)]}'
{
  "log": [
    {
      "commit": "57e4efcc84112abcd426dc5ec248b5339bee04f2",
      "tree": "91f9b020395e61fd3832b4d2393f1b898ad7dac8",
      "parents": [
        "5c372ffb4346af8d254f19489379f1b4d4d64fd5"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 16:27:32 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 16:28:49 2026"
      },
      "message": "MSL: Inherit expression dependencies for QuantToFP16.\n\nMissed this, GLSL did it correctly.\n"
    },
    {
      "commit": "5c372ffb4346af8d254f19489379f1b4d4d64fd5",
      "tree": "e140b93d6d97701701f2dd81cf3a5371f8bacb65",
      "parents": [
        "b364148ba9d8c6a52861fb4b274a3050a0556dd2"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 15:18:49 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 16:22:49 2026"
      },
      "message": "Ensure that control dependent expressions don\u0027t end up in a ternary.\n\nIt\u0027s been short-circuiting forever, so need to be extra careful.\n"
    },
    {
      "commit": "b364148ba9d8c6a52861fb4b274a3050a0556dd2",
      "tree": "cbee8451a5c683c5e4808ce4318934e112e686d2",
      "parents": [
        "8f6eac6a45bb0c5d0f228b68b3b6083aaae39de7"
      ],
      "author": {
        "name": "Abir Deol",
        "email": "abir.deol2006@gmail.com",
        "time": "Wed Oct 28 15:42:40 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 16:14:47 2026"
      },
      "message": "GLSL: Emit spec constant sized long vectors as vector\u003cT, N\u003e\n\nOpTypeVectorIdEXT shares its opcode with OpTypeCooperativeVectorNV, so a long vector with a spec constant size was emitted as coopvecNV\u003cT, N\u003e and required GL_NV_cooperative_vector. Recompiling that output turns LongVectorEXT into CooperativeVectorNV, which the device may not support. Emit vector\u003cT, N\u003e when the module declares LongVectorEXT without CooperativeVectorNV.\n"
    },
    {
      "commit": "8f6eac6a45bb0c5d0f228b68b3b6083aaae39de7",
      "tree": "c9c385b9fa738a36a674d378e6b457cafb0bf979",
      "parents": [
        "6c8e57dd855b6221ba8cf4be511828ef326371aa"
      ],
      "author": {
        "name": "Jean-Philippe",
        "email": "github@meunier.jp",
        "time": "Wed Oct 28 14:26:40 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:52:35 2026"
      },
      "message": "MSL: Reject PerVertexKHR on input block members\n\nThe MSL backend does not implement per-vertex fragment inputs yet, and a\nPerVertexKHR input variable is already rejected. A PerVertexKHR\ndecoration on a member of an input block was silently ignored instead,\nso the member was interpolated as an ordinary input. Reject it with the\nsame error as the variable decoration.\n\nCo-Authored-By: Claude Opus 5.5 \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "6c8e57dd855b6221ba8cf4be511828ef326371aa",
      "tree": "3621e9bb9035b28ddbf03b0c6e348fab5ff9f350",
      "parents": [
        "94f119993c4016e848821d4f05e4c743c3dadf07"
      ],
      "author": {
        "name": "DrDecki",
        "email": "289792529+DrDecki@users.noreply.github.com",
        "time": "Tue Oct 27 22:22:44 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:18:06 2026"
      },
      "message": "Fix spvc_parser_parse_spirv on targets where uint32_t is unsigned long\n\nSpvId is unsigned int while uint32_t is unsigned long on newlib ARM\ntoolchains such as the PS Vita SDK, so passing the SpvId pointer to the\nParser constructor does not compile there. Both types are 32 bits wide,\nso cast the pointer.\n"
    },
    {
      "commit": "94f119993c4016e848821d4f05e4c743c3dadf07",
      "tree": "a943bb7f83cab3d87f4673de3077fe132df1d2ba",
      "parents": [
        "98cbb7c02be35364ecacc23dc0c038d2360b0ed0"
      ],
      "author": {
        "name": "suraiya sulthana",
        "email": "suraiya@medios.digiscrypt.com",
        "time": "Fri Oct 23 09:56:17 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:17:54 2026"
      },
      "message": "Fix out-of-bounds read for OpTypeFloat width 8 without encoding\n"
    },
    {
      "commit": "98cbb7c02be35364ecacc23dc0c038d2360b0ed0",
      "tree": "fd1c11ce07d2b3a86ce6b11bfba235cd53c100de",
      "parents": [
        "e75e7dd8bd41c5447f63fdb95aa95ad64eb9e921"
      ],
      "author": {
        "name": "Timo Suoranta",
        "email": "tksuoran@gmail.com",
        "time": "Wed Oct 21 13:18:23 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:17:34 2026"
      },
      "message": "MSL: Inline precise arithmetic helpers on MSL 3.2\n\nNoContraction (precise) add, sub and mul are emitted as calls to helpers\ndeclared [[clang::optnone]], so the Metal compiler cannot inline them and\nevery precise operation becomes a real function call. In a fragment shader\nwith precise operations inside a loop (a shadow lookup per light) this costs\nseveral times the whole shader: on an Apple M4 Pro the forward pass of a\n43-light scene took 26.6 ms with the optnone helpers and 6.0 ms with them\ninlined, with bit-identical output.\n\nFrom MSL 3.2 the helpers are regular functions whose bodies start with\n#pragma metal fp math_mode(safe) and #pragma metal fp contract(off). The\nscoped pragmas keep every operation in the helper exact without preventing\ninlining. Earlier MSL versions keep the optnone helpers.\n"
    },
    {
      "commit": "e75e7dd8bd41c5447f63fdb95aa95ad64eb9e921",
      "tree": "d76dce0ac7fb97bc6dca356c5877cdd76574a5b7",
      "parents": [
        "27cbbbd935da2a5ff6d46ca2d7d9f180e3ef6699"
      ],
      "author": {
        "name": "Abir Deol",
        "email": "abir.deol2006@gmail.com",
        "time": "Tue Sep 29 01:23:30 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:17:19 2026"
      },
      "message": "HLSL: Ignore mesh PointSize writes with point_size_compat\n"
    },
    {
      "commit": "27cbbbd935da2a5ff6d46ca2d7d9f180e3ef6699",
      "tree": "b188c571fd7a5d9963ed2f1441ca5236817c3be4",
      "parents": [
        "738a4065b52de730774ddf36455c75d2788b189a"
      ],
      "author": {
        "name": "Faye Salwin",
        "email": "faye@zippysoft.com",
        "time": "Tue Oct 20 17:00:31 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:16:50 2026"
      },
      "message": "MSL: Don\u0027t remap a vertex shader\u0027s output block to a buffer.\n\nWith outputs captured to a buffer (vertex for tessellation), a vertex\nshader\u0027s output block is lowered to a thread-local copy that is written to\nthe output buffer on return from main, and get_type_address_space() gives it\nthe thread address space. variable_decl_is_remapped_storage() still reported\nit as StorageBuffer, so storing to a single component of one of its vectors\ncast a thread variable to a device pointer, which the Metal compiler rejects:\n\n  error: C-style cast from \u0027float4 *\u0027 to \u0027device float *\u0027 converts between\n  mismatching address spaces\n\nThe built-in gl_PerVertex block is not lowered that way and keeps its\nexisting handling. Seen with No Man\u0027s Sky\u0027s terrain shaders under MoltenVK.\n\nCo-Authored-By: Claude Opus 5.5 (1M context) \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "738a4065b52de730774ddf36455c75d2788b189a",
      "tree": "5a4d0f023e4c053346c83708b5776fdc6d7d0908",
      "parents": [
        "81ffeda4bf5babea7b6f9a7a59559e59fc8fa3b8"
      ],
      "author": {
        "name": "suraiya sulthana",
        "email": "suraiya@medios.digiscrypt.com",
        "time": "Sat Oct 17 19:48:50 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:16:34 2026"
      },
      "message": "Reject zero-constituent composite constants\n"
    },
    {
      "commit": "81ffeda4bf5babea7b6f9a7a59559e59fc8fa3b8",
      "tree": "7646af02b7ad161e15b6802f1ed656e4f733343b",
      "parents": [
        "b9701d430adb2917043c1bdc2d478be687d3fe2d",
        "ee54ddd87f9da0cdf40816c77ed85a755d93aff8"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:15:38 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Oct 28 14:15:38 2026"
      },
      "message": "Merge pull request #2716 from loewenmaehne/msl-alias-overlapping-combined-sampler\n\nMSL: Alias the sampler of an overlapping combined image sampler"
    },
    {
      "commit": "b9701d430adb2917043c1bdc2d478be687d3fe2d",
      "tree": "9edef1377e983533f5d7c459da284d1e09850f0d",
      "parents": [
        "008d2c9932e5e59009fd3ff6219f7745cbf15da7"
      ],
      "author": {
        "name": "suraiya-ctrl",
        "email": "suraiya@medios.digiscrypt.com",
        "time": "Wed Oct 28 14:14:37 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Oct 28 14:14:37 2026"
      },
      "message": "Avoid left shift of negative value in fe4m3_to_f32 (#2710)\n\n* Avoid left shift of negative value in fe4m3_to_f32\n\nint16_t(int8_t(v)) is negative for any E4M3 value with the sign bit set,\nand left shifting a negative value is undefined in C++11. Shift the\nsign-extended pattern as uint16_t instead, which yields the same bits."
