前言：

本筆記是基于對RISC-V ?DSP擴展指令集文檔總結的，《P-ext-proposal.pdf》文檔的關鍵內容如下：
主要介紹了RISC-V的P擴展指令集及其相關細節。

首先，對P擴展指令進行了概述，并列出了其與其他擴展重復的指令。

接著，詳細描述了P擴展的子集，包括Zbpbo擴展和Zpn擴展（適用于RV32和RV64）的指令。

此外，還提供了僅適用于RV64的詳細指令描述。
文檔還介紹了新的用戶控制和狀態寄存器，并提供了指令編碼表。最后，列出了因RVB重疊而被移除的指令。
這份文檔為RISC-V的P擴展指令集提供了全面而詳細的信息，包括指令的描述、編碼、以及與其他擴展的關系。這對于理解、開發和優化基于RISC-V架構的系統非常有價值。同時，文檔也提醒了開發者在使用P擴展時需要注意的兼容性和優化問題。

【RISC-V 指令集】RISC-V DSP 擴展指令集介紹(一)-CSDN博客

3.2. 部分SIMD數據處理指令

3.2.1. 16位組包指令

Table 12. 16-bit Packing Instructions

序號	指令	說明
1	PKBB16 rd, rs1, rs2	Pack two 16-bit data from Bottoms
2	PKBT16 rd, rs1, rs2	Pack two 16-bit data Bottom & Top
3	PKTB16 rd, rs1, rs2	Pack two 16-bit data Top & Bottom
4	PKTT16 rd, rs1, rs2	Pack two 16-bit data from Tops

3.2.2. ?最高有效字“32x32”乘法和加法指令

Table 13. Signed MSW 32x32 Multiply and Add Instructions

序號	指令	說明
1	SMMUL rd, rs1, rs2	MSW “32 x 32” Signed Multiplication (MSW 32 = 32x32)
2	SMMUL.u rd, rs1, rs2	MSW “32 x 32” Signed Multiplication with Rounding (MSW 32 = 32x32)
3	KMMAC rd, rs1, rs2	MSW “32 x 32” Signed Multiplication and Saturating Addition (MSW 32 = 32 + 32x32)
4	KMMAC.u rd, rs1, rs2	MSW “32 x 32” Signed Multiplication and Saturating Addition with Rounding (MSW 32 = 32 + 32x32)
5	KMMSB rd, rs1, rs2	MSW “32 x 32” Signed Multiplication and Saturating Subtraction (MSW 32 = 32 - 32x32)
6	KMMSB.u rd, rs1, rs2	MSW “32 x 32” Signed Multiplication and Saturating Subtraction with Rounding (MSW 32 = 32 - 32x32)
7	KWMMUL rd, rs1, rs2	MSW “32 x 32” Signed Multiplication & Double (MSW 32 = 32x32 << 1)
8	KWMMUL.u rd, rs1, rs2	MSW “32 x 32” Signed Multiplication & Double with Rounding (MSW 32 = 32x32 << 1)

3.2.3. ?最高有效字“32x16”乘法和加法指令

Table 14. Signed MSW 32x16 Multiply and Add Instructions

序號	指令	說明
1	SMMWB rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication (MSW 32 = 32x16)
2	SMMWB.u rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication with Rounding (MSW 32 = 32x16)
3	SMMWT rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication (MSW 32 = 32x16)
4	SMMWT.u rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication with Rounding (MSW 32 = 32x16)
5	KMMAWB rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication and Saturating Addition (MSW 32 = 32 + 32x16)
6	KMMAWB.u rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication and Saturating Addition with Rounding (MSW 32 = 32 + 32x16)
7	KMMAWT rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication and Saturating Addition (MSW 32 = 32 + 32x16)
8	KMMAWT.u rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication and Saturating Addition with Rounding (MSW 32 = 32 + 32x16)
9	KMMWB2 rd, rs1, rs2	MSW “32 x Bottom 16” Saturating Signed Multiplication and double (MSW 32 = (32x16) << 1)
10	KMMWB2.u rd, rs1, rs2	MSW “32 x Bottom 16” Saturating Signed Multiplication and double with Rounding (MSW 32 = (32x16) << 1)
11	KMMWT2 rd, rs1, rs2	MSW “32 x Top 16” Saturating Signed Multiplication and double (MSW 32 = (32x16) << 1)
12	KMMWT2.u rd, rs1, rs2	MSW “32 x Top 16” Saturating Signed Multiplication and double with Rounding (MSW 32 = (32x16) << 1)
13	KMMAWB2 rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication & double and Saturating Addition (MSW 32 = 32 + (32x16)<<1)
14	KMMAWB2.u rd, rs1, rs2	MSW “32 x Bottom 16” Signed Multiplication & double and Saturating Addition with Rounding (MSW 32 = 32 + (32x16)<<1)
15	KMMAWT2 rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication & double and Saturating Addition (MSW 32 = 32 + (32x16)<<1)
16	KMMAWT2.u rd, rs1, rs2	MSW “32 x Top 16” Signed Multiplication & double and Saturating Addition with Rounding (MSW 32 = 32 + (32x16)<<1)

