcr & 0xffff);
	if (words_copy > words_max) {
		log_unhandled("bad dma4 madr %x bcr %x\n", madr, bcr);
		words_copy = words_max;
	}

	switch (chcr) {
		case 0x01000201: //cpu to spu transfer
			PSXDMA_LOG("*** DMA4 SPU - mem2spu *** %x addr = %x size = %x\n", chcr, madr, bcr);
			if (ptr == INVALID_PTR)
				break;
			SPU_writeDMAMem(ptr, words_copy * 2, psxRegs.cycle);
			HW_DMA4_MADR = SWAPu32(madr + words_copy * 2);
			// This should be much slower, like 12+ cycles/byte, it's like
			// that because the CPU runs too fast and fifo is not emulated.
			// See also set_dma_end().
			set_event(PSXINT_SPUDMA, words * 4 * 4);
			return;

		case 0x01000200: //spu to cpu transfer
			PSXDMA_LOG("*** DMA4 SPU - spu2mem *** %x addr = %x size = %x\n", chcr, madr, bcr);
			if (ptr == INVALID_PTR)
				break;
			SPU_readDMAMem(ptr, words_copy * 2, psxRegs.cycle);
			psxCpu->Clear(madr, words_copy);

			HW_DMA4_MADR = SWAPu32(madr + words_copy * 4);
			set_event(PSXINT_SPUDMA, words * 4 * 4);
			return;

		default:
			log_unhandled("*** DMA4 SPU - unknown *** %x addr = %x size = %x\n", chcr, madr, bcr);
			break;
	}

	HW_DMA4_CHCR &= SWAP32(~0x01000000);
	DMA_INTERRUPT(4);
}

#if 0
// Taken from PEOPS SOFTGPU
static inline boolean CheckForEndlessLoop(u32 laddr, u32 *lUsedAddr) {
	if (laddr == lUsedAddr[1]) return TRUE;
	if (laddr == lUsedAddr[2]) return TRUE;

	if (laddr < lUsedAddr[0]) lUsedAddr[1] = laddr;
	else lUsedAddr[2] = laddr;

	lUsedAddr[0] = laddr;

	return FALSE;
}

static u32 gpuDmaChainSize(u32 addr) {
	u32 size;
	u32 DMACommandCounter = 0;
	u32 lUsedAddr[3];

	lUsedAddr[0] = lUsedAddr[1] = lUsedAddr[2] = 0xffffff;

	// initial linked list ptr (word)
	size = 1;

	do {
		addr &= 0x1ffffc;

		if (DMACommandCounter++ > 2000000) break;
		if (CheckForEndlessLoop(addr, lUsedAddr)) break;

		// # 32-bit blocks to transfer
		size += psxMu8( addr + 3 );

		// next 32-bit pointer
		addr = psxMu32( addr & ~0x3 ) & 0xffffff;
		size += 1;
	} while (!(addr & 0x800000)); // contrary to some documentation, the end-of-linked-list marker is not actually 0xFF'FFFF
                                  // any pointer with bit 23 set will do.

	return size;
}
#endif

void psxDma2(u32 madr, u32 bcr, u32 chcr) { // GPU
	u32 *ptr, madr_next, *madr_next_p;
	u32 words, words_left, words_max, words_copy;
	int cycles_sum, cycles_last_cmd = 0, do_walking;

	madr &= ~3;
	switch (chcr) {
		case 0x01000200: // vram2mem
			PSXDMA_LOG("*** DMA2 GPU - vram2mem *** %lx addr = %lx size = %lx\n", chcr, madr, bcr);
			ptr = getDmaRam(madr, &words_max);
			if (ptr == INVALID_PTR) {
				log_unhandled("bad dma2 madr %x\n", madr);
				break;
			}
			// BA blocks * BS words (word = 32-bits)
			words = words_copy = (bcr >> 16) * (bcr & 0xffff);
			if (words > words_max) {
				log_unhandled("bad dma2 madr %x bcr %x\n", madr, bcr);
				words_copy = words_max;
			}
			GPU_readDataMem(ptr, words_copy);
			psxCpu->Clear(madr, words_copy);