    },
    {
      "commit": "008d2c9932e5e59009fd3ff6219f7745cbf15da7",
      "tree": "fe5b4cfc2960ef88f81ed157a2a4c73c50a3ba2b",
      "parents": [
        "89a33616f1cf3f7892bb7a4a36c0997e704bbc5f"
      ],
      "author": {
        "name": "Abir Deol",
        "email": "abir.deol2006@gmail.com",
        "time": "Wed Sep 30 01:36:09 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 14:14:12 2026"
      },
      "message": "Handle structs that refer to themselves through buffer references\n\nA struct that contains a physical storage buffer pointer to a block that\ncontains the struct again made two functions recurse forever:\n\n- Parser::types_are_logically_equivalent, when a struct with the same\n  name was already seen, followed the pointer member into the pointee\n  and crashed with a stack overflow.\n- CompilerGLSL::emit_copy_logical_type unrolled OpCopyLogical into the\n  pointee of the pointer member, building ever longer access chains.\n\nStop comparing a pair of types that is already being compared, and\ncopy physical pointers as values in emit_copy_logical_type.\n"
    },
    {
      "commit": "89a33616f1cf3f7892bb7a4a36c0997e704bbc5f",
      "tree": "d557f4b6f2e0ac7ad68253373038c8ddc52cddb2",
      "parents": [
        "d704cd760d9c4453beff4048534b1f2343e7d206"
      ],
      "author": {
        "name": "Laela Zorana",
        "email": "zoranalaela9@gmail.com",
        "time": "Wed Oct 14 14:25:46 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:16:51 2026"
      },
      "message": "Do not fold OpPtrAccessChain into the chain of a loaded pointer\n\nThe base of an OpPtrAccessChain can already hold an index. The new\nindex was then added to the last ] in the base text. A pointer loaded\nfrom memory is a new pointer, so that ] belongs to the chain that\nproduced it. For pc.table[0].items the index moved onto table. For an\narray of pointers such as items[2], it became items[2 + 1].\n\nThe fold now only happens when the base is an access chain. A loaded\npointer gets its own index.\n\nFixes #2705.\n"
    },
    {
      "commit": "d704cd760d9c4453beff4048534b1f2343e7d206",
      "tree": "8b4823cadcca226ed199db8800fb0fd65af5dd17",
      "parents": [
        "090562cd6af7e756225c663661c1ac106841eb24"
      ],
      "author": {
        "name": "Jean-Philippe",
        "email": "github@meunier.jp",
        "time": "Fri Oct 02 20:57:07 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:16:32 2026"
      },
      "message": "MSL: Fix multiview view and instance indices of vertex shaders for tessellation\n\nWith vertex_for_tessellation and multiview, the vertex kernel derived\nInstanceIndex from a dispatch base that was not always declared, then\nassigned a new value to it, which produced invalid MSL in many shaders.\nDerive both indices from the grid position instead: the view index is\nthe grid y coordinate modulo the view count, and the instance index is\nthe grid y coordinate divided by the view count, plus the dispatch base.\nDeclare the vertex and grid base parameters whenever multiview or the\ncaptured vertex parameters need them.\n\nWithout tessellation, the multiview remapping of InstanceIndex under\nenable_base_index_zero assigned to the zero-based expression instead of\nthe builtin; emit it as a builtin declaration.\n\nCo-Authored-By: Claude Opus 5.5 \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "090562cd6af7e756225c663661c1ac106841eb24",
      "tree": "c9619da269065f5b8e7ec1e51bea9c6649c80326",
      "parents": [
        "736016d76a4c1b23c29d0e2f8b65ed888095762d"
      ],
      "author": {
        "name": "Jean-Philippe",
        "email": "github@meunier.jp",
        "time": "Fri Oct 02 20:52:45 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:15:56 2026"
      },
      "message": "MSL: Honor explicit member Locations in flattened interface blocks\n\nVulkan assigns locations to the members of an interface block in\ndeclaration order, but a member with its own Location decoration takes\nthat location, and the members after it continue from there. The\nflattened MSL interface kept counting from the previous member instead,\nso a block whose member Locations were out of order or had gaps got\nother [[user(locnN)]] attributes than the ones its Locations select.\nRestart the location range at each explicit member Location, for plain\nand composite members.\n\nArray members with a Component decoration now keep that component in\nthe flattened interface instead of being rejected. They stay rejected in\ntessellation shaders.\n\nCo-Authored-By: Claude Opus 5.5 \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "736016d76a4c1b23c29d0e2f8b65ed888095762d",
      "tree": "81850d910df2886efb03e00dabc935cf85a3ec0d",
      "parents": [
        "33ba33a4deb3ad3e76f162d8c931dd045bdba3dc"
      ],
      "author": {
        "name": "Jean-Philippe",
        "email": "github@meunier.jp",
        "time": "Fri Oct 02 21:05:13 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:15:48 2026"
      },
      "message": "MSL: Resolve interpolant access chains that start from another access chain\n\nAn interpolateAt* operand can be an access chain whose base is itself an\naccess chain into an input block. fix_up_interpolant_access_chain()\nwalked the remaining indices from the type of the whole input variable\ninstead of the type its base points to, and SPIRV-Cross stopped with an\ninternal error. Start from the pointee type of the base, and carry the\ninterpolant component expression of the base over to the new chain.\n\nCo-Authored-By: Claude Opus 5.5 \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "ee54ddd87f9da0cdf40816c77ed85a755d93aff8",
      "tree": "56ab34bf1efe38e10a86be5ca289e2401801d9e9",
      "parents": [
        "d8f6641eaa6902ba6905eb8fafa2240c8cd8b725"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:07:33 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 13:07:33 2026"
      },
      "message": "MSL: Don\u0027t reinterpret cast into the same type for no reason.\n"
    },
    {
      "commit": "33ba33a4deb3ad3e76f162d8c931dd045bdba3dc",
      "tree": "c5b1457080764d84985552f6dfe5b14cadf5c95f",
      "parents": [
        "d68058508096d3790883009020e5c5872dd1cf70"
      ],
      "author": {
        "name": "Jean-Philippe",
        "email": "github@meunier.jp",
        "time": "Fri Oct 02 21:02:52 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:40:39 2026"
      },
      "message": "MSL: Pass the SampleMask input to helper functions by value\n\nA fragment shader that reads gl_SampleMaskIn from a helper function gets\nthe builtin as an extra argument of that function. The argument was\ndeclared as an array reference, while the entry point holds the mask as\na single uint, so the generated MSL did not compile. Declare it as a\nuint, which the helper reads as the first word of the mask.\n\nCo-Authored-By: Claude Opus 5.5 \u003cnoreply@anthropic.com\u003e\n"
    },
    {
      "commit": "d68058508096d3790883009020e5c5872dd1cf70",
      "tree": "35875ead79faf8f4c0d7de4d8f67488f883da912",
      "parents": [
        "9c41cf91726a245305c65c49612656587b27fb17"
      ],
      "author": {
        "name": "Stanisław Zagórowski",
        "email": "duckonaut@gmail.com",
        "time": "Tue Sep 29 23:36:42 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:40:26 2026"
      },
      "message": "Fix semantics on fragment block output structs\n"
    },
    {
      "commit": "9c41cf91726a245305c65c49612656587b27fb17",
      "tree": "b0809cadd2a3eab3c9469f74ab874c7871b909bb",
      "parents": [
        "e99dc7a38daa790eff5425cee18129d9d5b26176",
        "f4294631323796a3a64bf1a1480bf176ce9f6489"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:40:07 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Oct 28 12:40:07 2026"
      },
      "message": "Merge pull request #2687 from HopeSuffers/fix/msl-bda-convert-utoptr-aliasing\n\nfix: msl bda convert utoptr aliasing"
    },
    {
      "commit": "e99dc7a38daa790eff5425cee18129d9d5b26176",
      "tree": "23fd46cf8e8bcdf171f17992f91c4d372f18ad42",
      "parents": [
        "5997436b9bcdd5ce778a32a901e447303bfc74e3",
