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 <stdbool.h>
#include "Gamecube/DEBUG.h"
#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);
			STORE_SWAP32p(psxHAddr(0x10c0), madr + (words << 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);
			STORE_SWAP32p(psxHAddr(0x10c0), madr + (words << 2));
			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);
}

// Taken from PEOPS SOFTGPU
static inline bool 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;
}

void psxDma2(u32 madr, u32 bcr, u32 chcr) { // GPU
	u32 *ptr, madr_next, *madr_next_p;
	u32 words, words_left, words_max, words_copy;
	s32 cycles_last_cmd = 0;
	int do_walking;
	long cycles_sum;

	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;
			}
			gpuPtr->readDataMem(ptr, words_copy);
			psxCpu->Clear(madr, words_copy);

			//HW_DMA2_MADR = SWAPu32(madr + words_copy * 4);
			STORE_SWAP32p(psxHAddr(0x10a0), madr + (words << 2));

			// 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);
				gpuPtr->writeDataMem(ptr, words_copy);
				words_left -= words_copy;
				madr += words_copy * 4;
			}

			//HW_DMA2_MADR = SWAPu32(madr);
			STORE_SWAP32p(psxHAddr(0x10a0), 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
			// (some games abort the dma and read madr so break out of that logic)
			madr_next = 0xffffff;

			do_walking = Config.hacks.gpu_slow_list_walking;
			madr_next_p = do_walking ? &madr_next : NULL;

			cycles_sum = gpuPtr->dmaChain((u32 *)psxM, madr & 0x1fffff,
					madr_next_p, &cycles_last_cmd);
			// Old Soft Gpu check
			if (gpuPtr == &oldSoftGpu)
			{
				if ((int)cycles_sum <= 0)
				    cycles_sum = gpuDmaChainSize(madr);
			    HW_GPU_STATUS &= SWAP32(~PSXGPU_nBUSY);
			}

			HW_DMA2_MADR = SWAPu32(madr_next);

			// timing hack with some lame heuristics
			if (Config.gpu_timing_override && (do_walking || cycles_sum > 64))
				cycles_sum = Config.gpu_timing_override;

			psxRegs.gpuIdleAfter = psxRegs.cycle + cycles_sum + cycles_last_cmd;
			set_event(PSXINT_GPUDMA, cycles_sum);
			//printf("%u dma2cf: %d,%d %08x\n", psxRegs.cycle, cycles_sum,
			//  cycles_last_cmd, 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 (gpuPtr == &newSoftGpu && HW_DMA2_CHCR == SWAP32(0x01000401) && !(HW_DMA2_MADR & SWAP32(0x800000)))
	{
		u32 madr_next = 0xffffff, madr = SWAPu32(HW_DMA2_MADR);
		s32 cycles_last_cmd = 0;
		long cycles_sum;

		cycles_sum = gpuPtr->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: %d,%d %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);
	}
	// Old Soft Gpu check
	if (gpuPtr == &oldSoftGpu)
	{
		HW_GPU_STATUS |= SWAP32(PSXGPU_nBUSY); // GPU no longer busy
	}
}

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);
			STORE_SWAP32p2(mem, (madr - 4) & 0xffffff);
			mem--;
			madr -= 4;
		}
		*++mem = SWAP32(0xffffff);

		//GPUOTCDMA_INT(size);
		// 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);
	}
}