nir/lower_idiv: add new llvm-based path

v2: make variable names snake_case
v2: minor cleanups in emit_udiv()
v2: fix Panfrost build failure
v3: use an enum instead of a boolean flag in nir_lower_idiv()'s signature
v4: remove nir_op_urcp
v5: drop nv50 path
v5: rebase
v6: add back nv50 path
v6: add comment for nir_lower_idiv_path enum
v7: rename _nv50/_llvm to _fast/_precise
v8: fix etnaviv build failure

Signed-off-by: Rhys Perry <[email protected]>
Reviewed-by: Daniel Schürmann <[email protected]>

1 files changed, 13 insertions, 1 deletions

diff --git a/src/compiler/nir/nir.h b/src/compiler/nir/nir.h
index d002102cad8..6b437040dcc 100644
--- a/src/compiler/nir/nir.h
+++ b/src/compiler/nir/nir.h
@@ -3917,7 +3917,19 @@ enum nir_lower_non_uniform_access_type {
 bool nir_lower_non_uniform_access(nir_shader *shader,
                                   enum nir_lower_non_uniform_access_type);
 
-bool nir_lower_idiv(nir_shader *shader);
+enum nir_lower_idiv_path {
+   /* This path is based on NV50LegalizeSSA::handleDIV(). It is the faster of
+    * the two but it is not exact in some cases (for example, 1091317713u /
+    * 1034u gives 5209173 instead of 1055432) */
+   nir_lower_idiv_fast,
+   /* This path is based on AMDGPUTargetLowering::LowerUDIVREM() and
+    * AMDGPUTargetLowering::LowerSDIVREM(). It requires more instructions than
+    * the nv50 path and many of them are integer multiplications, so it is
+    * probably slower. It should always return the correct result, though. */
+   nir_lower_idiv_precise,
+};
+
+bool nir_lower_idiv(nir_shader *shader, enum nir_lower_idiv_path path);
 
 bool nir_lower_input_attachments(nir_shader *shader, bool use_fragcoord_sysval);