        "3220300a3da593cf46e45cecdee526b6d974e1d1"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:39:50 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Oct 28 12:39:50 2026"
      },
      "message": "Merge pull request #2686 from dttdrv/msl-mesh-fixes\n\nMSL: Fix several mesh and task shader issues"
    },
    {
      "commit": "5997436b9bcdd5ce778a32a901e447303bfc74e3",
      "tree": "33682c02bed7aba175c9848212c413292f3b759e",
      "parents": [
        "80184462ecc55804026c9f9a05c34bebad7d52af"
      ],
      "author": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Sat Oct 17 03:53:57 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:39:10 2026"
      },
      "message": "HLSL: Declare the input struct of a geometry shader that reads no input\n"
    },
    {
      "commit": "80184462ecc55804026c9f9a05c34bebad7d52af",
      "tree": "de9e892074751ad6112d70b23bc15e43a8a02af0",
      "parents": [
        "85138b2daf020665952445d1eae2c146f0d907d6"
      ],
      "author": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Sat Oct 17 03:52:20 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:38:22 2026"
      },
      "message": "HLSL: Give a byte address buffer\u0027s atomic min / max value the signedness of its opcode\n"
    },
    {
      "commit": "85138b2daf020665952445d1eae2c146f0d907d6",
      "tree": "d41db77f9a6bd1c2074c4a4149e2531cda2882c2",
      "parents": [
        "900d698920d6fda33dbf570a8c80250dd29573fa"
      ],
      "author": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Sat Oct 17 03:50:08 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:36:46 2026"
      },
      "message": "HLSL: Don\u0027t declare an atomic\u0027s result twice when a loop\u0027s continue block reads it\n"
    },
    {
      "commit": "900d698920d6fda33dbf570a8c80250dd29573fa",
      "tree": "9bc9d98792522e72cbadfd585bf021e208b266e2",
      "parents": [
        "300681bbcb8365a8d1eacdacdbdd0249bbf1d6b1"
      ],
      "author": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Fri Oct 16 17:16:09 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 12:35:46 2026"
      },
      "message": "HLSL: Emit the two-argument GetDimensions for OpArrayLength on a preserved structured buffer\n"
    },
    {
      "commit": "f4294631323796a3a64bf1a1480bf176ce9f6489",
      "tree": "c8044a0ccd4cccf22d8b8d57c0cd5ba6fc3505af",
      "parents": [
        "38abe362c57311a77808b7ec4d1d0d9ceb8d9370"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 11:33:46 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 11:33:46 2026"
      },
      "message": "MSL: Don\u0027t forward pointer bitcasts at all.\n"
    },
    {
      "commit": "3220300a3da593cf46e45cecdee526b6d974e1d1",
      "tree": "fb5706f0ea136644614ad3c55e91c0176c7e860c",
      "parents": [
        "f378fd8896746a35012a98197e3de3111bccf93c"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 28 10:58:16 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Oct 28 10:58:16 2026"
      },
      "message": "Apply suggestion from @HansKristian-Work"
    },
    {
      "commit": "d8f6641eaa6902ba6905eb8fafa2240c8cd8b725",
      "tree": "776936d05e67436abadabe8bf9248b9b299be0a0",
      "parents": [
        "6189c476bccb3163101fc451769e44b76e5e24b8"
      ],
      "author": {
        "name": "Julian Zienert",
        "email": "187679608+loewenmaehne@users.noreply.github.com",
        "time": "Sat Oct 17 08:54:31 2026"
      },
      "committer": {
        "name": "Julian Zienert",
        "email": "187679608+loewenmaehne@users.noreply.github.com",
        "time": "Sat Oct 17 08:54:31 2026"
      },
      "message": "MSL: Alias the sampler of an overlapping combined image sampler\n\nWhen combined image samplers share a binding in an argument buffer, the\ntexture of each extra one becomes a local alias of the first, but its\nsampler had no alias, so the MSL used a sampler name that was never\ndeclared. Declare the sampler alias next to the texture alias.\n"
    },
    {
      "commit": "6189c476bccb3163101fc451769e44b76e5e24b8",
      "tree": "4d923498d4b3c558fee2e46b7f1ed3dda543dad4",
      "parents": [
        "300681bbcb8365a8d1eacdacdbdd0249bbf1d6b1"
      ],
      "author": {
        "name": "Julian Zienert",
        "email": "187679608+loewenmaehne@users.noreply.github.com",
        "time": "Sat Oct 17 08:54:31 2026"
      },
      "committer": {
        "name": "Julian Zienert",
        "email": "187679608+loewenmaehne@users.noreply.github.com",
        "time": "Sat Oct 17 08:54:31 2026"
      },
      "message": "MSL: Add CLI option for explicit resource bindings\n\n--msl-resource-binding maps a descriptor binding of the entry point to\nexplicit MSL buffer, texture and sampler indices. The .resource-binding.\ntag in test_shaders.py uses it to give a combined image sampler different\ntexture and sampler indices, as MoltenVK does with argument buffers.\n"
    },
    {
      "commit": "300681bbcb8365a8d1eacdacdbdd0249bbf1d6b1",
      "tree": "7e346369cb58546807c800d2d43f8d5a6cbd40b0",
      "parents": [
        "aa217aeb6c9f0ace7a0ab233b28807edf45eb165"
      ],
      "author": {
        "name": "Zheorgyan",
        "email": "zheorgyan@gmail.com",
        "time": "Thu Oct 08 14:00:19 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Oct 14 11:31:50 2026"
      },
      "message": "MSL: Preserve negative zero in precise multiplication\n\nspvFMul computed l * r as fma(l, r, T(0)). When the product is -0,\nadding +0 gives +0, so the sign was lost. Use -0.0 as the addend\ninstead, which leaves every product unchanged. The matrix helpers\nstart their fma accumulators at -0.0 for the same reason.\n"
    },
    {
      "commit": "38abe362c57311a77808b7ec4d1d0d9ceb8d9370",
      "tree": "f6c8d0fe490f37a9c62d542d41d82c64495de84b",
      "parents": [
        "aa217aeb6c9f0ace7a0ab233b28807edf45eb165"
      ],
      "author": {
        "name": "hopesuffers",
        "email": "maurice.ach@icloud.com",
        "time": "Sat Sep 26 10:56:58 2026"
      },
      "committer": {
        "name": "hopesuffers",
        "email": "maurice.ach@icloud.com",
        "time": "Sat Sep 26 10:56:58 2026"
      },
      "message": "fix: msl bda convert utoptr aliasing\n"
    },
    {
      "commit": "f378fd8896746a35012a98197e3de3111bccf93c",
      "tree": "20bc15359e86748d7aa5c1da1f1234abe29c9791",
      "parents": [
        "672a097eaafa0d9c56bf9bb97e2520f863fc80e2"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Pass 64-bit integer stage inputs and outputs as 32-bit words\n\nMSL does not allow long or ulong in stage IO structs. 64-bit integer inputs and outputs now go through the interface as uint vectors, two words per value and two values per location, so long3 and long4 take two locations as in Vulkan. The value is kept in a local variable and converted with as_type when it is copied in or out, including in the mesh output flush.\n"
    },
    {
      "commit": "672a097eaafa0d9c56bf9bb97e2520f863fc80e2",
      "tree": "2e169df2f86b63df6a25a6032f87ba10375ef966",
      "parents": [
        "44b7ff2773d564dda83472501ce5e323ecfcce38"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Copy array-typed mesh outputs element by element\n\nMesh outputs that are arrays are flattened into one member per element, but the flush assigned the whole array to each member. The flush now indexes the element for per-vertex and per-primitive outputs.\n"
    },
    {
      "commit": "44b7ff2773d564dda83472501ce5e323ecfcce38",
      "tree": "f10198a2af5ed3fd3ab71b26161beae3a625b643",
      "parents": [
        "0b39ee9a19a534fc43989c9e4d218c2a37dd5c7d"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Give unwritten mesh Layer, ViewportIndex and PointSize outputs their defaults\n\nA mesh shader that declares these outputs without writing them now still emits them, with zero or the default point size. Metal hands undefined values to the fragment shader when a mesh function omits them.\n"
    },
    {
      "commit": "0b39ee9a19a534fc43989c9e4d218c2a37dd5c7d",
      "tree": "2d770ed7547299ac8a04b78cbf6903ee749ab271",
      "parents": [
