diff options
Diffstat (limited to '')
-rw-r--r-- | src/common/cityhash.cpp | 178 | ||||
-rw-r--r-- | src/common/cityhash.h | 34 | ||||
-rw-r--r-- | src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp | 3 | ||||
-rw-r--r-- | src/core/hle/service/nvdrv/devices/nvhost_vic.cpp | 7 | ||||
-rw-r--r-- | src/input_common/sdl/sdl_impl.cpp | 7 | ||||
-rw-r--r-- | src/tests/CMakeLists.txt | 1 | ||||
-rw-r--r-- | src/tests/common/cityhash.cpp | 22 | ||||
-rw-r--r-- | src/video_core/cdma_pusher.cpp | 63 | ||||
-rw-r--r-- | src/video_core/cdma_pusher.h | 33 | ||||
-rw-r--r-- | src/video_core/command_classes/codecs/codec.cpp | 7 | ||||
-rw-r--r-- | src/video_core/command_classes/nvdec.cpp | 8 | ||||
-rw-r--r-- | src/video_core/command_classes/nvdec.h | 2 | ||||
-rw-r--r-- | src/video_core/command_classes/vic.cpp | 45 | ||||
-rw-r--r-- | src/video_core/command_classes/vic.h | 51 | ||||
-rw-r--r-- | src/video_core/gpu.cpp | 6 | ||||
-rw-r--r-- | src/video_core/gpu_thread.cpp | 3 |
16 files changed, 200 insertions, 270 deletions
diff --git a/src/common/cityhash.cpp b/src/common/cityhash.cpp index 4e1d874b5..66218fc21 100644 --- a/src/common/cityhash.cpp +++ b/src/common/cityhash.cpp @@ -28,8 +28,10 @@ // compromising on hash quality. #include <algorithm> -#include <string.h> // for memcpy and memset -#include "cityhash.h" +#include <cstring> +#include <utility> + +#include "common/cityhash.h" #include "common/swap.h" // #include "config.h" @@ -42,21 +44,17 @@ using namespace std; -typedef uint8_t uint8; -typedef uint32_t uint32; -typedef uint64_t uint64; - namespace Common { -static uint64 UNALIGNED_LOAD64(const char* p) { - uint64 result; - memcpy(&result, p, sizeof(result)); +static u64 unaligned_load64(const char* p) { + u64 result; + std::memcpy(&result, p, sizeof(result)); return result; } -static uint32 UNALIGNED_LOAD32(const char* p) { - uint32 result; - memcpy(&result, p, sizeof(result)); +static u32 unaligned_load32(const char* p) { + u32 result; + std::memcpy(&result, p, sizeof(result)); return result; } @@ -76,64 +74,64 @@ static uint32 UNALIGNED_LOAD32(const char* p) { #endif #endif -static uint64 Fetch64(const char* p) { - return uint64_in_expected_order(UNALIGNED_LOAD64(p)); +static u64 Fetch64(const char* p) { + return uint64_in_expected_order(unaligned_load64(p)); } -static uint32 Fetch32(const char* p) { - return uint32_in_expected_order(UNALIGNED_LOAD32(p)); +static u32 Fetch32(const char* p) { + return uint32_in_expected_order(unaligned_load32(p)); } // Some primes between 2^63 and 2^64 for various uses. -static const uint64 k0 = 0xc3a5c85c97cb3127ULL; -static const uint64 k1 = 0xb492b66fbe98f273ULL; -static const uint64 k2 = 0x9ae16a3b2f90404fULL; +static constexpr u64 k0 = 0xc3a5c85c97cb3127ULL; +static constexpr u64 k1 = 0xb492b66fbe98f273ULL; +static constexpr u64 k2 = 0x9ae16a3b2f90404fULL; // Bitwise right rotate. Normally this will compile to a single // instruction, especially if the shift is a manifest constant. -static uint64 Rotate(uint64 val, int shift) { +static u64 Rotate(u64 val, int shift) { // Avoid shifting by 64: doing so yields an undefined result. return shift == 0 ? val : ((val >> shift) | (val << (64 - shift))); } -static uint64 ShiftMix(uint64 val) { +static u64 ShiftMix(u64 val) { return val ^ (val >> 47); } -static uint64 HashLen16(uint64 u, uint64 v) { - return Hash128to64(uint128(u, v)); +static u64 HashLen16(u64 u, u64 v) { + return Hash128to64(u128{u, v}); } -static uint64 HashLen16(uint64 u, uint64 v, uint64 mul) { +static u64 HashLen16(u64 u, u64 v, u64 mul) { // Murmur-inspired hashing. - uint64 a = (u ^ v) * mul; + u64 a = (u ^ v) * mul; a ^= (a >> 47); - uint64 b = (v ^ a) * mul; + u64 b = (v ^ a) * mul; b ^= (b >> 47); b *= mul; return b; } -static uint64 HashLen0to16(const char* s, std::size_t len) { +static u64 HashLen0to16(const char* s, size_t len) { if (len >= 8) { - uint64 mul = k2 + len * 2; - uint64 a = Fetch64(s) + k2; - uint64 b = Fetch64(s + len - 8); - uint64 c = Rotate(b, 37) * mul + a; - uint64 d = (Rotate(a, 25) + b) * mul; + u64 mul = k2 + len * 2; + u64 a = Fetch64(s) + k2; + u64 b = Fetch64(s + len - 8); + u64 c = Rotate(b, 37) * mul + a; + u64 d = (Rotate(a, 25) + b) * mul; return HashLen16(c, d, mul); } if (len >= 4) { - uint64 mul = k2 + len * 2; - uint64 a = Fetch32(s); + u64 mul = k2 + len * 2; + u64 a = Fetch32(s); return HashLen16(len + (a << 3), Fetch32(s + len - 4), mul); } if (len > 0) { - uint8 a = s[0]; - uint8 b = s[len >> 1]; - uint8 c = s[len - 1]; - uint32 y = static_cast<uint32>(a) + (static_cast<uint32>(b) << 8); - uint32 z = static_cast<uint32>(len) + (static_cast<uint32>(c) << 2); + u8 a = s[0]; + u8 b = s[len >> 1]; + u8 c = s[len - 1]; + u32 y = static_cast<u32>(a) + (static_cast<u32>(b) << 8); + u32 z = static_cast<u32>(len) + (static_cast<u32>(c) << 2); return ShiftMix(y * k2 ^ z * k0) * k2; } return k2; @@ -141,22 +139,21 @@ static uint64 HashLen0to16(const char* s, std::size_t len) { // This probably works well for 16-byte strings as well, but it may be overkill // in that case. -static uint64 HashLen17to32(const char* s, std::size_t len) { - uint64 mul = k2 + len * 2; - uint64 a = Fetch64(s) * k1; - uint64 b = Fetch64(s + 8); - uint64 c = Fetch64(s + len - 8) * mul; - uint64 d = Fetch64(s + len - 16) * k2; +static u64 HashLen17to32(const char* s, size_t len) { + u64 mul = k2 + len * 2; + u64 a = Fetch64(s) * k1; + u64 b = Fetch64(s + 8); + u64 c = Fetch64(s + len - 8) * mul; + u64 d = Fetch64(s + len - 16) * k2; return HashLen16(Rotate(a + b, 43) + Rotate(c, 30) + d, a + Rotate(b + k2, 18) + c, mul); } // Return a 16-byte hash for 48 bytes. Quick and dirty. // Callers do best to use "random-looking" values for a and b. -static pair<uint64, uint64> WeakHashLen32WithSeeds(uint64 w, uint64 x, uint64 y, uint64 z, uint64 a, - uint64 b) { +static pair<u64, u64> WeakHashLen32WithSeeds(u64 w, u64 x, u64 y, u64 z, u64 a, u64 b) { a += w; b = Rotate(b + a + z, 21); - uint64 c = a; + u64 c = a; a += x; a += y; b += Rotate(a, 44); @@ -164,34 +161,34 @@ static pair<uint64, uint64> WeakHashLen32WithSeeds(uint64 w, uint64 x, uint64 y, } // Return a 16-byte hash for s[0] ... s[31], a, and b. Quick and dirty. -static pair<uint64, uint64> WeakHashLen32WithSeeds(const char* s, uint64 a, uint64 b) { +static pair<u64, u64> WeakHashLen32WithSeeds(const char* s, u64 a, u64 b) { return WeakHashLen32WithSeeds(Fetch64(s), Fetch64(s + 8), Fetch64(s + 16), Fetch64(s + 24), a, b); } // Return an 8-byte hash for 33 to 64 bytes. -static uint64 HashLen33to64(const char* s, std::size_t len) { - uint64 mul = k2 + len * 2; - uint64 a = Fetch64(s) * k2; - uint64 b = Fetch64(s + 8); - uint64 c = Fetch64(s + len - 24); - uint64 d = Fetch64(s + len - 32); - uint64 e = Fetch64(s + 16) * k2; - uint64 f = Fetch64(s + 24) * 9; - uint64 g = Fetch64(s + len - 8); - uint64 h = Fetch64(s + len - 16) * mul; - uint64 u = Rotate(a + g, 43) + (Rotate(b, 30) + c) * 9; - uint64 v = ((a + g) ^ d) + f + 1; - uint64 w = swap64((u + v) * mul) + h; - uint64 x = Rotate(e + f, 42) + c; - uint64 y = (swap64((v + w) * mul) + g) * mul; - uint64 z = e + f + c; +static u64 HashLen33to64(const char* s, size_t len) { + u64 mul = k2 + len * 2; + u64 a = Fetch64(s) * k2; + u64 b = Fetch64(s + 8); + u64 c = Fetch64(s + len - 24); + u64 d = Fetch64(s + len - 32); + u64 e = Fetch64(s + 16) * k2; + u64 f = Fetch64(s + 24) * 9; + u64 g = Fetch64(s + len - 8); + u64 h = Fetch64(s + len - 16) * mul; + u64 u = Rotate(a + g, 43) + (Rotate(b, 30) + c) * 9; + u64 v = ((a + g) ^ d) + f + 1; + u64 w = swap64((u + v) * mul) + h; + u64 x = Rotate(e + f, 42) + c; + u64 y = (swap64((v + w) * mul) + g) * mul; + u64 z = e + f + c; a = swap64((x + z) * mul + y) + b; b = ShiftMix((z + a) * mul + d + h) * mul; return b + x; } -uint64 CityHash64(const char* s, std::size_t len) { +u64 CityHash64(const char* s, size_t len) { if (len <= 32) { if (len <= 16) { return HashLen0to16(s, len); @@ -204,15 +201,15 @@ uint64 CityHash64(const char* s, std::size_t len) { // For strings over 64 bytes we hash the end first, and then as we // loop we keep 56 bytes of state: v, w, x, y, and z. - uint64 x = Fetch64(s + len - 40); - uint64 y = Fetch64(s + len - 16) + Fetch64(s + len - 56); - uint64 z = HashLen16(Fetch64(s + len - 48) + len, Fetch64(s + len - 24)); - pair<uint64, uint64> v = WeakHashLen32WithSeeds(s + len - 64, len, z); - pair<uint64, uint64> w = WeakHashLen32WithSeeds(s + len - 32, y + k1, x); + u64 x = Fetch64(s + len - 40); + u64 y = Fetch64(s + len - 16) + Fetch64(s + len - 56); + u64 z = HashLen16(Fetch64(s + len - 48) + len, Fetch64(s + len - 24)); + pair<u64, u64> v = WeakHashLen32WithSeeds(s + len - 64, len, z); + pair<u64, u64> w = WeakHashLen32WithSeeds(s + len - 32, y + k1, x); x = x * k1 + Fetch64(s); // Decrease len to the nearest multiple of 64, and operate on 64-byte chunks. - len = (len - 