video_core: flush GPU-written code when capturing native shaders

10e7063905 moved native program capture from one ReadBlock per instruction to
ReadBlockUnsafe per 4 KiB page. ReadBlockUnsafe skips the rasterizer flush, so
code that a GPU copy had written and not yet flushed to guest memory was
captured stale. The speed-up comes from reading a page at a time, not from
skipping the flush: read each page (and the odd single instruction outside a
fully mapped page) with ReadBlock, which flushes once per page instead of once
per instruction.

Follow-up to 10e7063905.

Assisted-by: Claude Sonnet 5.5 (implementation and commit message)

1 files changed, 5 insertions, 5 deletions

diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp
index 8d340ba881..0c7d8a0d42 100644
--- a/src/video_core/shader_environment.cpp
+++ b/src/video_core/shader_environment.cpp
@@ -328,9 +328,9 @@ Tegra::Texture::TICEntry GenericEnvironment::ReadTextureInfo(GPUVAddr tic_addr,
 namespace {
 
 // Reads a native program a 4 KiB page at a time instead of one 16-byte instruction at a time.
-// The reads do not flush the GPU caches first (ReadBlockUnsafe): the previous per-instruction
-// ReadBlock did, once per instruction. An instruction that is not inside a fully mapped page is
-// left to the caller.
+// Each page is read with ReadBlock, so GPU-written code is flushed first, once per page instead
+// of once per instruction. An instruction that is not inside a fully mapped page is left to the
+// caller.
 class NativeInstructionReader {
 public:
     using Instruction = Shader::Ampere::Instruction;
@@ -365,7 +365,7 @@ private:
         page_begin = address;
         page_end = address;
         if (length >= sizeof(Instruction) && gpu_memory.IsFullyMappedRange(address, length)) {
-            gpu_memory.ReadBlockUnsafe(address, page.data(), length);
+            gpu_memory.ReadBlock(address, page.data(), length);
             page_end = next_page;
         }
     }
@@ -415,7 +415,7 @@ std::shared_ptr<const LoweredNativeCode> GenericEnvironment::LowerNativeCode(u32
                 if (!native.empty()) break;
                 throw Shader::RuntimeError("Unmapped native shader instruction at {:#x}", address);
             }
-            gpu_memory->ReadBlockUnsafe(address, &instruction, sizeof(instruction));
+            gpu_memory->ReadBlock(address, &instruction, sizeof(instruction));
         }
         native.push_back(instruction);
         const auto status = capture.Observe(instruction, index);