3.2.4. 帶32位加/減法的有符號16位乘法指令

Table 15. Signed 16-bit Multiply 32-bit Add/Subtract Instructions

序號	指令	說明
1	SMBB16 rd, rs1, rs2	Signed Multiply Bottom 16 & Bottom 16 (32 = 16x16)
2	SMBT16 rd, rs1, rs2	Signed Multiply Bottom 16 & Top 16 (32 = 16x16)
3	SMTT16 rd, rs1, rs2	Signed Multiply Top 16 & Top 16 (32 = 16x16)
4	KMDA rd, rs1, rs2	Two “16x16” and Signed Addition (32 = 16x16 + 16x16)
5	KMXDA rd, rs1, rs2	Two Crossed “16x16” and Signed Addition (32 = 16x16 + 16x16)
6	SMDS rd, rs1, rs2	Two “16x16” and Signed Subtraction (32 = 16x16 - 16x16)
7	SMDRS rd, rs1, rs2	Two “16x16” and Signed Reversed Subtraction (32 = 16x16 - 16x16)
8	SMXDS rd, rs1, rs2	Two Crossed “16x16” and Signed Subtraction (32 = 16x16 - 16x16)
9	KMABB rd, rs1, rs2	“Bottom 16 x Bottom 16” with 32-bit Signed Addition (32 = 32 + 16x16)
10	KMABT rd, rs1, rs2	“Bottom 16 x Top 16” with 32-bit Signed Addition (32 = 32 + 16x16)
11	KMATT rd, rs1, rs2	“Top 16 x Top 16” with 32-bit Signed Addition (32 = 32 + 16x16)
12	KMADA rd, rs1, rs2	Two “16x16” with 32-bit Signed Double Addition (32 = 32 + 16x16 + 16x16)
13	KMAXDA rd, rs1, rs2	Two Crossed “16x16” with 32-bit Signed Double Addition (32 = 32 + 16x16 + 16x16)
14	KMADS rd, rs1, rs2	Two “16x16” with 32-bit Signed Addition and Subtraction (32 = 32 + 16x16 - 16x16)
15	KMADRS rd, rs1, rs2	Two “16x16” with 32-bit Signed Addition and Reversed Subtraction (32 = 32 + 16x16 - 16x16)
16	KMAXDS rd, rs1, rs2	Two Crossed “16x16” with 32-bit Signed Addition and Subtraction (32 = 32 + 16x16 - 16x16)
17	KMSDA rd, rs1, rs2	Two “16x16” with 32-bit Signed Double Subtraction (32 = 32 - 16x16 - 16x16)
18	KMSXDA rd, rs1, rs2	Two Crossed “16x16” with 32-bit Signed Double Subtraction (32 = 32 - 16x16 - 16x16)

3.2.5. 帶64位加/減法的有符號16位乘法指令

Table 16. Signed 16-bit Multiply 64-bit Add/Subtract Instructions

序號	指令	說明
1	SMAL rd, rs1, rs2	“16 x 16” with 64-bit Signed Addition (64 = 64 + 16x16)

3.2.6. 其他指令

Table 17. Partial-SIMD Miscellaneous Instructions

序號	指令	說明
1	SCLIP32 rd, rs1, imm5u	Signed Clip Value
2	UCLIP32 rd, rs1, imm5u	Unsigned Clip Value
3	CLRS32 rd, rs1	32-bit Count Leading Redundant Sign
4	CLZ32 rd, rs1	32-bit Count Leading Zero
5	PBSAD rd, rs1, rs2	Parallel Byte Sum of Absolute Difference
6	PBSADA rd, rs1, rs2	Parallel Byte Sum of Absolute Difference Accumulation

3.2.7. 帶32位加法的8位乘法指令

Table 18. 8-bit Multiply with 32-bit Add Instructions

序號	指令	說明
1	SMAQA rd, rs1, rs2	Four signed “8x8” with 32-bit Signed Addition (32 = 32 + 8x8 + 8x8 + 8x8 + 8x8)
2	UMAQA rd, rs1, rs2	Four unsigned “8x8” with 32-bit Unsigned Addition (32 = 32 + 8x8 + 8x8 + 8x8 + 8x8)
3	SMAQA.SU rd, rs1, rs2	Four “signed 8 x unsigned 8” with 32- bit Signed Addition (32 = 32 + 8x8 + 8x8 + 8x8 + 8x8)