1) & ~static_cast<std::size_t>(63); + len = (len - 1) & ~static_cast<size_t>(63); do { x = Rotate(x + y + v.first + Fetch64(s + 8), 37) * k1; y = Rotate(y + v.second + Fetch64(s + 48), 42) * k1; @@ -229,21 +226,21 @@ uint64 CityHash64(const char* s, std::size_t len) { HashLen16(v.second, w.second) + x); } -uint64 CityHash64WithSeed(const char* s, std::size_t len, uint64 seed) { +u64 CityHash64WithSeed(const char* s, size_t len, u64 seed) { return CityHash64WithSeeds(s, len, k2, seed); } -uint64 CityHash64WithSeeds(const char* s, std::size_t len, uint64 seed0, uint64 seed1) { +u64 CityHash64WithSeeds(const char* s, size_t len, u64 seed0, u64 seed1) { return HashLen16(CityHash64(s, len) - seed0, seed1); } // A subroutine for CityHash128(). Returns a decent 128-bit hash for strings // of any length representable in signed long. Based on City and Murmur. -static uint128 CityMurmur(const char* s, std::size_t len, uint128 seed) { - uint64 a = Uint128Low64(seed); - uint64 b = Uint128High64(seed); - uint64 c = 0; - uint64 d = 0; +static u128 CityMurmur(const char* s, size_t len, u128 seed) { + u64 a = seed[0]; + u64 b = seed[1]; + u64 c = 0; + u64 d = 0; signed long l = static_cast<long>(len) - 16; if (l <= 0) { // len <= 16 a = ShiftMix(a * k1) * k1; @@ -266,20 +263,20 @@ static uint128 CityMurmur(const char* s, std::size_t len, uint128 seed) { } a = HashLen16(a, c); b = HashLen16(d, b); - return uint128(a ^ b, HashLen16(b, a)); + return u128{a ^ b, HashLen16(b, a)}; } -uint128 CityHash128WithSeed(const char* s, std::size_t len, uint128 seed) { +u128 CityHash128WithSeed(const char* s, size_t len, u128 seed) { if (len < 128) { return CityMurmur(s, len, seed); } // We expect len >= 128 to be the common case. Keep 56 bytes of state: // v, w, x, y, and z. - pair<uint64, uint64> v, w; - uint64 x = Uint128Low64(seed); - uint64 y = Uint128High64(seed); - uint64 z = len * k1; + pair<u64, u64> v, w; + u64 x = seed[0]; + u64 y = seed[1]; + u64 z = len * k1; v.first = Rotate(y ^ k1, 49) * k1 + Fetch64(s); v.second = Rotate(v.first, 42) * k1 + Fetch64(s + 8); w.first = Rotate(y + z, 35) * k1 + x; @@ -313,7 +310,7 @@ uint128 CityHash128WithSeed(const char* s, std::size_t len, uint128 seed) { w.first *= 9; v.first *= k0; // If 0 < len < 128, hash up to 4 chunks of 32 bytes each from the end of s. - for (std::size_t tail_done = 0; tail_done < len;) { + for (size_t tail_done = 0; tail_done < len;) { tail_done += 32; y = Rotate(x + y, 42) * k0 + v.second; w.first += Fetch64(s + len - tail_done + 16); @@ -328,13 +325,12 @@ uint128 CityHash128WithSeed(const char* s, std::size_t len, uint128 seed) { // different 56-byte-to-8-byte hashes to get a 16-byte final result. x = HashLen16(x, v.first); y = HashLen16(y + z, w.first); - return uint128(HashLen16(x + v.second, w.second) + y, HashLen16(x + w.second, y + v.second)); + return u128{HashLen16(x + v.second, w.second) + y, HashLen16(x + w.second, y + v.second)}; } -uint128 CityHash128(const char* s, std::size_t len) { - return len >= 16 - ? CityHash128WithSeed(s + 16, len - 16, uint128(Fetch64(s), Fetch64(s + 8) + k0)) - : CityHash128WithSeed(s, len, uint128(k0, k1)); +u128 CityHash128(const char* s, size_t len) { + return len >= 16 ? CityHash128WithSeed(s + 16, len - 16, u128{Fetch64(s), Fetch64(s + 8) + k0}) + : CityHash128WithSeed(s, len, u128{k0, k1}); } } // namespace Common diff --git a/src/common/cityhash.h b/src/common/cityhash.h index a00804e01..022d0f7cb 100644 --- a/src/common/cityhash.h +++ b/src/common/cityhash.h @@ -61,50 +61,38 @@ #pragma once -#include <cstddef> -#include <cstdint> -#include <utility> +#include "common/common_types.h" namespace Common { -using uint128 = std::pair<uint64_t, uint64_t>; - -[[nodiscard]] inline uint64_t Uint128Low64(const uint128& x) { - return x.first; -} -[[nodiscard]] inline uint64_t Uint128High64(const uint128& x) { - return x.second; -} - // Hash function for a byte array. -[[nodiscard]] uint64_t CityHash64(const char* buf, std::size_t len); +[[nodiscard]] u64 CityHash64(const char* buf, size_t len); // Hash function for a byte array. For convenience, a 64-bit seed is also // hashed into the result. -[[nodiscard]] uint64_t CityHash64WithSeed(const char* buf, std::size_t len, uint64_t seed); +[[nodiscard]] u64 CityHash64WithSeed(const char* buf, size_t len, u64 seed); // Hash function for a byte array. For convenience, two seeds are also // hashed into the result. -[[nodiscard]] uint64_t