gallivm/nir: fix big-endian 64-bit splitting/merging.

The shuffles need to be swapped to do this properly on big-endian

Cc: <mesa-stable@lists.freedesktop.org>

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5679>
This commit is contained in:
Dave Airlie 2020-06-29 14:59:20 +10:00
parent 9286605276
commit 3aeb61da49
2 changed files with 20 additions and 0 deletions

View file

@ -353,8 +353,13 @@ static LLVMValueRef split_64bit(struct lp_build_nir_context *bld_base,
LLVMValueRef shuffles2[LP_MAX_VECTOR_WIDTH/32];
int len = bld_base->base.type.length * 2;
for (unsigned i = 0; i < bld_base->base.type.length; i++) {
#if UTIL_ARCH_LITTLE_ENDIAN
shuffles[i] = lp_build_const_int32(gallivm, i * 2);
shuffles2[i] = lp_build_const_int32(gallivm, (i * 2) + 1);
#else
shuffles[i] = lp_build_const_int32(gallivm, (i * 2) + 1);
shuffles2[i] = lp_build_const_int32(gallivm, (i * 2));
#endif
}
src = LLVMBuildBitCast(gallivm->builder, src, LLVMVectorType(LLVMInt32TypeInContext(gallivm->context), len), "");
@ -378,8 +383,13 @@ merge_64bit(struct lp_build_nir_context *bld_base,
assert(len <= (2 * (LP_MAX_VECTOR_WIDTH/32)));
for (i = 0; i < bld_base->base.type.length * 2; i+=2) {
#if UTIL_ARCH_LITTLE_ENDIAN
shuffles[i] = lp_build_const_int32(gallivm, i / 2);
shuffles[i + 1] = lp_build_const_int32(gallivm, i / 2 + bld_base->base.type.length);
#else
shuffles[i] = lp_build_const_int32(gallivm, i / 2 + bld_base->base.type.length);
shuffles[i + 1] = lp_build_const_int32(gallivm, i / 2);
#endif
}
return LLVMBuildShuffleVector(builder, input, input2, LLVMConstVector(shuffles, len), "");
}

View file

@ -69,8 +69,13 @@ emit_fetch_64bit(
assert(len <= (2 * (LP_MAX_VECTOR_WIDTH/32)));
for (i = 0; i < bld_base->base.type.length * 2; i+=2) {
#if UTIL_ARCH_LITTLE_ENDIAN
shuffles[i] = lp_build_const_int32(gallivm, i / 2);
shuffles[i + 1] = lp_build_const_int32(gallivm, i / 2 + bld_base->base.type.length);
#else
shuffles[i] = lp_build_const_int32(gallivm, i / 2 + bld_base->base.type.length);
shuffles[i + 1] = lp_build_const_int32(gallivm, i / 2);
#endif
}
res = LLVMBuildShuffleVector(builder, input, input2, LLVMConstVector(shuffles, len), "");
@ -91,8 +96,13 @@ emit_store_64bit_split(struct lp_build_nir_context *bld_base,
value = LLVMBuildBitCast(gallivm->builder, value, LLVMVectorType(LLVMFloatTypeInContext(gallivm->context), len), "");
for (i = 0; i < bld_base->base.type.length; i++) {
#if UTIL_ARCH_LITTLE_ENDIAN
shuffles[i] = lp_build_const_int32(gallivm, i * 2);
shuffles2[i] = lp_build_const_int32(gallivm, (i * 2) + 1);
#else
shuffles[i] = lp_build_const_int32(gallivm, i * 2 + 1);
shuffles2[i] = lp_build_const_int32(gallivm, i * 2);
#endif
}
split_values[0] = LLVMBuildShuffleVector(builder, value,