Skip to content

Commit 0644bf3

Browse files
authored
[SYCL][Doc] Add support for WCL in joint matrix (#20795)
1 parent 0811ef2 commit 0644bf3

File tree

1 file changed

+29
-15
lines changed

1 file changed

+29
-15
lines changed

sycl/doc/extensions/experimental/sycl_ext_matrix/sycl_ext_oneapi_matrix.asciidoc

Lines changed: 29 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1110,7 +1110,8 @@ This is currently available in devices with the architecture
11101110
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
11111111
`architecture::intel_gpu_dg2_g10`, `architecture::intel_gpu_dg2_g11`,
11121112
`architecture::intel_gpu_dg2_g12`, `architecture::intel_gpu_arl_h`,
1113-
`architecture::intel_gpu_ptl_h`, and `architecture::intel_gpu_ptl_u`.
1113+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1114+
and `architecture::intel_gpu_wcl`.
11141115

11151116
[frame="none",options="header"]
11161117
|======================
@@ -1119,39 +1120,44 @@ This is currently available in devices with the architecture
11191120
`matrix_type::sint32` .2+| `matrix_type::sint32` .2+| +<=+ 8 | 16 .2+| 32
11201121
|`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11211122
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1122-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1123+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1124+
`architecture::intel_gpu_wcl`
11231125
|8|`architecture::intel_gpu_dg2_g10,
11241126
architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11251127
`architecture::intel_gpu_arl_h`
11261128
.2+| `matrix_type::uint8` .2+| `matrix_type::sint8` .2+|
11271129
`matrix_type::sint32` .2+|`matrix_type::sint32` .2+| +<=+ 8 | 16 .2+| 32 |
11281130
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11291131
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1130-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1132+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1133+
`architecture::intel_gpu_wcl`
11311134
|8|`architecture::intel_gpu_dg2_g10,
11321135
architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11331136
`architecture::intel_gpu_arl_h`
11341137
.2+| `matrix_type::sint8` .2+| `matrix_type::uint8` .2+|
11351138
`matrix_type::sint32` .2+|`matrix_type::sint32` .2+| +<=+ 8 | 16 .2+| 32 |
11361139
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11371140
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1138-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1141+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1142+
`architecture::intel_gpu_wcl`
11391143
|8|`architecture::intel_gpu_dg2_g10,
11401144
architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11411145
`architecture::intel_gpu_arl_h`
11421146
.2+| `matrix_type::sint8` .2+| `matrix_type::sint8` .2+|
11431147
`matrix_type::sint32` .2+| `matrix_type::sint32` .2+| +<=+ 8 | 16 .2+| 32 |
11441148
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11451149
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1146-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1150+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1151+
`architecture::intel_gpu_wcl`
11471152
|8|`architecture::intel_gpu_dg2_g10,
11481153
architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11491154
`architecture::intel_gpu_arl_h`
11501155
.8+|`matrix_type::fp16` .8+| `matrix_type::fp16` .8+|
11511156
`matrix_type::fp32` .8+|`matrix_type::fp32` .1+| 16 .1+| 16 | 16
11521157
.6+|`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11531158
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1154-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1159+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1160+
`architecture::intel_gpu_wcl`
11551161
.2+| 1 .2+| 64 | 16 |32
11561162
.2+| 32 .2+| 64 | 16 |32
11571163
.2+| +<=+ 8 | 16 .2+| 16
@@ -1163,27 +1169,31 @@ architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11631169
`matrix_type::fp16` .6+|`matrix_type::fp32` .1+| +<=+ 8 | 16 .1+| 16
11641170
.6+| `architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11651171
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1166-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1172+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1173+
`architecture::intel_gpu_wcl`