			HW_DMA2_MADR = SWAPu32(madr + words_copy * 4);

			// careful: gpu_state_change() also messes with this
			psxRegs.gpuIdleAfter = psxRegs.cycle + words / 4 + 16;
			// already 32-bit word size ((size * 4) / 4)
			set_event(PSXINT_GPUDMA, words / 4);
			return;

		case 0x01000201: // mem2vram
			PSXDMA_LOG("*** DMA 2 - GPU mem2vram *** %lx addr = %lx size = %lx\n", chcr, madr, bcr);
			words = words_left = (bcr >> 16) * (bcr & 0xffff);
			while (words_left > 0) {
				ptr = getDmaRam(madr, &words_max);
				if (ptr == INVALID_PTR) {
					log_unhandled("bad2 dma madr %x\n", madr);
					break;
				}
				words_copy = min(words_left, words_max);
				GPU_writeDataMem(ptr, words_copy);
				words_left -= words_copy;
				madr += words_copy * 4;
			}

			HW_DMA2_MADR = SWAPu32(madr);

			// careful: gpu_state_change() also messes with this
			psxRegs.gpuIdleAfter = psxRegs.cycle + words / 4 + 16;
			// already 32-bit word size ((size * 4) / 4)
			set_event(PSXINT_GPUDMA, words / 4);
			return;

		case 0x01000401: // dma chain
			PSXDMA_LOG("*** DMA 2 - GPU dma chain *** %lx addr = %lx size = %lx\n", chcr, madr, bcr);
			// when not emulating walking progress, end immediately
			madr_next = 0xffffff;

			do_walking = Config.GpuListWalking;
			if (do_walking < 0 || Config.hacks.gpu_timing1024)
				do_walking = Config.hacks.gpu_slow_list_walking;
			madr_next_p = do_walking ? &madr_next : NULL;

			cycles_sum = GPU_dmaChain((u32 *)psxM, madr & 0x1fffff,
					madr_next_p, &cycles_last_cmd);

			HW_DMA2_MADR = SWAPu32(madr_next);

			// a hack for Judge Dredd which is annoyingly sensitive to timing
			if (Config.hacks.gpu_timing1024)
				cycles_sum = 1024;

			psxRegs.gpuIdleAfter = psxRegs.cycle + cycles_sum + cycles_last_cmd;
			set_event(PSXINT_GPUDMA, cycles_sum);
			//printf("%u dma2cf: %6d,%4d %08x %08x %08x %08x\n", psxRegs.cycle,
			//  cycles_sum, cycles_last_cmd, madr, bcr, chcr, HW_DMA2_MADR);
			return;

		default:
			log_unhandled("*** DMA 2 - GPU unknown *** %x addr = %x size = %x\n", chcr, madr, bcr);
			break;
	}

	HW_DMA2_CHCR &= SWAP32(~0x01000000);
	DMA_INTERRUPT(2);
}

void gpuInterrupt() {
	if (HW_DMA2_CHCR == SWAP32(0x01000401) && !(HW_DMA2_MADR & SWAP32(0x800000)))
	{
		u32 madr_next = 0xffffff, madr = SWAPu32(HW_DMA2_MADR);
		int cycles_sum, cycles_last_cmd = 0;
		cycles_sum = GPU_dmaChain((u32 *)psxM, madr & 0x1fffff,
				&madr_next, &cycles_last_cmd);
		HW_DMA2_MADR = SWAPu32(madr_next);
		if ((s32)(psxRegs.gpuIdleAfter - psxRegs.cycle) > 0)
			cycles_sum += psxRegs.gpuIdleAfter - psxRegs.cycle;
		psxRegs.gpuIdleAfter = psxRegs.cycle + cycles_sum + cycles_last_cmd;
		set_event(PSXINT_GPUDMA, cycles_sum);
		//printf("%u dma2cn: %6d,%4d %08x\n", psxRegs.cycle, cycles_sum,
		//	cycles_last_cmd, HW_DMA2_MADR);
		return;
	}
	if (HW_DMA2_CHCR & SWAP32(0x01000000))
	{
		HW_DMA2_CHCR &= SWAP32(~0x01000000);
		DMA_INTERRUPT(2);
	}
}

