48 #define LD_B(RTYPE, psrc) *((RTYPE*)(psrc))  argument
52 #define LD_H(RTYPE, psrc) *((RTYPE*)(psrc))  argument
56 #define LD_W(RTYPE, psrc) *((RTYPE*)(psrc))  argument
73   static inline TYPE FUNC_NAME(const void* const psrc) {  \
74     const uint8_t* const psrc_m = (const uint8_t*)psrc;   \
83 #define MSA_LOAD(psrc, FUNC_NAME)  FUNC_NAME(psrc)  argument
99   #define LH(psrc)  MSA_LOAD(psrc, msa_lh)  argument
101   #define LW(psrc)  MSA_LOAD(psrc, msa_lw)  argument
104     #define LD(psrc)  MSA_LOAD(psrc, msa_ld)  argument
106     #define LD(psrc)  ((((uint64_t)MSA_LOAD(psrc + 4, msa_lw)) << 32) | \  argument
107                        MSA_LOAD(psrc, msa_lw))
118   #define LH(psrc)  MSA_LOAD(psrc, msa_ulh)  argument
120   #define LW(psrc)  MSA_LOAD(psrc, msa_ulw)  argument
123     #define LD(psrc)  MSA_LOAD(psrc, msa_uld)  argument
125     #define LD(psrc)  ((((uint64_t)MSA_LOAD(psrc + 4, msa_ulw)) << 32) | \  argument
126                         MSA_LOAD(psrc, msa_ulw))
150 #define LW4(psrc, stride, out0, out1, out2, out3) do {  \  argument
151   const uint8_t* ptmp = (const uint8_t*)psrc;           \
220 #define LD_B2(RTYPE, psrc, stride, out0, out1) do {  \  argument
221   out0 = LD_B(RTYPE, psrc);                          \
222   out1 = LD_B(RTYPE, psrc + stride);                 \
227 #define LD_B3(RTYPE, psrc, stride, out0, out1, out2) do {  \  argument
228   LD_B2(RTYPE, psrc, stride, out0, out1);                  \
229   out2 = LD_B(RTYPE, psrc + 2 * stride);                   \
234 #define LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3) do {  \  argument
235   LD_B2(RTYPE, psrc, stride, out0, out1);                        \
236   LD_B2(RTYPE, psrc + 2 * stride , stride, out2, out3);          \
241 #define LD_B8(RTYPE, psrc, stride,                                  \  argument
243   LD_B4(RTYPE, psrc, stride, out0, out1, out2, out3);               \
244   LD_B4(RTYPE, psrc + 4 * stride, stride, out4, out5, out6, out7);  \
255 #define LD_H2(RTYPE, psrc, stride, out0, out1) do {  \  argument
256   out0 = LD_H(RTYPE, psrc);                          \
257   out1 = LD_H(RTYPE, psrc + stride);                 \
270 #define LD_W2(RTYPE, psrc, stride, out0, out1) do {  \  argument
271   out0 = LD_W(RTYPE, psrc);                          \
272   out1 = LD_W(RTYPE, psrc + stride);                 \
277 #define LD_W3(RTYPE, psrc, stride, out0, out1, out2) do {  \  argument
278   LD_W2(RTYPE, psrc, stride, out0, out1);                  \
279   out2 = LD_W(RTYPE, psrc + 2 * stride);                   \
284 #define LD_W4(RTYPE, psrc, stride, out0, out1, out2, out3) do {  \  argument
285   LD_W2(RTYPE, psrc, stride, out0, out1);                        \
286   LD_W2(RTYPE, psrc + 2 * stride, stride, out2, out3);           \