        "16e2dd8fc63dd7cc34744bbe1610e06013c1b3c2"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Take mesh task counts from the first task invocation\n\nEmitMeshTasksEXT sets the mesh grid only from the first invocation of the task workgroup. Every invocation used to call set_threadgroups_per_grid, so the last one won.\n"
    },
    {
      "commit": "16e2dd8fc63dd7cc34744bbe1610e06013c1b3c2",
      "tree": "2785f1bd6ad60d27474d22824ba23fb5b9f811bd",
      "parents": [
        "aa76954b1311ba10cf49c0334f114f6a8ea82cef"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Use builtin types for mesh shader input arguments\n\nInput builtins passed to the wrapped mesh entry function use their builtin type, as in other stages. gl_DrawID in a mesh shader was passed as int\u0026 against the uint read from the draw index buffer.\n"
    },
    {
      "commit": "aa76954b1311ba10cf49c0334f114f6a8ea82cef",
      "tree": "627688a7d245e395d875bf2106d1275d7e3bcda6",
      "parents": [
        "aa217aeb6c9f0ace7a0ab233b28807edf45eb165"
      ],
      "author": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "committer": {
        "name": "dttdrv",
        "email": "dttdrv@users.noreply.github.com",
        "time": "Fri Sep 25 14:05:39 2026"
      },
      "message": "MSL: Emit mesh shader workgroup size from LocalSizeId\n\nMesh shaders now request the implicit workgroup size builtin, so gl_WorkGroupSize comes from LocalSizeId (and its specialization constants) as well as LocalSize. With LocalSizeId the output flush loop strided by zero.\n"
    },
    {
      "commit": "aa217aeb6c9f0ace7a0ab233b28807edf45eb165",
      "tree": "fdfdf68bae75fbf87e872afb74ccdbaf5a5226e9",
      "parents": [
        "d029329bd164a3f38338d95ab56c74f27128a029"
      ],
      "author": {
        "name": "KhronosWebservices",
        "email": "webservices@khronosgroup.org",
        "time": "Mon Sep 21 18:03:43 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 22 08:57:59 2026"
      },
      "message": "This PR adds a CONTRIBUTING file which includes a legally mandated AI Contributions block of text and adds CLA pre-disclosure with a link to the CLA text\n"
    },
    {
      "commit": "d029329bd164a3f38338d95ab56c74f27128a029",
      "tree": "34a5fe981939a96888c98344ff9596652f15434f",
      "parents": [
        "8af715398cfe1ccc337710865aecf80959bd442f",
        "d979ee513cdbc21df5a91af5bbbc155979aad96b"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 14:05:31 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Sep 18 14:05:31 2026"
      },
      "message": "Merge pull request #2684 from KhronosGroup/shader-abort\n\nGLSL: Implement GL_EXT_abort"
    },
    {
      "commit": "d979ee513cdbc21df5a91af5bbbc155979aad96b",
      "tree": "34a5fe981939a96888c98344ff9596652f15434f",
      "parents": [
        "07cc23e582cdb415e1765a9d2a2f78079fe0479d"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 13:54:27 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 13:54:27 2026"
      },
      "message": "Fix narrowing warning.\n"
    },
    {
      "commit": "07cc23e582cdb415e1765a9d2a2f78079fe0479d",
      "tree": "ab63e0cb7df3275b7b79a52295c3365ae2c8d870",
      "parents": [
        "8af715398cfe1ccc337710865aecf80959bd442f"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 13:52:15 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 13:52:15 2026"
      },
      "message": "GLSL: Implement GL_EXT_abort\n"
    },
    {
      "commit": "8af715398cfe1ccc337710865aecf80959bd442f",
      "tree": "430fc2a743c9b0c15bcc8fedef8a8fbe161770ec",
      "parents": [
        "00ae9b44c9c16561bdff2e094c518a89f756944b",
        "da0328a19083fc94e63f664e6ff727b5454e8640"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 12:07:45 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Sep 18 12:07:45 2026"
      },
      "message": "Merge pull request #2683 from Karpov-Ruslan/function_attributes\n\nGLSL: Implement GL_EXT_maximal_reconvergence, GL_EXT_subgroup_uniform_control_flow and GL_EXT_function_control_attributes"
    },
    {
      "commit": "da0328a19083fc94e63f664e6ff727b5454e8640",
      "tree": "430fc2a743c9b0c15bcc8fedef8a8fbe161770ec",
      "parents": [
        "191119cee258885555db26de2cf05f7afeaa2610"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 18 11:58:49 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Sep 18 11:58:49 2026"
      },
      "message": "Apply batched suggestions from code review\n\nCo-authored-by: Hans-Kristian Arntzen \u003cpost@arntzen-software.no\u003e"
    },
    {
      "commit": "191119cee258885555db26de2cf05f7afeaa2610",
      "tree": "c51dc310768d2638e5b74904edef2e1527806688",
      "parents": [
        "157149537be6693b4e002dfc65dc3de8fcf9da93"
      ],
      "author": {
        "name": "ifrfor",
        "email": "karpov.rv@phystech.edu",
        "time": "Fri Sep 18 08:05:23 2026"
      },
      "committer": {
        "name": "ifrfor",
        "email": "karpov.rv@phystech.edu",
        "time": "Fri Sep 18 08:38:20 2026"
      },
      "message": "GLSL: Add support for function control hints\n"
    },
    {
      "commit": "157149537be6693b4e002dfc65dc3de8fcf9da93",
      "tree": "23b5f21b071914b4bce7d95ad41bb7ff44a187f7",
      "parents": [
        "00ae9b44c9c16561bdff2e094c518a89f756944b"
      ],
      "author": {
        "name": "ifrfor",
        "email": "karpov.rv@phystech.edu",
        "time": "Fri Sep 18 07:37:53 2026"
      },
      "committer": {
        "name": "ifrfor",
        "email": "karpov.rv@phystech.edu",
        "time": "Fri Sep 18 07:37:53 2026"
      },
      "message": "GLSL: Add subgroup reconvergence support\n"
    },
    {
      "commit": "00ae9b44c9c16561bdff2e094c518a89f756944b",
      "tree": "dc3eed74504b588639be6507d1444ab2cb868f74",
      "parents": [
        "71055a8144bef210e43cc6519900079629ca7f89",
        "8d0e00a4ffa0824c0573e5cb75f83656e4cd8990"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 13:23:26 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Thu Sep 17 13:23:26 2026"
      },
      "message": "Merge pull request #2682 from KhronosGroup/roll-deps\n\nRoll deps"
    },
    {
      "commit": "8d0e00a4ffa0824c0573e5cb75f83656e4cd8990",
      "tree": "dc3eed74504b588639be6507d1444ab2cb868f74",
      "parents": [
        "bdd3479438f3394ab42395c17f5730cc581473e5"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:58:53 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:58:53 2026"
      },
      "message": "GLSL: Prefer EXT_texture_shadow_lod on Vulkan GLSL.\n"
    },
    {
      "commit": "bdd3479438f3394ab42395c17f5730cc581473e5",
      "tree": "ff279dec59fa1d3711a587912ca23ad69d1e5dc0",
      "parents": [
        "1c7b0be304fcb63b9b9da69d15bd06e332aaf525"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:03:05 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:46:49 2026"
      },
      "message": "Roll deps.\n"
    },
    {
      "commit": "1c7b0be304fcb63b9b9da69d15bd06e332aaf525",
      "tree": "04c9aba414a37de8df609e03ce262463fab761ec",
      "parents": [
        "47587f99b8887b5ecc969c561346f179a3d5ce2c"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:46:11 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:46:49 2026"
      },
      "message": "Handle more NonSemantic.DebugInfo versions.\n\nDepending on how glslang is invoked it can arbitrarily use 102 version\ninstead of 100, which broke test suite.\n"
    },
    {
      "commit": "47587f99b8887b5ecc969c561346f179a3d5ce2c",
      "tree": "0c647728884061aeaff7af4864e90b950d3674bc",
      "parents": [
        "71055a8144bef210e43cc6519900079629ca7f89"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:02:19 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 12:02:19 2026"
      },
      "message": "Handle codegen for new glslang descriptor heap.\n\nNeed to consider that storage images can be emitted inside a block type\nnow to signal nonwritable/nonreadable.\n"
    },
    {
      "commit": "71055a8144bef210e43cc6519900079629ca7f89",
      "tree": "99c9e5ae718f5ac2fe12256fb6ac0d5e4de00468",
      "parents": [
        "a9193e76134de63810ee1a57b5343f236d6ff102",
        "cd06139cbbe732faddde181882f8324baa24ec15"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 09:55:42 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Thu Sep 17 09:55:42 2026"