3.3?64位數據計算指令

3.3.1 64位加減指令

Table 19. 64-bit Add/Subtract Instructions

序號	指令	說明
1	ADD64 rd, rs1, rs2	64-bit Addition
2	RADD64 rd, rs1, rs2	64-bit Signed Halving Addition
3	URADD64 rd, rs1, rs2	64-bit Unsigned Halving Addition
4	KADD64 rd, rs1, rs2	64-bit Signed Saturating Addition
5	UKADD64 rd, rs1, rs2	64-bit Unsigned Saturating Addition
6	SUB64 rd, rs1, rs2	64-bit Subtraction
7	RSUB64 rd, rs1, rs2	64-bit Signed Halving Subtraction
8	URSUB64 rd, rs1, rs2	64-bit Unsigned Halving Subtraction
9	KSUB64 rd, rs1, rs2	64-bit Signed Saturating Subtraction
10	UKSUB64 rd, rs1, rs2	64-bit Unsigned Saturating Subtraction

3.3.2 32位乘法與64位加減指令

Table 20. 32-bit Multiply 64-bit Add/Subtract Instructions

序號	指令	說明
1	SMAR64 rd, rs1, rs2	32x32 with 64-bit Signed Addition
2	SMSR64 rd, rs1, rs2	32x32 with 64-bit Signed Subtraction
3	UMAR64 rd, rs1, rs2	32x32 with 64-bit Unsigned Addition
4	UMSR64 rd, rs1, rs2	32x32 with 64-bit Unsigned Subtraction
5	KMAR64 rd, rs1, rs2	32x32 with Saturating 64-bit Signed Addition
6	KMSR64 rd, rs1, rs2	32x32 with Saturating 64-bit Signed Subtraction
7	UKMAR64 rd, rs1, rs2	32x32 with Saturating 64-bit Unsigned Addition
8	UKMSR64 rd, rs1, rs2	32x32 with Saturating 64-bit Unsigned Subtraction

3.3.3 帶符號16位乘法與64位加減指令

Table 21. Signed 16-bit Multiply 64-bit Add/Subtract Instructions

序號	指令	說明
1	SMALBB rd, rs1, rs2	“Bottom 16 x Bottom 16” with 64-bit Signed Addition (64 = 64 + 16x16)
2	SMALBT rd, rs1, rs2	“Bottom 16 x Top 16” with 64-bit Signed Addition (64 = 64 + 16x16)
3	SMALTT rd, rs1, rs2	“Top 16 x Top 16” with 64-bit Signed Addition (64 = 64 + 16x16)
4	SMALDA rd, rs1, rs2	Two “16x16” with 64-bit Signed Double Addition (64 = 64 + 16x16 + 16x16)
5	SMALXDA rd, rs1, rs2	Two Crossed “16x16” with 64-bit Signed Double Addition (64 = 64 + 16x16 + 16x16)
6	SMALDS rd, rs1, rs2	Two “16x16” with 64-bit Signed Addition and Subtraction (64 = 64 + 16x16 - 16x16)
7	SMALDRS rd, rs1, rs2	Two “16x16” with 64-bit Signed Addition and Reversed Subtraction (64 = 64 + 16x16 - 16x16)
8	SMALXDS rd, rs1, rs2	Two Crossed “16x16” with 64-bit Signed Addition and Subtraction (64 = 64 + 16x16 - 16x16)
9	SMSLDA rd, rs1, rs2	Two “16x16” with 64-bit Signed Double Subtraction (64 = 64 - 16x16 - 16x16)
10	SMSLXDA rd, rs1, rs2	Two Crossed “16x16” with 64-bit Signed Double Subtraction (64 = 64 - 16x16 - 16x16)

3.4?非SIMD指令

3.4.1 Q15飽和說明

Table 22. Non-SIMD Q15 saturation ALU Instructions

序號	指令	說明
1	KADDH rd, rs1, rs2	Add with Q15 saturation
2	KSUBH rd, rs1, rs2	Subtract with Q15 saturation
3	KHMBB rd, rs1, rs2	Multiply the first 16- bit Q15 elements of two registers and transform the Q30 result into a saturated Q15 number.
4	KHMBT rd, rs1, rs2	Multiply the first 16- bit Q15 element of one register with the second 16-bit Q15 element of another register and transform the Q30 result into a saturated Q15 number.
5	KHMTT rd, rs1, rs2	Multiply the second 16-bit Q15 elements of two registers and transform the Q30 result into a saturated Q15 number.
6	UKADDH rd, rs1, rs2	Add with I16 saturation
7	UKSUBH rd, rs1, rs2	Subtract with I16 saturation

