X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=blobdiff_plain;f=plugins%2Fgpulib%2Fgpu.c;h=24fbf3bfc3dda146e9b6703f582028b14b9fafcf;hb=5bd33f52586479b4e4b96418522ceeb330df87f2;hp=49d63eeb0be0ef245b26b9b282a455708c673a21;hpb=80bc142638988fa7ba87aa17af1ccfcfc21d1e99;p=pcsx_rearmed.git diff --git a/plugins/gpulib/gpu.c b/plugins/gpulib/gpu.c index 49d63eeb..24fbf3bf 100644 --- a/plugins/gpulib/gpu.c +++ b/plugins/gpulib/gpu.c @@ -14,8 +14,12 @@ #include /* for calloc */ #include "gpu.h" +#include "../../libpcsxcore/gpu.h" // meh +#include "../../frontend/plugin_lib.h" +#ifndef ARRAY_SIZE #define ARRAY_SIZE(x) (sizeof(x) / sizeof((x)[0])) +#endif #ifdef __GNUC__ #define unlikely(x) __builtin_expect((x), 0) #define preload __builtin_prefetch @@ -74,7 +78,10 @@ static noinline void update_width(void) int hres = hres_all[(gpu.status >> 16) & 7]; int pal = gpu.status & PSX_GPU_STATUS_PAL; int sw = gpu.screen.x2 - gpu.screen.x1; + int type = gpu.state.screen_centering_type; int x = 0, x_auto; + if (type == C_AUTO) + type = gpu.state.screen_centering_type_default; if (sw <= 0) /* nothing displayed? */; else { @@ -83,10 +90,10 @@ static noinline void update_width(void) x = (x + 1) & ~1; // blitter limitation sw /= hdiv; sw = (sw + 2) & ~3; // according to nocash - switch (gpu.state.screen_centering_type) { - case 1: + switch (type) { + case C_INGAME: break; - case 2: + case C_MANUAL: x = gpu.state.screen_centering_x; break; default: @@ -128,9 +135,12 @@ static noinline void update_height(void) /* nothing displayed? */; else { switch (gpu.state.screen_centering_type) { - case 1: + case C_INGAME: break; - case 2: + case C_BORDERLESS: + y = 0; + break; + case C_MANUAL: y = gpu.state.screen_centering_y; break; default: @@ -212,8 +222,9 @@ static noinline void get_gpu_info(uint32_t data) } } -// double, for overdraw guard -#define VRAM_SIZE ((1024 * 512 * 2 * 2) + 4096) +#ifndef max +#define max(a, b) (((a) > (b)) ? (a) : (b)) +#endif // Minimum 16-byte VRAM alignment needed by gpu_unai's pixel-skipping // renderer/downscaler it uses in high res modes: @@ -225,18 +236,31 @@ static noinline void get_gpu_info(uint32_t data) #define VRAM_ALIGN 16 #endif +// double, for overdraw guard + at least 1 page before +#define VRAM_SIZE ((1024 * 512 * 2 * 2) + max(VRAM_ALIGN, 4096)) + // vram ptr received from mmap/malloc/alloc (will deallocate using this) static uint16_t *vram_ptr_orig = NULL; -#ifdef GPULIB_USE_MMAP +#ifndef GPULIB_USE_MMAP +# ifdef __linux__ +# define GPULIB_USE_MMAP 1 +# else +# define GPULIB_USE_MMAP 0 +# endif +#endif static int map_vram(void) { - gpu.vram = vram_ptr_orig = gpu.mmap(VRAM_SIZE + (VRAM_ALIGN-1)); - if (gpu.vram != NULL) { - // 4kb guard in front +#if GPULIB_USE_MMAP + gpu.vram = vram_ptr_orig = gpu.mmap(VRAM_SIZE); +#else + gpu.vram = vram_ptr_orig = calloc(VRAM_SIZE, 1); +#endif + if (gpu.vram != NULL && gpu.vram != (void *)(intptr_t)-1) { + // 4kb guard in front gpu.vram += (4096 / 2); - // Align - gpu.vram = (uint16_t*)(((uintptr_t)gpu.vram + (VRAM_ALIGN-1)) & ~(VRAM_ALIGN-1)); + // Align + gpu.vram = (uint16_t*)(((uintptr_t)gpu.vram + (VRAM_ALIGN-1)) & ~(VRAM_ALIGN-1)); return 0; } else { @@ -244,54 +268,9 @@ static int map_vram(void) return -1; } } -#else -static int map_vram(void) -{ - gpu.vram = vram_ptr_orig = (uint16_t*)calloc(VRAM_SIZE + (VRAM_ALIGN-1), 1); - if (gpu.vram != NULL) { - // 4kb guard in front - gpu.vram += (4096 / 2); - // Align - gpu.vram = (uint16_t*)(((uintptr_t)gpu.vram + (VRAM_ALIGN-1)) & ~(VRAM_ALIGN-1)); - return 0; - } else { - fprintf(stderr, "could not allocate vram, expect crashes\n"); - return -1; - } -} - -static int allocate_vram(void) -{ - gpu.vram = vram_ptr_orig = (uint16_t*)calloc(VRAM_SIZE + (VRAM_ALIGN-1), 1); - if (gpu.vram != NULL) { - // 4kb guard in front - gpu.vram += (4096 / 2); - // Align - gpu.vram = (uint16_t*)(((uintptr_t)gpu.vram + (VRAM_ALIGN-1)) & ~(VRAM_ALIGN-1)); - return 0; - } else { - fprintf(stderr, "could not allocate vram, expect crashes\n"); - return -1; - } -} -#endif long GPUinit(void) { -#ifndef GPULIB_USE_MMAP - if (gpu.vram == NULL) { - if (allocate_vram() != 0) { - printf("ERROR: could not allocate VRAM, exiting..