void psxAbortDma2() {
	psxRegs.gpuIdleAfter = psxRegs.cycle + 32;
}

void psxDma6(u32 madr, u32 bcr, u32 chcr) {
	u32 words, words_max;
	u32 *mem;

	PSXDMA_LOG("*** DMA6 OT *** %x addr = %x size = %x\n", chcr, madr, bcr);

	if (chcr == 0x11000002) {
		madr &= ~3;
		mem = getDmaRam(madr, &words_max);
		if (mem == INVALID_PTR) {
			log_unhandled("bad6 dma madr %x\n", madr);
			HW_DMA6_CHCR &= SWAP32(~0x11000000);
			DMA_INTERRUPT(6);
			return;
		}

		// already 32-bit size
		words = bcr;

		while (bcr-- && mem > (u32 *)psxM) {
			*mem-- = SWAP32((madr - 4) & 0xffffff);
			madr -= 4;
		}
		*++mem = SWAP32(0xffffff);

		// halted
		psxRegs.cycle += words;
		set_event(PSXINT_GPUOTCDMA, 16);
		return;
	}
	else {
		// Unknown option
		log_unhandled("*** DMA6 OT - unknown *** %x addr = %x size = %x\n", chcr, madr, bcr);
	}

	HW_DMA6_CHCR &= SWAP32(~0x11000000);
	DMA_INTERRUPT(6);
}

void gpuotcInterrupt()
{
	if (HW_DMA6_CHCR & SWAP32(0x01000000))
	{
		HW_DMA6_CHCR &= SWAP32(~0x11000000);
		DMA_INTERRUPT(6);
	}
}
Commit	Line	Data
	1	/***************************************************************************
	2	* Copyright (C) 2007 Ryan Schultz, PCSX-df Team, PCSX team *
	3	* *
	4	* This program is free software; you can redistribute it and/or modify *
	5	* it under the terms of the GNU General Public License as published by *
	6	* the Free Software Foundation; either version 2 of the License, or *
	7	* (at your option) any later version. *
	8	* *
	9	* This program is distributed in the hope that it will be useful, *
	10	* but WITHOUT ANY WARRANTY; without even the implied warranty of *
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the *
	12	* GNU General Public License for more details. *
	13	* *
	14	* You should have received a copy of the GNU General Public License *
	15	* along with this program; if not, write to the *
	16	* Free Software Foundation, Inc., *
	17	* 51 Franklin Street, Fifth Floor, Boston, MA 02111-1307 USA. *
	18	***************************************************************************/
	19
	20	/*
	21	* Handles PSX DMA functions.
	22	*/
	23
	24	#include "psxdma.h"
	25	#include "gpu.h"
	26
	27	#ifndef min
	28	#define min(a, b) ((b) < (a) ? (b) : (a))
	29	#endif
	30	#ifndef PSXDMA_LOG
	31	#define PSXDMA_LOG(...)
	32	#endif
	33
	34	// Dma0/1 in Mdec.c
	35	// Dma3 in CdRom.c
	36
	37	void spuInterrupt() {
	38	if (HW_DMA4_CHCR & SWAP32(0x01000000))
	39	{
	40	HW_DMA4_CHCR &= SWAP32(~0x01000000);
	41	DMA_INTERRUPT(4);
	42	}
	43	}
	44
	45	void psxDma4(u32 madr, u32 bcr, u32 chcr) { // SPU
