Commit Graph

  • 783264f91c libretro: "Enable RTT (Render To Texture) Buffer" is gone, and the table of games that forced it LibretroAdmin 2026-10-07 16:51:21 +00:00
  • d359c881bb vk: a render to a texture reaches video memory when the game touches it there LibretroAdmin 2026-10-07 16:36:48 +00:00
  • 49ac112e31 vk: a render is drawn with its own matrices, not the last render's LibretroAdmin 2026-10-07 16:36:29 +00:00
  • 5f7c11942a rend: a render to a texture reaches video memory when the game touches it there (OpenGL) LibretroAdmin 2026-10-07 16:10:40 +00:00
  • 82f65906ba rend: 37 more games get their renders to a texture copied to video memory, and keep it LibretroAdmin 2026-10-07 15:44:35 +00:00
  • 74b9d8f8f5 savestate: loading one no longer drops a sound about to start, or restarts CD audio that had stopped LibretroAdmin 2026-10-07 15:38:01 +00:00
  • 889a89de56 vk: whether a pipeline clamps colours is the polygon's say alone LibretroAdmin 2026-10-07 14:58:53 +00:00
  • 4f4db17cf2 vk oit: the picture is put together over all of it, not the part the game clips to LibretroAdmin 2026-10-07 14:58:53 +00:00
  • e867692ef1 rend: per-pixel blending onto the second accumulation buffer uses that buffer's colour LibretroAdmin 2026-10-07 14:58:53 +00:00
  • e710e6b848 gl: mipmaps whatever the filter, no anisotropy left on, no blending of levels for punch-through LibretroAdmin 2026-10-07 14:48:56 +00:00
  • 77ecf7cadc libretro: texture upscaling off is 1, and palettes go to the GPU again LibretroAdmin 2026-10-07 14:11:34 +00:00
  • 9d64036cdc rend: a second volume's paletted texture is made colours on the CPU LibretroAdmin 2026-10-07 14:11:34 +00:00
  • abfe80b94a libretro: the Mipmapping option becomes Force Texture LOD0 LibretroAdmin 2026-10-07 14:01:02 +00:00
  • 6a4c68dfb4 tools/threads: a polygon with two textures on the test disc LibretroAdmin 2026-10-07 13:47:02 +00:00
  • 873658e514 rend: the picture is cleared when the game moves to a framebuffer it has not been drawing to LibretroAdmin 2026-10-07 13:26:06 +00:00
  • 95ba9e9147 rend: YUV textures converted four and eight pixels at a time LibretroAdmin 2026-10-07 13:16:31 +00:00
  • a092eec4e5 rend: 240 lines without interlace, the alpha threshold itself counts, no allocation to show the framebuffer LibretroAdmin 2026-10-07 13:16:31 +00:00
  • a2ad6ed32b rend: a modifier volume is clipped to the rectangle it was sent under LibretroAdmin 2026-10-07 13:16:31 +00:00
  • 2e0844f1bf rend: the texture converters, up to four times as fast LibretroAdmin 2026-10-07 12:47:58 +00:00
  • de7d1101e5 rend: a texture is converted in memory that is kept, and a wide stride has room LibretroAdmin 2026-10-07 12:31:10 +00:00
  • 7b373116ad rend: a compressed texture's name is hashed without an allocation LibretroAdmin 2026-10-07 12:24:04 +00:00
  • 61c1f535f7 deps: glm goes LibretroAdmin 2026-10-07 12:12:14 +00:00
  • 2d161a2983 rend: the transforms without glm LibretroAdmin 2026-10-07 12:12:13 +00:00
  • 8bcb23920a pvr: the background is the plane through its vertices, and three fixes from upstream LibretroAdmin 2026-10-07 12:02:27 +00:00
  • 424941222c tools/rend: the texture converters, byte for byte against upstream's LibretroAdmin 2026-10-07 11:49:21 +00:00
  • 1dfb89b24b rend: five fixes to the texture cache, from upstream LibretroAdmin 2026-10-07 11:49:21 +00:00
  • a9f605a1f2 maple: three fixes to the devices, from upstream LibretroAdmin 2026-10-07 11:31:36 +00:00
  • 01d2997624 rec-x64: a block reaches the table of blocks from the context register LibretroAdmin 2026-10-07 11:20:57 +00:00
  • c2df42caf5 reios: PAL for Windows CE games, and the DMA controller left on LibretroAdmin 2026-10-07 11:07:57 +00:00
  • e0a53e7a1e gdrom: while music plays, the drive's status and subcode are the music's LibretroAdmin 2026-10-07 11:07:57 +00:00
  • 42f30dbca9 Makefile: the Switch build assembles again LibretroAdmin 2026-10-07 10:54:41 +00:00
  • 958e5ba74b reios: a boot file that fits, a flash read that says 0, and four games' quirks LibretroAdmin 2026-10-07 10:46:16 +00:00