\n"); - exit(1); - } - } -#endif - - //extern uint32_t hSyncCount; // in psxcounters.cpp - //extern uint32_t frame_counter; // in psxcounters.cpp - //gpu.state.hcnt = &hSyncCount; - //gpu.state.frame_count = &frame_counter; - int ret; ret = vout_init(); ret |= renderer_init(); @@ -319,7 +298,7 @@ long GPUshutdown(void) ret = vout_finish(); if (vram_ptr_orig != NULL) { -#ifdef GPULIB_USE_MMAP +#if GPULIB_USE_MMAP gpu.munmap(vram_ptr_orig, VRAM_SIZE); #else free(vram_ptr_orig); @@ -333,6 +312,7 @@ long GPUshutdown(void) void GPUwriteStatus(uint32_t data) { uint32_t cmd = data >> 24; + int src_x, src_y; if (cmd < ARRAY_SIZE(gpu.regs)) { if (cmd > 1 && cmd != 5 && gpu.regs[cmd] == data) @@ -362,14 +342,17 @@ void GPUwriteStatus(uint32_t data) gpu.status |= PSX_GPU_STATUS_DMA(data & 3); break; case 0x05: - gpu.screen.src_x = data & 0x3ff; - gpu.screen.src_y = (data >> 10) & 0x1ff; - renderer_notify_scanout_x_change(gpu.screen.src_x, gpu.screen.hres); - if (gpu.frameskip.set) { - decide_frameskip_allow(gpu.ex_regs[3]); - if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) { - decide_frameskip(); - gpu.frameskip.last_flip_frame = *gpu.state.frame_count; + src_x = data & 0x3ff; src_y = (data >> 10) & 0x1ff; + if (src_x != gpu.screen.src_x || src_y != gpu.screen.src_y) { + gpu.screen.src_x = src_x; + gpu.screen.src_y = src_y; + renderer_notify_scanout_change(src_x, src_y); + if (gpu.frameskip.set) { + decide_frameskip_allow(gpu.ex_regs[3]); + if (gpu.frameskip.last_flip_frame != *gpu.state.frame_count) { + decide_frameskip(); + gpu.frameskip.last_flip_frame = *gpu.state.frame_count; + } } } break; @@ -517,15 +500,21 @@ static void start_vram_transfer(uint32_t pos_word, uint32_t size_word, int is_re log_io("start_vram_transfer %c (%d, %d) %dx%d\n", is_read ? 'r' : 'w', gpu.dma.x, gpu.dma.y, gpu.dma.w, gpu.dma.h); + if (gpu.gpu_state_change) + gpu.gpu_state_change(PGS_VRAM_TRANSFER_START); } static void finish_vram_transfer(int is_read) { if (is_read) gpu.status &= ~PSX_GPU_STATUS_IMG; - else + else { + gpu.state.fb_dirty = 1; renderer_update_caches(gpu.dma_start.x, gpu.dma_start.y, gpu.dma_start.w, gpu.dma_start.h, 0); + } + if (gpu.gpu_state_change) + gpu.gpu_state_change(PGS_VRAM_TRANSFER_END); } static void do_vram_copy(const uint32_t *params) @@ -674,6 +663,7 @@ static noinline int do_cmd_buffer(uint32_t *data, int count) break; } do_vram_copy(data + pos + 1); + vram_dirty = 1; pos += 4; continue; } @@ -703,12 +693,16 @@ static noinline int do_cmd_buffer(uint32_t *data, int count) return count - pos; } -static void flush_cmd_buffer(void) +static noinline void flush_cmd_buffer(void) { int left = do_cmd_buffer(gpu.cmd_buffer, gpu.cmd_len); if (left > 0) memmove(gpu.cmd_buffer, gpu.cmd_buffer + gpu.cmd_len - left, left * 4); - gpu.cmd_len = left; + if (left != gpu.cmd_len) { + if (!gpu.dma.h && gpu.gpu_state_change) + gpu.gpu_state_change(PGS_PRIMITIVE_START); + gpu.cmd_len = left; + } } void GPUwriteDataMem(uint32_t *mem, int count) @@ -892,7 +886,7 @@ long GPUfreeze(uint32_t type, struct GPUFreeze *freeze) GPUwriteStatus((i << 24) | (gpu.regs[i] ^ 1)); } renderer_sync_ecmds(gpu.ex_regs); - renderer_update_caches(0, 0, 1024, 512, 1); + renderer_update_caches(0, 0, 1024, 512, 0); break; } @@ -969,8 +963,6 @@ void GPUgetScreenInfo(int *y, int *base_hres) *base_hres >>= 1; } -#include "../../frontend/plugin_lib.h" - void GPUrearmedCallbacks(const struct rearmed_cbs *cbs) { gpu.frameskip.set = cbs->frameskip; @@ -983,6 +975,7 @@ void GPUrearmedCallbacks(const struct rearmed_cbs *cbs) gpu.state.frame_count = cbs->gpu_frame_count; gpu.state.allow_interlace = cbs->gpu_neon.allow_interlace; gpu.state.enhancement_enable = cbs->gpu_neon.enhancement_enable; + gpu.state.screen_centering_type_default = cbs->screen_centering_type_default; if (gpu.state.screen_centering_type != cbs->screen_centering_type || gpu.state.screen_centering_x != cbs->screen_centering_x || gpu.state.screen_centering_y != cbs->screen_centering_y) { @@ -995,6 +988,7 @@ void GPUrearmedCallbacks(const struct rearmed_cbs *cbs) gpu.mmap = cbs->mmap; gpu.munmap = cbs->munmap; + gpu.gpu_state_change = cbs->gpu_state_change; // delayed vram mmap if (gpu.vram == NULL)