      },
      "message": "Merge pull request #2681 from MouseChannel/merge\n\nMSL: Fix invalid pointer constructor when unrolling row-major column …"
    },
    {
      "commit": "cd06139cbbe732faddde181882f8324baa24ec15",
      "tree": "99c9e5ae718f5ac2fe12256fb6ac0d5e4de00468",
      "parents": [
        "c6c94d7fbed04f41f925a23f486f5a2c5bdb4011"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 09:43:55 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 17 09:44:40 2026"
      },
      "message": "Never pass down pointer types to matrix conversions in the first place.\n"
    },
    {
      "commit": "c6c94d7fbed04f41f925a23f486f5a2c5bdb4011",
      "tree": "88c7cb66d3beeea9212f4c4901236adaa5e032dd",
      "parents": [
        "a9193e76134de63810ee1a57b5343f236d6ff102"
      ],
      "author": {
        "name": "mousechannel",
        "email": "mochenghh@gmail.com",
        "time": "Thu Sep 17 05:59:47 2026"
      },
      "committer": {
        "name": "mousechannel",
        "email": "mochenghh@gmail.com",
        "time": "Thu Sep 17 05:59:47 2026"
      },
      "message": "MSL: Fix invalid pointer constructor when unrolling row-major column load\n"
    },
    {
      "commit": "a9193e76134de63810ee1a57b5343f236d6ff102",
      "tree": "4685ebaf3bae3fe1d42211f406772ca65d522226",
      "parents": [
        "f11ba9f0b21ba8fc15153d50a2a1ae31ab1cf8f7",
        "e7496abeaf8acd5c4f36ef585416325f1bff5f33"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Sep 16 12:07:24 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Sep 16 12:07:24 2026"
      },
      "message": "Merge pull request #2645 from zainsharief/msl-coopmat-elementwise\n\nMSL: Add cooperative matrix element-wise emulation"
    },
    {
      "commit": "f11ba9f0b21ba8fc15153d50a2a1ae31ab1cf8f7",
      "tree": "85495d43955e3764051db37c9c974e298583de21",
      "parents": [
        "94d59e564363a08b7d053692e9499d0f909c106b",
        "0706157ef54d37a653681d6a0ce6af05350b0e3f"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Sep 16 11:27:19 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Sep 16 11:27:19 2026"
      },
      "message": "Merge pull request #2666 from anshulnoori/feat/depth-clip-enable-moltenvk\n\nfeat(msl): add depth clip enable emulation"
    },
    {
      "commit": "94d59e564363a08b7d053692e9499d0f909c106b",
      "tree": "8582ba5b6d327fc546238aa1ad82e0b2b403c5e3",
      "parents": [
        "2ca01986900b7c0f86ccc98bf0da9080332790cd",
        "96f76bcc7ae133e4daf45c3b94663f0b836e61c5"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Sep 16 11:24:27 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Sep 16 11:24:27 2026"
      },
      "message": "Merge pull request #2678 from hmennen90/msl-struct-array-stride-padding\n\nMSL: Fix layout of struct arrays whose ArrayStride exceeds the packed struct size"
    },
    {
      "commit": "2ca01986900b7c0f86ccc98bf0da9080332790cd",
      "tree": "38f735e558440028a07b0de737aa7cfd04aefb4d",
      "parents": [
        "00afed5079da75bd8d4ca96f1b8ff2130a96b7f4",
        "90588e91d57e2ed7d50bfe92e4c750aa135d9316"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Sep 16 11:12:56 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Sep 16 11:12:56 2026"
      },
      "message": "Merge pull request #2679 from arcusbuilds/fix/2677-cxx11\n\nGLSL: Fix C++11 lambda compatibility"
    },
    {
      "commit": "00afed5079da75bd8d4ca96f1b8ff2130a96b7f4",
      "tree": "57c7327ab9aafd13bb816bd59e3c374a56205185",
      "parents": [
        "be71ee8c12cd7dc5ca8fa9581f708c2e8561fe2a",
        "b16f65747a097c0e9247b67374c9b03fb90c4376"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Wed Sep 16 11:12:36 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Wed Sep 16 11:12:36 2026"
      },
      "message": "Merge pull request #2680 from xen2/fix/hlsl-undef-phi-double-zero-init\n\nHLSL: Don\u0027t emit a second zero initializer for OpUndef loop variables"
    },
    {
      "commit": "b16f65747a097c0e9247b67374c9b03fb90c4376",
      "tree": "57c7327ab9aafd13bb816bd59e3c374a56205185",
      "parents": [
        "be71ee8c12cd7dc5ca8fa9581f708c2e8561fe2a"
      ],
      "author": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Sun Sep 13 11:43:50 2026"
      },
      "committer": {
        "name": "Virgile Bello",
        "email": "virgile@stride3d.net",
        "time": "Sun Sep 13 11:43:50 2026"
      },
      "message": "HLSL: Don\u0027t emit a second zero initializer for OpUndef loop variables\n\nWith --force-zero-initialized-variables, variable_decl() already zero-initializes an OpUndef loop variable, so the suffix added for FXC produced invalid code such as \"uint _19 \u003d 0u \u003d 0u;\".\n"
    },
    {
      "commit": "e7496abeaf8acd5c4f36ef585416325f1bff5f33",
      "tree": "80e2194566c4bd034e885704a657830107ef7636",
      "parents": [
        "9538032a7985418f7e31988a8cffbfa00c3ae1ab"
      ],
      "author": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sat Sep 12 19:37:22 2026"
      },
      "committer": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sat Sep 12 19:37:22 2026"
      },
      "message": "MSL: Use thread elements for coopmat element-wise emulation\n"
    },
    {
      "commit": "90588e91d57e2ed7d50bfe92e4c750aa135d9316",
      "tree": "2c8ec72ead1eded78b611cf496e1aa80c6dba265",
      "parents": [
        "be71ee8c12cd7dc5ca8fa9581f708c2e8561fe2a"
      ],
      "author": {
        "name": "Srijan Keshri",
        "email": "Srijan.Keshri@mindbodyonline.com",
        "time": "Wed Sep 09 01:44:11 2026"
      },
      "committer": {
        "name": "Srijan Keshri",
        "email": "Srijan.Keshri@mindbodyonline.com",
        "time": "Wed Sep 09 02:20:55 2026"
      },
      "message": "GLSL: Fix C++11 lambda compatibility\n"
    },
    {
      "commit": "96f76bcc7ae133e4daf45c3b94663f0b836e61c5",
      "tree": "81fc6cd469c20a75ad666f2c1bd2b911b658cbbb",
      "parents": [
        "be71ee8c12cd7dc5ca8fa9581f708c2e8561fe2a"
      ],
      "author": {
        "name": "Hendrik Mennen",
        "email": "hmennen@gambio.net",
        "time": "Tue Sep 08 11:03:02 2026"
      },
      "committer": {
        "name": "Hendrik Mennen",
        "email": "hmennen@gambio.net",
        "time": "Tue Sep 08 11:03:02 2026"
      },
      "message": "MSL: Fix layout of struct arrays whose ArrayStride exceeds the packed struct size.\n\nA std140 struct array such as `SpotLight lights[4]` (7 members, ArrayStride 64)\npacks to 52 bytes in MSL. Two things went wrong when such an array is followed\nby further members:\n\n1. mark_scalar_layout_structs() runs before the element struct is packed by its\n   own align_struct() pass, so it compared the ArrayStride against the *unpacked*\n   struct size (64 \u003d\u003d 64), never set DecorationArrayStride on the struct and the\n   array was emitted as `SpotLight lights[4]` with a native stride of 52. Every\n   element after the first was read from the wrong offset.\n   align_struct() now re-checks the packed size after ensure_member_packing_rules_msl()\n   and routes the array through spvPaddedArrayElement when needed.\n\n2. When a later compilation pass did switch the array to spvPaddedArrayElement\n   (e.g. triggered by another recompile), the SPIRVCrossDecorationPaddingTarget\n   recorded for the *following* member in the first pass (48 bytes computed from\n   the 4 * 52 layout) was never cleared, so both the padded element and the stale\n   `_mN_pad[48]` were emitted and every member after the array shifted by 48 bytes.\n   align_struct() now recomputes the padding target from scratch on every pass.\n\nAdds a regression shader with a dynamically indexed struct array followed by\nscalar and vector members.\n"
    },
    {
      "commit": "be71ee8c12cd7dc5ca8fa9581f708c2e8561fe2a",
      "tree": "5b63b53d7158f13daf59205682e24edc4658de19",
      "parents": [
        "a4d940425a157242adb034f83c4ef2d77e5d9446"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 12:03:14 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 12:03:55 2026"
      },
      "message": "GLSL: Fix regression with long-vector + weird type mismatches.\n\nNot exposed in normal GLSL.\n"