  • 2a26dde694 reios: the BIOS replacement's drive calls, from upstream's fixes LibretroAdmin 2026-10-07 10:43:57 +00:00
  • 87a1fbb21a gdrom: a disc put in takes the drive a second to recognise LibretroAdmin 2026-10-07 10:41:06 +00:00
  • 59e07e0f26 sh4: the top of the address space is only partly the processor's own LibretroAdmin 2026-10-07 10:24:27 +00:00
  • fb7d9c8fd8 sh4 dmac: a transfer the program asks for goes by its unit, its count and its address modes LibretroAdmin 2026-10-07 10:22:31 +00:00
  • 6d95b4437c gdrom: the drive starts over on reset, and four more of upstream's fixes LibretroAdmin 2026-10-07 10:20:46 +00:00
  • 7f97e1c0fa pvr: render time by the size of the scene, and three ways a game could end the process LibretroAdmin 2026-10-07 09:59:45 +00:00
  • 200e20e967 gdrom: four of upstream's fixes for games LibretroAdmin 2026-10-07 09:57:30 +00:00
  • c308deb2c2 maple: a device's answer arrives when the transfer is over, and takes as long as it does LibretroAdmin 2026-10-07 09:55:57 +00:00
  • 86b3243ddc pvr: the scanline counts every line, and a line's interrupt comes on its line LibretroAdmin 2026-10-07 09:30:52 +00:00
  • 92e91df202 rec-x64: a block's cycles come off the counter once, where it ends LibretroAdmin 2026-10-07 09:24:00 +00:00
  • 09faaca9fe sh4 dynarec: no wait check in a loop that can never be found waiting LibretroAdmin 2026-10-07 09:24:00 +00:00
  • 33d505604f tools/threads: a benchmark to set two cores side by side LibretroAdmin 2026-10-07 09:18:34 +00:00
  • 3991d0b782 rec-x64: a constant goes in the arithmetic instruction itself LibretroAdmin 2026-10-07 09:18:34 +00:00
  • b1265af72a tools/threads: the headless test runs the Windows build, under wine LibretroAdmin 2026-10-07 08:55:13 +00:00
  • bd07bfb46d mmu: the translation tables for x86-64 without the host mapping, and a cheaper way to empty them LibretroAdmin 2026-10-07 08:33:18 +00:00
  • da92efd5bf tools/threads: the disc's MMU test covers protection, remapping and address spaces LibretroAdmin 2026-10-07 08:21:04 +00:00
  • 793c830d9c rec-arm: translations kept by, so that most accesses with the MMU on are not calls LibretroAdmin 2026-10-07 08:21:04 +00:00
  • afba48305f mmu: a write to a page that may not be written is an exception LibretroAdmin 2026-10-07 08:20:51 +00:00
  • 2a65988852 rec-arm: the 32-bit ARM recompiler does the MMU LibretroAdmin 2026-10-07 07:54:24 +00:00
  • ef7fd677d4 32-bit ARM: a game that needs the MMU is run by the interpreter LibretroAdmin 2026-10-07 07:37:13 +00:00
  • 2526433c00 tools/threads: the disc runs with the MMU on, where the core has one LibretroAdmin 2026-10-07 07:37:13 +00:00
  • eaa5e033ce mmu: a 64-bit access has to sit on a 4-byte boundary, not an 8-byte one LibretroAdmin 2026-10-07 07:15:58 +00:00
  • 1c234ad8d8 sh4: an instruction that is none does not end the process in recompiled code LibretroAdmin 2026-10-07 06:25:03 +00:00
  • 50aee88d7a reios: a boot sector that was not read is not booted from LibretroAdmin 2026-10-07 06:25:03 +00:00
  • b4b372200f sh4 dynarec: no offset where the displacement is zero LibretroAdmin 2026-10-07 06:03:13 +00:00
  • c1bc6a3d79 sh4 dynarec: moves, FLDS, FSTS and FNEG are recompiled in double-precision mode LibretroAdmin 2026-10-07 06:02:18 +00:00
  • 3693c63119 sh4 dynarec: a load of SR in a delay slot does not cost the branch LibretroAdmin 2026-10-07 06:01:47 +00:00
  • 7a893d4eff rec-arm: the wait check does not cost a block its branch condition LibretroAdmin 2026-10-07 05:51:39 +00:00
  • da9d91b7b4 rec-arm: the main loop gives the frontend its frame back without an interrupt LibretroAdmin 2026-10-07 05:51:39 +00:00
  • 2939e66784 vmem: shift the address by the shift, not by the whole table entry LibretroAdmin 2026-10-07 05:29:27 +00:00
  • 42edf901d0 sh4: each store queue goes by its own QACR LibretroAdmin 2026-10-07 04:58:39 +00:00
  • b2d17f76ac Makefile: a 32-bit ARM core can be cross-built, and the generic ARM target builds LibretroAdmin 2026-10-07 04:58:39 +00:00
