pan/bi: Fix shader prefetch size

The prefetch buffer size is larger than first thought, but includes
the final clause, so subtract the size of the final clause from the
prefetch size.

Reviewed-by: Boris Brezillon <boris.brezillon@collabora.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8354>

diff --git a/src/panfrost/bifrost/bifrost_compile.c b/src/panfrost/bifrost/bifrost_compile.c
index b300b7d32caf7da20f04db210e32425ccc707d06..26961abc3b8057ca1bb2fdc2f51a4318410da870 100644 (file)
--- a/src/panfrost/bifrost/bifrost_compile.c
+++ b/src/panfrost/bifrost/bifrost_compile.c
@@ -48,9 +48,9 @@ static const struct debug_named_value bifrost_debug_options[] = {
 
 DEBUG_GET_ONCE_FLAGS_OPTION(bifrost_debug, "BIFROST_MESA_DEBUG", bifrost_debug_options, 0)
 
-/* How many bytes are prefetched by the Bifrost shader core. Past the end of
- * the shader, this range must contain valid instructions or zero. */
-#define BIFROST_SHADER_PREFETCH 96
+/* How many bytes are prefetched by the Bifrost shader core. From the final
+ * clause of the shader, this range must be valid instructions or zero. */
+#define BIFROST_SHADER_PREFETCH 128
 
 /* TODO: This is not thread safe!! */
 static unsigned SHADER_DB_COUNT = 0;
@@ -2610,7 +2610,7 @@ bifrost_compile_shader_nir(void *mem_ctx, nir_shader *nir,
                 bi_print_shader(ctx, stdout);
 
         util_dynarray_init(&program->compiled, NULL);
-        bi_pack(ctx, &program->compiled);
+        unsigned final_clause = bi_pack(ctx, &program->compiled);
 
         /* If we need to wait for ATEST or BLEND in the first clause, pass the
          * corresponding bits through to the renderer state descriptor */
@@ -2630,8 +2630,10 @@ bifrost_compile_shader_nir(void *mem_ctx, nir_shader *nir,
         }
 
         /* Pad the shader with enough zero bytes to trick the prefetcher */
-        memset(util_dynarray_grow(&program->compiled, uint8_t, BIFROST_SHADER_PREFETCH),
-               0, BIFROST_SHADER_PREFETCH);
+        unsigned prefetch_size = BIFROST_SHADER_PREFETCH - final_clause;
+
+        memset(util_dynarray_grow(&program->compiled, uint8_t, prefetch_size),
+               0, prefetch_size);
 
         program->tls_size = ctx->tls_size;