    },
    {
      "commit": "a4d940425a157242adb034f83c4ef2d77e5d9446",
      "tree": "6fc2b9b0cb7af63b5ff6fbd325825ee7ebcdda03",
      "parents": [
        "83fa691cb8606ca4b3af7f13bfcbedd5668f2a3a",
        "67ac82532209ff929f81e799e670c5b558de2ff5"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 11:45:05 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Mon Sep 07 11:45:05 2026"
      },
      "message": "Merge pull request #2676 from KhronosGroup/long-vector-fix\n\nFix long vector bugs with 1 component vectors"
    },
    {
      "commit": "67ac82532209ff929f81e799e670c5b558de2ff5",
      "tree": "6fc2b9b0cb7af63b5ff6fbd325825ee7ebcdda03",
      "parents": [
        "621778eb79bbbcd101d9d2b848e198a2b9f3da63"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 11:36:16 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 11:36:16 2026"
      },
      "message": "GLSL: Add some pragmatic guards against implementation bugs.\n\nIt\u0027s unlikely we\u0027ve found all edge cases where false positive long\nvector would go through. Gate its emission based on SPIR-V caps.\n"
    },
    {
      "commit": "621778eb79bbbcd101d9d2b848e198a2b9f3da63",
      "tree": "d7ef1ec501fc4d065cb4a916ef2b30681344f6ed",
      "parents": [
        "83fa691cb8606ca4b3af7f13bfcbedd5668f2a3a"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 11:13:41 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Sep 07 11:30:50 2026"
      },
      "message": "GLSL: Fix issues with single component long vectors.\n\nNeed to rewrite a lot of old assumptions here.\nPossible there are still gremlins lurking.\n"
    },
    {
      "commit": "0706157ef54d37a653681d6a0ce6af05350b0e3f",
      "tree": "3853113911583bf53a3a45af910b2608f91d89c7",
      "parents": [
        "9c3c8e2cefdd8194b193bb8ed2fdff4d5527e382"
      ],
      "author": {
        "name": "Anshul Noori",
        "email": "anshulnoori@gmail.com",
        "time": "Mon Aug 24 23:37:16 2026"
      },
      "committer": {
        "name": "Anshul Noori",
        "email": "anshulnoori@gmail.com",
        "time": "Sun Sep 06 02:16:20 2026"
      },
      "message": "feat(msl): add depth clip enable emulation\n"
    },
    {
      "commit": "83fa691cb8606ca4b3af7f13bfcbedd5668f2a3a",
      "tree": "423782ba26a2a81b6c235dc3d1361340b8146a2b",
      "parents": [
        "fdfd4e5d9974e1e03530a5db8755b81b3f7c58d7"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 15:00:56 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 15:01:19 2026"
      },
      "message": "Fix accidental crash if we\u0027re running complex continue blocks.\n\nSmall regression from earlier fix.\n"
    },
    {
      "commit": "fdfd4e5d9974e1e03530a5db8755b81b3f7c58d7",
      "tree": "ce321c20007f32e10582016745fa1ead6d1eb1e4",
      "parents": [
        "a9d9b314472b425b196ec3794f373a0a2a964813",
        "be7e85fb68eaaca432a13a399bd7469dd065ac49"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 14:37:57 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Sep 04 14:37:57 2026"
      },
      "message": "Merge pull request #2675 from KhronosGroup/glsl-long-vector\n\nGLSL: Add support for EXT_long_vector"
    },
    {
      "commit": "be7e85fb68eaaca432a13a399bd7469dd065ac49",
      "tree": "ce321c20007f32e10582016745fa1ead6d1eb1e4",
      "parents": [
        "f60071c2d53a7f45d51bb49fa54d3b9eaee010e3"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 14:19:59 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 14:19:59 2026"
      },
      "message": "GLSL: Handle OpConstantNull long vectors.\n"
    },
    {
      "commit": "f60071c2d53a7f45d51bb49fa54d3b9eaee010e3",
      "tree": "9fa4144a4b85ce76f2e5031c368b93e44c9368e4",
      "parents": [
        "42d22908659f83436bd2ffed1ba4e883aca59336"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 12:18:29 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 12:18:29 2026"
      },
      "message": "GLSL: Add test for SSA-based long vector opcodes.\n"
    },
    {
      "commit": "42d22908659f83436bd2ffed1ba4e883aca59336",
      "tree": "22f97339f7dda7a6e2fa84c3b46153dab022a9fd",
      "parents": [
        "a9d9b314472b425b196ec3794f373a0a2a964813"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 12:04:39 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Sep 04 12:04:39 2026"
      },
      "message": "GLSL: Add basic EXT_long_vector support.\n\nThe blast radius for this extension is extremely large.\nIt\u0027s not impossible I missed some edge case.\n"
    },
    {
      "commit": "a9d9b314472b425b196ec3794f373a0a2a964813",
      "tree": "18788c2dd72fbe9cde3c6a39f9a84410e344f849",
      "parents": [
        "bb1f908ae0b532aeadb9fdecd8a01ba219733fb8",
        "667ac82feccb0ed255701649dd7509afccb566eb"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:48:55 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:48:55 2026"
      },
      "message": "Merge branch \u0027fix-hlsl-scalar-to-vector-expr\u0027\n"
    },
    {
      "commit": "bb1f908ae0b532aeadb9fdecd8a01ba219733fb8",
      "tree": "875e2e47034681225046a8e773851ebc45f5bb06",
      "parents": [
        "2a5a3ed1ce14dc3bc8b7372df38f9dd2d1991a07",
        "ca7cca0573f50fa159e1b4423d29b027ccc94d2c"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:47:57 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:47:57 2026"
      },
      "message": "Merge branch \u0027fix-glsl-row-major-matrix-order\u0027\n"
    },
    {
      "commit": "ca7cca0573f50fa159e1b4423d29b027ccc94d2c",
      "tree": "8d8e065c6b6fbabd13022b1f6abfb48add27cd3d",
      "parents": [
        "ba7c9f44688680c72f428db662d4b331023bfe0b"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:35:44 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:35:44 2026"
      },
      "message": "Add more test coverage for other backends.\n"
    },
    {
      "commit": "667ac82feccb0ed255701649dd7509afccb566eb",
      "tree": "0d8a4b1e54054a6a96949093e3958e6807468a57",
      "parents": [
        "c6fb52cd3494e8bdc2825e1998eab7d153bc0f13"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:22:49 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:22:49 2026"
      },
      "message": "HLSL: Properly enclose expression when splatting with C-cast.\n"
    },
    {
      "commit": "2a5a3ed1ce14dc3bc8b7372df38f9dd2d1991a07",
      "tree": "5d301fe845786106bb645cadcc2b4d69bc1574f4",
      "parents": [
        "bf0e342fffc7f3a5ddf025796590ee8d9e9e859b"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:20:57 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Thu Sep 03 10:20:57 2026"
      },
      "message": "MSL: Add missing test shader for compilation loop edge case.\n"
    },
    {
      "commit": "c6fb52cd3494e8bdc2825e1998eab7d153bc0f13",
      "tree": "bd658ce862db26714232c53a5856a44dab212742",
      "parents": [
        "cd3fcb2603ede297edb90ab5a679e4ac814055e2"
      ],
      "author": {
        "name": "Laura Hermanns",
        "email": "lukas.hermanns90@gmail.com",
        "time": "Tue Sep 01 23:49:46 2026"
      },
      "committer": {
        "name": "Laura Hermanns",
        "email": "lukas.hermanns90@gmail.com",
        "time": "Wed Sep 02 15:56:32 2026"
      },
      "message": "HLSL: Fix HLSL tests (FXC isn\u0027t running in CI anymore).\n\n- Update HLSL references for \u0027replicated-composites.spv16.vk.asm.comp\u0027;\n  Their scalar-to-vector constructors are not supported by FXC nor DXC,\n  it needs to be a type cast, e.g. `(float4)(0.0f)` instead of `float4(0.0f)`.\n- Don\u0027t run FXC in test_shaders.py script for shader models \u003e 5;\n  Mesh shaders for instance are not supported by FXC.\n- Fix ssbo-store-array.comp test shader;\n  `Data d1` variable was read uninitialized and that fails with FXC\n  while glslang seems to ignore it.\n- There\u0027s one single `[[fallthrough]]` statement in the entire code base;\n  Replace it with a /* fallthrough */ comment like at the other places,\n  to fix warning about being a C++17 feature on MSVC17.\n"
    },
    {
      "commit": "ba7c9f44688680c72f428db662d4b331023bfe0b",
      "tree": "b60790e8189ed040b8fc077cea65bef49b6f7bea",
      "parents": [
        "cd3fcb2603ede297edb90ab5a679e4ac814055e2"