CityHash64WithSeeds(const char* buf, std::size_t len, uint64_t seed0, - uint64_t seed1); +[[nodiscard]] u64 CityHash64WithSeeds(const char* buf, size_t len, u64 seed0, u64 seed1); // Hash function for a byte array. -[[nodiscard]] uint128 CityHash128(const char* s, std::size_t len); +[[nodiscard]] u128 CityHash128(const char* s, size_t len); // Hash function for a byte array. For convenience, a 128-bit seed is also // hashed into the result. -[[nodiscard]] uint128 CityHash128WithSeed(const char* s, std::size_t len, uint128 seed); +[[nodiscard]] u128 CityHash128WithSeed(const char* s, size_t len, u128 seed); // Hash 128 input bits down to 64 bits of output. // This is intended to be a reasonably good hash function. -[[nodiscard]] inline uint64_t Hash128to64(const uint128& x) { +[[nodiscard]] inline u64 Hash128to64(const u128& x) { // Murmur-inspired hashing. - const uint64_t kMul = 0x9ddfea08eb382d69ULL; - uint64_t a = (Uint128Low64(x) ^ Uint128High64(x)) * kMul; + const u64 mul = 0x9ddfea08eb382d69ULL; + u64 a = (x[0] ^ x[1]) * mul; a ^= (a >> 47); - uint64_t b = (Uint128High64(x) ^ a) * kMul; + u64 b = (x[1] ^ a) * mul; b ^= (b >> 47); - b *= kMul; + b *= mul; return b; } diff --git a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp index 36970f828..ecba1dba1 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_nvdec.cpp @@ -34,8 +34,7 @@ NvResult nvhost_nvdec::Ioctl1(Ioctl command, const std::vector<u8>& input, case 0xa: { if (command.length == 0x1c) { LOG_INFO(Service_NVDRV, "NVDEC video stream ended"); - Tegra::ChCommandHeaderList cmdlist(1); - cmdlist[0] = Tegra::ChCommandHeader{0xDEADB33F}; + Tegra::ChCommandHeaderList cmdlist{{0xDEADB33F}}; system.GPU().PushCommandBuffer(cmdlist); } return UnmapBuffer(input, output); diff --git a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp index 72499654c..70849a9bd 100644 --- a/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp +++ b/src/core/hle/service/nvdrv/devices/nvhost_vic.cpp @@ -28,8 +28,13 @@ NvResult nvhost_vic::Ioctl1(Ioctl command, const std::vector<u8>& input, std::ve return GetWaitbase(input, output); case 0x9: return MapBuffer(input, output); - case 0xa: + case 0xa: { + if (command.length == 0x1c) { + Tegra::ChCommandHeaderList cmdlist{{0xDEADB33F}}; + system.GPU().PushCommandBuffer(cmdlist); + } return UnmapBuffer(input, output); + } default: break; } diff --git a/src/input_common/sdl/sdl_impl.cpp b/src/input_common/sdl/sdl_impl.cpp index a88ae452f..f67de37e3 100644 --- a/src/input_common/sdl/sdl_impl.cpp +++ b/src/input_common/sdl/sdl_impl.cpp @@ -717,13 +717,6 @@ SDLState::SDLState() { if (SDL_SetHint(SDL_HINT_JOYSTICK_ALLOW_BACKGROUND_EVENTS, "1") == SDL_FALSE) { LOG_ERROR(Input, "Failed to set hint for background events with: {}", SDL_GetError()); } -// these hints are only defined on sdl2.0.9 or higher -#if SDL_VERSION_ATLEAST(2, 0, 9) -#if !SDL_VERSION_ATLEAST(2, 0, 12) - // There are also hints to toggle the individual drivers if needed. - SDL_SetHint(SDL_HINT_JOYSTICK_HIDAPI, "0"); -#endif -#endif SDL_AddEventWatch(&SDLEventWatcher, this); diff --git a/src/tests/CMakeLists.txt b/src/tests/CMakeLists.txt index 6a5c18945..4ea0076e9 100644 --- a/src/tests/CMakeLists.txt +++ b/src/tests/CMakeLists.txt @@ -1,5 +1,6 @@ add_executable(tests common/bit_field.cpp + common/cityhash.cpp common/fibers.cpp common/param_package.cpp common/ring_buffer.cpp diff --git a/src/tests/common/cityhash.cpp b/src/tests/common/cityhash.cpp new file mode 100644 index 000000000..7a40b6c4a --- /dev/null +++ b/src/tests/common/cityhash.cpp @@ -0,0 +1,22 @@ +// Copyright 2021 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <catch2/catch.hpp> + +#include "common/cityhash.h" + +constexpr char msg[] = "The blue frogs are singing under the crimson sky.