From ed2068fb03e6c14b03bc7a379acf008a47da50f3 Mon Sep 17 00:00:00 2001
From: kd-11 <karokidii@gmail.com>
Date: Wed, 25 May 2022 23:49:08 +0300
Subject: [PATCH] gl: Rewrite buffer mapping

---
 rpcs3/Emu/RSX/GL/GLGSRender.cpp |  8 +++-----
 rpcs3/Emu/RSX/GL/GLHelpers.h    | 24 +++++++++++++++---------
 rpcs3/Emu/RSX/GL/GLProcTable.h  |  3 +++
 rpcs3/Emu/RSX/GL/GLTexture.cpp  |  2 +-
 4 files changed, 22 insertions(+), 15 deletions(-)

diff --git a/rpcs3/Emu/RSX/GL/GLGSRender.cpp b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
index db27d87056..83ae9dcf2d 100644
--- a/rpcs3/Emu/RSX/GL/GLGSRender.cpp
+++ b/rpcs3/Emu/RSX/GL/GLGSRender.cpp
@@ -264,17 +264,15 @@ void GLGSRender::on_init_thread()
 
 	if (gl_caps.vendor_AMD)
 	{
-		m_identity_index_buffer = std::make_unique<gl::buffer>();
-		m_identity_index_buffer->create(gl::buffer::target::element_array, 1 * 0x100000, nullptr, gl::buffer::memory_type::host_visible);
-
 		// Initialize with 256k identity entries
-		auto* dst = reinterpret_cast<u32*>(m_identity_index_buffer->map(gl::buffer::access::write));
+		std::vector<u32> dst(256 * 1024);
 		for (u32 n = 0; n < (0x100000 >> 2); ++n)
 		{
 			dst[n] = n;
 		}
 
-		m_identity_index_buffer->unmap();
+		m_identity_index_buffer = std::make_unique<gl::buffer>();
+		m_identity_index_buffer->create(gl::buffer::target::element_array,dst.size() * sizeof(u32), dst.data(), gl::buffer::memory_type::local);
 	}
 	else if (gl_caps.vendor_NVIDIA)
 	{
diff --git a/rpcs3/Emu/RSX/GL/GLHelpers.h b/rpcs3/Emu/RSX/GL/GLHelpers.h
index 797f0fd187..7e314d3a66 100644
--- a/rpcs3/Emu/RSX/GL/GLHelpers.h
+++ b/rpcs3/Emu/RSX/GL/GLHelpers.h
@@ -51,11 +51,16 @@ using namespace ::rsx::format_class_;
 namespace gl
 {
 	//Function call wrapped in ARB_DSA vs EXT_DSA compat check
-#define DSA_CALL(func, texture_name, target, ...)\
+#define DSA_CALL(func, object_name, target, ...)\
 	if (::gl::get_driver_caps().ARB_dsa_supported)\
-		gl##func(texture_name, __VA_ARGS__);\
+		gl##func(object_name, __VA_ARGS__);\
 	else\
-		gl##func##EXT(texture_name, target, __VA_ARGS__);
+		gl##func##EXT(object_name, target, __VA_ARGS__);
+
+#define DSA_CALL2(func, object_name, ...)\
+	(::gl::get_driver_caps().ARB_dsa_supported) ?\
+		gl##func(object_name, __VA_ARGS__) :\
+		gl##func##EXT(object_name, __VA_ARGS__)
 
 	class fence;
 
@@ -520,9 +525,9 @@ namespace gl
 
 		enum class access
 		{
-			read = GL_READ_ONLY,
-			write = GL_WRITE_ONLY,
-			read_write = GL_READ_WRITE
+			read = GL_MAP_READ_BIT,
+			write = GL_MAP_WRITE_BIT,
+			read_write = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT
 		};
 
 		enum class memory_type
@@ -735,12 +740,13 @@ namespace gl
 			m_size = size;
 		}
 
-		GLubyte* map(access access_)
+		GLubyte* map(GLsizeiptr offset, GLsizeiptr length, access access_)
 		{
 			ensure(m_memory_type == memory_type::host_visible);
 
-			bind(current_target());
-			return reinterpret_cast<GLubyte*>(glMapBuffer(static_cast<GLenum>(current_target()), static_cast<GLenum>(access_)));
+			const GLenum access_bits = static_cast<GLenum>(access_);
+			auto raw_data = DSA_CALL2(MapNamedBufferRange, id(), offset, length, access_bits | GL_MAP_UNSYNCHRONIZED_BIT);
+			return reinterpret_cast<GLubyte*>(raw_data);
 		}
 
 		void unmap()
diff --git a/rpcs3/Emu/RSX/GL/GLProcTable.h b/rpcs3/Emu/RSX/GL/GLProcTable.h
index 0a2a1a34eb..65bf950b51 100644
--- a/rpcs3/Emu/RSX/GL/GLProcTable.h
+++ b/rpcs3/Emu/RSX/GL/GLProcTable.h
@@ -174,6 +174,9 @@ OPENGL_PROC(PFNGLMAPBUFFERRANGEPROC, MapBufferRange);
 OPENGL_PROC(PFNGLBINDBUFFERRANGEPROC, BindBufferRange);
 OPENGL_PROC(PFNGLBINDBUFFERBASEPROC, BindBufferBase);
 
+OPENGL_PROC(PFNGLMAPNAMEDBUFFERRANGEPROC, MapNamedBufferRange);
+OPENGL_PROC(PFNGLMAPNAMEDBUFFERRANGEEXTPROC, MapNamedBufferRangeEXT);
+
 OPENGL_PROC(PFNGLMULTIDRAWELEMENTSPROC, MultiDrawElements);
 OPENGL_PROC(PFNGLMULTIDRAWARRAYSPROC, MultiDrawArrays);
 
diff --git a/rpcs3/Emu/RSX/GL/GLTexture.cpp b/rpcs3/Emu/RSX/GL/GLTexture.cpp
index 864b39afc0..4dec07ed88 100644
--- a/rpcs3/Emu/RSX/GL/GLTexture.cpp
+++ b/rpcs3/Emu/RSX/GL/GLTexture.cpp
@@ -706,7 +706,7 @@ namespace gl
 				{
 					const u64 row_pitch = rsx::align2<u64, u64>(layout.width_in_block * block_size_in_bytes, caps.alignment);
 					image_linear_size = row_pitch * layout.height_in_block * layout.depth;
-					dst_buffer = { reinterpret_cast<std::byte*>(upload_scratch_mem.map(buffer::access::write)), image_linear_size };
+					dst_buffer = { reinterpret_cast<std::byte*>(upload_scratch_mem.map(0, image_linear_size, gl::buffer::access::write)), image_linear_size };
 				}
 
 				auto op = upload_texture_subresource(dst_buffer, layout, format, is_swizzled, caps);