After the branch, the expected SEW/LMUL ratio is 1 byte/vector. So we have to set the same ratio before branching (QEMU does not care, but real hardware does).
Unified · Split
| @@ -644,7 +644,7 @@ func ff_h264_idct4_add\type\()_\depth\()_rvv, zve32x · 643 unmodified lines | |||
| 644 | sd a4, 24(sp) | 644 | sd a4, 24(sp) |
| 645 | ld a0, 0(a0) # dest[0] | 645 | ld a0, 0(a0) # dest[0] |
| 646 | addi a1, a1, 16 * 4 # &block_offset[16] | 646 | addi a1, a1, 16 * 4 # &block_offset[16] |
| 647 | vsetivli zero, 4, e8, mf4, ta, ma | 647 | vsetivli zero, 4, e8, m1, ta, ma |
| 648 | jal .Lidct4_add4_\depth\()_rvv | 648 | jal .Lidct4_add4_\depth\()_rvv |
| 649 | 649 | ||
| 650 | ld a4, 24(sp) # nnzc | 650 | ld a4, 24(sp) # nnzc |
| @@ -655,7 +655,7 @@ func ff_h264_idct4_add\type\()_\depth\()_rvv, zve32x · 4 unmodified lines | |||
| 655 | ld a0, 8(a0) # dest[1] | 655 | ld a0, 8(a0) # dest[1] |
| 656 | lla t0, ff_h264_scan8 + 32 | 656 | lla t0, ff_h264_scan8 + 32 |
| 657 | .ifc \type, 8_422 | 657 | .ifc \type, 8_422 |
| 658 | vsetivli zero, 4, e8, mf4, ta, ma | 658 | vsetivli zero, 4, e8, m1, ta, ma |
| 659 | jal .Lidct4_add4_\depth\()_rvv | 659 | jal .Lidct4_add4_\depth\()_rvv |
| 660 | 660 | ||
| 661 | ld a4, 24(sp) # nnzc | 661 | ld a4, 24(sp) # nnzc |
| @@ -665,7 +665,7 @@ func ff_h264_idct4_add\type\()_\depth\()_rvv, zve32x · 3 unmodified lines | |||
| 665 | addi a1, t5, (-8 - 4) * 4 # &block_offset[24] | 665 | addi a1, t5, (-8 - 4) * 4 # &block_offset[24] |
| 666 | ld a0, 0(a0) # dest[0] | 666 | ld a0, 0(a0) # dest[0] |
| 667 | lla t0, ff_h264_scan8 + 24 | 667 | lla t0, ff_h264_scan8 + 24 |
| 668 | vsetivli zero, 4, e8, mf4, ta, ma | 668 | vsetivli zero, 4, e8, m1, ta, ma |
| 669 | jal .Lidct4_add4_\depth\()_rvv | 669 | jal .Lidct4_add4_\depth\()_rvv |
| 670 | 670 | ||
| 671 | ld a4, 24(sp) # nnzc | 671 | ld a4, 24(sp) # nnzc |
| @@ -679,7 +679,7 @@ func ff_h264_idct4_add\type\()_\depth\()_rvv, zve32x · 7 unmodified lines | |||
| 679 | ld ra, 8(sp) | 679 | ld ra, 8(sp) |
| 680 | ld s0, 0(sp) | 680 | ld s0, 0(sp) |
| 681 | addi sp, sp, 32 | 681 | addi sp, sp, 32 |
| 682 | vsetivli zero, 4, e8, mf4, ta, ma | 682 | vsetivli zero, 4, e8, m1, ta, ma |
| 683 | j .Lidct4_add4_\depth\()_rvv | 683 | j .Lidct4_add4_\depth\()_rvv |
| 684 | endfunc | 684 | endfunc |
| 685 | .endm | 685 | .endm |