intel/fs: Implement extended strides greater than 4 for IR source regions.

Strides up to 32B can be implemented for the source regions of most
instructions by leveraging either the vertical or the horizontal
stride of the hardware Align1 region.  The main motivation for this is
that currently the lower_integer_multiplication() pass will happily
double the stride of one of the 32-bit sources, which can blow up if
the stride of the original source was already the maximum value
allowed by the hardware.

An alternative would be to use the regioning legalization pass in
order to lower such strides into the composition of multiple legal
strides, but that would be somewhat less efficient.

This showed up as a regression from my commit cbea91eb57a501bebb1ca2
in Vulkan 1.1 CTS tests on CHV/BXT platforms, however it was really a
pre-existing problem that had affected conformance on other platforms
without native support for integer multiplication.  CHV/BXT were
getting around it because the code I removed in that commit had the
"fortunate" side effect of emitting narrower regions that didn't hit
the hardware stride limit after lowering.  Beyond fixing the
regression this fixes ~90 additional Vulkan 1.1 subgroup CTS tests on
ICL (that's why this patch is marked for inclusion in mesa-stable even
though the original regressing patch was not).

According to Jason, a nearly equivalent change had been committed
previously as e8c9e65185de3e821e1 and then (mistakenly?) reverted as
a31d0382084c8aa8.

Cc: [email protected]
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=109328
Reported-by: Mark Janes <[email protected]>
Tested-by: Anuj Phogat <[email protected]>
Reviewed-by: Jason Ekstrand <[email protected]>

1 files changed, 10 insertions, 3 deletions

diff --git a/src/intel/compiler/brw_fs_generator.cpp b/src/intel/compiler/brw_fs_generator.cpp
index 64872fc9487..1b9ef490502 100644
--- a/src/intel/compiler/brw_fs_generator.cpp
+++ b/src/intel/compiler/brw_fs_generator.cpp
@@ -90,9 +90,16 @@ brw_reg_from_fs_reg(const struct gen_device_info *devinfo, fs_inst *inst,
           *       different execution size when the number of components
           *       written to each destination GRF is not the same.
           */
-         const unsigned width = MIN2(reg_width, phys_width);
-         brw_reg = brw_vecn_reg(width, brw_file_from_reg(reg), reg->nr, 0);
-         brw_reg = stride(brw_reg, width * reg->stride, width, reg->stride);
+         if (reg->stride > 4) {
+            assert(reg != &inst->dst);
+            assert(reg->stride * type_sz(reg->type) <= REG_SIZE);
+            brw_reg = brw_vecn_reg(1, brw_file_from_reg(reg), reg->nr, 0);
+            brw_reg = stride(brw_reg, reg->stride, 1, 0);
+         } else {
+            const unsigned width = MIN2(reg_width, phys_width);
+            brw_reg = brw_vecn_reg(width, brw_file_from_reg(reg), reg->nr, 0);
+            brw_reg = stride(brw_reg, width * reg->stride, width, reg->stride);
+         }
 
          if (devinfo->gen == 7 && !devinfo->is_haswell) {
             /* From the IvyBridge PRM (EU Changes by Processor Generation, page 13):