3.4.2 Q31飽和指令

Table 23. Non-SIMD Q31 saturation ALU Instructions

序號	指令	說明
1	KADDW rd, rs1, rs2	Add with Q31 saturation
2	UKADDW rd, rs1, rs2	Unsigned Add with U32 saturation
3	KSUBW rd, rs1, rs2	Subtract with Q31 saturation
4	UKSUBW rd, rs1, rs2	Unsigned Subtract with U32 saturation
5	KDMBB rd, rs1, rs2	Multiply the first 16- bit Q15 elements of two registers and transform the Q30 result into a saturated Q31 number.
6	KDMBT rd, rs1, rs2	Multiply the first 16- bit Q15 element of one register with the second 16-bit Q15 element of another register and transform the Q30 result into a saturated Q31 number.
7	KDMTT rd, rs1, rs2	Multiply the second 16-bit Q15 elements of two registers and transform the Q30 result into a saturated Q31 number.
8	KSLRAW rd, rs1, rs2	Shift Left Logical with Q31 Saturation or Shift Right Arithmetic
9	KSLRAW.u rd, rs1, rs2	Shift Left Logical with Q31 Saturation or Rounding Shift Right Arithmetic
10	KSLLW rd, rs1, rs2	Saturating Shift Left Logical for 32-bit Word
11	KSLLIW rd, rs1, imm5u	Saturating Shift Left Logical Immediate for 32-bit Word
12	KDMABB rd, rs1, rs2	Multiply the first 16- bit Q15 elements of two registers and transform the Q30 result into a saturated Q31 number. Add the Q31 number with a 32-bit accumulator.
13	KDMABT rd, rs1, rs2	Multiply the first 16- bit Q15 element of one register with the second 16-bit Q15 element of another register and transform the Q30 result into a saturated Q31 number. Add the Q31 number with a 32-bit accumulator.
14	KDMATT rd, rs1, rs2	Multiply the second 16-bit Q15 elements of two registers and transform the Q30 result into a saturated Q31 number. Add the Q31 number with a 32-bit accumulator.
15	KABSW rd, rs1	32-bit Absolute Value (scalar version)

3.4.3 32位計算指令

Table 24. 32-bit Computation Instructions

序號	指令	說明
1	RADDW rd, rs1, rs2	32-bit Signed Halving Addition
2	URADDW rd, rs1, rs2	32-bit Unsigned Halving Addition
3	RSUBW rd, rs1, rs2	32-bit Signed Halving Subtraction
4	URSUBW rd, rs1, rs2	32-bit Unsigned Halving Subtraction
5	MULR64 rd, rs1, rs2	Multiply Word Unsigned to 64-bit data
6	MULSR64 rd, rs1, rs2	Multiply Word Signed to 64-bit data
7	MSUBR32 rd, rs1, rs2	Multiply and Subtract from 32-bit Word

3.4.4 溢流/飽和狀態操作說明

Table 25. OV (Overflow) flag Set/Clear Instructions

序號	指令	說明
1	RDOV rd	Read vxsat.OV to rd.
2	CLROV	Clear vsat.OV flag

3.4.5 其他指令

Table 22. Non-SIMD Q15 saturation ALU Instructions

序號	指令	說明
1	AVE rd, rs1, rs2	Average with rounding
2	SRA.u rd, rs1, rs2	Rounding Shift Right Arithmetic
3	SRAI.u rd, rs1, imm5u/imm6u	Rounding Shift Right Arithmetic Immediate
4	BITREV rd, rs1, rs2	Bit Reverse
5	BITREVI rd, rs1, imm5u/imm6u	Bit Reverse Immediate
6	WEXT rd, rs1, rs2	Extract 32-bit from a 64-bit value
7	WEXTI rd, rs1, imm5u	Extract 32-bit from a 64-bit value Immediate
8	CMIX rd, rs2, rs1, rs3	Conditional Mix
9	INSB rd, rs1, imm2u/imm3u	Insert Byte
10	MADDR32 rd, rs1, tb	Multiply and Add to 32-bit Word
11	MSUBR32 rd, rs1, tb	Multiply and Subtract from 32-bit Word
12	MAX rd, rs1, rs2	Signed Word Maximum
13	MIN rd, rs1, rs2	Signed Word Minimum

?RISC-V? DSP擴展指令集文檔：

https://download.csdn.net/download/u011376987/88898800