blender
/
blender
129
398
Fork 572
Code Issues 6.2k Pull Requests 748 Projects 19 Wiki Activity

Gawain: improve immediate mode performance 

This was already fast on Apple, but @Severin and @dfelinto noticed slowdowns in user prefs, which is text heavy.

The problem was immBeginAtMost not being smart about VBO write flushing. immBeginAtMost can use all of its allocated range or only a subrange. The previous code was forcing back-to-back draw calls and buffer writes to serialize. This commit lets OpenGL know that our VBO writes never overlap, so there's no need to wait.

Should be much faster now!
This commit is contained in:
Mike Erwin 2016-10-14 20:50:17 -04:00
parent 5e428b3b3f
commit cf5750a85e
1 changed files with 22 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
source/blender/gpu/gawain/immediate.c
View File

@ -211,7 +211,8 @@ void immBegin(GLenum primitive, unsigned vertex_ct)
#if APPLE_LEGACY
imm.buffer_data = glMapBuffer(GL_ARRAY_BUFFER, GL_WRITE_ONLY) + imm.buffer_offset;
#else
imm.buffer_data = glMapBufferRange(GL_ARRAY_BUFFER, imm.buffer_offset, bytes_needed, GL_MAP_WRITE_BIT | GL_MAP_UNSYNCHRONIZED_BIT);
imm.buffer_data = glMapBufferRange(GL_ARRAY_BUFFER, imm.buffer_offset, bytes_needed,
GL_MAP_WRITE_BIT | GL_MAP_UNSYNCHRONIZED_BIT | (imm.strict_vertex_ct ? 0 : GL_MAP_FLUSH_EXPLICIT_BIT));
#endif
#if TRUST_NO_ONE
@ -341,13 +342,27 @@ void immEnd()
{
#if TRUST_NO_ONE
assert(imm.vertex_idx <= imm.vertex_ct);
assert(imm.vertex_idx == 0 || vertex_count_makes_sense_for_primitive(imm.vertex_idx, imm.primitive));
#endif
// printf("used %u of %u verts,", imm.vertex_idx, imm.vertex_ct);
imm.vertex_ct = imm.vertex_idx;
buffer_bytes_used = vertex_buffer_size(&imm.vertex_format, imm.vertex_ct);
// unused buffer bytes are available to the next immBegin
// printf(" %u of %u bytes\n", buffer_bytes_used, imm.buffer_bytes_mapped);
if (imm.vertex_idx == imm.vertex_ct)
{
buffer_bytes_used = imm.buffer_bytes_mapped;
}
else
{
#if TRUST_NO_ONE
assert(imm.vertex_idx == 0 || vertex_count_makes_sense_for_primitive(imm.vertex_idx, imm.primitive));
#endif
imm.vertex_ct = imm.vertex_idx;
buffer_bytes_used = vertex_buffer_size(&imm.vertex_format, imm.vertex_ct);
// unused buffer bytes are available to the next immBegin
// printf(" %u of %u bytes\n", buffer_bytes_used, imm.buffer_bytes_mapped);
}
#if !APPLE_LEGACY
// tell OpenGL what range was modified so it doesn't copy the whole mapped range
// printf("flushing %u to %u\n", imm.buffer_offset, imm.buffer_offset + buffer_bytes_used - 1);
glFlushMappedBufferRange(GL_ARRAY_BUFFER, 0, buffer_bytes_used);
#endif
}
#if IMM_BATCH_COMBO
@ -367,8 +382,8 @@ void immEnd()
{
#if APPLE_LEGACY
// tell OpenGL what range was modified so it doesn't copy the whole buffer
// printf("flushing %u to %u\n", imm.buffer_offset, imm.buffer_offset + buffer_bytes_used - 1);
glFlushMappedBufferRangeAPPLE(GL_ARRAY_BUFFER, imm.buffer_offset, buffer_bytes_used);
// printf("flushing %u to %u\n", imm.buffer_offset, imm.buffer_offset + buffer_bytes_used - 1);
#endif
glUnmapBuffer(GL_ARRAY_BUFFER);