[Mesa-dev] [PATCH] radeonsi: don't use allocas for arrays with LLVM 3.8
Marek Olšák
maraeo at gmail.com
Thu Aug 25 19:20:38 UTC 2016
FYI, I've already committed this because our other team needs this bug fix.
Marek
On Thu, Aug 25, 2016 at 8:26 PM, Marek Olšák <maraeo at gmail.com> wrote:
> From: Marek Olšák <marek.olsak at amd.com>
>
> It crashes.
>
> Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=97413
> ---
> src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
> index 71f52fd..4643e6d 100644
> --- a/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
> +++ b/src/gallium/drivers/radeon/radeon_setup_tgsi_llvm.c
> @@ -552,22 +552,24 @@ static void emit_declaration(struct lp_build_tgsi_context *bld_base,
> *
> * Note that 16 is the number of vector elements that
> * LLVM will store in a register, so theoretically an
> * array with up to 4 * 16 = 64 elements could be
> * handled this way, but whether that's a good idea
> * depends on VGPR register pressure elsewhere.
> *
> * FIXME: We shouldn't need to have the non-alloca
> * code path for arrays. LLVM should be smart enough to
> * promote allocas into registers when profitable.
> + *
> + * LLVM 3.8 crashes with this.
> */
> - if (array_size > 16) {
> + if (HAVE_LLVM >= 0x0309 && array_size > 16) {
> array_alloca = LLVMBuildAlloca(builder,
> LLVMArrayType(bld_base->base.vec_type,
> array_size), "array");
> ctx->temp_array_allocas[id] = array_alloca;
> }
> }
>
> if (!ctx->temps_count) {
> ctx->temps_count = bld_base->info->file_max[TGSI_FILE_TEMPORARY] + 1;
> ctx->temps = MALLOC(TGSI_NUM_CHANNELS * ctx->temps_count * sizeof(LLVMValueRef));
> --
> 2.7.4
>
More information about the mesa-dev
mailing list