\n" + "It is time to run, Robert."; + +using namespace Common; + +TEST_CASE("CityHash", "[common]") { + // These test results were built against a known good version. + REQUIRE(CityHash64(msg, sizeof(msg)) == 0x92d5c2e9cbfbbc01); + REQUIRE(CityHash64WithSeed(msg, sizeof(msg), 0xdead) == 0xbfbe93f21a2820dd); + REQUIRE(CityHash64WithSeeds(msg, sizeof(msg), 0xbeef, 0xcafe) == 0xb343317955fc8a06); + REQUIRE(CityHash128(msg, sizeof(msg)) == u128{0x98e60d0423747eaa, 0xd8694c5b6fcaede9}); + REQUIRE(CityHash128WithSeed(msg, sizeof(msg), {0xdead, 0xbeef}) == + u128{0xf0307dba81199ebe, 0xd77764e0c4a9eb74}); +} diff --git a/src/video_core/cdma_pusher.cpp b/src/video_core/cdma_pusher.cpp index 33b3c060b..a3fda1094 100644 --- a/src/video_core/cdma_pusher.cpp +++ b/src/video_core/cdma_pusher.cpp @@ -37,59 +37,43 @@ CDmaPusher::CDmaPusher(GPU& gpu_) CDmaPusher::~CDmaPusher() = default; -void CDmaPusher::Push(ChCommandHeaderList&& entries) { - cdma_queue.push(std::move(entries)); -} - -void CDmaPusher::DispatchCalls() { - while (!cdma_queue.empty()) { - Step(); - } -} - -void CDmaPusher::Step() { - const auto entries{cdma_queue.front()}; - cdma_queue.pop(); - - std::vector<u32> values(entries.size()); - std::memcpy(values.data(), entries.data(), entries.size() * sizeof(u32)); - - for (const u32 value : values) { +void CDmaPusher::ProcessEntries(ChCommandHeaderList&& entries) { + for (const auto& value : entries) { if (mask != 0) { const auto lbs = static_cast<u32>(std::countr_zero(mask)); mask &= ~(1U << lbs); - ExecuteCommand(static_cast<u32>(offset + lbs), value); + ExecuteCommand(offset + lbs, value.raw); continue; } else if (count != 0) { --count; - ExecuteCommand(static_cast<u32>(offset), value); + ExecuteCommand(offset, value.raw); if (incrementing) { ++offset; } continue; } - const auto mode = static_cast<ChSubmissionMode>((value >> 28) & 0xf); + const auto mode = value.submission_mode.Value(); switch (mode) { case ChSubmissionMode::SetClass: { - mask = value & 0x3f; - offset = (value >> 16) & 0xfff; - current_class = static_cast<ChClassId>((value >> 6) & 0x3ff); + mask = value.value & 0x3f; + offset = value.method_offset; + current_class = static_cast<ChClassId>((value.value >> 6) & 0x3ff); break; } case ChSubmissionMode::Incrementing: case ChSubmissionMode::NonIncrementing: - count = value & 0xffff; - offset = (value >> 16) & 0xfff; + count = value.value; + offset = value.method_offset; incrementing = mode == ChSubmissionMode::Incrementing; break; case ChSubmissionMode::Mask: - mask = value & 0xffff; - offset = (value >> 16) & 0xfff; + mask = value.value; + offset = value.method_offset; break; case ChSubmissionMode::Immediate: { - const u32 data = value & 0xfff; - offset = (value >> 16) & 0xfff; - ExecuteCommand(static_cast<u32>(offset), data); + const u32 data = value.value & 0xfff; + offset = value.method_offset; + ExecuteCommand(offset, data); break; } default: @@ -102,8 +86,8 @@ void CDmaPusher::Step() { void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) { switch (current_class) { case ChClassId::NvDec: - ThiStateWrite(nvdec_thi_state, state_offset, {data}); - switch (static_cast<ThiMethod>(state_offset)) { + ThiStateWrite(nvdec_thi_state, offset, data); + switch (static_cast<ThiMethod>(offset)) { case ThiMethod::IncSyncpt: { LOG_DEBUG(Service_NVDRV, "NVDEC Class IncSyncpt Method"); const auto syncpoint_id = static_cast<u32>(data & 0xFF); @@ -120,7 +104,7 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) { LOG_DEBUG(Service_NVDRV, "NVDEC method 0x{:X}", static_cast<u32>(nvdec_thi_state.method_0)); nvdec_processor->ProcessMethod(static_cast<Nvdec::Method>(nvdec_thi_state.method_0), - {data}); + data); break; default: break; @@ -144,7 +128,7 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) { case ThiMethod::SetMethod1: LOG_DEBUG(Service_NVDRV, "VIC method 0x{:X}, Args=({})", static_cast<u32>(vic_thi_state.method_0), data); - vic_processor->ProcessMethod(static_cast<Vic::Method>(vic_thi_state.method_0), {data}); + vic_processor->ProcessMethod(static_cast<Vic::Method>(vic_thi_state.method_0), data); break; default: break; @@ -153,7 +137,7 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) { case ChClassId::Host1x: // This device is mainly for syncpoint synchronization LOG_DEBUG(Service_NVDRV, "Host1X Class Method"); - host1x_processor->ProcessMethod(static_cast<Host1x::Method>(state_offset), {data}); + host1x_processor->ProcessMethod(static_cast<Host1x::Method>(offset), data); break; default: UNIMPLEMENTED_MSG("Current class not implemented {:X}", static_cast<u32>(current_class)); @@ -161,10 +145,9 @@ void CDmaPusher::ExecuteCommand(u32 state_offset, u32 data) { } } -void CDmaPusher::ThiStateWrite(ThiRegisters& state, u32 state_offset, - const std::vector<u32>& arguments) { - u8* const state_offset_ptr = reinterpret_cast<u8*>(&state) + sizeof(u32) * state_offset; - std::memcpy(state_offset_ptr, arguments.data(), sizeof(u32) * arguments.size()); +void CDmaPusher::ThiStateWrite(ThiRegisters& state, u32 state_offset, u32 argument) { + u8* const offset_ptr = reinterpret_cast<u8*>(&state) + sizeof(u32) * state_offset; + std::memcpy(offset_ptr, &argument, sizeof(u32)); } } // namespace Tegra diff --git a/src/video_core/cdma_pusher.h b/src/video_core/cdma_pusher.h index e5f212c1a..1bada44dd 100644 --- a/src/video_core/cdma_pusher.h +++ b/src/video_core/cdma_pusher.h @@ -5,9 +5,7 @@ #pragma once #include <memory> -#include <unordered_map> #include <vector> -#include <queue> #include "common/bit_field.h" #include "common/common_types.h" @@ -16,9 +14,9 @@ namespace Tegra { class GPU; +class Host1x; class Nvdec; class Vic; -class Host1x; enum class ChSubmissionMode : u32 { SetClass = 0, @@ -48,16 +46,10 @@ enum class ChClassId : u32 { NvDec = 0xf0 }; -enum class ChMethod : u32 { - Empty = 0, - SetMethod = 0x10, - SetData = 0x11, -}; - union ChCommandHeader { u32 raw; BitField<0, 16, u32> value; - BitField<16, 12, ChMethod> method_offset; + BitField<16, 12, u32> method_offset; BitField<28, 4, ChSubmissionMode> submission_mode; }; static_assert(sizeof(ChCommandHeader) == sizeof(u32), "ChCommand header is an invalid size"); @@ -99,21 +91,15 @@ public: explicit CDmaPusher(GPU& gpu_); ~CDmaPusher(); - /// Push NVDEC command buffer entries into queue - void Push(ChCommandHeaderList&& entries); - - /// Process queued command buffer entries - void DispatchCalls(); - - /// Process one queue element - void Step(); + /// Process the command entry + void ProcessEntries(ChCommandHeaderList&& entries); +private: /// Invoke command class devices to execute the command based on the current state void ExecuteCommand(u32 state_offset, u32 data); -private: /// Write arguments value to the ThiRegisters member at the specified offset - void ThiStateWrite(ThiRegisters& state, u32 state_offset, const std::vector<u32>& arguments); + void ThiStateWrite(ThiRegisters& state, u32 offset, u32 argument); GPU& gpu; std::shared_ptr<Tegra::Nvdec> nvdec_processor; @@ -124,13 +110,10 @@ private: ThiRegisters vic_thi_state{}; ThiRegisters nvdec_thi_state{}; - s32 count{}; - s32 offset{}; + u32 count{}; + u32 offset{}; u32 mask{}; bool incrementing{}; - - // Queue of command lists to be processed - std::queue<ChCommandHeaderList> cdma_queue; }; } // namespace Tegra diff --git a/src/video_core/command_classes/codecs/codec.cpp b/src/video_core/command_classes/codecs/codec.cpp index 39bc923a5..d02dc6260 100644 --- a/src/video_core/command_classes/codecs/codec.cpp +++ b/src/video_core/command_classes/codecs/codec.cpp @@ -44,8 +44,10 @@ Codec::~Codec() { } void Codec::SetTargetCodec(NvdecCommon::VideoCodec codec) { - LOG_INFO(Service_NVDRV, "NVDEC video codec initialized to {}", codec); - current_codec = codec; + if (current_codec != codec) { + LOG_INFO(Service_NVDRV, "NVDEC video codec initialized to {}", static_cast<u32>(codec)); + current_codec = codec; + } } void Codec::StateWrite(u32 offset, u64 arguments) { @@ -55,7 +57,6 @@ void Codec::StateWrite(u32 offset, u64 arguments) { void Codec::Decode() { bool is_first_frame = false; - if (!initialized) { if (current_codec == NvdecCommon::VideoCodec::H264) { av_codec = avcodec_find_decoder(AV_CODEC_ID_H264); diff --git a/src/video_core/command_classes/nvdec.cpp b/src/video_core/command_classes/nvdec.cpp index 79e1f4e13..e4f919afd 100644 --- a/src/video_core/command_classes/nvdec.cpp +++ b/src/video_core/command_classes/nvdec.cpp @@ -12,16 +12,16 @@ Nvdec::Nvdec(GPU& gpu_) : gpu(gpu_), codec(std::make_unique<Codec>(gpu)) {} Nvdec::~Nvdec() = default; -void Nvdec::ProcessMethod(Method method, const std::vector<u32>& arguments) { +void Nvdec::ProcessMethod(Method method, u32 argument) { if (method == Method::SetVideoCodec) { - codec->StateWrite(static_cast<u32>(method), arguments[0]); + codec->StateWrite(static_cast<u32>(method), argument); } else { - codec->StateWrite(static_cast<u32>(method), static_cast<u64>(arguments[0]) << 8); + codec->StateWrite(static_cast<u32>(method), static_cast<u64>(argument) << 8); } switch (method) { case Method::SetVideoCodec: - codec->SetTargetCodec(static_cast<NvdecCommon::VideoCodec>(arguments[0])); + codec->SetTargetCodec(static_cast<NvdecCommon::VideoCodec>(argument)); break; case Method::Execute: Execute(); diff --git a/src/video_core/command_classes/nvdec.h b/src/video_core/command_classes/nvdec.h index e4877c533..e66be80b8 100644 --- a/src/video_core/command_classes/nvdec.h +++ b/src/video_core/command_classes/nvdec.h @@ -23,7 +23,7 @@ public: ~Nvdec(); /// Writes the method into the state, Invoke