	46	u32 words, words_max = 0, words_copy;
	47	u16 *ptr;
	48
	49	madr &= ~3;
	50	ptr = getDmaRam(madr, &words_max);
	51	if (ptr == INVALID_PTR)
	52	log_unhandled("bad dma4 madr %x\n", madr);
	53
	54	words = words_copy = (bcr >> 16) * (bcr & 0xffff);
	55	if (words_copy > words_max) {
	56	log_unhandled("bad dma4 madr %x bcr %x\n", madr, bcr);
	57	words_copy = words_max;
	58	}
	59
	60	switch (chcr) {
	61	case 0x01000201: //cpu to spu transfer
	62	PSXDMA_LOG("* DMA4 SPU - mem2spu * %x addr = %x size = %x\n", chcr, madr, bcr);
	63	if (ptr == INVALID_PTR)
	64	break;
	65	SPU_writeDMAMem(ptr, words_copy * 2, psxRegs.cycle);
	66	HW_DMA4_MADR = SWAPu32(madr + words_copy * 2);
	67	// This should be much slower, like 12+ cycles/byte, it's like
	68	// that because the CPU runs too fast and fifo is not emulated.
	69	// See also set_dma_end().
	70	set_event(PSXINT_SPUDMA, words * 4 * 4);
	71	return;
	72
	73	case 0x01000200: //spu to cpu transfer
	74	PSXDMA_LOG("* DMA4 SPU - spu2mem * %x addr = %x size = %x\n", chcr, madr, bcr);
	75	if (ptr == INVALID_PTR)
	76	break;
	77	SPU_readDMAMem(ptr, words_copy * 2, psxRegs.cycle);
	78	psxCpu->Clear(madr, words_copy);
	79
	80	HW_DMA4_MADR = SWAPu32(madr + words_copy * 4);
	81	set_event(PSXINT_SPUDMA, words * 4 * 4);
	82	return;
	83
	84	default:
	85	log_unhandled("* DMA4 SPU - unknown * %x addr = %x size = %x\n", chcr, madr, bcr);
	86	break;
	87	}
	88
	89	HW_DMA4_CHCR &= SWAP32(~0x01000000);
	90	DMA_INTERRUPT(4);
	91	}
	92
	93	#if 0
	94	// Taken from PEOPS SOFTGPU
	95	static inline boolean CheckForEndlessLoop(u32 laddr, u32 *lUsedAddr) {
	96	if (laddr == lUsedAddr[1]) return TRUE;
	97	if (laddr == lUsedAddr[2]) return TRUE;
	98
	99	if (laddr < lUsedAddr[0]) lUsedAddr[1] = laddr;
	100	else lUsedAddr[2] = laddr;
	101
	102	lUsedAddr[0] = laddr;
	103
	104	return FALSE;
	105	}
	106
	107	static u32 gpuDmaChainSize(u32 addr) {
	108	u32 size;
	109	u32 DMACommandCounter = 0;
	110	u32 lUsedAddr[3];
	111
	112	lUsedAddr[0] = lUsedAddr[1] = lUsedAddr[2] = 0xffffff;
	113
	114	// initial linked list ptr (word)
	115	size = 1;
	116
	117	do {
	118	addr &= 0x1ffffc;
	119
	120	if (DMACommandCounter++ > 2000000) break;
	121	if (CheckForEndlessLoop(addr, lUsedAddr)) break;
	122
	123	// # 32-bit blocks to transfer
	124	size += psxMu8( addr + 3 );
	125
	126	// next 32-bit pointer
	127	addr = psxMu32( addr & ~0x3 ) & 0xffffff;
	128	size += 1;
	129	} while (!(addr & 0x800000)); // contrary to some documentation, the end-of-linked-list marker is not actually 0xFF'FFFF
	130	// any pointer with bit 23 set will do.
	131
	132	return size;
	133	}
