46 #define LD_B(RTYPE, psrc) *((RTYPE*)(psrc))  argument
50 #define LD_H(RTYPE, psrc) *((RTYPE*)(psrc))  argument
54 #define LD_W(RTYPE, psrc) *((RTYPE*)(psrc))  argument
71   static inline TYPE FUNC_NAME(const void* const psrc) {  \
72     const uint8_t* const psrc_m = (const uint8_t*)psrc;   \
81 #define MSA_LOAD(psrc, FUNC_NAME)  FUNC_NAME(psrc)  argument
97   #define LH(psrc)  MSA_LOAD(psrc, msa_lh)  argument
99   #define LW(psrc)  MSA_LOAD(psrc, msa_lw)  argument
102     #define LD(psrc)  MSA_LOAD(psrc, msa_ld)  argument
104     #define LD(psrc)  ((((uint64_t)MSA_LOAD(psrc + 4, msa_lw)) << 32) | \  argument
105                        MSA_LOAD(psrc, msa_lw))
116   #define LH(psrc)  MSA_LOAD(psrc, msa_ulh)  argument
118   #define LW(psrc)  MSA_LOAD(psrc, msa_ulw)  argument
121     #define LD(psrc)  MSA_LOAD(psrc, msa_uld)  argument
123     #define LD(psrc)  ((((uint64_t)MSA_LOAD(psrc + 4, msa_ulw)) << 32) | \  argument
124                         MSA_LOAD(psrc, msa_ulw))
148 #define LW4(psrc, stride, out0, out1, out2, out3) do {  \  argument
149   const uint8_t* ptmp = (const uint8_t*)psrc;           \
218 #define LD_B2(RTYPE, psrc, stride, out0, out1) do {  \  argument
219   out0 = LD_B(RTYPE, psrc);                          \
220   out1 = LD_B(RTYPE, psrc + stride);                 \
225 #define LD_B3(RTYPE, psrc, stride, out0, out1, out2) do {  \  argument
226   LD_B2(RTYPE, psrc, stride, out0, out1);                  \
227   out2 = LD_B(RTYPE, psrc + 2 * stride);                   \
232 #define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) do {  \  argument
233   LD_B2(RTYPE, psrc, stride, out0, out1);                        \
234   LD_B2(RTYPE, psrc + 2 * stride , stride, out2, out3);          \
239 #define LD_B8(RTYPE, psrc, stride,                                  \  argument
241   LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3);               \
242   LD_B4(RTYPE, psrc + 4 * stride, stride, out4, out5, out6, out7);  \
253 #define LD_H2(RTYPE, psrc, stride, out0, out1) do {  \  argument
254   out0 = LD_H(RTYPE, psrc);                          \
255   out1 = LD_H(RTYPE, psrc + stride);                 \
268 #define LD_W2(RTYPE, psrc, stride, out0, out1) do {  \  argument
269   out0 = LD_W(RTYPE, psrc);                          \
270   out1 = LD_W(RTYPE, psrc + stride);                 \
275 #define LD_W3(RTYPE, psrc, stride, out0, out1, out2) do {  \  argument
276   LD_W2(RTYPE, psrc, stride, out0, out1);                  \
277   out2 = LD_W(RTYPE, psrc + 2 * stride);                   \
282 #define LD_W4(RTYPE, psrc, stride, out0, out1, out2, out3) do {  \  argument
283   LD_W2(RTYPE, psrc, stride, out0, out1);                        \
284   LD_W2(RTYPE, psrc + 2 * stride, stride, out2, out3);           \