Execute() if encountered - void ProcessMethod(Method method, const std::vector<u32>& arguments); + void ProcessMethod(Method method, u32 argument); /// Return most recently decoded frame [[nodiscard]] AVFramePtr GetFrame(); diff --git a/src/video_core/command_classes/vic.cpp b/src/video_core/command_classes/vic.cpp index 2b7569335..0a8b82f2b 100644 --- a/src/video_core/command_classes/vic.cpp +++ b/src/video_core/command_classes/vic.cpp @@ -18,18 +18,14 @@ extern "C" { namespace Tegra { Vic::Vic(GPU& gpu_, std::shared_ptr<Nvdec> nvdec_processor_) - : gpu(gpu_), nvdec_processor(std::move(nvdec_processor_)) {} -Vic::~Vic() = default; + : gpu(gpu_), + nvdec_processor(std::move(nvdec_processor_)), converted_frame_buffer{nullptr, av_free} {} -void Vic::VicStateWrite(u32 offset, u32 arguments) { - u8* const state_offset = reinterpret_cast<u8*>(&vic_state) + offset * sizeof(u32); - std::memcpy(state_offset, &arguments, sizeof(u32)); -} +Vic::~Vic() = default; -void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) { - LOG_DEBUG(HW_GPU, "Vic method 0x{:X}", method); - VicStateWrite(static_cast<u32>(method), arguments[0]); - const u64 arg = static_cast<u64>(arguments[0]) << 8; +void Vic::ProcessMethod(Method method, u32 argument) { + LOG_DEBUG(HW_GPU, "Vic method 0x{:X}", static_cast<u32>(method)); + const u64 arg = static_cast<u64>(argument) << 8; switch (method) { case Method::Execute: Execute(); @@ -53,8 +49,7 @@ void Vic::ProcessMethod(Method method, const std::vector<u32>& arguments) { void Vic::Execute() { if (output_surface_luma_address == 0) { - LOG_ERROR(Service_NVDRV, "VIC Luma address not set. Received 0x{:X}", - vic_state.output_surface.luma_offset); + LOG_ERROR(Service_NVDRV, "VIC Luma address not set."); return; } const VicConfig config{gpu.MemoryManager().Read<u64>(config_struct_address + 0x20)}; @@ -89,8 +84,10 @@ void Vic::Execute() { // Get Converted frame const std::size_t linear_size = frame->width * frame->height * 4; - using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>; - AVMallocPtr converted_frame_buffer{static_cast<u8*>(av_malloc(linear_size)), av_free}; + // Only allocate frame_buffer once per stream, as the size is not expected to change + if (!converted_frame_buffer) { + converted_frame_buffer = AVMallocPtr{static_cast<u8*>(av_malloc(linear_size)), av_free}; + } const int converted_stride{frame->width * 4}; u8* const converted_frame_buf_addr{converted_frame_buffer.get()}; @@ -104,12 +101,12 @@ void Vic::Execute() { const u32 block_height = static_cast<u32>(config.block_linear_height_log2); const auto size = Tegra::Texture::CalculateSize(true, 4, frame->width, frame->height, 1, block_height, 0); - std::vector<u8> swizzled_data(size); + luma_buffer.resize(size); Tegra::Texture::SwizzleSubrect(frame->width, frame->height, frame->width * 4, - frame->width, 4, swizzled_data.data(), + frame->width, 4, luma_buffer.data(), converted_frame_buffer.get(), block_height, 0, 0); - gpu.MemoryManager().WriteBlock(output_surface_luma_address, swizzled_data.data(), size); + gpu.MemoryManager().WriteBlock(output_surface_luma_address, luma_buffer.data(), size); } else { // send pitch linear frame gpu.MemoryManager().WriteBlock(output_surface_luma_address, converted_frame_buf_addr, @@ -132,15 +129,15 @@ void Vic::Execute() { const auto stride = frame->linesize[0]; const auto half_stride = frame->linesize[1]; - std::vector<u8> luma_buffer(aligned_width * surface_height); - std::vector<u8> chroma_buffer(aligned_width * half_height); + luma_buffer.resize(aligned_width * surface_height); + chroma_buffer.resize(aligned_width * half_height); // Populate luma buffer for (std::size_t y = 0; y < surface_height - 1; ++y) { - std::size_t src = y * stride; - std::size_t dst = y * aligned_width; + const std::size_t src = y * stride; + const std::size_t dst = y * aligned_width; - std::size_t size = surface_width; + const std::size_t size = surface_width; for (std::size_t offset = 0; offset < size; ++offset) { luma_buffer[dst + offset] = luma_ptr[src + offset]; @@ -151,8 +148,8 @@ void Vic::Execute() { // Populate chroma buffer from both channels with interleaving. for (std::size_t y = 0; y < half_height; ++y) { - std::size_t src = y * half_stride; - std::size_t dst = y * aligned_width; + const std::size_t src = y * half_stride; + const std::size_t dst = y * aligned_width; for (std::size_t x = 0; x < half_width; ++x) { chroma_buffer[dst + x * 