      ],
      "author": {
        "name": "Laura Hermanns",
        "email": "lukas.hermanns90@gmail.com",
        "time": "Wed Sep 02 00:05:05 2026"
      },
      "committer": {
        "name": "Laura Hermanns",
        "email": "lukas.hermanns90@gmail.com",
        "time": "Wed Sep 02 15:21:22 2026"
      },
      "message": "GLSL: Fix row_major matrix order for push_constant to struct translation in OpenGL semantic.\n\nHLSL shaders with the `[[vk::push_constant]]` attribute that are compiled to SPIR-V,\ntheir matrix layout uses row_major by default but these push constants are translated to\nbare structs in GLSL for OpenGL semantics, which do not support `layout(row_major)` qualifiers.\nIn this case, allow the respective expression in `member_is_non_native_row_major_matrix()`\nto be marked with the `need_transpose` bit.\n\nThe expected behavior is presented in the new test \u0027push-constant-row-major-matrix.asm.vert\u0027:\n`gl_Position \u003d matrix_constants.transform * in_var_POSITION;`.\n\nWithout this change, SPIRV-Cross compiles this to the wrong expression:\n`gl_Position \u003d in_var_POSITION * matrix_constants.transform;`\n"
    },
    {
      "commit": "bf0e342fffc7f3a5ddf025796590ee8d9e9e859b",
      "tree": "3d3bb0603eae8db95e36ae1c680ad0c64f5769c9",
      "parents": [
        "763015417e47139b7d397121e727841d49317073"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 14:00:52 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 14:35:49 2026"
      },
      "message": "Don\u0027t skip running continue block code.\n\nThe continue block might be doing transformations on its own that should\ntrigger new compilation iterations.\nAvoids hitting the limit in some extreme cases.\n"
    },
    {
      "commit": "763015417e47139b7d397121e727841d49317073",
      "tree": "8a91c5ece26afe35d89b1d18f232c6293640b2ba",
      "parents": [
        "df0828ca34d6e94ce23ccf211abe6967591dfdda"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 12:48:38 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 13:43:55 2026"
      },
      "message": "MSL: Bump the descriptor set limit to 16.\n"
    },
    {
      "commit": "df0828ca34d6e94ce23ccf211abe6967591dfdda",
      "tree": "3fa9d4a792e0d4b395486d10bd2437efb037ad07",
      "parents": [
        "cd3fcb2603ede297edb90ab5a679e4ac814055e2"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 13:03:25 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Tue Sep 01 13:43:41 2026"
      },
      "message": "MSL: Fix AABBOpaque check. Needs to be inverted.\n"
    },
    {
      "commit": "cd3fcb2603ede297edb90ab5a679e4ac814055e2",
      "tree": "94ee221eef810a64a998434a9fa6b8f126977baf",
      "parents": [
        "d6b093e02cdc5eeaafff6592a6365b317a9dd74b"
      ],
      "author": {
        "name": "walkbackgod",
        "email": "69391547+darrinperry83@users.noreply.github.com",
        "time": "Mon Aug 31 15:13:37 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Mon Aug 31 15:13:37 2026"
      },
      "message": "MSL: Reserve Metal type names so resources cannot shadow them. (#2664)\n\nA SPIR-V variable or struct member named after an MSL type emits a declaration\nthat shadows the type, and the next declaration using that type fails to compile:\n\n    texture2d\u003cfloat\u003e sampler [[texture(0)]];\n    sampler samplerSmplr [[sampler(1)]];\n    // error: must use \u0027struct\u0027 tag to refer to type \u0027sampler\u0027 in this scope\n\nThis is reachable from ordinary shaders. Combined image samplers always emit a\n\u0027sampler \u003cname\u003eSmplr\u0027 parameter immediately after the texture, so a resource named\n\u0027sampler\u0027 always breaks. SPIR-V carries arbitrary OpName strings, so the name need\nnot be expressible in GLSL, where \u0027sampler\u0027 is a keyword.\n\nReserve the names emitted by sampler_type() and image_type_glsl(). The existing\ntexture-access.swizzle references show the same shadowing already in the suite\n(a variable named depth2d of type depth2d\u003cfloat\u003e); those compiled only because the\ntype is not used again afterwards."
    },
    {
      "commit": "d6b093e02cdc5eeaafff6592a6365b317a9dd74b",
      "tree": "2be100bd3833115dbaa7d798bc1917fd5df05483",
      "parents": [
        "88960ab8600ad90b0914cabf63171b9c95ccc1c9"
      ],
      "author": {
        "name": "Jackson Yarger",
        "email": "50038825+JackTYM@users.noreply.github.com",
        "time": "Mon Aug 31 14:10:49 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Mon Aug 31 14:10:49 2026"
      },
      "message": "MSL: don\u0027t type a resource as depth/comparison from provably dead specialization branches (#2655)\n\n* MSL: declare Dref-sampled resources as their non-depth type, cast at the call site\n\nCombinedImageSamplerUsageHandler previously force-typed any resource touched\nby an OpImageSampleDrefImplicitLod-family instruction as depth2d\u003cT\u003e (or the\ncube/array equivalents), regardless of whether that Dref usage was actually\nreachable. This broke resources that are genuinely sampled both ways (e.g.\nvia a spec-constant-gated branch, or a bindless array indexed by both\ndepth-compare and regular samplers), corrupting the non-depth sample\u0027s\noutput since depth2d\u003cT\u003e only returns real data on the depth channel.\n\nThis replaces that resource-level, usage-derived typing with a strategy that\nnever needs reachability or dead-code analysis at all: sampled-image\nresources always declare as their plain type (texture2d\u003cT\u003e, texturecube\u003cT\u003e,\ntexture2d_array\u003cT\u003e, texturecube_array\u003cT\u003e), and each Dref sample call site\ncasts the resource via a new spvDepthCast() helper immediately before\ncalling sample_compare(). Genuine static depth attachments/render\ntargets/subpass inputs are untouched -- image_type_glsl() now keys off the\nSPIR-V image type\u0027s own depth flag (type.image.depth with\nImageFormatUnknown) rather than the Dref-usage-derived is_depth_image(),\nwhich no longer influences declaration type at all.\n\nVerified against SPIRV-Cross\u0027s own shaders-msl and shaders-msl-no-opt\nsuites (with and without --opt): identical output everywhere except the\nresources that genuinely go through this path, plus three new regression\ntests (2D, texture2d_array, texturecube) covering the pattern, mirroring a\nreal-world case found in Call of Duty: Modern Warfare 2\u0027s reflection-probe\nand sky cubemap shaders under sogen (github.com/momo5502/sogen) -- the same\nspec-constant-gated Dref/plain dual-sample pattern as the original 2D bug,\njust on cube resources instead of 2D, and confirmed via a real end-to-end\nMoltenVK build driving actual Metal hardware.\n\nNo depth2d_ms/texture2d_ms overload: confirmed via spirv-val that any Dref\nsampling operation (Implicit or Explicit Lod) against a multisample image\nis invalid SPIR-V (\"Dref sampling operation is invalid for multisample\nimage\"), so this path is not reachable for MS resources at all.\n\n* Fix various test suite fallouts.\n\nCo-authored-by: Hans-Kristian Arntzen \u003cpost@arntzen-software.no\u003e"
    },
    {
      "commit": "88960ab8600ad90b0914cabf63171b9c95ccc1c9",
      "tree": "7823a980622760214bd33cc477b188157e6e8993",
      "parents": [
        "9c3c8e2cefdd8194b193bb8ed2fdff4d5527e382"
      ],
      "author": {
        "name": "baggins183",
        "email": "baggins31084@proton.me",
        "time": "Mon Aug 31 12:23:21 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Mon Aug 31 12:23:21 2026"
      },
      "message": "Merge pull request #2652 from baggins183/shared_memory_blocks\n\n(GLSL) Support SPV_KHR_workgroup_memory_explicit_layout"
    },
    {
      "commit": "9c3c8e2cefdd8194b193bb8ed2fdff4d5527e382",
      "tree": "8cd8df24c729d52ad600640bfdc405c061004963",
      "parents": [
        "1f152ab88b426c8cb1558096db2e41be9d934fa6",
        "a159fc7a046569fe7b77325e24c6990a870df252"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 13:06:55 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Jul 31 13:06:55 2026"
      },
      "message": "Merge pull request #2661 from KhronosGroup/fix-2653\n\nGLSL: Implement conservative depth properly."
