diff options
| author | Marek Olšák <[email protected]> | 2018-07-13 00:23:36 -0400 |
|---|---|---|
| committer | Marek Olšák <[email protected]> | 2018-07-23 20:23:52 -0400 |
| commit | 86b52d42368ac496fe24bc6674e754c323381635 (patch) | |
| tree | 755cd4b6dad029f8882473bca8fdcdc78a42f78e /src/gallium/drivers/radeonsi/si_shader.c | |
| parent | 0866edede0116e33b4bed28737e4d242ad0da2ad (diff) | |
radeonsi: reduce LDS stalls by 40% for tessellation
40% is the decrease in the LGKM counter (which includes SMEM too)
for the GFX9 LSHS stage.
This will make the LDS size slightly larger, but I wasn't able to increase
the patch stride without corruption, so I'm increasing the vertex stride.
Diffstat (limited to 'src/gallium/drivers/radeonsi/si_shader.c')
| -rw-r--r-- | src/gallium/drivers/radeonsi/si_shader.c | 8 |
1 files changed, 4 insertions, 4 deletions
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c index 5dc12d87243..43ba23ff494 100644 --- a/src/gallium/drivers/radeonsi/si_shader.c +++ b/src/gallium/drivers/radeonsi/si_shader.c @@ -417,14 +417,14 @@ static LLVMValueRef get_tcs_in_vertex_dw_stride(struct si_shader_context *ctx) switch (ctx->type) { case PIPE_SHADER_VERTEX: - stride = util_last_bit64(ctx->shader->selector->outputs_written); - return LLVMConstInt(ctx->i32, stride * 4, 0); + stride = ctx->shader->selector->lshs_vertex_stride / 4; + return LLVMConstInt(ctx->i32, stride, 0); case PIPE_SHADER_TESS_CTRL: if (ctx->screen->info.chip_class >= GFX9 && ctx->shader->is_monolithic) { - stride = util_last_bit64(ctx->shader->key.part.tcs.ls->outputs_written); - return LLVMConstInt(ctx->i32, stride * 4, 0); + stride = ctx->shader->key.part.tcs.ls->lshs_vertex_stride / 4; + return LLVMConstInt(ctx->i32, stride, 0); } return si_unpack_param(ctx, ctx->param_vs_state_bits, 24, 8); |