On 09/02/18 21:54, Pierre Moreau wrote:
On 2018-02-09 — 12:06, Timothy Arceri wrote:
PIPE_SHADER_CAP_PREFERRED_IR was conflicting with PIPE_SHADER_IR_NIR
for compute shaders, so we let clover pick the one it wants to use.
---
  src/gallium/state_trackers/clover/core/device.cpp | 11 +++++++++--
  1 file changed, 9 insertions(+), 2 deletions(-)

diff --git a/src/gallium/state_trackers/clover/core/device.cpp 
b/src/gallium/state_trackers/clover/core/device.cpp
index 9dd7eed3f1..71cf4bf60a 100644
--- a/src/gallium/state_trackers/clover/core/device.cpp
+++ b/src/gallium/state_trackers/clover/core/device.cpp
@@ -243,8 +243,15 @@ device::vendor_name() const {
enum pipe_shader_ir
  device::ir_format() const {
-   return (enum pipe_shader_ir) pipe->get_shader_param(
-      pipe, PIPE_SHADER_COMPUTE, PIPE_SHADER_CAP_PREFERRED_IR);
+   int supported_irs =
+      pipe->get_shader_param(pipe, PIPE_SHADER_COMPUTE,
+                             PIPE_SHADER_CAP_SUPPORTED_IRS);
+
+   if (supported_irs & (1 << PIPE_SHADER_IR_NATIVE)) {
+      return PIPE_SHADER_IR_NATIVE;
+   }
+
+   return PIPE_SHADER_IR_TGSI;

This looks fine, but I was wondering whether we should instead:
* check that TGSI is in supported_irs before returning it, else throw an
   exception;

I'm happy to add the following above the return:

assert(supported_irs & (1 << PIPE_SHADER_IR_TGSI));

or
* swap TGSI and NATIVE (so check for TGSI, if not found return NATIVE), which
   would match the existing behaviour (see for example
   
https://cgit.freedesktop.org/mesa/mesa/tree/src/gallium/state_trackers/clover/core/program.cpp#n54).

That would break radeonsi and r600 as it would always return TGSI IR as its supported by the drivers.



  }
std::string
--
2.14.3

_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/mesa-dev

Reply via email to