[Mesa-dev] [PATCH] ac, radv: fix removing the vec3 restriction on SI

Samuel Pitoiset samuel.pitoiset at gmail.com
Mon Jun 3 19:42:00 UTC 2019


On 6/3/19 9:33 PM, Marek Olšák wrote:
> SI doesn't support buffer_load_dwordx3 and buffer_store_dwordx3, but 
> it supports buffer_load_format_xyz and buffer_store_format_xyz.
OK, I will update.
>
> Marek
>
> On Mon, Jun 3, 2019 at 9:09 AM Samuel Pitoiset 
> <samuel.pitoiset at gmail.com <mailto:samuel.pitoiset at gmail.com>> wrote:
>
>     I thought LLVM was able to handle that itself but actually it
>     does not. That means we shouldn't try to emit vec3 on SI because
>     it's unsupported.
>
>     Fixes: 6970a9a6ca9 ("ac,radv: remove the vec3 restriction with
>     LLVM 9+")"
>     Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com
>     <mailto:samuel.pitoiset at gmail.com>>
>     ---
>      src/amd/common/ac_llvm_build.c    | 12 ++++++------
>      src/amd/common/ac_llvm_util.h     |  9 +++++++++
>      src/amd/common/ac_nir_to_llvm.c   |  3 ++-
>      src/amd/vulkan/radv_nir_to_llvm.c |  2 +-
>      4 files changed, 18 insertions(+), 8 deletions(-)
>
>     diff --git a/src/amd/common/ac_llvm_build.c
>     b/src/amd/common/ac_llvm_build.c
>     index 613c1eef942..7f5c8ef873c 100644
>     --- a/src/amd/common/ac_llvm_build.c
>     +++ b/src/amd/common/ac_llvm_build.c
>     @@ -1167,7 +1167,7 @@ ac_build_llvm8_buffer_store_common(struct
>     ac_llvm_context *ctx,
>             args[idx++] = voffset ? voffset : ctx->i32_0;
>             args[idx++] = soffset ? soffset : ctx->i32_0;
>             args[idx++] = LLVMConstInt(ctx->i32, (glc ? 1 : 0) + (slc
>     ? 2 : 0), 0);
>     -       unsigned func = HAVE_LLVM < 0x900 && num_channels == 3 ? 4
>     : num_channels;
>     +       unsigned func = !ac_has_vec3_support(ctx->chip_class) &&
>     num_channels == 3 ? 4 : num_channels;
>             const char *indexing_kind = structurized ? "struct" : "raw";
>             char name[256], type_name[8];
>
>     @@ -1227,7 +1227,7 @@ ac_build_buffer_store_dword(struct
>     ac_llvm_context *ctx,
>      {
>             /* Split 3 channel stores, because only LLVM 9+ support
>     3-channel
>              * intrinsics. */
>     -       if (num_channels == 3 && HAVE_LLVM < 0x900) {
>     +       if (num_channels == 3 &&
>     !ac_has_vec3_support(ctx->chip_class)) {
>                     LLVMValueRef v[3], v01;
>
>                     for (int i = 0; i < 3; i++) {
>     @@ -1354,7 +1354,7 @@ ac_build_llvm8_buffer_load_common(struct
>     ac_llvm_context *ctx,
>             args[idx++] = voffset ? voffset : ctx->i32_0;
>             args[idx++] = soffset ? soffset : ctx->i32_0;
>             args[idx++] = LLVMConstInt(ctx->i32, (glc ? 1 : 0) + (slc
>     ? 2 : 0), 0);
>     -       unsigned func = HAVE_LLVM < 0x900 && num_channels == 3 ? 4
>     : num_channels;
>     +       unsigned func = !ac_has_vec3_support(ctx->chip_class) &&
>     num_channels == 3 ? 4 : num_channels;
>             const char *indexing_kind = structurized ? "struct" : "raw";
>             char name[256], type_name[8];
>
>     @@ -1420,7 +1420,7 @@ ac_build_buffer_load(struct ac_llvm_context
>     *ctx,
>                     if (num_channels == 1)
>                             return result[0];
>
>     -               if (num_channels == 3 && HAVE_LLVM < 0x900)
>     +               if (num_channels == 3 &&
>     !ac_has_vec3_support(ctx->chip_class))
>                             result[num_channels++] =
>     LLVMGetUndef(ctx->f32);
>                     return ac_build_gather_values(ctx, result,
>     num_channels);
>             }
>     @@ -1512,7 +1512,7 @@ ac_build_llvm8_tbuffer_load(struct
>     ac_llvm_context *ctx,
>             args[idx++] = soffset ? soffset : ctx->i32_0;