2] = chroma_b_ptr[src + x]; diff --git a/src/video_core/command_classes/vic.h b/src/video_core/command_classes/vic.h index 8c4e284a1..f5a2ed100 100644 --- a/src/video_core/command_classes/vic.h +++ b/src/video_core/command_classes/vic.h @@ -15,43 +15,6 @@ namespace Tegra { class GPU; class Nvdec; -struct PlaneOffsets { - u32 luma_offset{}; - u32 chroma_u_offset{}; - u32 chroma_v_offset{}; -}; - -struct VicRegisters { - INSERT_PADDING_WORDS(64); - u32 nop{}; - INSERT_PADDING_WORDS(15); - u32 pm_trigger{}; - INSERT_PADDING_WORDS(47); - u32 set_application_id{}; - u32 set_watchdog_timer{}; - INSERT_PADDING_WORDS(17); - u32 context_save_area{}; - u32 context_switch{}; - INSERT_PADDING_WORDS(43); - u32 execute{}; - INSERT_PADDING_WORDS(63); - std::array<std::array<PlaneOffsets, 8>, 8> surfacex_slots{}; - u32 picture_index{}; - u32 control_params{}; - u32 config_struct_offset{}; - u32 filter_struct_offset{}; - u32 palette_offset{}; - u32 hist_offset{}; - u32 context_id{}; - u32 fce_ucode_size{}; - PlaneOffsets output_surface{}; - u32 fce_ucode_offset{}; - INSERT_PADDING_WORDS(4); - std::array<u32, 8> slot_context_id{}; - INSERT_PADDING_WORDS(16); -}; -static_assert(sizeof(VicRegisters) == 0x7A0, "VicRegisters is an invalid size"); - class Vic { public: enum class Method : u32 { @@ -67,14 +30,11 @@ public: ~Vic(); /// Write to the device state. - void ProcessMethod(Method method, const std::vector<u32>& arguments); + void ProcessMethod(Method method, u32 argument); private: void Execute(); - void VicStateWrite(u32 offset, u32 arguments); - VicRegisters vic_state{}; - enum class VideoPixelFormat : u64_le { RGBA8 = 0x1f, BGRA8 = 0x20, @@ -88,8 +48,6 @@ private: BitField<9, 2, u64_le> chroma_loc_vert; BitField<11, 4, u64_le> block_linear_kind; BitField<15, 4, u64_le> block_linear_height_log2; - BitField<19, 3, u64_le> reserved0; - BitField<22, 10, u64_le> reserved1; BitField<32, 14, u64_le> surface_width_minus1; BitField<46, 14, u64_le> surface_height_minus1; }; @@ -97,6 +55,13 @@ private: GPU& gpu; std::shared_ptr<Tegra::Nvdec> nvdec_processor; + /// Avoid reallocation of the following buffers every frame, as their + /// size does not change during a stream + using AVMallocPtr = std::unique_ptr<u8, decltype(&av_free)>; + AVMallocPtr converted_frame_buffer; + std::vector<u8> luma_buffer; + std::vector<u8> chroma_buffer; + GPUVAddr config_struct_address{}; GPUVAddr output_surface_luma_address{}; GPUVAddr output_surface_chroma_u_address{}; diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp index 2a9bd4121..51c63af4a 100644 --- a/src/video_core/gpu.cpp +++ b/src/video_core/gpu.cpp @@ -30,8 +30,7 @@ MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192)); GPU::GPU(Core::System& system_, bool is_async_, bool use_nvdec_) : system{system_}, memory_manager{std::make_unique<Tegra::MemoryManager>(system)}, - dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)}, - cdma_pusher{std::make_unique<Tegra::CDmaPusher>(*this)}, use_nvdec{use_nvdec_}, + dma_pusher{std::make_unique<Tegra::DmaPusher>(system, *this)}, use_nvdec{use_nvdec_}, maxwell_3d{std::make_unique<Engines::Maxwell3D>(system, *memory_manager)}, fermi_2d{std::make_unique<Engines::Fermi2D>()}, kepler_compute{std::make_unique<Engines::KeplerCompute>(system, *memory_manager)}, @@ -494,8 +493,7 @@ void GPU::PushCommandBuffer(Tegra::ChCommandHeaderList& entries) { // TODO(ameerj): RE proper async nvdec operation // gpu_thread.SubmitCommandBuffer(std::move(entries)); - cdma_pusher->Push(std::move(entries)); - cdma_pusher->DispatchCalls(); + cdma_pusher->ProcessEntries(std::move(entries)); } void GPU::SwapBuffers(const Tegra::FramebufferConfig* framebuffer) { diff --git a/src/video_core/gpu_thread.cpp b/src/video_core/gpu_thread.cpp index 50319f1d5..eb0e43c0c 100644 --- a/src/video_core/gpu_thread.cpp +++ b/src/video_core/gpu_thread.cpp @@ -48,8 +48,7 @@ static void RunThread(Core::System& system, VideoCore::RendererBase& renderer, dma_pusher.DispatchCalls(); } else if (auto* command_list = std::get_if<SubmitChCommandEntries>(&next.data)) { // NVDEC - cdma_pusher.Push(std::move(command_list->entries)); - cdma_pusher.DispatchCalls(); + cdma_pusher.ProcessEntries(std::move(command_list->entries)); } else if (const auto* data = std::get_if<SwapBuffersCommand>(&next.data)) { renderer.SwapBuffers(data->framebuffer ? &*data->framebuffer : nullptr); } else if (std::holds_alternative<OnCommandListEndCommand>(next.data)) { |