	134	#endif
	135
	136	void psxDma2(u32 madr, u32 bcr, u32 chcr) { // GPU
	137	u32 ptr, madr_next, madr_next_p;
	138	u32 words, words_left, words_max, words_copy;
	139	int cycles_sum, cycles_last_cmd = 0, do_walking;
	140
	141	madr &= ~3;
	142	switch (chcr) {
	143	case 0x01000200: // vram2mem
	144	PSXDMA_LOG("* DMA2 GPU - vram2mem * %lx addr = %lx size = %lx\n", chcr, madr, bcr);
	145	ptr = getDmaRam(madr, &words_max);
	146	if (ptr == INVALID_PTR) {
	147	log_unhandled("bad dma2 madr %x\n", madr);
	148	break;
	149	}
	150	// BA blocks * BS words (word = 32-bits)
	151	words = words_copy = (bcr >> 16) * (bcr & 0xffff);
	152	if (words > words_max) {
	153	log_unhandled("bad dma2 madr %x bcr %x\n", madr, bcr);
	154	words_copy = words_max;
	155	}
	156	GPU_readDataMem(ptr, words_copy);
	157	psxCpu->Clear(madr, words_copy);
	158
	159	HW_DMA2_MADR = SWAPu32(madr + words_copy * 4);
	160
	161	// careful: gpu_state_change() also messes with this
	162	psxRegs.gpuIdleAfter = psxRegs.cycle + words / 4 + 16;
	163	// already 32-bit word size ((size * 4) / 4)
	164	set_event(PSXINT_GPUDMA, words / 4);
	165	return;
	166
	167	case 0x01000201: // mem2vram
	168	PSXDMA_LOG("* DMA 2 - GPU mem2vram * %lx addr = %lx size = %lx\n", chcr, madr, bcr);
	169	words = words_left = (bcr >> 16) * (bcr & 0xffff);
	170	while (words_left > 0) {
	171	ptr = getDmaRam(madr, &words_max);
	172	if (ptr == INVALID_PTR) {
	173	log_unhandled("bad2 dma madr %x\n", madr);
	174	break;
	175	}
	176	words_copy = min(words_left, words_max);
	177	GPU_writeDataMem(ptr, words_copy);
	178	words_left -= words_copy;
	179	madr += words_copy * 4;
	180	}
	181
	182	HW_DMA2_MADR = SWAPu32(madr);
	183
	184	// careful: gpu_state_change() also messes with this
	185	psxRegs.gpuIdleAfter = psxRegs.cycle + words / 4 + 16;
	186	// already 32-bit word size ((size * 4) / 4)
	187	set_event(PSXINT_GPUDMA, words / 4);
	188	return;
	189
	190	case 0x01000401: // dma chain
	191	PSXDMA_LOG("* DMA 2 - GPU dma chain * %lx addr = %lx size = %lx\n", chcr, madr, bcr);
	192	// when not emulating walking progress, end immediately
	193	madr_next = 0xffffff;
	194
	195	do_walking = Config.GpuListWalking;
	196	if (do_walking < 0 \|\| Config.hacks.gpu_timing1024)
	197	do_walking = Config.hacks.gpu_slow_list_walking;
	198	madr_next_p = do_walking ? &madr_next : NULL;
	199
	200	cycles_sum = GPU_dmaChain((u32 *)psxM, madr & 0x1fffff,
	201	madr_next_p, &cycles_last_cmd);
	202
	203	HW_DMA2_MADR = SWAPu32(madr_next);
	204
	205	// a hack for Judge Dredd which is annoyingly sensitive to timing
	206	if (Config.hacks.gpu_timing1024)