    },
    {
      "commit": "a159fc7a046569fe7b77325e24c6990a870df252",
      "tree": "8cd8df24c729d52ad600640bfdc405c061004963",
      "parents": [
        "1f152ab88b426c8cb1558096db2e41be9d934fa6"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:40:45 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:59:37 2026"
      },
      "message": "GLSL: Implement conservative depth properly.\n"
    },
    {
      "commit": "1f152ab88b426c8cb1558096db2e41be9d934fa6",
      "tree": "3d9fd0206401bc0a52734afd0baa4d6abf914855",
      "parents": [
        "54bb613d9c47d05853f2a09319b783f9f225acb4",
        "3b14d1c66bb4b4b9669716fc66f5d8e44821e122"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:44:36 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Jul 31 12:44:36 2026"
      },
      "message": "Merge pull request #2660 from KhronosGroup/fix-2653\n\nHLSL: Don\u0027t declare undef block types."
    },
    {
      "commit": "54bb613d9c47d05853f2a09319b783f9f225acb4",
      "tree": "2aaafd6259360035c063319e49d486fb74bddee7",
      "parents": [
        "5ff3bea0136c5cdc5ae133d5f751e492e6a55f5c",
        "090bb6b7d6fad49f04327d4b54c16ad5c697a345"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:44:25 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Jul 31 12:44:25 2026"
      },
      "message": "Merge pull request #2659 from KhronosGroup/fix-2651\n\nNever forward synthesized BDA loads."
    },
    {
      "commit": "3b14d1c66bb4b4b9669716fc66f5d8e44821e122",
      "tree": "7ff00e09cd10a2664e3ece1315785d9d4a573724",
      "parents": [
        "5ff3bea0136c5cdc5ae133d5f751e492e6a55f5c"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:24:30 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:24:30 2026"
      },
      "message": "HLSL: Don\u0027t declare undef block types.\n\nOther backends deal with this in other ways.\n"
    },
    {
      "commit": "090bb6b7d6fad49f04327d4b54c16ad5c697a345",
      "tree": "2aaafd6259360035c063319e49d486fb74bddee7",
      "parents": [
        "5ff3bea0136c5cdc5ae133d5f751e492e6a55f5c"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 11:55:12 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 12:10:24 2026"
      },
      "message": "Never forward synthesized BDA loads.\n"
    },
    {
      "commit": "5ff3bea0136c5cdc5ae133d5f751e492e6a55f5c",
      "tree": "fcead151c4b8c066d99467d643f528851cdd719c",
      "parents": [
        "6c09849fe88c48eaed08413aa022aaa136a3a057",
        "40b96db2ee1691d89ce5b1dd243ac8bed5f9876d"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 10:54:49 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Fri Jul 31 10:54:49 2026"
      },
      "message": "Merge pull request #2657 from GustavoSilvera/fix-MSL-matmul\n\nFix spvFMulMatrixMatrix type mismatch for non-square matrix multiply"
    },
    {
      "commit": "40b96db2ee1691d89ce5b1dd243ac8bed5f9876d",
      "tree": "fcead151c4b8c066d99467d643f528851cdd719c",
      "parents": [
        "7a8061390241c15126319c40f6f547c78d395187"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 10:39:57 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Fri Jul 31 10:41:57 2026"
      },
      "message": "MSL: Add some test coverage for non-square precise matmul.\n"
    },
    {
      "commit": "7a8061390241c15126319c40f6f547c78d395187",
      "tree": "b70eb01df2cb27188f924802acc5c7892a654d8f",
      "parents": [
        "6c09849fe88c48eaed08413aa022aaa136a3a057"
      ],
      "author": {
        "name": "Gustavo Silvera",
        "email": "grs5382@gmail.com",
        "time": "Sun Jul 26 05:45:01 2026"
      },
      "committer": {
        "name": "Gustavo Silvera",
        "email": "grs5382@gmail.com",
        "time": "Sun Jul 26 05:45:04 2026"
      },
      "message": "Fix spvFMulMatrixMatrix type mismatch for non-square matrix multiply\n"
    },
    {
      "commit": "9538032a7985418f7e31988a8cffbfa00c3ae1ab",
      "tree": "3d66e1754595584130ad3b038ec6476d51866764",
      "parents": [
        "e8f95f5e69f6ced30717b1dd1c3ec39a50f92973"
      ],
      "author": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sun Jul 19 20:54:42 2026"
      },
      "committer": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sun Jul 19 20:54:42 2026"
      },
      "message": "MSL: Hardcode coop-mat dimensions and scope with invalid warning\n"
    },
    {
      "commit": "e8f95f5e69f6ced30717b1dd1c3ec39a50f92973",
      "tree": "0e8f246328d56a1d1a6272b7ad2c3127072df6af",
      "parents": [
        "07e26094a72ea230dcb4ac716e842677e926d65c"
      ],
      "author": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sat Jul 18 17:13:07 2026"
      },
      "committer": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Sat Jul 18 17:13:07 2026"
      },
      "message": "MSL: Fix cooperative matrix validation priority\n"
    },
    {
      "commit": "07e26094a72ea230dcb4ac716e842677e926d65c",
      "tree": "02c77939826f33bbb65a0324ca7bf2236ba6c0c9",
      "parents": [
        "a432b60822a6c5fda982dd50d1908643af516418"
      ],
      "author": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Fri Jul 10 20:40:48 2026"
      },
      "committer": {
        "name": "Zain Sharief",
        "email": "zain.sharief21@gmail.com",
        "time": "Fri Jul 10 20:40:48 2026"
      },
      "message": "MSL: Use unified fixed-size threadgroup buffer for coop-mat emulation\n"
    },
    {
      "commit": "6c09849fe88c48eaed08413aa022aaa136a3a057",
      "tree": "3477246459ac20f083011431fe861fe6e1ae0cb2",
      "parents": [
        "32cf9a45d6062d01268b0108aac26b1c8951a3df",
        "06b95c405df81ea4ae5b3d135961a47e0c29cad8"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Jul 06 12:43:32 2026"
      },
      "committer": {
        "name": "GitHub",
        "email": "noreply@github.com",
        "time": "Mon Jul 06 12:43:32 2026"
      },
      "message": "Merge pull request #2647 from KhronosGroup/fix-2642\n\nHLSL: Support baseline SM 6.4 idot."
    },
    {
      "commit": "06b95c405df81ea4ae5b3d135961a47e0c29cad8",
      "tree": "3477246459ac20f083011431fe861fe6e1ae0cb2",
      "parents": [
        "32cf9a45d6062d01268b0108aac26b1c8951a3df"
      ],
      "author": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Jul 06 12:32:27 2026"
      },
      "committer": {
        "name": "Hans-Kristian Arntzen",
        "email": "post@arntzen-software.no",
        "time": "Mon Jul 06 12:33:16 2026"
      },
      "message": "HLSL: Support baseline SM 6.4 idot.\n"
    }
  ],
  "next": "32cf9a45d6062d01268b0108aac26b1c8951a3df"
}
