From: Georgi Gerganov <redacted>
Date: Tue, 7 Nov 2023 09:02:16 +0000 (+0200)
Subject: metal : fix asserts for setThreadgroupMemoryLength (close #1435)
X-Git-Tag: upstream/1.7.4~1283
X-Git-Url: https://git.djapps.eu/?a=commitdiff_plain;h=3bd7d48f518d679a479a156cdd6c0fd8dadf8e47;p=pkg%2Fggml%2Fsources%2Fwhisper.cpp

metal : fix asserts for setThreadgroupMemoryLength (close #1435)
---

diff --git a/ggml-metal.m b/ggml-metal.m
index 43d0dff0..3bee8397 100644
--- a/ggml-metal.m
+++ b/ggml-metal.m
@@ -1030,7 +1030,7 @@ void ggml_metal_graph_compute(
                             [encoder setBytes:&ne00 length:sizeof(ne00) atIndex:2];
                             [encoder setBytes:&ne01 length:sizeof(ne01) atIndex:3];
                             [encoder setBytes:&ne02 length:sizeof(ne02) atIndex:4];
-                            [encoder setThreadgroupMemoryLength:nth/32*sizeof(float) atIndex:0];
+                            [encoder setThreadgroupMemoryLength:GGML_PAD(nth/32*sizeof(float), 16) atIndex:0];
 
                             [encoder dispatchThreadgroups:MTLSizeMake(ne01*ne02*ne03, 1, 1) threadsPerThreadgroup:MTLSizeMake(nth, 1, 1)];
                         } break;
@@ -1342,7 +1342,7 @@ void ggml_metal_graph_compute(
                             [encoder setBytes:&ne00 length:sizeof( int64_t) atIndex:2];
                             [encoder setBytes:&nb01 length:sizeof(uint64_t) atIndex:3];
                             [encoder setBytes:&eps  length:sizeof(   float) atIndex:4];
-                            [encoder setThreadgroupMemoryLength:nth/32*sizeof(float) atIndex:0];
+                            [encoder setThreadgroupMemoryLength:GGML_PAD(nth/32*sizeof(float), 16) atIndex:0];
 
                             const int64_t nrows = ggml_nrows(src0);
 
@@ -1361,7 +1361,7 @@ void ggml_metal_graph_compute(
                             [encoder setBytes:&ne00    length:sizeof( int64_t) atIndex:2];
                             [encoder setBytes:&nb01    length:sizeof(uint64_t) atIndex:3];
                             [encoder setBytes:&eps     length:sizeof(   float) atIndex:4];
-                            [encoder setThreadgroupMemoryLength:nth*sizeof(float) atIndex:0];
+                            [encoder setThreadgroupMemoryLength:GGML_PAD(nth*sizeof(float), 16) atIndex:0];
 
                             const int64_t nrows = ggml_nrows(src0);