[Mesa-dev] [PATCH v4 16/28] i965/vec4: fix VEC4_OPCODE_FROM_DOUBLE for IVB/BYT
Samuel Iglesias Gonsálvez
siglesias at igalia.com
Mon Mar 20 09:17:13 UTC 2017
From: "Juan A. Suarez Romero" <jasuarez at igalia.com>
In the generator we must generate slightly different code for
Ivybridge/Baytrail, because of the way the stride works in
this hardware.
v2:
- Use stride and don't need to fix dst (Curro)
Signed-off-by: Samuel Iglesias Gonsálvez <siglesias at igalia.com>
---
src/intel/compiler/brw_vec4_generator.cpp | 26 +++++++++++++++++++-------
1 file changed, 19 insertions(+), 7 deletions(-)
diff --git a/src/intel/compiler/brw_vec4_generator.cpp b/src/intel/compiler/brw_vec4_generator.cpp
index 0d70690b057..34bcd32c958 100644
--- a/src/intel/compiler/brw_vec4_generator.cpp
+++ b/src/intel/compiler/brw_vec4_generator.cpp
@@ -1950,16 +1950,28 @@ generate_code(struct brw_codegen *p,
brw_set_default_access_mode(p, BRW_ALIGN_1);
- dst.hstride = BRW_HORIZONTAL_STRIDE_2;
- dst.width = BRW_WIDTH_4;
+ /* When converting from DF->F, we set destination's stride as 2 as an
+ * aligment requirement. But in IVB/BYT, each DF implicitly writes
+ * two floats, being the first one the converted value. So we don't
+ * need to explicitly set stride 2, but 1.
+ */
+ struct brw_reg spread_dst;
+ if (devinfo->gen == 7 && !devinfo->is_haswell)
+ spread_dst = stride(dst, 8, 4, 1);
+ else
+ spread_dst = stride(dst, 8, 4, 2);
+
src[0].vstride = BRW_VERTICAL_STRIDE_4;
src[0].width = BRW_WIDTH_4;
- brw_MOV(p, dst, src[0]);
+ brw_MOV(p, spread_dst, src[0]);
- struct brw_reg dst_as_src = dst;
- dst.hstride = BRW_HORIZONTAL_STRIDE_1;
- dst.width = BRW_WIDTH_8;
- brw_MOV(p, dst, dst_as_src);
+ /* As we have set horizontal stride 1 instead of 2 in IVB/BYT, we
+ * need to fix it here to have the expected value.
+ */
+ if (devinfo->gen == 7 && !devinfo->is_haswell)
+ spread_dst = stride(dst, 8, 4, 2);
+
+ brw_MOV(p, dst, spread_dst);
brw_set_default_access_mode(p, BRW_ALIGN_16);
break;
--
2.11.0
More information about the mesa-dev
mailing list