	207	cycles_sum = 1024;
	208
	209	psxRegs.gpuIdleAfter = psxRegs.cycle + cycles_sum + cycles_last_cmd;
	210	set_event(PSXINT_GPUDMA, cycles_sum);
	211	//printf("%u dma2cf: %6d,%4d %08x %08x %08x %08x\n", psxRegs.cycle,
	212	// cycles_sum, cycles_last_cmd, madr, bcr, chcr, HW_DMA2_MADR);
	213	return;
	214
	215	default:
	216	log_unhandled("* DMA 2 - GPU unknown * %x addr = %x size = %x\n", chcr, madr, bcr);
	217	break;
	218	}
	219
	220	HW_DMA2_CHCR &= SWAP32(~0x01000000);
	221	DMA_INTERRUPT(2);
	222	}
	223
	224	void gpuInterrupt() {
	225	if (HW_DMA2_CHCR == SWAP32(0x01000401) && !(HW_DMA2_MADR & SWAP32(0x800000)))
	226	{
	227	u32 madr_next = 0xffffff, madr = SWAPu32(HW_DMA2_MADR);
	228	int cycles_sum, cycles_last_cmd = 0;
	229	cycles_sum = GPU_dmaChain((u32 *)psxM, madr & 0x1fffff,
	230	&madr_next, &cycles_last_cmd);
	231	HW_DMA2_MADR = SWAPu32(madr_next);
	232	if ((s32)(psxRegs.gpuIdleAfter - psxRegs.cycle) > 0)
	233	cycles_sum += psxRegs.gpuIdleAfter - psxRegs.cycle;
	234	psxRegs.gpuIdleAfter = psxRegs.cycle + cycles_sum + cycles_last_cmd;
	235	set_event(PSXINT_GPUDMA, cycles_sum);
	236	//printf("%u dma2cn: %6d,%4d %08x\n", psxRegs.cycle, cycles_sum,
	237	// cycles_last_cmd, HW_DMA2_MADR);
	238	return;
	239	}
	240	if (HW_DMA2_CHCR & SWAP32(0x01000000))
	241	{
	242	HW_DMA2_CHCR &= SWAP32(~0x01000000);
	243	DMA_INTERRUPT(2);
	244	}
	245	}
	246
	247	void psxAbortDma2() {
	248	psxRegs.gpuIdleAfter = psxRegs.cycle + 32;
	249	}
	250
	251	void psxDma6(u32 madr, u32 bcr, u32 chcr) {
	252	u32 words, words_max;
	253	u32 *mem;
	254
	255	PSXDMA_LOG("* DMA6 OT * %x addr = %x size = %x\n", chcr, madr, bcr);
	256
	257	if (chcr == 0x11000002) {
	258	madr &= ~3;
	259	mem = getDmaRam(madr, &words_max);
	260	if (mem == INVALID_PTR) {
	261	log_unhandled("bad6 dma madr %x\n", madr);
	262	HW_DMA6_CHCR &= SWAP32(~0x11000000);
	263	DMA_INTERRUPT(6);
	264	return;
	265	}
	266
	267	// already 32-bit size
	268	words = bcr;
	269
	270	while (bcr-- && mem > (u32 *)psxM) {
	271	*mem-- = SWAP32((madr - 4) & 0xffffff);
	272	madr -= 4;
	273	}
	274	*++mem = SWAP32(0xffffff);
	275
	276	// halted
	277	psxRegs.cycle += words;
	278	set_event(PSXINT_GPUOTCDMA, 16);
	279	return;
	280	}
	281	else {
	282	// Unknown option
	283	log_unhandled("* DMA6 OT - unknown * %x addr = %x size = %x\n", chcr, madr, bcr);
	284	}
	285
	286	HW_DMA6_CHCR &= SWAP32(~0x11000000);
	287	DMA_INTERRUPT(6);
	288	}
	289
	290	void gpuotcInterrupt()
	291	{
	292	if (HW_DMA6_CHCR & SWAP32(0x01000000))
	293	{
	294	HW_DMA6_CHCR &= SWAP32(~0x11000000);
	295	DMA_INTERRUPT(6);
	296	}
	297	}