>             args[idx++] = LLVMConstInt(ctx->i32, dfmt | (nfmt << 4), 0);
>             args[idx++] = LLVMConstInt(ctx->i32, (glc ? 1 : 0) + (slc
>     ? 2 : 0), 0);
>     -       unsigned func = HAVE_LLVM < 0x900 && num_channels == 3 ? 4
>     : num_channels;
>     +       unsigned func = !ac_has_vec3_support(ctx->chip_class) &&
>     num_channels == 3 ? 4 : num_channels;
>             const char *indexing_kind = structurized ? "struct" : "raw";
>             char name[256], type_name[8];
>
>     @@ -2011,7 +2011,7 @@ ac_build_llvm8_tbuffer_store(struct
>     ac_llvm_context *ctx,
>             args[idx++] = soffset ? soffset : ctx->i32_0;
>             args[idx++] = LLVMConstInt(ctx->i32, dfmt | (nfmt << 4), 0);
>             args[idx++] = LLVMConstInt(ctx->i32, (glc ? 1 : 0) + (slc
>     ? 2 : 0), 0);
>     -       unsigned func = HAVE_LLVM < 0x900 && num_channels == 3 ? 4
>     : num_channels;
>     +       unsigned func = !ac_has_vec3_support(ctx->chip_class) &&
>     num_channels == 3 ? 4 : num_channels;
>             const char *indexing_kind = structurized ? "struct" : "raw";
>             char name[256], type_name[8];
>
>     diff --git a/src/amd/common/ac_llvm_util.h
>     b/src/amd/common/ac_llvm_util.h
>     index ca00540da80..a45647a3360 100644
>     --- a/src/amd/common/ac_llvm_util.h
>     +++ b/src/amd/common/ac_llvm_util.h
>     @@ -146,6 +146,15 @@ bool ac_compile_module_to_binary(struct
>     ac_compiler_passes *p, LLVMModuleRef mod
>      void ac_llvm_add_barrier_noop_pass(LLVMPassManagerRef passmgr);
>      void ac_enable_global_isel(LLVMTargetMachineRef tm);
>
>     +static inline bool
>     +ac_has_vec3_support(enum chip_class chip)
>     +{
>     +       if (chip == GFX6)
>     +               return false;
>     +
>     +       return HAVE_LLVM >= 0x900;
>     +}
>     +
>      #ifdef __cplusplus
>      }
>      #endif
>     diff --git a/src/amd/common/ac_nir_to_llvm.c
>     b/src/amd/common/ac_nir_to_llvm.c
>     index 51f92a6b062..429dac63d63 100644
>     --- a/src/amd/common/ac_nir_to_llvm.c
>     +++ b/src/amd/common/ac_nir_to_llvm.c
>     @@ -1576,7 +1576,8 @@ static void visit_store_ssbo(struct
>     ac_nir_context *ctx,
>
>                     /* Due to an LLVM limitation with LLVM < 9, split
>     3-element
>                      * writes into a 2-element and a 1-element write. */
>     -               if (count == 3 && (elem_size_bytes != 4 ||
>     HAVE_LLVM < 0x900)) {
>     +               if (count == 3 &&
>     +                   (elem_size_bytes != 4 ||
>     !ac_has_vec3_support(ctx->ac.chip_class))) {
>                             writemask |= 1 << (start + 2);
>                             count = 2;
>                     }
>     diff --git a/src/amd/vulkan/radv_nir_to_llvm.c
>     b/src/amd/vulkan/radv_nir_to_llvm.c
>     index dca4bebcdd1..ab552ae34ab 100644
>     --- a/src/amd/vulkan/radv_nir_to_llvm.c
>     +++ b/src/amd/vulkan/radv_nir_to_llvm.c
>     @@ -2768,7 +2768,7 @@ radv_emit_stream_output(struct
>     radv_shader_context *ctx,
>                     /* fall through */
>             case 4: /* as v4i32 */
>                     vdata = ac_build_gather_values(&ctx->ac, out,
>     -                                              HAVE_LLVM < 0x900 ?
>     + !ac_has_vec3_support(ctx->ac.chip_class) ?
>      util_next_power_of_two(num_comps) :
>                                                    num_comps);
>                     break;
>     -- 
>     2.21.0
>
>     _______________________________________________
>     mesa-dev mailing list
>     mesa-dev at lists.freedesktop.org <mailto:mesa-dev at lists.freedesktop.org>
>     https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20190603/6250e968/attachment-0001.html>


More information about the mesa-dev mailing list