11671174
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 | 32
11681175
.2+| 32 .2+| 64 | 16 | 32
11691176
.6+|`matrix_type::fp16` .6+| `matrix_type::fp16` .6+|
11701177
`matrix_type::fp32` .6+|`matrix_type::fp16` .1+| +<=+ 8 | 16 .1+| 16
11711178
.6+|`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11721179
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1173-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1180+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1181+
`architecture::intel_gpu_wcl`
11741182
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 | 32
11751183
.2+| 32 .2+| 64 |16 | 32
11761184
.6+|`matrix_type::fp16` .6+| `matrix_type::fp16` .6+|
11771185
`matrix_type::fp16` .6+|`matrix_type::fp16` .1+| +<=+ 8 | 16 .1+| 16
11781186
.6+|`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11791187
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1180-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1188+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1189+
`architecture::intel_gpu_wcl`
11811190
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 |32 .2+| 32 .2+| 64 | 16 | 32
11821191
.8+| `matrix_type::bf16` .8+| `matrix_type::bf16` .8+|
11831192
`matrix_type::fp32` .8+| `matrix_type::fp32` | 16 | 16 | 16
11841193
.6+|`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11851194
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1186-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1195+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1196+
`architecture::intel_gpu_wcl`
11871197
.2+| 1 .2+| 64 | 16 | 32
11881198
.2+| 32 .2+| 64 | 16 |32
11891199
.2+| +<=+ 8 | 16 .2+| 16
@@ -1195,34 +1205,38 @@ architecture::intel_gpu_dg2_g11, architecture::intel_gpu_dg2_g12`,
11951205
`matrix_type::bf16` .6+|`matrix_type::fp32` .1+| +<=+ 8 | 16 .1+| 16 .6+|
11961206
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
11971207
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1198-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1208+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1209+
`architecture::intel_gpu_wcl`
11991210
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 | 32
12001211
.2+| 32 .2+| 64 |16 | 32
12011212
.6+|`matrix_type::bf16` .6+| `matrix_type::bf16` .6+|
12021213
`matrix_type::fp32` .6+|`matrix_type::bf16` .1+| +<=+ 8 | 16 .1+| 16 .6+|
12031214
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
12041215
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1205-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1216+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1217+
`architecture::intel_gpu_wcl`
12061218
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 | 32
12071219
.2+| 32 .2+| 64 |16 | 32
12081220
.6+|`matrix_type::bf16` .6+| `matrix_type::bf16` .6+|
12091221
`matrix_type::bf16` .6+|`matrix_type::bf16` .1+| +<=+ 8 | 16 .1+| 16 .6+|
12101222
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
12111223
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1212-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1224+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1225+
`architecture::intel_gpu_wcl`
12131226
| 16 | 16 | 16 .2+| 1 .2+| 64 | 16 | 32
12141227
.2+| 32 .2+| 64 |16 | 32
12151228
| `matrix_type::tf32` | `matrix_type::tf32` |
12161229
`matrix_type::fp32` .2+| `matrix_type::fp32` | +<=+ 8 | 16 | 8 |
12171230
`architecture::intel_gpu_pvc`, `architecture::intel_gpu_bmg_g21`,
12181231
`architecture::intel_gpu_bmg_g31`, `architecture::intel_gpu_lnl_m`,
1219-
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`
1232+
`architecture::intel_gpu_ptl_h`, `architecture::intel_gpu_ptl_u`,
1233+
`architecture::intel_gpu_wcl`
12201234
|======================
12211235

12221236
===== Restrictions on `architecture::intel_gpu_pvc`,
12231237
`architecture::intel_gpu_bmg_g21`, `architecture::intel_gpu_bmg_g31`,
12241238
`architecture::intel_gpu_lnl_m`, `architecture::intel_gpu_ptl_h`,
1225-
and `architecture::intel_gpu_ptl_u`
1239+
`architecture::intel_gpu_ptl_u`, and `architecture::intel_gpu_wcl`
12261240

12271241
- The `stride` parameter to `joint_matrix_load` and
12281242
`joint_matrix_store` has the following restrictions:

0 commit comments

Comments
 (0)