gallivm: (trivial) handle non-aligned fetch for lp_build_fetch_rgba_soa

soa fetch so far always assumed that data was aligned. However, we want to
use this for vertex fetch, and data might not be aligned there, so handle
it in this path too (basically just pass through alignment through to other
functions). (It looks like it wouldn't work for for cached s3tc but this is
no different than with AoS fetch.)

Reviewed-by: Jose Fonseca <[email protected]>

3 files changed, 12 insertions, 8 deletions

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format.h b/src/gallium/auxiliary/gallivm/lp_bld_format.h
index 5c866f420bd..6540caaa293 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format.h
@@ -143,6 +143,7 @@ void
 lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
                         const struct util_format_description *format_desc,
                         struct lp_type type,
+                        boolean aligned,
                         LLVMValueRef base_ptr,
                         LLVMValueRef offsets,
                         LLVMValueRef i,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
index 7444c518e42..389bfa05d4b 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
@@ -349,6 +349,7 @@ lp_build_rgba8_to_fi32_soa(struct gallivm_state *gallivm,
  *
  * \param type  the desired return type for 'rgba'.  The vector length
  *              is the number of texels to fetch
+ * \param aligned if the offset is guaranteed to be aligned to element width
  *
  * \param base_ptr  points to the base of the texture mip tree.
  * \param offset    offset to start of the texture image block.  For non-
@@ -365,6 +366,7 @@ void
 lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
                         const struct util_format_description *format_desc,
                         struct lp_type type,
+                        boolean aligned,
                         LLVMValueRef base_ptr,
                         LLVMValueRef offset,
                         LLVMValueRef i,
@@ -402,7 +404,7 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
                                type.length,
                                format_desc->block.bits,
                                type.width,
-                               TRUE,
+                               aligned,
                                base_ptr, offset, FALSE);
 
       /*
@@ -428,7 +430,7 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
 
       packed = lp_build_gather(gallivm, type.length,
                                format_desc->block.bits,
-                               type.width, TRUE,
+                               type.width, aligned,
                                base_ptr, offset, FALSE);
       if (format_desc->format == PIPE_FORMAT_R11G11B10_FLOAT) {
          lp_build_r11g11b10_to_float(gallivm, packed, rgba_out);
@@ -456,14 +458,14 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
          LLVMValueRef s_offset = lp_build_const_int_vec(gallivm, type, 4);
          offset = LLVMBuildAdd(builder, offset, s_offset, "");
          packed = lp_build_gather(gallivm, type.length, 32, type.width,
-                                  TRUE, base_ptr, offset, FALSE);
+                                  aligned, base_ptr, offset, FALSE);
          packed = LLVMBuildAnd(builder, packed,
                                lp_build_const_int_vec(gallivm, type, mask), "");
       }
       else {
          assert (format_desc->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT);
          packed = lp_build_gather(gallivm, type.length, 32, type.width,
-                                  TRUE, base_ptr, offset, TRUE);
+                                  aligned, base_ptr, offset, TRUE);
          packed = LLVMBuildBitCast(builder, packed,
                                    lp_build_vec_type(gallivm, type), "");
       }
@@ -489,7 +491,7 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
       tmp_type.norm = TRUE;
 
       tmp = lp_build_fetch_rgba_aos(gallivm, format_desc, tmp_type,
-                                    TRUE, base_ptr, offset, i, j, cache);
+                                    aligned, base_ptr, offset, i, j, cache);
 
       lp_build_rgba8_to_fi32_soa(gallivm,
                                 type,
@@ -509,6 +511,7 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
       const struct util_format_description *flinear_desc;
       LLVMValueRef packed;
       flinear_desc = util_format_description(util_format_linear(format_desc->format));
+      /* This probably only works with aligned data */
       packed = lp_build_fetch_cached_texels(gallivm,
                                             flinear_desc,
                                             type.length,
@@ -572,7 +575,7 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
 
          /* Get a single float[4]={R,G,B,A} pixel */
          tmp = lp_build_fetch_rgba_aos(gallivm, format_desc, tmp_type,
-                                       TRUE, base_ptr, offset_elem,
+                                       aligned, base_ptr, offset_elem,
                                        i_elem, j_elem, cache);
 
          /*
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index 939cd43ac48..161a03f3524 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -159,7 +159,7 @@ lp_build_sample_texel_soa(struct lp_build_sample_context *bld,
 
    lp_build_fetch_rgba_soa(bld->gallivm,
                            bld->format_desc,
-                           bld->texel_type,
+                           bld->texel_type, TRUE,
                            data_ptr, offset,
                            i, j,
                            bld->cache,
@@ -2406,7 +2406,7 @@ lp_build_fetch_texel(struct lp_build_sample_context *bld,
 
    lp_build_fetch_rgba_soa(bld->gallivm,
                            bld->format_desc,
-                           bld->texel_type,
+                           bld->texel_type, TRUE,
                            bld->base_ptr, offset,
                            i, j,
                            bld->cache,