  • 10b55bf5e8 The wait-loop detector on the ARM recompilers, and a way to run them here LibretroAdmin 2026-10-07 04:32:52 +00:00
  • 2eb108167d rec-x64: a game that is only waiting gives up its time slice LibretroAdmin 2026-10-07 04:00:05 +00:00
  • 01346d4644 sh4: accurate instruction timing, as a core option LibretroAdmin 2026-10-07 03:30:05 +00:00
  • f978f952f4 sh4: upstream's issue cycles, units and groups in the opcode table LibretroAdmin 2026-10-07 03:30:05 +00:00
  • 22efbded7a NaN tests that a fast-math build cannot remove LibretroAdmin 2026-10-07 02:41:52 +00:00
  • b94e2d267a rec-arm, rec-arm64: FTRC of a NaN is 0x80000000 LibretroAdmin 2026-10-07 02:29:01 +00:00
  • d41e069590 sh4: the recompilers do DIV1, TAS.B and the SR and FPSCR loads themselves LibretroAdmin 2026-10-07 02:20:43 +00:00
  • efd9d865ec rec-x64: eight floating-point host registers on Linux and macOS LibretroAdmin 2026-10-07 01:57:24 +00:00
  • 43cee05688 rec-x64: the SH4's context is reached from a register LibretroAdmin 2026-10-07 01:33:50 +00:00
  • a98dbc889e rec-x64: a rewritten memory access keeps the floating-point registers LibretroAdmin 2026-10-07 01:39:53 +00:00
  • 292a22a3bc rec-x64: loads and stores use the host's mapping of guest memory LibretroAdmin 2026-10-07 01:11:07 +00:00
  • 87afb81b8e rec-x64: a block goes on to the next one itself LibretroAdmin 2026-10-07 01:02:38 +00:00
  • 1dcde44d4b rec-x64: FIPR and FTRV in line LibretroAdmin 2026-10-07 00:34:49 +00:00
  • d2fabed9fc sh4: FIPR and FTRV in plain doubles, as upstream has them LibretroAdmin 2026-10-07 00:34:49 +00:00
  • 386bdbbcd9 sh4: FTRC, FMAC and FIPR as the processor and upstream have them LibretroAdmin 2026-10-06 18:18:56 +00:00
  • 3258b64e09 sh4: DIV1 takes its divisor before the dividend is shifted LibretroAdmin 2026-10-06 18:18:56 +00:00
  • b30528d4b3 arm7: a core option to run the ARM7 through the interpreter on x86-64 LibretroAdmin 2026-10-06 17:59:33 +00:00
  • c6a039855b arm7: a block's cycles come off at its end LibretroAdmin 2026-10-06 17:26:36 +00:00
  • 58682dda08 arm7: pending register values are written back before conditional code LibretroAdmin 2026-10-06 17:18:53 +00:00
  • 01c2c99963 arm7: loads and stores to the ARM's memory are done in line LibretroAdmin 2026-10-06 16:37:39 +00:00
  • 7550fe985d arm7: each recompiled block has its own jump to the next LibretroAdmin 2026-10-06 16:37:39 +00:00
  • 7bec4a4054 tools/arm7: the ARM7, recompiled against interpreted LibretroAdmin 2026-10-06 16:08:40 +00:00
  • 890ee535bf arm7: a recompiler for x86-64 LibretroAdmin 2026-10-06 16:08:40 +00:00
  • de4aaaab0a tools/aica: the DSP, compiled against interpreted LibretroAdmin 2026-10-06 15:15:48 +00:00
  • c9e8884ede aica dsp: two faults in the compiled code LibretroAdmin 2026-10-06 15:15:48 +00:00
  • b650e75135 tools/threads: a pitched channel, and the sound across loaded states LibretroAdmin 2026-10-06 15:00:50 +00:00
  • ab205dbc58 aica: nothing is told to the ARM when there is nothing to tell LibretroAdmin 2026-10-06 15:00:50 +00:00
  • bb8a2d2cbd aica: the filter's coefficients are kept between samples LibretroAdmin 2026-10-06 15:00:50 +00:00
  • f16c11db20 aica dsp: the empty steps after the program's last are not run LibretroAdmin 2026-10-06 15:00:50 +00:00
  • 6244733b14 tools/threads: take Python's compiled files out of the repository LibretroAdmin 2026-10-06 14:46:11 +00:00
  • 396d9a22f3 aica: interpolate once, at 20 bits LibretroAdmin 2026-10-06 14:37:06 +00:00
  • c1aba3c6dd tools/threads: a filtered channel on the live test's disc LibretroAdmin 2026-10-06 14:37:06 +00:00
  • b6083369e3 aica: the low-pass filter is upstream's two-pole filter, on 20-bit samples LibretroAdmin 2026-10-06 14:37:06 +00:00
  • f85cc4e4c5 aica: the low-pass filter carries the fraction it used to drop LibretroAdmin 2026-10-06 14:15:50 +00:00
  • 5f015ea4ce tools/threads: the live test starts the program as an ELF LibretroAdmin 2026-10-06 14:08:32 +00:00
  • 828a0671f9 an ELF file given as content is loaded and run LibretroAdmin 2026-10-06 14:08:32 +00:00