Module: Mesa
Branch: staging/23.3
Commit: f63ea1a1303dcd25adb115eb695c6198e4606e19
URL:    
http://cgit.freedesktop.org/mesa/mesa/commit/?id=f63ea1a1303dcd25adb115eb695c6198e4606e19

Author: Rhys Perry <[email protected]>
Date:   Fri Oct 13 19:56:24 2023 +0100

nir/loop_analyze: skip if basis/limit/comparison is vector

Replacement in try_eval_const_alu() doesn't work because the replacements
are always scalar. The callers also always give a scalar dest.

This is encountered when compiling a Redout shader under ASan.

Signed-off-by: Rhys Perry <[email protected]>
Acked-by: Timothy Arceri <[email protected]>
Fixes: bc170e895f38 ("nir/loop_analyze: Use try_eval_const_alu and induction 
variable basis info")
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/26225>
(cherry picked from commit 796058841378c4d5e04ff8b5cbe5adef5aae326a)

---

 .pick_status.json                   | 2 +-
 src/compiler/nir/nir_loop_analyze.c | 4 ++++
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/.pick_status.json b/.pick_status.json
index e5bb7c2000c..4e85130d7ef 100644
--- a/.pick_status.json
+++ b/.pick_status.json
@@ -2154,7 +2154,7 @@
         "description": "nir/loop_analyze: skip if basis/limit/comparison is 
vector",
         "nominated": true,
         "nomination_type": 1,
-        "resolution": 0,
+        "resolution": 1,
         "main_sha": null,
         "because_sha": "bc170e895f384acf24c849798fcff092db440135",
         "notes": null
diff --git a/src/compiler/nir/nir_loop_analyze.c 
b/src/compiler/nir/nir_loop_analyze.c
index 4c39de9d949..f468d2316e5 100644
--- a/src/compiler/nir/nir_loop_analyze.c
+++ b/src/compiler/nir/nir_loop_analyze.c
@@ -1050,6 +1050,10 @@ calculate_iterations(nir_def *basis, nir_def 
*limit_basis,
              induction_base_type);
    }
 
+   if (cond.def->num_components != 1 || basis->num_components != 1 ||
+       limit_basis->num_components != 1)
+      return -1;
+
    /* do-while loops can increment the starting value before the condition is
     * checked. e.g.
     *

Reply via email to